Seeed Studio Grove Air Quality Sensor BME AI-Studio Manual

Download

BME AI-Studio Manual

Document revision 1.6.0

Document release date June 2021

Document number BST-BME688-AN001-00

Sales Part Number 0 273 017 016

Notes

Data and descriptions in this document are subject to change without notice. Product photos and pictures are for illustration purposes only and may diﬀer from the real product appearance.

Bosch Sensortec BME AI-Studio Manual

Table of Contents

1 Overview 6

1.1 Getting started . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

1.1.1 The BME688 Gas Sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

1.1.2 Example: Coﬀee vs. Normal Air . . . . . . . . . . . . . . . . . . . . . . . . . . 6

1.1.3 A Few Things To Keep In Mind . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

1.1.4 Step 1: Record Normal Air . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

1.1.5 Step 2: Record Espresso Coﬀee . . . . . . . . . . . . . . . . . . . . . . . . . . 7

1.1.6 Step 3: Record Normal Air Again . . . . . . . . . . . . . . . . . . . . . . . . . . 8

1.1.7 Step 4: Record Filter Coﬀee . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

1.1.8 Step 5: Import & Label The Data . . . . . . . . . . . . . . . . . . . . . . . . . . 8

1.1.9 Step 6: Create New Algorithm and Classes . . . . . . . . . . . . . . . . . . . . 9

1.1.10 Step 7: Train And Evaluate The Algorithm . . . . . . . . . . . . . . . . . . . . . 9

1.1.11 Step 8: Export The Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1.1.12 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

1.2.1 What is it about? – An analogy . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

1.2.2 Why the BME688? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

1.2.3 What is a use case for a gas sensor? . . . . . . . . . . . . . . . . . . . . . . . 13

1.2.4 What is special about the BME688 gas sensor? . . . . . . . . . . . . . . . . . 13

1.2.5 How can I evaluate BME688 performance for a speciﬁc use cases? . . . . . . 14

1.2.6 Phase I: Doing the ﬁrst measurements . . . . . . . . . . . . . . . . . . . . . . . 14

1.2.7 Phase II: Heater Proﬁle Exploration – the space of possible heater proﬁles . . 15

1.2.8

1.2.9 How can I use the results for my product development? . . . . . . . . . . . . . 16

1.3 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

1.3.1 Sensor Board . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

1.3.2 Measurement Session . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

1.3.3 Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

Phase III: Duty Cycle Exploration – balancing power consumption and perfor-

mance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2 Process Steps 23

2.1 Conﬁgure Board . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

2.1.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

2.1.2 Board Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

2.1.3 Board Mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

2.1.4 Heater Proﬁle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

2.1.5 Duty Cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

2.1.6 Board Layout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

2.2 Record Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

2.2.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

2.2.2 Start recording . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

2.2.3 During recording . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

2.2.4 End recording . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

2.3 Import Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

2.3.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

2.3.2 Data Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

2.3.3 Board ID . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

2.3.4 Board Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

2.3.5 Board Mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

2.3.6 Session Name . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

2.3.7 Session Date . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

2.3.8 Specimen Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

2.4 Collect Specimens . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

2.4.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

2.4.2 Label . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

2.4.3 Comment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

2.4.4 Session . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

2.4.5 Start & End Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

2.4.6 Duration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

2.4.7 Cycles Total . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

2.4.8 Cycles Dropped . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

2.4.9 Remaining Cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

2.4.10 Board Conﬁguration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

2.4.11 Board ID . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

2.4.12 Board Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

2.4.13 Board Mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

2.4.14 Show Conﬁguration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

2.5 Train Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

2.5.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

2.5.2 Name . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

2.5.3 Created . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

2.5.4 Classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

2.5.5 Class Name & Color . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

2.5.6 Common Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

2.5.7 Data Balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

2.5.8 Data Channels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

2.5.9 Neural Net . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

2.5.10 Training Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

2.5.11 Max. Training Rounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

2.5.12 Data Splitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

2.6 Evaluate Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

2.6.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

2.6.2 Confusion Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

2.6.3 Accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

2.6.4 Macro-averaged F1 Score . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

2.6.5 Macro-averaged False Positive rate . . . . . . . . . . . . . . . . . . . . . . . . 50

2.6.6 Training Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

2.6.7 Test Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

2.6.8 BSEC Export . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

2.6.9 Additional Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

3 Technical Speciﬁcation 52

3.1 Raw Data Format . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

3.1.1 File Name . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

3.1.2 Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

3.1.3 conﬁgHeader . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

3.1.4 conﬁgBody . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

3.1.5 rawDataHeader . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

3.1.6 rawDataBody . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

3.1.7 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

3.2 Specimen Data Format . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

3.2.1 File Name . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

3.2.2 Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

3.2.3 Specimen Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

3.2.4 measurementSession . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

3.2.5 boardConﬁg . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

3.2.6 boardType . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

3.2.7 heaterProﬁles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

3.2.8 dutyCycleProﬁles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

3.2.9 sensorConﬁgs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

3.2.10 sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

3.2.11 cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

3.2.12 dataColumns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

3.2.13 specimenDataPoints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

3.2.14 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

List of Tables

1 An analogy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2 Steps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

3 Use Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

4 File Name . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

5 conﬁg Header . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

6 conﬁg Body . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

7 Raw Data Header . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

8 File Name . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

9 Specimen Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

10 measurementSession . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

11 boardConﬁg . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

12 boardType . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

13 heaterProﬁles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

14 dutyCycleProﬁles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

15 sensorConﬁgs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

16 sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

17 cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

List of Figures

1 BME688 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2 Use case Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3 Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

4 BME688 Board . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

5 Heater Proﬁle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

6 Duty Cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

7 Measurement Session . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

8 BME board x8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

9 BME board x8 Back . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

10 Inserting SD Card . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

11 Start Recording . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

12 During Recording . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

13 End Recording . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

14 Start Recording . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

15 Specimen Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

16 Train Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

17 Confusion Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

18 Macro-averaged F1 Score . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

1 Overview

1.1 Getting started

Welcome to our “Getting Started” section of the Bosch BME AI-Studio. The BME AI-Studio brings together our BME Board x8 (equipped with the new BME688 gas sensor) and our BSEC Software (BME Library).

This section will give you a short introduction and a hands-on example on how to use the application.

Click to Watch Getting Started Video

1.1.1 The BME688 Gas Sensor

See: "0:25 - The BME688 Gas Sensor"

The BME688 gas sensor is like a digital nose. It can detect gases by measuring their unique electric ﬁngerprint and therefore distinguish diﬀerent gas compositions.

However, the sensor has to be teached about these diﬀerent gases ﬁrst. That means, we have to train the sensor using Machine Learning.

And that’s where the BME AI-Studio comes in: It let’s you explore and verify your speciﬁc use cases: With the application, you can collect your gas measurements, train Machine Learning algorithms and export a ﬁnal algorithm to be used with the BSEC Software (BME Library) in your project.

1.1.2 Example: Coﬀee vs. Normal Air

See: "1:03 - Example: Coﬀee vs. Normal Air"

Let’s take a look at a real world example.

Assume, you want to use the BME688 to detect coﬀee beans. Or in other words: distinguish the gas composition of coﬀee from the gas composition of normal air and other smells.

So in order to train an algorithm for that, we ﬁrst have to record gas measurements of coﬀee and normal air.

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

1.1.3 A Few Things To Keep In Mind

See: "1:26 - A Few things To Keep In Mind"

Before we start with the measurements, there are three things to keep in mind:

If you have a brand new BME Board, you should stabilize the sensors on the board before taking any measurements. To do so, connect the board to power with a Micro USB cable and let it run for at least 24 hours. This procedure is neccessary only once and your board is then ready to take reliable measurents.

Make sure that your BME Board is conﬁgured correctly. New BME Boards come already preconﬁgured with a standard setting – so if you have a brand new board, you don’t have to worry about the conﬁguration for now.

· Make sure that an SD card is inserted in the board.

1.1.4 Step 1: Record Normal Air

See: "2:08 - Step 1: Record Normal Air"

OK – let’s start the Measurement Session for our Coﬀee Detection Example.

Let’s record normal air as our ﬁrst Specimen. We are just going to record the gas composition of the room. Additional smells to make the detection more robust should be added later. But to keep it simple in this example we just stick to normal air for now.

Connect the BME Board to power and place the board on your table. Always be careful not to touch the sensors with your ﬁngers while handling the board.

If everything is OK and the board is recording data, the red LED is ﬂashing every two seconds. Let the board record for about 30 minutes.

1.1.5 Step 2: Record Espresso Coﬀee

See: "2:48 - Step 2: Record Espresso Coﬀee"

Now, let’s move on to the coﬀee. For this example, we prepared two types of coﬀee in order to have a bit of variance in our measurements: Espresso and ﬁlter coﬀee. Let’s measure the espresso coﬀee ﬁrst.

Place the espresso coﬀee in a relatively air-tight container. Press “Button 1” (labeled “S1”) on the board to mark this moment in the data. Place the board in the container together with the coﬀee, and record the gas composition for 30 minutes.

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

1.1.6 Step 3: Record Normal Air Again

See: "3:19 - Step 3: Record Normal Air Again"

After that, let’s measure normal air again. Take the board out of the container. This time, press

“Button 2” (labeled “S2”) on the board to set another marker in the data.

Place the board again on your table, and record the gas composition for another 30 minutes. Put the container with the coﬀee far away, so it doesn’t interfere with the recording.

1.1.7 Step 4: Record Filter Coﬀee

See: "3:36 - Step 4: Record Filter Coﬀee"

Finally, let’s record the last Specimen: Filter coﬀee. Same procedure as with the espresso coﬀee: Place the ﬁlter coﬀee in a second container, press “Button 1” on the board and place it in the container along with the coﬀee. Again, record the gas composition for 30 minutes.

Once you are ﬁnished with your Measurement Session, you can just power oﬀ the board by unplugging the USB cable.

1.1.8 Step 5: Import & Label The Data

See: "4:04 - Step 5: Import And Label The Data"

Now it’s time to open BME AI-Studio on your computer. Create a new project and select a location, where you want to save it on your hard drive. Let’s call the project “Coﬀee Example”. Then select

Import data

. Remove the SD card from the BME Board and select the ﬁle on the card with the

ending .bmerawdata .

The application now gives you an overview of your recorded data. There’s a plot showing you the data on a timeline. First, give your Measurement Session a name. Let’s call it “Coﬀee Session”.

As you can see below the plot, there are already four Specimens deﬁned – one that started with powering on the board, and three that started when we pressed the buttons during measurment. On the left, there are input ﬁelds to label each Specimen. We measured normal air ﬁrst, then espresso coﬀee, then again normal air, and ﬁnally ﬁlter coﬀee.

If you forgot to press a button during the measurement, you could deﬁne Specimens manually here as well.

Select Import Data to ﬁnish. The data is now copied to the project folder on your hard drive.

Once the import is ﬁnished, you can see your four Specimens in your

Specimen Collection

. The

Specimen Collection is where you can collect and manage all of your recorded data for a speciﬁc

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

project. If you select one of the Specimens, you can see all available details about this Specimen on the right hand side.

1.1.9 Step 6: Create New Algorithm and Classes

See: "5:37 - Step 6: create New Algorithm And Classes"

So now that we have recorded and imported the data into the Specimen Collection, it is ﬁnally time to create and train an algorithm.

Switch to

My Algorithms

and select

New algorithm

. We can give this algorithm a name, let’s call it

“Coﬀee or Not”.

As you remember, in our example use case we wanted to detect when coﬀee is present. That means, we need to teach our algorithm the diﬀerence between two so called “classes”: Coﬀee and other smells. So let’s create these two classes. Select

Add class

and give this class a name. Let’s call it

“Coﬀee”.

Next, we have to select the Specimens from our Specimen Collection that should be part of that class. In our example, that would be the two Specimens “Espresso Coﬀee” and “Filter Coﬀee”.

Now, let’s create the second class. We call this one “Normal Air” and select the two “Normal Air” Specimens to be associated with this class.

1.1.10 Step 7: Train And Evaluate The Algorithm

See: "6:38 - Step 7: Train And Evaluate The Algorithm"

Now, that we have deﬁned our two classes with their respective Specimens, we can train our algorithm. There are additional settings for the algorithm and the training, but for now you can leave these settings at their default values. Just hit

Train Neural Net

and BME AI-Studio will start to train

the algorithm.

Once the training is ﬁnished, you can see the results. The most important value to check is the

Accuracy . The higher the Accuracy the better the algorithm’s performance.

One thing you should be aware of though: The algorithm training results can be too perfect. If you have recorded too little data, it might happen that the algorithm adapts too closely to that speciﬁc data – and gives perfect results. However, it won’t be able to predict other data that is slightly diﬀerent. To make the algorithm more robust, you could record additional coﬀee blends or other smells, like tea or juices. In short: always make sure to record enough data with a bit of variance.

A second thing to check is the so called

Confusion Matrix

, which you will ﬁnd under

Details

. It

shows the algorithm’s performance to distinguish our two classes “Coﬀee” and “Normal Air”.

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

A part of the Specimen’s data has been reserved just for this perfomance test. The rows show what the actual classes of these test elements were, while the columns show which classes were predicted by the algorithm. That means, for a good result, the upper left quadrant and the lower right quadrant should be dark blue and have a high number – indicating how often the algorithm made the right prediction for both classes. At the same time, the upper right and lower left quadrants should be light blue and have a low number, since these numbers indicate how often the algorithm made the wrong prediction. The upper right quadrant tells you how often the algorithm falsely predicted “Normal Air”, when in fact it was “Coﬀee”. And the lower left quadrant tells you how often the algorithm predicted “Coﬀee”, when it actually was “Normal Air”.

In our example, the results look pretty good. More than 90% accuracy and very few false predictions in the Confusion Matrix. That’s a good result for a ﬁrst training.

1.1.11 Step 8: Export The Algorithm

See: "8:55 - Step 8: Export The Algorithm"

As a last step, if you are happy with the result, you can export the algorithm as a

BSEC Conﬁguration File

for the BSEC Software (BME Library). This conﬁg ﬁle can be used

together with the BME Library (BSEC) in your project.

1.1.12 Conclusion

See: "9:09 - Conclusion"

If you have any questions or if you want to get further information, for example on how to use the BSEC Software (BME Library), please write us an email or check out the Bosch Sensortec Community.

Happy training!

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

1.2 Introduction

The combination of BME AI-Studio with the BME Board allows you to explore the great possibilities of the BME688 and the BSEC Software (BME Library). It helps you validate gas sensor use cases for your product or application. Try and ﬁnd out how well your gas sensing application can be fulﬁlled by the BME688.

1.2.1 What is it about? – An analogy

In a nutshell the BME AI-Studio can be explained within the following analogy:

Using the BME AI-Studio can be compared to creating a picture book for children. Within that analogy training an algorithm can be compared to teaching a child how diﬀerent animals look like. This analogy should help you grasp the essence of the BME AI-Studio with it’s two major parts: the

Specimen Collection and My Algorithms .

BME world Child analogy

Goal: Teaching the sensor to distinguish

diﬀerent gas compositions with labeled

Specimen Data, recorded with the BME

Board

Specimen Collection: Collection of

recorded Measurement Sessions and

labeled Specimens

My Algorithms: Machine learning process to teach the algorithm which gases are which (or which classes of

gasses are which), so that they can later

be distinguished by the sensor

Example Algorithms: A trained

algorithm that can distinguish apples

from oranges. This works because

apples and oranges produce diﬀerent

gases, which produce diﬀerent sensor

signals. The algorithm was trained to

detect these diﬀerences and is now

able to classify the two

Goal: Teaching a child to distinguish diﬀerent animals

with a picture book, that has photos captured with

your photo camera

Picture book: Collection of photos labeled with the

name of the animals

Teaching a child: Process of showing the labeled photos to a child to teach him/her which animals are which, so that they can later be distinguished by the

child

Example: A child that can distinguish giraﬀes from

elephants. This works because giraﬀes and

elephants have diﬀerent appearances, which look

diﬀerent on a photograph. The child learned to detect

these diﬀerences and is now able to classify the two

Table 1: An analogy

Within that analogy the individual steps are:

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

BME AI-Studio BME board Photo software Camera

1 Conﬁgure the BME Board with BME

2 Record data with the BME Board Take photos with the camera

3 Import data into the BME AI-Studio Import photos into photo editing

4 Label diﬀerent Specimens within the data Label diﬀerent animals in the

5 Train algorithm with labeled Specimens Teach a child with labeled photos

6 Evaluate and test algorithm Ask the child what he/she has

1.2.2 Why the BME688?

Choose the settings in the camera

AI-Studio

software

photos

learned

Table 2: Steps

Fig. 1: BME688

With the BME688 gas sensor you experience the world’s most advanced gas sensor with unique capabilities:

High sensitivity and selectivity towards various gases (a variety of gases have already been studied in the lab of Bosch Sensortec. For detailed information please refer to the

Bosch Sensortec Website or the Bosch Sensortec Community)

· Sensitivity and selectivity can be adjusted during operation

· Low power consumption

· Small footprint

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

· Additional sensing capabilities for temperature, relative humidity and barometric pressure

With this unique capabilities the BME688 gas sensor can be applied to numerous use cases. A gas sensor can be regarded as a digital nose, which has a lot of advantages if compared to the human nose. The following table may inspire you to ﬁnd unique use cases:

Human nose Sensor cabability Digital nose

attached to the body location independent extended nose

needs recovery time always ready for use hardworking nose

blind for slow and steady

reproducible precision absolute nose

changes

sensitive to special gases sensitive for diﬀerent gases complementary nose

individual sensitivity calibrated sensitivity objective nose

highly emotional sober and rational not my own nose

Table 3: Use Cases

1.2.3 What is a use case for a gas sensor?

The magic lies within the question itself and it is up to your explorations to fully cover the possible space of use cases for the BME688. This is exactly where the BME AI-Studio is aiming for.

A use case is a potential application for the BME688. From a gas sensing perspective it is the ability to classify diﬀerent gas compositions (with diﬀerent combinations and concentrations of gases) that occur in diﬀerent situations. The goal is to ﬁnd a conﬁguration in which the sensor is able to reliably distinguish the diﬀerent gas compositions.

Interestingly, it is not necessarily important to know the exact composition of gases, as long as you have deﬁned speciﬁc cases and you are able to reproduce these cases and their corresponding gas compositions.

A speciﬁc gas sensor use case may be a part of product premise you are aiming for. It can detect diﬀerent situations and thus provide real value added to your product. Typically use cases are the detection of various pleasant or unpleasant smells or the detection of gases which are odorless.

1.2.4 What is special about the BME688 gas sensor?

The gas sensing element of the BME688 is mounted on a micro-hotplate, so its temperature can be changed individually. This happens during

Scanning Cycles

. Depending on the

Heater Proﬁle (HP)

the sensor cycles through a list of 10 predeﬁned temperatures. The exact time sequence of these temperatures deﬁnes the surface chemistry of the gas sensing element and thus the sensitivity and selectivity of the sensor.

Therefore the

Heater Proﬁle

has a substantial inﬂuence on the performance of the gas sensor.

Choosing the right Heater Proﬁle for your use case can be the key to getting a good performance in

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

classifying diﬀerent situations.

These Heater Proﬁles are one of key advantages of the BME688. This allows the sensor to adopt to individual gases by adopting the measurement scheme to the gases’ individual ﬁngerprints.

1.2.5 How can I evaluate BME688 performance for a speciﬁc use cases?

The BME AI-Studio is the perfect tool to support you in the quest of ﬁnding out whether a speciﬁc use case can be tackled with the BME688 gas sensor. Basically there are the following degrees of freedom that you can vary within the testing conditions:

· Data recording and testing environment

· Heater Proﬁle conﬁguration

· Duty Cycle conﬁguration

· Algorithm training parameters

In general the process can be viewed as a careful exploration and an iterative approach. The following procedure has proven to be very eﬃcient:

Fig. 2: Use case Evaluation

1.2.6 Phase I: Doing the ﬁrst measurements

The best thing is to just start recording whatever use case you have in mind. On your way of exploring the use case more and more, you will gain knowledge and sometimes you have to redo some of the measurements. That’s very normal when exploring new use cases.

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

Please note

Fabric-new BME Boards need to undergo an initial stabilization procedure. Please let the board run for at least 24 hours before recording any meaningful data.

A factory-new BME Board comes preconﬁgured with the

RDC-5-10

. This means you can start right away capturing the ﬁrst data. As a more advanced user,

Heater Proﬁle HP-354

and the

Duty Cycle

you might want to conﬁgure the board with a selection of diﬀerent Heater Proﬁles.

Before starting the measurement it is wise to spend some thoughts on the right testing environment. This also means choosing how controlled your setup should be. The two kinds of measurements are:

· Lab recording

where the gas atmosphere is fully controlled. These measurements make it easy for the algorithm to distinguish between diﬀerent gas compositions and are therefore ideal to test ﬁrst. If the algorithm has a hard time distinguishing data from lab recordings, it is most likely even more diﬃcult to distinguish data later within the application. You may also want to think about certain nuisances to be added to the test to be more stable towards external inﬂuences later on.

· Field recordings

where changing backgrounds of the gas composition are not controlled but taken into account as a statistical ﬂuctuation. These changing conditions help the algorithm to get more robust. Try to get a variance of the situations that is a little bit worse than it will be later within the application. If the algorithm was trained in diﬃcult conditions, it will have an easier time later. However, do not go to extremes while changing the situations – otherwise you make the training process more complicated than necessary.

Once you ﬁnished recording, you can import the data. Build up a small collection of specimens and try an early training of an algorithm.

Evaluating the results

of the ﬁrst measurements the performance indicators should look very good. Small amounts of data should make it very easy for the algorithm to ﬁnd features that allow for distinguishing the diﬀerent gas compositions.

If you do not see good results, please go back and check whether previous steps are properly done. If so it may be the case that the sensitivity and selectivity of the BME688 does not match your use case. For example carbon dioxide is known to be hard to measure with the BME688.

If you see good results, please proceed with a Heater Proﬁle Exploration.

1.2.7 Phase II: Heater Proﬁle Exploration – the space of possible heater proﬁles

Within the Heater Proﬁle Exploration the goal is to ﬁnd the Heater Proﬁle that best suits your use case. Therefore try and

conﬁgure multiple Heater Proﬁles

and record data to compare the results of

diﬀerent Heater Proﬁles.

You may want to record as much data as possible during Heater Proﬁle Exploration by choosing the Duty Cycle

RDC-1-0 Continuous

, where the sensors are continuously recording data without any

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

Sleeping Cycles in between. You can record multiple sessions with multiple Specimens and import them all into your Specimen Collection.

When you train algorithms for data with multiple

HP/RDC Combinations

(Heater Proﬁle / Duty Cycle Combinations) an independent neural net will be trained for each instance, giving you the opportunity to compare the performance of diﬀerent HP/RDCs. By iterating that process in terms of conﬁguring the BME Board with diﬀerent Heater Proﬁles, recording data and then comparing the performances, you can determine the best suitable Heater Proﬁle for your speciﬁc use case.

After you found the best Heater Proﬁle, you may want to proceed with a Duty Cycle Exploration, if power consumption is an issue for your use case. If power consumption is irrelevant you can directly export your BSEC conﬁguration to be used the BSEC Software (BME Library).

1.2.8 Phase III: Duty Cycle Exploration – balancing power consumption and performance

A Duty Cycle Exploration is important if you want to optimize power consumption of the BME688 for your application. Therefore the purpose of Duty Cycle Exploration is to ﬁnd the best Duty Cycle for your use case – it is the one that uses a minimum amount of energy while still providing very good performance.

Try

conﬁguring multiple Duty Cycles

, while keeping the Heater Proﬁle that was perfoming best and that you determined within the Heater Proﬁle Exploration. While comparing diﬀerent Duty Cycles, keep in mind that the ratio of Scanning Cycles to Sleeping Cycles determines the average power consumption of the sensor. You ﬁnd an estimated power consumption within the

board conﬁguration

in your Specimen Collection.

Once you found your optimal Duty Cycle, you can export your

BSEC conﬁguration

to be used with

the BSEC Software (BME Library).

1.2.9 How can I use the results for my product development?

The BME AI-Studio with BME Board and the BME Library (BSEC) are build around the BME688 gas sensor. Together they form an ecosystem to explore, validate and prototype gas sensor use cases. With the BME Board you can record data to be used within the AI-Studio for training neural nets. These can be exported to be used with the BME Library (BSEC) and inside an embedded application.

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

Fig. 3: Integration

For any trained algorithm you can export a BSEC conﬁguration ﬁle that can be used with the BSEC Software (BME Library).

Please refer to the website of Bosch Sensortec BME Software for further information.

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

1.3 Glossary

1.3.1 Sensor Board

BME Board x8 / Development Kit

BME Board x8 (which is also referred to as BME Development Kit) is an experimental board with eight BME688 sensors.

Fig. 4: BME688 Board

BME Board Conﬁguration

The Board Conﬁguration states which sensor runs which sensor mode.

HP/RDC Combination

Each sensor can operate with a speciﬁc setting which consists of

· Heater Proﬁle (HP)

· Duty Cycle (RDC)

Hence, the term HP/RDC Combination.

Heater Proﬁle

Temperature proﬁle with 10 steps which is run through by the sensor during a Scanning Cycle. Each of the 10 steps consists of:

· A temperature

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

· A duration (timebase is 140 ms)

At the end of each step, a measurement is recorded and stored in the data.

Fig. 5: Heater Proﬁle

Please note

The duration of a step of a Heater Proﬁle is calculated with a timebase of 140 ms. That means e.g. a duration of 2 will result in a duration of 2 * 140 ms = 280 ms .

Duty Cycle

Repeated sequence of a deﬁned number of Scanning Cycles followed by a deﬁned number of Sleeping Cycles.

E.g. Duty Cycle with 5 Scanning Cycles and 10 Sleeping Cycles (RDC-5-10):

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

Fig. 6: Duty Cycle

Each Scanning Cycle can be viewed as one iteration of a Heater Proﬁle. Therefore the duration of each Scanning Cycle is derived from the duration of the Heater Proﬁle used. The duration of each Sleeping Cycle is always equal to the duration of a Scanning Cycle.

The duration of the whole Duty Cycle deﬁnes the algorithms response time.

Sleeping Cycle

During a Sleeping Cycle the sensor is in sleep mode and does not perform any measurements.

Scanning Cycles

During a Scanning Cycle the sensor uses the internal micro-hotplate to run through a sequence of temperature steps according to a speciﬁc Heater Proﬁle.

1.3.2 Measurement Session

Measurement Session

Data acquisition within one “session“ with one BME Board Conﬁguration. The Measurement Session begins with powering on the board and ends with powering oﬀ the board. Multiple sessions can be conducted one after another using the same SD card. Typically, multiple Specimens are measured during a Measurement Session.

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

Fig. 7: Measurement Session

Specimen

One speciﬁc gas composition that the sensors on the BME Board are exposed to during measurement (e.g. an object, a speciﬁc situation, a smell, gas, etc.).

Session Raw Data / Raw Data

The raw data recorded during a Measurement Session which is saved to the SD card. This includes data from all 8 sensors on the BME Board x8.

Session Raw Data is split into multiple ﬁles when board power is switched on/oﬀ or the ﬁle limit (297 MB) is exceeded.

Session Raw Data can be marked during measurement for easy splitting into Specimen Data by pressing one of the buttons S1 or S2 on the BME Board x8.

Specimen Data

A part (a set of cycles that belong to one Specimen) of the Session Raw Data of one Measurement Session. The Specimen Data can be curated and labeled with respect to the measured Specimen within the application.

Specimen Data includes four data channels (for each sensor on the board):

· Gas Data Channel (10 data points)

· Humidity Data Channel (1 data point)

· Temperature Data Channel (1 data point)

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

· Barometric Pressure Data Channel (1 data point)

1.3.3 Algorithm

Algorithm

A neural net for classiﬁcation, including multiple classes with added Specimens, as well as machine learning settings and results.

The Specimen Data of the added Specimens are split and used for training and veriﬁcation of the the neural net (Training and Test Data).

Training result consists of a Confusion Matrix and performance indicators, e.g. Accuracy.

Classes

Classes are the possible outcomes of the classiﬁcation algorithm. Each class has added Specimens. The corresponding Specimen Data is used for the classiﬁcation training and veriﬁcation (Training and Test Data).

Training and Test Data

The Specimen Data of all classes of a Algorithm is split into two parts:

· Training Data: Used to train the neural net

· Test Data: Used to test the performance of the neural net after training

BSEC Conﬁguration File

A conﬁguration ﬁle that can be exported from the application for each algorithm. It contains all parameters (weights and biases) of the trained neural net. It is meant to be used with the

BSEC Software

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

2 Process Steps

2.1 Conﬁgure Board

2.1.1 Overview

The ﬁrst step in the process of using a BME Board with the BME-AI Studio is to conﬁgure the board. This step prepares the boards with a proper conﬁguration for recording. The conﬁguration is ﬁrst deﬁned within the application and then saved as a conﬁguration ﬁle, that has to be transferred onto the SD card of the BME Board.

Good to know

If you have a factory-new BME Board and just want to get started, you can also skip this step. New boards are already conﬁgured with a default conﬁguration (HP-354 / RDC-5-10). In this case, check out our Getting started section.

Create a conﬁguration ﬁle

Open the application and create a new project by clicking screen. Alternatively, you can open the menu and select

Create new project

New project...

on the Welcome

. Next, a ﬁle dialogue opens.

Please select, where you want to save your new project.

If you already have an existing project, just open the project and switch to your

To set up a conﬁguration, click

Conﬁgure Board

. A new overlay opens that lets you deﬁne the

Specimen Collection

conﬁguration and save it as a conﬁguration ﬁle to the SD card.

Three types of conﬁgurations

The operation with the boards is divided into three categories, which are intented for diﬀerent phases with diﬀerent goals:

1. Default conﬁguration

BME Boards that are straight from the factory come with a default conﬁguration, that suits most use cases. This mode can also be used for an initial stabilization which is needed before you can use them for data recording.

2. Finding the right Heater Proﬁles

The power of adapting the BME688 gas sensor to a speciﬁc use case lies within the concept of Heater Proﬁles. Finding the right Heater Proﬁle that suits best for your use case is key to getting peak performance results. Therefore we advise you to spend some time ﬁnding the optimal Heater Proﬁle, which delivers the best results speciﬁcally for your use case. This involves sensitivity and selectivity. To explore a variety of Heater Proﬁles and ﬁnd the best one for your use case, please use the board mode Heater Proﬁle Exploration .

3. Optimizing the Duty Cycle

The BME688 allows you to adopt its mode of operation in order to save energy. During a

Sleeping Cycle

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

the sensor is in low-power mode and will be automatically woken up after a

Bosch Sensortec BME AI-Studio Manual

certain amount of time to enter a

Scanning Cycle

. The balance between sleeping and scanning is deﬁned by the Duty Cycle. To explore a variety of Duty Cycles and ﬁnd the best one for your use case, please use the board mode Duty Cycle Exploration .

It is important to pay attention to individual combinations of Heater Proﬁles (HP) and Duty Cylces (RDC), so called

HP/RDC combinations

. Each sensor runs in a speciﬁc

HP/RDC combination

that determines the behavior how data is recorded. Not only is this important to the conﬁguration of the boards and the recording of the data, but also for creating and training algorithms. One algorithm can use only identical HP/RDC combinations for training.

2.1.2 Board Type

Here you can choose the type of BME Board you want to create a conﬁguration ﬁle for. In this version of BME AI-Studio, you can choose one board type:

· BME Board with 8 BME688 sensors (BME Board x8)

Fig. 8: BME board x8

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

Fig. 9: BME board x8 Back

Other board types Coming soon

In a future release of BME AI-Studio we plan to have the possibility to chose other BME Boards.

2.1.3 Board Mode

The board mode can be set to diﬀerent modes, depending on your goal.

Warning

Fabric-new boards need to undergo a stabilization procedure. For stabilization you can use any board mode as long as you let it run for 24 hours.

· Sensorboard Default HP/RDC

is used to perform standard measurements. It uses the HP-354

heater proﬁle, a versatile and short proﬁle, which is best to begin with.

· Heater Proﬁle Exploration

is used to scan a variety of Heater Proﬁles in order to ﬁnd the one best suitable for a speciﬁc use case. In this mode you are able to choose multiple Heater Proﬁles, but only one Duty Cycle.

· Duty Cycle Exploration

is used if you want to optimize the Duty Cycling for a given Heater Proﬁle. In this mode you are able to choose multiple Duty Cycles, but only one Heater Proﬁle. Usually this mode is used for optimizing the energy eﬃciency, when you already found the best Heater Proﬁle for your use case.

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

2.1.4 Heater Proﬁle

In this section you can choose the Heater Proﬁles to be used in the Sensor Board conﬁguration.

Select Heater Proﬁles

In order to make a selection of Heater Proﬁles, click on

Select Heater Proﬁles ...

and choose the

desired Heater Proﬁles from the list. You can select multiple Heater Proﬁles, if you have chosen

“Heater Proﬁle Exploration” as

chosen “Duty Cycle Exploration” as

Board Mode

. You can select only one Heater Proﬁle, if you have

. Each Heater Proﬁle can be viewed with its

temperature sequence (graph and table) as well as a description of its purpose.

Diﬀerent groups of Heater Proﬁles

The pre-deﬁned Heater Proﬁles can be viewed as proﬁles that belong to diﬀerent proﬁle groups. This is also indicated by the numbering system within the name of the Heater Proﬁles. While

exploring new use cases it is good advise to try Heater Proﬁles from various groups in the ﬁrst place. Once it is clear which group of Heater Proﬁles work best you may explore this group in depth, by recording data with various Heater Proﬁles from this group and compare the results.

Maximum number of Heater Proﬁles

The maximum number of Heater Proﬁles (in

Heater Proﬁle Exploration

) is given by the board type:

· BME Board x8 can be conﬁgured with a maximum of 4 Heater Proﬁles

Review Heater Proﬁles

Once you have selected your desired Heater Proﬁles, you will see them in the corresponding table. To quickly remove Heater Proﬁles from the list, you can click on X .

Please note

If you have chosen “Sensorboard Default HP/RDC” for

Board Mode

you can not choose any

Heater Proﬁle. The standard proﬁle is selected automatically.

Individual Heater Proﬁles

In the current version of BME AI-Studio you can choose pre-deﬁned Heater Proﬁles. In a future release we plan to have the possibility of deﬁning individual Heater Proﬁles.

2.1.5 Duty Cycle

In this section you can choose the Duty Cycles to be used in the Sensor Board conﬁguration.

Select Duty Cycles

In order to make a selection of Duty Cycles, click on

Select Duty Cycles ...

and choose the desired

Duty Cycles from the list. You can select multiple Duty Cycles, if you have chosen “Duty Cycle

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

Exploration” as Board Mode. You can select only one Duty Cycle, if you have chosen “Heater Proﬁle Exploration” as Board Mode. Each Duty Cycle can be viewed with its speciﬁc number of Scanning Cycles and Sleeping Cycles and a schematic.

Additionally, you can see the percentage of Scanning Cycles within one iteration of a Duty Cycle. E.g. a 25% value means that for a quarter of the overall time the sensor is in scanning mode. This ratio also translates to the estimated energy consumption.

Maximum number of Duty Cycles

The maximum number of Duty Cycles (in Duty Cycle Exploration ) is given by the board type:

· BME Board x8 can be conﬁgured with maximum 4 Duty Cycles

The table shows you the selected Duty Cycles. By clicking onXyou can easily remove Duty Cycles from the list.

Individual Duty Cycles

In the current version of BME AI-Studio you can choose pre-deﬁned Duty Cycles. In a future release we plan to have the possibility of deﬁning individual Duty Cycles.

2.1.6 Board Layout

This option determines how the HP/RDC combinations are distributed over the available sensors on the BME Board. You can choose between three options:

· Grouped

places the combinations in a grouped way on the board. This can be thought of sequentially picking the sensors combination by combination and placing them on the board. It is the most obvious and easy to understand distribution.

· Shuﬄed

places the combinations in a shuﬄed or alternated way. This can be thought of sequentially picking sensors from each group and placing them on the board. This option gives a reproducible and evenly distribution. You may want to use this combination if you want to avoid grouping of sensors at certain positions on the BME Board. This may come in handy when you ﬂush the BME Board with a focused stream of gas and you want to avoid the risk of having a systemic error in the gas concentrations per sensor conﬁguration.

· Random

places the combinations in a randomized way. This can be thought of mixing sensors from each group and placing them on the board. This option gives an unintentional and evenly distribution. You may want to try selecting this option multiple times to get a new and independent randomization each time.

If the number of sensors on the BME Board is not divisible by the number of combinations, there will be residual sensors. These are ﬁlled up in an unbalanced manner, independent of the chosen board layout.

The actual distribution is shown in the corresponding schematic.

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

2.2 Record Data

2.2.1 Overview

Once you conﬁgured your BME Board, you are ready to make some measurements and record data. During this step, you don’t need the BME AI-Studio application. You can record data with just the BME Board.

In order to produce meaningful data, you should pay attention to the following important points:

· Try to capture realistic situations

The data that you record will later be used to train algorithms. In order to evaluate whether the algorithm will work under realistic conditions it is important to record data under realistic conditions as well.This means: if you want to distinguish apples from oranges that are lying on a table with the BME688 gassensor, do not record data where you cut the fruit and put it in a closed box with the sensor. Gas concentrations will be very diﬀerent from the use case you intend to have later.

· Take some time

Depending on the Heater Proﬁles and Duty Cylces you are using the sensor takes between 15 seconds and several minutes for each measurement. Try to collect at least 50 data points (which is a recording duration of 50 times the Duty Cycle length) per situation. This means: if you use a Duty Cycle with a duration of 30 seconds, try to record each situation for at least 30 minutes.

· Vary the situations

Try recording the use case you intend to do in various situations and across multiple days. The more you vary the more robust your algorithm will be. Temperature and humidity have a substantial inﬂuence on the sensitivity and selectivity of the sensor. Measuring across diﬀerent situations and with diﬀerent background gases will increase the variety of measurements that go into the training of the algorithm, which will give you a better feedback on how robust the algorithm is across diﬀerent situations

Some questions you could ask yourself

Depending on what you want to measure, it is important to prepare the appropriate setup and context. Some questions you should ask yourself:

· What do I want to measure? What is my Specimen?

· Can I reproduce the Specimen and if yes, how?

· Does the measurement take place inside or outside?

· How could the sourroundings aﬀect the measurement?

· Which nuisances could inﬂuence the use case and need to be included?

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

Warning

Do not touch the sensors with your ﬁngers. Fat on your skin will stick to the surface of the sensor enclosure and outgas for a long time. This will alter your gas recordings and lead to the training of unrealistic algorithms.

2.2.2 Start recording

Fig. 10: Inserting SD Card

Insert the SD card into the BME Board. The SD card slot is on the top side of the BME Board. Insert the card with its contacts facing towards the upper circuit board.

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

Fig. 11: Start Recording

Connect the BME Board to power with a Micro USB cable. Once the board is connected to power it will automatically turn on and start recording. The red LED next to the USB connector should blink every two seconds. If it blinks rapidly, the board is not recording.

Best practice

The BME Board x8 has an internal battery to keep track of time, but it is good prac-

tice to note down the exact date and time when you start your recording, so you can verify the data later.

Using BME Board with a Battery Pack

For recordings in closed environments you can use a battery pack to power the sensor board. A battery pack with 10.000 mAh should power your board for at least 24 hours, depending on the Duty cycle used and the surrounding temperatures.

2.2.3 During recording

You can record multiple Specimens during your Measurement Session without the need to turn oﬀ the BME Board. To mark diﬀerent Specimens during your recording on the ﬂy, you can either press the buttons on the BME Board or use the BME mobile App:

Deﬁne Specimens during measurement via on-board buttons

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

During measurement you can use either Button 1 (labeled

) or Button 2 (labeled

) on the BME Board x8 to mark speciﬁc moments as a new Specimen. If you do, these Specimen deﬁnitions will appear as Specimens during data import.

Fig. 12: During Recording

Deﬁne Specimens during measurement via BME mobile app Coming soon

Documentation for the BME mobile app will be added soon.

Deﬁne Specimens manually during data import

You can also add Specimen deﬁnitions manually during import. See Import Data for further details.

Best practice

As with the start of your recording, we recommend keeping notes of which Speci-

mens were recorded with the exact date and time for each Specimen.

2.2.4 End recording

Once you are ﬁnished recording your Specimens, you can simply unplug the power. The BME Board will turn oﬀ and the recording will stop.

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

Fig. 13: End Recording

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

2.3 Import Data

2.3.1 Overview

After you recorded data with the BME Board, it is now time to import the data into the application. During import, you can verify the data and deﬁne your Specimens. You can import several measurements and

train Algorithms. You can import data into a new project or into existing projects.

Access Files on SD Card

After your measurements, take the SD card out of the BME Board. Access the ﬁles on the SD card from your computer via an SD card reader.

collect everything in your Specimen Collection

, before later proceeding to

Fig. 14: Start Recording

Please note

If you do not see any data on the SD card, it is likely your board needs to be conﬁgured. To create a conﬁguration ﬁle for your board, please see Conﬁgure Board.

Importing Raw Data

Open the application and create a new project by clicking screen. Alternatively, you can open the menu and select

Create new project

New project...

on the Welcome

. Next, a ﬁle dialogue opens.

Please select, where you want to save your new project.

If you want to import data into an existing project, just open the project and switch to your

Specimen Collection .

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

Now, to import data, click Import Data . A ﬁle dialogue opens. You can

· select the Raw Data File directly from the SD card.

ﬁrst copy the Raw Data File from the SD card onto your computer hard drive, and then select the ﬁle from your hard drive.

Once you selected the Raw Data File, the data is being imported into your project and stored within the project folder of your project.

A new overlay opens that gives you an overview over the recorded data and lets you deﬁne your Specimens within the recording.

Importing Specimen Data

You can also import Specimen Data, that you exported from another project. The Specimens will be imported including their respective Measurement Sessions.

Please note

If you import Specimens that are already part of your project, the Specimens will imported as second duplicate. This can be interesting, if you want to weight speciﬁc Specimens as double or triple within algorithm training.

Importing whole projects

You can also import a whole project, by selecting the respective project folder. All Measurement Sessions and their Specimens will be imported.

2.3.2 Data Overview

This section gives you a brief summary of the data you just imported into your project. Verify the information to make sure everything went ﬁne during your measurement:

· Adjacent raw data ﬁles:

If the data was automatically split into multiple parts during measurement, BME AI-Studio automatically recognizes all adjacent ﬁles and imports them as well, no matter if you choose the ﬁrst or any other of these ﬁles. Automatic splitting of Raw Data ﬁles can happen when the BME Board power is switched on/oﬀ or the ﬁle limit (297 MB) is exceeded during measurement.

· Number of sensors found within raw data:

This number should correspond with the number of sensors on the BME Board used to record the data. If you used the BME Board x8, which has eight sensors, the imported Raw Data should include data from eight sensors as well. However, in some cases the number of sensors found within the data does not match with the number of sensors on the board. This usually indicates that one or more sensors were malfunctioning or not recording any data for some reason.

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

· Number of HP/RDC Combinations found within raw data:

This number should correspond with the speciﬁc BME Board Conﬁguration used to record the data. E.g. if your conﬁguration included four Heater Proﬁles and one Duty Cycle (so 4 HP/RDC Combinations in total), the imported Raw Data should include data from four HP/RDC Combinations as well. However, in some cases the number of HP/RDC Combinations found within the data does not match with the original BME Board Conﬁguration. This usually indicates that one or more sensors were malfunctioning or not recording any data for some reason.

2.3.3 Board ID

This indicates which BME Board has been used to record the data. Each board has a unique ID that is permanently stored on the board itself, which makes it uniquely identiﬁable.

2.3.4 Board Type

This indicates the type of BME Board used to record the data.

2.3.5 Board Mode

This indicates the Board Mode in which the BME Board was conﬁgured in when used to record the data.

2.3.6 Session Name

Give your Measurement Session a name – this could be a topic, a speciﬁc location or special occasion. Choose your Session name in a way that helps you identify and distinguish your Measurement Session later on.

2.3.7 Session Date

This indicates the date and time the Measurement Session has been recorded. Within the Raw Data, the time information is stored in Coordinated Universal Time (UTC) and is here automatically displayed in your respective local time zone. You can manually overwrite the time information if needed.

Start Date

The time the BME Board was powered on.

End Date

The time the BME Board was powered oﬀ.

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

Please note

If your BME Board lost its time information (e.g. if you disconnected the battery of BME Board x8), the date and time of the Measurement Session will always be January 1,

1970 by default. In this case, please overwrite the date and time with the actual correct date

and time of your measurement.

2.3.8 Specimen Data

This section gives you a visual representation (data plot) of the Measurement Raw Data and an overview of the Specimens in this session.

Specimens within a Measurement Session reﬂect what you have measured during that session. This can be an object, a speciﬁc situation, a smell, gas, etc. – or in other words: One speciﬁc gas composition the sensors on the BME Board were exposed to during the measurement.

Deﬁned Specimens are shown in the data plot as well as a list below.

Data Plot

You can view the Specimen Data in a visual plot. You can switch between:

· Four Data Channels

· Gas Data Channel

· Humidity Data Channel

· Temperature Data Channel

· Barometric Pressure Data Channel

· each Sensor on the BME

Deﬁne Specimens during measurement via on-board buttons

During measurement you can use the two buttons on the BME Board x8 to mark speciﬁc moments as a new Specimen. If you did, these Specimen deﬁnitions will appear in here this section (in the data plot and in the list of Specimens below). See Record Data for further details.

Deﬁne Specimens during measurement via BME mobile app Coming soon

Documentation for the BME mobile app will be added soon.

Deﬁne Specimens manually during data import

You can also add new Specimen deﬁnitions manually during import by clicking

Add Specimen

. This

creates a new Specimen deﬁnition in the list and you can specify the respective start and end time

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

within the Measurement Session. You can also add the corresponding Label here.

Edit Specimen Deﬁnitions during data import

You can edit the start and end time of Specimen deﬁnitions manually. When editing a Specimen deﬁnition, the respective area is highlighted in the data plot above.

Good to know

Although it is possible to deﬁne “overlapping” Specimens within a Session where the

end time of the ﬁrst Specimen lies after the start time of the second Specimen (so the same data is used in two Specimens), we recommend not doing so.

Please note

The start and end time of Specimens can not exceed the start and end time of the

overall session.

Label Specimens

You can label your recorded Specimens either here during Import, or later within the

Specimen Collection (see Collect Specimens).

Delete a Specimen deﬁnition

You can easily delete a Specimen deﬁnition by clicking the X button.

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

2.4 Collect Specimens

2.4.1 Overview

Once you individual Specimens in your

imported the Raw Data ﬁles

Specimen Collection

from your measurements, you ﬁnd the imported data as

. The Specimens in this list are grouped by Measurement Session and sorted by import date by default. You can organize all your data here, before later proceeding to train Algorithms.

Fig. 15: Specimen Collection

The purpose of the Specimen Collection is to collect various Specimens that belong to one use case. You may collect a lot of diﬀerent Specimens over time and use diﬀerent combinations of Specimens to try out the training of diﬀerent algorithms.

Good to know

It is recommended to start a new project (with a fresh Specimen Collection) for each use case you want to investigate. Therefore you can collect all Specimens that you want to use for training of use-case-speciﬁc algorithms within one project. You can easily export Specimens and import them in other projects.

Sort Specimens

You can change the sorting of Specimens by clicking repeatedly on the respective column header on top of the list. A little caret icon indicates the sorting direction:

· ∧ means ascending order (A-Z, 0-9, early-late)

· ∨ means descending order (Z-A, 9-0, late-early)

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

Search Specimens

You can search for speciﬁc Specimens within the Specimen Collection via the search ﬁeld above the list. Once you entered a search term, the list below will be ﬁltered according to your search and show only Specimens matching your search.

Measurement Session Settings

After importing Measurement Raw Data into your Specimen Collection you can always come back to the settings of each Measurement Session. Click the Session in your Specimen Collection to access these settings.

Here you can see information about the session, change its name and date, as well as edit Specimen deﬁnitions, add additional Specimen deﬁnitions or remove them.

See Specimen details

Settings

button next to each Measurement

Once you select a Specimen within the Specimen Collection, detail information of the selected Specimen is shown on the right hand side. Here you can add and change the label of the Specimen, add and edit a corresponding comment, and see all additional details regarding the Specimen.

Select multiple Specimens

You can also select multiple Specimens by clicking on the respective checkbox on the right hand side of each list entry. The detail information on the right now reﬂects which information has equal values and which information has multiple values within your selection.

If you select the checkbox of a whole Measurement Session, all Specimens within that session are selected. If you select the checkbox inside the column header of the Specimen Collection, all Specimens of all Measurement Sessions are selected.

Export Specimen Data

Once you select one or multiple Specimens, you can export these Specimens as individual ﬁles. Each ﬁle will include everything related to that speciﬁc Specimen, including its Label, Comment, etc.

Export Specimen Data if you want to use these Specimens in another BME project, if you want to share the data with others or if you want to manually back up that data.

Delete Specimens

Once you select one or multiple Specimens, you can delete these Specimens from your Specimen Collection. BME AI-Studio will ask you to conﬁrm the action before deleting the Specimens.

The action can not be undone, however only the deﬁnition of the Specimen (e.g. its start and end time, its Label, etc.) is deleted. The underlying Measurement Raw Data will be kept in the background, as

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

long as you have one or more other Specimens from the same Measurement Session in your collection. If you delete all Specimens from a session, the Measurement Raw Data will be deleted as well.

Please note

Be careful with deleting Specimens that you associated with a class as part of an algorithm. After deleting the Specimen, it will be missing from your class and therefore aﬀect the outcome of the algorithm training. If you try to delete a Specimen which is or has been associated with a Class as part of an

Algorithm, BME AI-Studio will give you warning.

2.4.2 Label

The label of each Specimen speciﬁes what exactly has been measured – this could be an object, a speciﬁc situation, a smell, gas, etc. You can conﬁrm your edit by pressing

blue√button. If you press

ESC

or select the greyXbutton, your edit will be discarded and the

original value of the ﬁeld will be kept.

Enter

or selecting the

Good to know

Label your Specimens as explicitly and clear as possible. “Apple Granny Smith, 1 day old” is better than “Apple 1”.

Please note

Labels are not to be confused with classes. Labels can include additional information and diﬀerentiate Specimens, that will later go into the same class. E.g. you might want to label your Specimens “Red apple”, “Green apple” and “Apple bio”, but use all three of them in one Class called “Apples”.

2.4.3 Comment

Add a comment to your Specimen to add extra information or notes. This might help you identify and distinguish your Specimen later on.

2.4.4 Session

This indicates the Measurement Session the selected Specimen belongs to.

2.4.5 Start & End Time

This indicates the date and time the selected Specimen has been recorded within the respective Measurement Session. Within the Raw Data, the time information is stored in Coordinated Universal

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

Time (UTC) and is here automatically displayed in your respective local time zone.

2.4.6 Duration

This indicates the length of the measurement of the selected Specimen, according to its start and end time.

2.4.7 Cycles Total

This indicates the total number of Scanning Cycles in the Measurement Raw Data of the selected Specimen.

Please note

Since each Heater Proﬁle has slightly diﬀerent cycle lengths, the number of Scanning Cycles within a Specimen can vary and might be diﬀerent for equal measurement durations.

2.4.8 Cycles Dropped

This indicates the number of Scanning Cycles that have been dropped during import. When importing Raw Measurement Data, BME AI-Studio automatically checks the data of every Scanning Cycle and excludes incomplete cycles from the import.

The number of dropped Scanning Cycles is also indicated as a percentage of the total number of Scanning Cycles.

2.4.9 Remaining Cycles

This indicates the number of remaining Scanning Cycles of the selected Specimen after dropping incomplete cycles.

2.4.10 Board Conﬁguration

This section provides detail information about the BME Board which was used to record the data of the selected Specimen. You can see the unique board ID, the board type and access the BME Board Conﬁguration used by clicking

Show conﬁguration

. This opens up an overlay with the following

information:

Board Type

This indicates the type of BME Board used to record the data of the selected Specimen.

Board Mode

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

This indicates the board mode in which the BME Board was conﬁgured in during recording of the data of the selected Specimen.

Heater Proﬁles

The Heater Proﬁle or list of Heater Proﬁles used to record the data of the selected Specimen. If multiple Heater Proﬁles have been used, you can click details of each Heater Proﬁle.

Duty Cycle

The Duty Cycle or list of Duty Cycles used to record the data of the selected Specimen. If multiple Duty Cycles have been used, you can click

Show Duty Cycles

Duty Cycle.

Board Layout

Show Heater Proﬁles

to see additional details of each

to see additional

This indicates how the HP/RDC combinations were distributed over the available sensors on the BME Board during recording of the data of the selected Specimen.

2.4.11 Board ID

This indicates which BME Board has been used to record the data of the selected Specimen. Each board has a unique ID that is permanently stored on the board itself, which makes it uniquely identiﬁable.

2.4.12 Board Type

This indicates the type of BME Board used to record the data of the selected Specimen.

2.4.13 Board Mode

This indicates the board mode in which the BME Board was conﬁgured in when used to record the data of the selected Specimen.

2.4.14 Show Conﬁguration

You can see the Board Conﬁguration of the BME Board which was used to record the data by selecting Show Conﬁguration .

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

2.5 Train Algorithms

2.5.1 Overview

Within the section

My Algorithms

you are able to use the data collected in your Specimen Collection to train (classiﬁcation) algorithms. You are able to create multiple algorithms with a diﬀerent data basis as well as various settings. Once trained you can compare the performance of the diﬀerent algorithms and choose the one algorithm that ﬁts best your use case (see

Evaluate Algorithms

). An algorithm can be exported as ﬁle for the BSEC Software (BME Library), which is basically a set of speciﬁc settings for the sensor, that enables it to distinguish diﬀerent gas compositions, once provided with these settings.

Create new Algorithm

First switch to

My Algorithms

, then select

New Algorithm

to create a new Algorithm. On the left

you will see all your algorithms in a list view.

See Algorithm details

Once you select an algorithm within that list of algorithms, detail information of the selected algorithm is shown on the right hand side. Make sure you select

Algorithm Settings

. Here you can set up

everything for the training of your Algorithm.

Train Algorithm

After setting up your algorithm, you can click

Train Algorithm

to start the training. Depending on the amount of data, the training can take from a few seconds up to several minutes. After the training, you can see the training results under

Training Results

. Jump to

Evaluate Algorithms

for further

information.

Duplicate Algorithm

Once you select an algorithm, you can duplicate that algorithm, if you want to repeat your training with slightly diﬀerent settings. Just click

Duplicate Algorithm

on the bottom of the screen or via the

context menu by clicking ... .

Delete Algorithm

Once you select an algorithm, you can delete that algorithm by clicking

Delete Algorithm

on the

bottom of the screen or via the context menu by clicking ... .

2.5.2 Name

Give your algorithm a name. Choose a name that might help you to identify and distinguish your Algorithms later on.

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

2.5.3 Created

The date and time you created that algorithm.

2.5.4 Classes

Classes are the most important part of your algorithm. Each class represents one of the diﬀerent gas compositions you want to distinguish in your use case. For example, if you want to distinguish apples and oranges, you would create two classes – one class for apples, one class for oranges.

Once you created a class, you can then select Specimens from your Specimen Collection to add to that class. For example, if you recorded data from ﬁve diﬀerent apples, you would add these Specimens to your apple class. And those Specimens where oranges were recorded, you would add to your oranges class.

Fig. 16: Train Algorithms

2.5.5 Class Name & Color

Give your class a name that describes the gas composition you want to distinguish. Additionally, you can give your class a color of your choice.

2.5.6 Common Data

After adding Specimens to your classes, BME AI-Studio will check the underlying data and the corresponding Heater Proﬁles and Duty Cycles, which were used to record that data. This section indicates, how many diﬀerent combinations of Heater Proﬁles and Duty Cycles were found within all the data, and how many of these combinations are shared with all selected Specimens. If there is a little checkmark, everything is okay and there are common HP/RDC combinations found for training.

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

2.5.7 Data Balance

This indicates how the data for your algorithm training is distributed over your classes. Ideally the distribution should be balanced. If there is a little checkmark, everything is okay and the data is balanced enough for training.

How Data Balance is calculated

Each class consists of multiple Specimens and each Specimen has a duration. The application checks if the total duration (sum of all Specimen durations) of any of the classes is below the following threshold:

threshold = 24% / number of classes

E.g.

· threshold for 2 classes: 12%

· threshold for 3 classes: 8%

· threshold for 4 classes: 6%

2.5.8 Data Channels

Specimen Data includes four data channels. You can choose, which Data Channel of each Specimen should be part of the training:

· Gas Data Channel (10 data points)

· Humidity Data Channel (1 data point)

· Temperature Data Channel (1 data point)

· Barometric Pressure Data Channel (1 data point)

By default, only the Gas Data Channel is used for training. If other channels play a key role in your use case, you can try to include additional channels in the training and compare the training results.

Please note

Be careful with

using additional channels

for your training. Using additional Data Channels does not automatically mean better training results. The algorithm might focus only on one of the additional Data Channels during training, which might not be what you actually

want. E.g. if one of your recorded Specimens has a strong impact on humidity, the algorithm

might only focus on the correlation between the Specimen and the corresponding rise in humidity, ignoring all other data. Your training results may look very good, but the algorithm only “learned” to distinguish your Specimen by looking at the humidity, completely ignoring the respective gas data.

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

Please note

Be aware that including the

nel

needs careful attention, since the temperature and relative humidity may not directly

Temperature Data Channel

and

Humidity Data Chan-

reﬂect the environmental conditions. This is due to the fact that both values are measured inside the metal packaging, which can be aﬀected by self-heating eﬀects of the BME688. For example running the sensor in a continuous cycling leads to a signiﬁcant temperature increase inside the metal packaging and thus to a decrease of the relative humidity. This introduces transient eﬀects on the recorded data, that do not originate from the actual sample you are measuring and therefore may mislead the training of the algorithm.

Please note

Be aware that including the

Barometric Pressure Data Channel

may introduce un-

wanted side eﬀects. Since the barometric pressure signal is changing with your local weather,

this may have an uncontrolled inﬂuence on the training process. Only include the Barometric Pressure Data Channel if you are sure that diﬀerent Specimen or situations diﬀer in their barometric pressure signal.

2.5.9 Neural Net

Here you can see the Neural Net Architecture used for the training of the algorithm.

Choose other Neural Nets Coming soon

In the current version of BME AI-Studio you can choose one pre-deﬁned Neural Net Architecture. In a future release we plan to have the possibility of chose other architectures.

2.5.10 Training Method

Choose the training method for the training. In this version of BME AI-Studio, only one optimizer function is available (ADAM optimizer). You can choose between diﬀerent batch sizes:

· ADAM optimizer, batch size 4

· ADAM optimizer, batch size 16

· ADAM optimizer, batch size 32

· ADAM optimizer, batch size 64

The batch size tells you how many data-points are taken into account for the optimization at once. This has inﬂuence on the training speed and stability.

As a rule of thumb, one can say the larger the batch size is the faster and more stable the training will be. The disadvantage of having larger batch sizes and the reason to go to smaller batch sizes is that it is more likely for the algorithm to be stuck in a local minima.

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

Please note

This is an expert setting. If you wan’t to know more about Training Rounds, please

get in touch with us.

2.5.11 Max. Training Rounds

Choose the maximum number of training rounds for the training.

Please note

This is an expert setting. If you wan’t to know more about Training Rounds, please

get in touch with us.

2.5.12 Data Splitting

Choose how much data should be used for training and testing respectively.

Please note

This is an expert setting. If you wan’t to know more about Training Rounds, please

get in touch with us.

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

2.6 Evaluate Algorithms

2.6.1 Overview

The evaluation process of Machine Learning algorithms is maybe the most important part of all. It is about evaluating the performance of the diﬀerent neural nets trained for each Not only does the evaluation tell you how good the predictions of the algorithm are, but you also get information on how robust and stable the algorithm will operate in the ﬁeld.

Good to know

The performance of an algorithm should always be evaluated as a compromise of

two factors:

· Variance of the input data (the more variance the better)

· Accuracy of the neural net (the higher the better)

HP/RDC combination

Too little variance

An example: You want to distinguish coﬀee from normal air and you trained your algorithm with data from freshly opened coﬀee only. So that means, the neural net only knows the smell of fresh coﬀee compared to normal air and you are probably getting great results with high accuracies and F1 Scores, as well as low False Positive rates. However, this does not necessarily imply that the algorithm would give you a great performance in all coﬀee situations. It only means that in this special situation (with freshly opened coﬀee) you are achieving good performance. So in order to get meaningful performance evaluations as well as eﬀective training, it is important to reﬂect the variance of diﬀerent situations also within the data recording used for training.

Too much variance

If your neural net performance is very low, you might try to reduce the variance of the data and experiment with diﬀerent subsets of the original data used for training. This might lead to better performance results, which give you some indication on which situations are very easy to handle for the algorithm and which are not.

2.6.2 Confusion Matrix

The Confusion Matrix describes the performance of a (classﬁcation) algorithm. Therefore it uses the

Test Data

, which was deﬁned before the training and was not used for training (see

Data Splitting

The individual matrix elements represent how many samples of the actual test have been predicted to a certain class.

For a perfect algorithm all test data samples have been predicted to their actual class, which are the diagonal elements of the Confusion Matrix. Oﬀ-diagonal elements represent false predictions and you may ﬁnd it interesting which kind of false predictions are present.

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

Fig. 17: Confusion Matrix

The Confusion Matrix also helps you evaluate how severe errors are for your speciﬁc use case. False alarms in one or the other direction may be very diﬀerent in their impact on your use case. For example, if you want to detect coﬀee to warn the user, who maybe has an allergy against coﬀee, the case where the algorithm missed to detect coﬀee (so called “False Negative case”) is much more severe than the opposite way, where you have an alarm, although there is no coﬀee involved (so called “False Positive case”).

2.6.3 Accuracy

The accuracy is the standard performance measurement for a (classiﬁcation) algorithm. It tells you how likely it is that the algorithm will give the correct classiﬁcation.

In other words it gives the ratio between correct and incorrect predictions. It is calculated by dividing the sum of the diagonal elements of the

Confusion Matrix

by the total sum of all elements (which is

the amount of Test Data used).

2.6.4 Macro-averaged F1 Score

The F1 Score is another performance indicator on how well the prediction of test data samples is working. It comes in handy when class distributions are unevenly distributed and therefore accuracy values are harder to interpret.

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

Fig. 18: Macro-averaged F1 Score

It is based on the recall and precision, two quantities that relate the True Positive case with its false alarms. Recall relates between all correctly predicted elements to a given class and all elements that actually belong to this class. While Precision relates between all correctly predicted elements to a given class and all elements that were predicted to this class. Both values are calculated for each class and then averaged (so “called macro-averaging”). The shown F1 Score represents the harmonic mean of macro-averaged recall and precision.

Further information on performance measures

We follow the ﬁndings of Sokolova, Marina, and Guy Lapalme. “A systematic analy-

sis of performance measures for classiﬁcation tasks.” Information processing & management

45.4 (2009): 427-437 in order to use F1 Score as a harmonic mean of macro-averaged recall

and precision rather than calculating the macro-average of class independent F1 Scores.

2.6.5 Macro-averaged False Positive rate

The False Positive rate (also called miss rate) gives you an indication of how many false alarms (or more general: how many false predictions) will happen in your algorithm. Therefore a lower False Positive rate indicates a better performance of the algorithm.

The False Positive rate is deﬁned as the relation between elements that belong to a diﬀerent class, but have been falsely predicted to the given class (often called false alarms) and all elements predicted to the given class. The shown False Positive rate is calculated as the macro-average of the class-speciﬁc False Positive rates.

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

2.6.6 Training Data

This indicates which fraction of the data samples were used as training data.

2.6.7 Test Data

This indicates which fraction of the data samples were used as test data.

2.6.8 BSEC Export

To export a

BSEC Conﬁguration File

, ﬁrst select the BSEC Version you want to use on your micro-

controller and then click on Export as BSEC Conﬁg File .

2.6.9 Additional Testing

After training your Algorithm, you can perform further tests of your Algorithm using additional data from your Specimen Collection. These tests are meant to be conducted with Specimen Data, which has not been used during the original training of the Algorithm, in order to give insight on how the Algorithm performs with previously unknown or completely diﬀerent data, that you do not want to inlcude in the training process. This helps you validate the robustness of the Algorithm, when facing exotic and unknown data.

To perfom an additional test, simply select

Test Algorithm

. A new overlay opens, showing you the same Classes which where originally used for Algorithm training, but without any associated Specimens. You can now associate new, additional Specimen Data – the data you want to test – to each Class. Once you associated Specimen Data to each Class, you can select Run Tests .

Please note

As of this version of BME AI-Studio, additional tests and their results are not stored within your project. You can, however, save the results as a CSV-File by selecting

Save Results .

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

3 Technical Speciﬁcation

3.1 Raw Data Format

The Raw Data Format developed for the Bosch Sensortec BME688 is used to store the Board Conﬁguration as well as the raw data. It is recommended to use this data format whenever data with the BME688 is recorded. The BME Board x8 also uses this data format.

3.1.1 File Name

The ﬁlename consists of the following parts (each seperated by a _):

· <date>

· <boardId>

· <counterPowerOnOﬀ>

· <seedPowerOnOﬀ>

· <counterFileLimit>

· .bmerawdata

Example:

2021_03_06_09_15_Board_2462ABD29048_PowerOnOﬀ_1_pfqncv1hjow4xm4t_File_0.bmerawdata

part Description

<date> Date created with the format:

<counterPowerOnOﬀ> Counter for power interruptions preﬁxed with

<seedPowerOnOﬀ> Unique seed that labels raw data that belongs to the

<counterFileLimit> Counter for consecutive raw data ﬁles preﬁxed with

.bmerawdata BME AI-Studio speciﬁc ending with JSON syntax

"yyyy_mm_dd_hh_mm"

Board-speciﬁc ID preﬁxed with "

Board_

" (each board

has a unique ID)

"PowerOnOﬀ_" (a new ﬁle is created every time the

board is powered on)

same Measurement Session

"File_" (after a ﬁle limit of 299 MB is reached a new

ﬁle is created)

Table 4: File Name

3.1.2 Structure

The ﬁle is structured into 4 parts:

· conﬁgHeader

· conﬁgBody

· rawDataHeader

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

· rawDataBody

The parts conﬁgHeader and conﬁgBody represent the board conﬁguration. They are also saved within the

.bmeconﬁg ﬁle

The following subsections give a description of each parameter used.

3.1.3 conﬁgHeader

. The parts rawDataHeader and rawDataBody store the actual recorded raw data.

Parameter Description

dateCreated Creation date and time of the board conﬁguration,

according to ISO 8601 in Coordinated Universal Time

(UTC)

appVersion Application version number of BME AI-Studio that

created the board conﬁguration

boardType Identiﬁer for the type of hardware used. board_8

refers to the BME Board x8 (see Board type). Other

boardType are also accepted, and can be used

when data is recorded with other hardware devices.

boardMode Identiﬁer for Board Mode. burn-in refers to the

sensorboard’s default HP/RDC, heater_proﬁle_exploration refers to Heater Proﬁle Exploration, duty_cycle_proﬁle_exploration refers

to Duty Cycle Exploration. Other boardMode are

also accepted, and can be used when data is

recorded with speciﬁc settings.

boardLayout Identiﬁer for Board Layout. grouped refers to

Grouped Board Layout, shuﬄed refers to Shuﬄed

Board Layout, random refers to Random Board

Layout.

Table 5: conﬁg Header

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

3.1.4 conﬁgBody

Parameter Description

heaterProﬁles Deﬁnition of Heater Proﬁles.

id can be any identiﬁer string.

timeBase is stated in milliseconds and deﬁnes the

base unit of time used for temperatureTimeVectors.

In order to work properly it should be set to 140.

temperatureTimeVectors are deﬁned as

[temperature,time], where temperature is stated in

degree Celsius and time spans are stated as

dutyCycleProﬁles Deﬁnition of Duty Cycles.

id can be any identiﬁer string.

numberScanningCycles derlineScanning Cycles.

numberSleepingCycles states the number of

multiples of timeBase.

Sleeping Cycles.

sensorConﬁgurations Allocation of sensors to HP/RDC Combinations

referencing to already deﬁned heaterProﬁles and

dutyCycleProﬁles by their id.

sensorIndex states the index of one of the sensors

on the board (usually counting starts at 0).

heaterProﬁle states the id of one of the deﬁned

heaterProﬁles.

dutyCycleProﬁle states the id of one of the deﬁned

dutyCycleProﬁles.

Table 6: conﬁg Body

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

3.1.5 rawDataHeader

part Description

counterPowerOnOﬀ Counter for power interruptions preﬁxed with

PowerOnOﬀ_ (a new ﬁle is created every time the

board is powered on). This parameter is also

reﬂected within the File Name.

seedPowerOnOﬀ Unique seed that labels raw data that belongs to the

same Measurement Session. This parameter is also

reﬂected within the File Name.

counterFileLimit Counter for consecutive rawdata ﬁles preﬁxed with

File_

(after a ﬁle limit of 299 MB is reached a new ﬁle

is created). This parameter is also reﬂected within the

dateCreated Creation date and time of the start of raw data

recording, according to UnixTime (seconds since 1

January 1970 in Coordinated Universal Time (UTC))

File Name.

dateCreated_ISO Creation date and time of the start of raw data

recording, according to ISO 8601 in Coordinated

Universal Time (UTC)

ﬁrmwareVersion Firmware version of the BME Board x8

boardId Unique board id for the BME Board x8 used for

recording

Table 7: Raw Data Header

3.1.6 rawDataBody

This section is divided into two parts. The ﬁrst part (

dataColumns

) acts as a column description for the second part ( dataBlock ), which actually carries the raw data in a condensed form. Every block within dataColumns describes one of the 12 data columns.

Here are some additional remarks concerning the column descriptions:

· sensor_id

is a unique id per sensor, that can be used to trace back data to an individual sensor

element

· heater_proﬁle_step_index

is a counter for 10 heater steps that will be cycled continously (see

Heater Proﬁle ). Counting starts at 0 up to 9.

· label_tag records when the buttons on the BME Board x8 are used. The encoding is:

0 if no button was pressed so far, 1 if Button 1 (SW2) was pressed, 2 if Button 2 (SW3) was pressed, and 3 if both buttons (SW2+SW3) were pressed simultaneously.

The current state is carried on until another button is pressed. (see During recording).

· error_code states errors that happened during recording with the follwing encoding:

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

1 for invalid Heater Proﬁles (aﬀects all data points that follow) 2

for time stamp mismatch or data loss (aﬀects only a single data point, dropping the whole cycle

is recommended)

for communication errors (aﬀects only a single data point, dropping the whole cycle is recom-

mended).

3.1.7 Example

{

"configHeader": {

"dateCreated": "2020-09-29T13:00:55.202Z", "appVersion": "1.6.0", "boardType": "board_8", "boardMode": "heater_profile_exploration",

"boardLayout": "grouped" }, "configBody": {

"heaterProfiles": [

{

"id": "heater_412", "timeBase": 140, "temperatureTimeVectors": [

[100,64], [320,2], [170,64], [320,2], [240,2], [240,31], [240,32], [320,2], [320,31], [320,32]

]

} ], "dutyCycleProfiles": [

{

"id": "duty_1", "numberScanningCycles": 1,

"numberSleepingCycles": 0 }, {

"id": "duty_5_10",

"numberScanningCycles": 5,

"numberSleepingCycles": 10 }

], "sensorConfigurations": [

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

{

"sensorIndex": 0,

"heaterProfile": "heater_412",

"dutyCycleProfile": "duty_1" }, {

"sensorIndex": 1,

"heaterProfile": "heater_412",

"dutyCycleProfile": "duty_5_10" }, {

"sensorIndex": 2,

"heaterProfile": "heater_412",

"dutyCycleProfile": "duty_1" }, {

"sensorIndex": 3,

"heaterProfile": "heater_412",

"dutyCycleProfile": "duty_5_10" }, {

"sensorIndex": 4,

"heaterProfile": "heater_412",

"dutyCycleProfile": "duty_1" }, {

"sensorIndex": 5,

"heaterProfile": "heater_412",

"dutyCycleProfile": "duty_5_10" }, {

"sensorIndex": 6,

"heaterProfile": "heater_412",

"dutyCycleProfile": "duty_1" }, {

"sensorIndex": 7,

"heaterProfile": "heater_412",

"dutyCycleProfile": "duty_5_10" }

] }, "rawDataHeader": {

"counterPowerOnOff": 1,

"seedPowerOnOff": "jecxzq530rhj2r5x",

"counterFileLimit": 1,

"dateCreated": "1601459722",

"dateCreated_ISO": "2020-09-30T09:55:22+02:00",

"firmwareVersion": "1.5.0",

"boardId": "1730555495",

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

}, "rawDataBody": {

"dataColumns": [

{

"name": "Sensor Index", "unit": "", "format": "integer",

"key": "sensor_index" }, {

"name": "Sensor ID",

"unit": "",

"format": "integer",

"key": "sensor_id" }, {

"name": "Time Since PowerOn",

"unit": "Milliseconds",

"format": "integer",

"key": "timestamp_since_poweron" }, {

"name": "Real time clock",

"unit": "Unix Timestamp: seconds since Jan 01 1970. (UTC); 0 =

missing",

"format": "integer",

"key": "real_time_clock" }, {

"name": "Temperature",

"unit": "DegreesCelcius",

"format": "float",

"key": "temperature" }, {

"name": "Pressure",

"unit": "Hectopascals",

"format": "float",

"key": "pressure" }, {

"name": "Relative Humidity",

"unit": "Percent",

"format": "float",

"key": "relative_humidity" }, {

"name": "Resistance Gassensor",

"unit": "Ohms",

"format": "float",

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

"key": "resistance_gassensor" }, {

"name": "Heater Profile Step Index",

"unit": "",

"format": "integer",

"key": "heater_profile_step_index" }, {

"name": "Scanning Mode Enabled",

"unit": "",

"format": "integer",

"key": "scanning_mode_enabled" }, {

"name": "Label Tag",

"unit": "",

"format": "integer",

"key": "label_tag" }, {

"name": "Error Code",

"unit": "",

"format": "integer",

"key": "error_code" } ], "dataBlock": [ [0,1730555495,4104,1615022156,27.862839,983.447632,40.376171,5684.846191,0,1,0,0], [1,1730565479,4126,1615022156,27.521061,983.286804,40.225910,5684.846191,0,1,0,0], [2,1730573669,4130,1615022157,27.673712,983.577271,40.541809,5684.846191,0,1,0,0], [3,1730557288,4134,1615022157,27.794174,983.404114,40.968414,5684.846191,0,1,0,0], [4,1730572647,4138,1615022157,27.552378,983.289185,42.118191,6240.249512,0,1,0,0], [5,1730567013,4141,1615022157,27.541058,983.432556,41.143063,5684.846191,0,1,0,0], [6,1730580836,4145,1615022157,27.639883,983.590149,43.335255,5684.846191,0,1,0,0], [7,1730556517,4149,1615022157,27.760021,983.735229,41.513172,5684.846191,0,1,0,0], [0,1730555495,4383,1615022157,27.792160,983.495789,40.256760,10213.852539,1,1,0,0], [1,1730565479,4406,1615022157,27.463169,983.366150,40.086674,33376.792969,1,1,0,0], [2,1730573669,4410,1615022157,27.615894,983.594910,40.396038,21850.460938,1,1,0,0], [3,1730557288,4414,1615022157,27.741306,983.456482,40.813091,29137.263672,1,1,0,0], [4,1730572647,4418,1615022157,27.519522,983.327698,42.050274,70213.929688,1,1,0,0], [5,1730567013,4421,1615022157,27.503132,983.463196,41.004383,28100.988281,1,1,0,0], [6,1730580836,4425,1615022157,27.599720,983.625610,42.526207,24601.191406,1,1,0,0], [7,1730556517,4429,1615022157,27.686945,983.767456,40.948986,26862.539062,1,1,0,0], [0,1730555495,5783,1615022158,28.203604,983.499878,39.452438,10352.636719,2,1,0,0], [1,1730565479,5806,1615022158,27.825611,983.341309,39.227486,33560.566406,2,1,0,0], [2,1730573669,5810,1615022158,27.992977,983.589050,39.522552,22284.123047,2,1,0,0], [3,1730557288,5814,1615022158,28.144110,983.448120,39.906544,29337.611328,2,1,0,0], [4,1730572647,5818,1615022158,27.944128,983.347229,41.179039,70796.460938,2,1,0,0], [5,1730567013,5821,1615022158,27.920330,983.499451,40.089077,28419.183594,2,1,0,0], [6,1730580836,5825,1615022158,28.016439,983.607971,41.731613,25048.923828,2,1,0,0],

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

[7,1730556517,5829,1615022158,28.052332,983.758484,40.062920,27432.490234,2,1,0,0], [0,1730555495,9983,1615022162,28.726116,983.515198,37.202606,10536.713867,3,1,0,0], [1,1730565479,10006,1615022162,28.381857,983.376282,36.991051,33272.679688,3,1,0,0], [2,1730573669,10010,1615022162,28.541010,983.613831,37.000546,22076.578125,3,1,0,0], [3,1730557288,10014,1615022162,28.687897,983.449951,37.366409,28822.337891,3,1,0,0], [4,1730572647,10018,1615022162,28.505222,983.394104,38.104702,70098.578125,3,1,0,0], [5,1730567013,10021,1615022162,28.504414,983.488953,37.240280,28343.666016,3,1,0,0], [6,1730580836,10025,1615022162,28.556171,983.658813,38.780487,24990.238281,3,1,0,0], [7,1730556517,10029,1615022162,28.591593,983.813904,37.237930,27898.867188,3,1,0,0], [3,1730557288,10573,1615022163,29.246792,983.437073,36.985844,8719.345703,4,1,0,0], [0,1730555495,10684,1615022163,29.289021,983.541565,36.938293,5684.846191,4,1,0,0], [1,1730565479,10706,1615022163,28.917973,983.321167,36.683632,10021.922852,4,1,0,0], [2,1730573669,10710,1615022163,29.086529,983.609131,36.590668,7426.316895,4,1,0,0], [4,1730572647,10717,1615022163,29.071375,983.355774,37.663334,19265.501953,4,1,0,0], [5,1730567013,10721,1615022163,29.032871,983.472229,36.849651,8961.075195,4,1,0,0], [6,1730580836,10725,1615022163,29.088375,983.648865,38.384171,7982.040527,4,1,0,0], [7,1730556517,10729,1615022163,29.148499,983.783630,36.861649,9052.333984,4,1,0,0], [3,1730557288,11273,1615022164,29.629461,983.469971,36.604248,9184.844727,5,1,0,0], [0,1730555495,11384,1615022164,29.647461,983.527710,36.560806,5684.846191,5,1,0,0], [1,1730565479,11406,1615022164,29.290485,983.375427,36.365276,10663.112305,5,1,0,0], [2,1730573669,11410,1615022164,29.463619,983.659912,36.612564,7744.433594,5,1,0,0], [4,1730572647,11417,1615022164,29.455547,983.397217,37.305729,19656.019531,5,1,0,0], [5,1730567013,11421,1615022164,29.260437,983.493042,36.240665,9329.446289,5,1,0,0], [6,1730580836,11425,1615022164,29.444851,983.617310,37.934799,8456.659180,5,1,0,0], [7,1730556517,11429,1615022164,29.513889,983.805298,36.431786,9386.916992,5,1,0,0], [3,1730557288,11974,1615022164,29.810726,983.466003,36.180058,9555.090820,6,1,0,0], [0,1730555495,12084,1615022164,29.829205,983.547729,36.165672,5684.846191,6,1,0,0], [1,1730565479,12106,1615022164,29.476742,983.376709,35.954807,11084.170898,6,1,0,0] ]

}

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

3.2 Specimen Data Format

The Specimen Data Format developed for the Bosch Sensortec BME688 is used to store data for a speciﬁc specimen. BME AI-Studio can be used to import Raw Data can be sliced into one or multiple specimens. Specimen Data ﬁles can be exchanged between diﬀerent BME AI-Studio projects. If curated correctly they should only contain gas data that belongs to one speciﬁc specimen or situation.

3.2.1 File Name

The ﬁlename consists of the following parts (each seperated by a _):

· <specimen_name>

· <id>

· .bmespecimen

Raw Data Format

, which then

Example: coﬀee_2.bmespecimen

part Description

<specimen_name> individual name that can be deﬁned within the BME

AI-Studio (strings are converted to lowercase and

spaces are replaced with dashes)

<id> project-speciﬁc id per specimen

.bmerawdata BME AI-Studio speciﬁc ending with JSON syntax

Table 8: File Name

3.2.2 Structure

The ﬁle is structured into a metadata part ( meta ) and data part ( data ).

The meta part consists of:

· appVersion

states application version number of BME AI-Studio that created the board conﬁgura-

tion

· exportedAt

states export date and time, according to ISO 8601 in Coordinated Universal Time

(UTC)

The data part consists of:

· specimenData

· measurementSession

· boardConﬁg

· boardType

· heaterProﬁles

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

· dutyCycleProﬁles

· sensorConﬁgs

· sensors

· cycles

· dataColumns

· specimenDataPoints

The following subsections give a description of each parameter used.

3.2.3 Specimen Data

Parameter Description

id Project-speciﬁc Specimen id (automatically deﬁned

by BME AI-Studio)

uuid Universally unique identiﬁer (UUID) per Specimen

name Specimen Label (see Specimen Label)

comment Comment that further describes the Specimen

startTime Start time of time interval that references to

real_time_clock within specimenDataPoints

endTime End time of time interval that references to

real_time_clock within specimenDataPoints

cloneOfUuid Used for BME AI-Studio internal data handling

measurementSessionId Reference to id of Measurement Session to which

this Specimen belongs to

createdAt Creation date and time of Specimen, according to

ISO 8601 in Coordinated Universal Time (UTC)

updatedAt

Update date and time of Specimen, according to ISO

8601 in Coordinated Universal Time (UTC)

Table 9: Specimen Data

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

3.2.4 measurementSession

part Description

id Project-speciﬁc Measurement Session id

(automatically deﬁned by BME AI-Studio)

uuid

Universally unique identiﬁer (UUID) per Measurement

Session

name Measurement Session label (see Session Name)

comment Comment that further describes the Specimen

startTime Recording start date and time of Measurement

Session, according to ISO 8601 in Coordinated

Universal Time (UTC)

endTime Recording end date and time of Measurement

Session, according to ISO 8601 in Coordinated

Universal Time (UTC)

boardId Unique board id for the BME Board x8 used for

recording

createdAt Creation date and time of Measurement Session,

according to ISO 8601 in Coordinated Universal Time

(UTC)

updatedAt Update date and time of Measurement Session,

according to ISO 8601 in Coordinated Universal Time

(UTC)

Table 10: measurementSession

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

3.2.5 boardConﬁg

part Description

Project-speciﬁc Board Conﬁguration id (automatically

deﬁned by BME AI-Studio)

boardMode Identiﬁer for Board Mode. burn-in refers to

Sensorboard Default HP/RDC, heater_proﬁle_exploration refers to Heater Proﬁle Exploration, duty_cycle_proﬁle_exploration refers

to Duty Cycle Exploration. Other boardMode are

also accepted, and can be used when data is

recorded with speciﬁc settings.

measurementSessionId Reference to id of Measurement Session to which

this Board Conﬁguration belongs t0

boardTypeId Reference to id of Board Type to which this Board

Conﬁguration belongs to

createdAt Creation date and time of Board Conﬁguration,

according to ISO 8601 in Coordinated Universal Time

(UTC)

updatedAt Update date and time of Board Conﬁguration,

according to ISO 8601 in Coordinated Universal Time

(UTC)

Table 11: boardConﬁg

3.2.6 boardType

numSensors Number of (BME688) sensors on the hardware

part Description

id Project-speciﬁc Board Type id (automatically deﬁned

by BME AI-Studio)

uid Unique identiﬁer for the type of hardware used.

board_8 refers to the BME Board x8. See

Board type. Other boardType are also accepted,

and can be used when data is recorded with other

hardware devices.

name Individual name of the hardware used (e.g. BME

Board x8)

createdAt Creation date and time of Board Type, according to

ISO 8601 in Coordinated Universal Time (UTC)

updatedAt Update date and time of Board Type, according to

ISO 8601 in Coordinated Universal Time (UTC)

Table 12: boardType

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

3.2.7 heaterProﬁles

Parameter speciﬁcation per Heater Proﬁle:

part Description

id Project-speciﬁc Heater Proﬁle id (automatically

uid Unique identiﬁer for the Heater Proﬁle

name Individual name of the Heater Proﬁle

useCase Additional information on speciﬁc use cases

timeBase Deﬁnes the base unit of time used for

temperatureTimeVectors (stated in milliseconds).

In order to work properly it should be set to 140.

steps Temperature-duration steps, where temperature is

stated in degree Celsius and duration is stated as

deﬁned by BME AI-Studio)

multiples of timeBase.

createdAt

Creation date and time of Heater Proﬁle, according to

updatedAt Update date and time of Heater Proﬁle, according to

Table 13: heaterProﬁles

3.2.8 dutyCycleProﬁles

Parameter speciﬁcation per Duty Cycle Proﬁle:

ISO 8601 in Coordinated Universal Time (UTC)

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

part Description

id Project-speciﬁc Duty Cycle id (automatically deﬁned

uid Unique identiﬁer for the Duty Cycle Proﬁle

name Individual name of the Duty Cycle Proﬁle

description Additional information

energyConsumption States the energy consumption of the Duty Cycle

scanningCycles States the number of Scanning Cycles

sleepingCycles States the number of Sleeping Cycles

createdAt Creation date and time of Duty Cycle Proﬁle,

according to ISO 8601 in Coordinated Universal Time

updatedAt

Update date and time of Duty Cycle Proﬁle, according

to ISO 8601 in Coordinated Universal Time (UTC)

by BME AI-Studio)

Proﬁle in mA

(UTC)

Table 14: dutyCycleProﬁles

3.2.9 sensorConﬁgs

The Sensor Conﬁguration attaches a Board Conﬁguration and individual Heater Proﬁles and Duty Cycle Proﬁles to a speciﬁc sensor.

Parameter speciﬁcation per Sensor Conﬁguration:

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

part Description

id Project-speciﬁc Sensor Conﬁguration id

(automatically deﬁned by BME AI-Studio)

boardConﬁgId Reference to id of Board Conﬁguration to which this

Sensor Conﬁguration belongs to

sensorId Reference to id of Sensors to which this Sensor

Conﬁguration belongs to

heaterProﬁleId

Reference toidof Heater Proﬁle to which this Sensor

Conﬁguration belongs to

dutyCycleProﬁleId Reference to id of Duty Cycle Proﬁle to which this

Sensor Conﬁguration belongs to

createdAt Creation date and time of Sensor Conﬁguration,

according to ISO 8601 in Coordinated Universal Time

(UTC)

updatedAt Update date and time of Sensor Conﬁguration,

according to ISO 8601 in Coordinated Universal Time

(UTC)

Table 15: sensorConﬁgs

3.2.10 sensors

The list of sensors connects their Id (used within BME-AI Studio) to their index (used in raw data logging).

part Description

id Project-speciﬁc sensor id (automatically deﬁned by

index Sensor index which is used within Raw Data Format

createdAt Creation date and time of sensor, according to ISO

updatedAt Unique seed that labels raw data that belongs to the

<counterFileLimit> Update date and time of sensor, according to ISO

3.2.11 cycles

The list of curated Scanning Cycles.

Parameter speciﬁcation per cycle:

BME AI-Studio)

8601 in Coordinated Universal Time (UTC)

same Measurement Session

8601 in Coordinated Universal Time (UTC)

Table 16: sensors

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

part Description

id Project-speciﬁc cycle id (automatically deﬁned by

uuid Universally unique identiﬁer (UUID) per cycle that

references to cycle_id within specimenDataPoints

startTime Start time of cycle that references to

real_time_clock within specimenDataPoints

BME AI-Studio)

endTime

dropped Decision (encoded as boolean), whether cycle was

measurementSessionId Reference to id of Measurement Session to which

sensorId Reference to id of Sensors to which this cycle

createdAt Creation date and time of cycle, according to ISO

updatedAt

3.2.12 dataColumns

Every block within

dataColumns

End time of cycle that references to

within specimenDataPoints

selected as "dropped" within import process (see

Cycles Dropped

this cycle belongs to

belongs to

8601 in Coordinated Universal Time (UTC)

Update date and time of cycle, according to ISO 8601

in Coordinated Universal Time (UTC)

Table 17: cycles

describes one of the 12 data columns within

real_time_clock

specimenDataPoints

Here are some additional remarks concerning the column descriptions:

· data_point_id

is a unique id per data point, that can be used to trace back data to an individual

sensor element

· error_code states errors that happened during recording with the follwing encoding:

1 for invalid Heater Proﬁles (aﬀects all data points that follow) 2

for time stamp mismatch or data loss (aﬀects only a single data point, dropping the whole cycle

is recommended)

for communication errors (aﬀects only a single data point, dropping the whole cycle is

recommended).

· cycle_step_index

is a counter for 10 heater steps that will be cycled continously (see Heater

Proﬁle ). Counting starts at 0 up to 9.

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

3.2.13 specimenDataPoints

The part carries the sensor data in a condensed form according to dataColumns .

3.2.14 Example

{

"meta": {

"appVersion": "1.6.0",

"exportedAt": "2021-05-20T09:07:11.922Z" }, "data": {

"specimenData": {

"id": 2, "uuid": "a5159d14-f0af-4f05-be13-fbf629d51c39", "name": "coffee", "comment": "specialty coffee", "startTime": 427208, "endTime": 1149010, "cloneOfUuid": null, "measurementSessionId": 1, "createdAt": "2021-05-20T08:59:24.000Z",

"updatedAt": "2021-05-20T08:59:50.000Z" }, "measurementSession": {

"id": 1,

"uuid": "137ebdea-3f9b-4cb4-a4e3-b5e00d36894c",

"name": "Coffee Recording",

"startTime": "2021-03-06T09:15:56.000Z",

"endTime": "2021-03-06T10:15:20.000Z",

"boardId": "2462ABD29048",

"createdAt": "2021-05-20T08:59:24.000Z",

"updatedAt": "2021-05-20T08:59:24.000Z" }, "boardConfig": {

"id": 1,

"boardMode": "duty_cycle_profile_exploration",

"measurementSessionId": 1,

"boardTypeId": 1,

"createdAt": "2021-05-20T08:59:24.000Z",

"updatedAt": "2021-05-20T08:59:24.000Z" }, "boardType": {

"id": 1,

"uid": "board_8",

"name": "BME Board x8",

"numSensors": 8,

"createdAt": "2021-05-19T12:29:24.274Z",

"updatedAt": "2021-05-19T12:29:24.274Z" },

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

"heaterProfiles": [ {

"id": 6,

"uid": "heater_354",

"name": "HP-354",

"useCase": "Standard BST profile",

"timeBase": 140,

"steps": [

{

"temperature": 320,

"duration": 5 }, {

"temperature": 100,

"duration": 2 }, {

"temperature": 100,

"duration": 10 }, {

"temperature": 100,

"duration": 30 }, {

"temperature": 200,

"duration": 5 }, {

"temperature": 200,

"duration": 5 }, {

"temperature": 200,

"duration": 5 }, {

"temperature": 320,

"duration": 5 }, {

"temperature": 320,

"duration": 5 }, {

"temperature": 320,

"duration": 5 } ], "createdAt": "2021-05-19T12:29:24.261Z", "updatedAt": "2021-05-19T12:29:24.261Z"

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

} ], "dutyCycleProfiles": [ {

"id": 1, "uid": "duty_1", "name": "RDC-1-0 Continuous", "description": "This is a continous duty cycle profile, which is best suited for maximum data acquisition, but with maximum power consumption. Please not that the sensor will significantly warm up in continous duty cycle profile.", "energyConsumption": 12, "scanningCycles": 1, "sleepingCycles": 0, "createdAt": "2021-05-19T12:29:24.267Z", "updatedAt": "2021-05-19T12:29:24.267Z"

}, {

"id": 12, "uid": "duty_5_10", "name": "RDC-5-10", "description": "", "energyConsumption": 4, "scanningCycles": 5, "sleepingCycles": 10, "createdAt": "2021-05-19T12:29:24.267Z", "updatedAt": "2021-05-19T12:29:24.267Z"

} ], "sensorConfigs": [ {

"id": 1, "boardConfigId": 1, "sensorId": 1, "heaterProfileId": 6, "dutyCycleProfileId": 1, "createdAt": "2021-05-20T08:59:24.000Z", "updatedAt": "2021-05-20T08:59:24.000Z"

}, {

"id": 2, "boardConfigId": 1, "sensorId": 2, "heaterProfileId": 6, "dutyCycleProfileId": 12, "createdAt": "2021-05-20T08:59:24.000Z", "updatedAt": "2021-05-20T08:59:24.000Z"

}, {

"id": 3, "boardConfigId": 1,

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

"sensorId": 3, "heaterProfileId": 6, "dutyCycleProfileId": 1, "createdAt": "2021-05-20T08:59:24.000Z", "updatedAt": "2021-05-20T08:59:24.000Z"

}, {

"id": 4, "boardConfigId": 1, "sensorId": 4, "heaterProfileId": 6, "dutyCycleProfileId": 12, "createdAt": "2021-05-20T08:59:24.000Z", "updatedAt": "2021-05-20T08:59:24.000Z"

}, {

"id": 5, "boardConfigId": 1, "sensorId": 5, "heaterProfileId": 6, "dutyCycleProfileId": 1, "createdAt": "2021-05-20T08:59:24.000Z", "updatedAt": "2021-05-20T08:59:24.000Z"

}, {

"id": 6, "boardConfigId": 1, "sensorId": 6, "heaterProfileId": 6, "dutyCycleProfileId": 12, "createdAt": "2021-05-20T08:59:24.000Z", "updatedAt": "2021-05-20T08:59:24.000Z"

}, {

"id": 7, "boardConfigId": 1, "sensorId": 7, "heaterProfileId": 6, "dutyCycleProfileId": 1, "createdAt": "2021-05-20T08:59:24.000Z", "updatedAt": "2021-05-20T08:59:24.000Z"

}, {

"id": 8, "boardConfigId": 1, "sensorId": 8, "heaterProfileId": 6, "dutyCycleProfileId": 12, "createdAt": "2021-05-20T08:59:24.000Z", "updatedAt": "2021-05-20T08:59:24.000Z"

}

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

], "sensors": [ {

"id": 1, "index": 0, "createdAt": "2021-05-20T08:59:24.000Z", "updatedAt": "2021-05-20T08:59:24.000Z"

}, {

"id": 2, "index": 1, "createdAt": "2021-05-20T08:59:24.000Z", "updatedAt": "2021-05-20T08:59:24.000Z"

}, {

"id": 3, "index": 2, "createdAt": "2021-05-20T08:59:24.000Z", "updatedAt": "2021-05-20T08:59:24.000Z"

}, {

"id": 4, "index": 3, "createdAt": "2021-05-20T08:59:24.000Z", "updatedAt": "2021-05-20T08:59:24.000Z"

}, {

"id": 5, "index": 4, "createdAt": "2021-05-20T08:59:24.000Z", "updatedAt": "2021-05-20T08:59:24.000Z"

}, {

"id": 6, "index": 5, "createdAt": "2021-05-20T08:59:24.000Z", "updatedAt": "2021-05-20T08:59:24.000Z"

}, {

"id": 7, "index": 6, "createdAt": "2021-05-20T08:59:24.000Z", "updatedAt": "2021-05-20T08:59:24.000Z"

}, {

"id": 8, "index": 7, "createdAt": "2021-05-20T08:59:24.000Z", "updatedAt": "2021-05-20T08:59:24.000Z"

} ],

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

"cycles": [ {

"id": 221, "uuid": "cf0b8cb0-1133-48a1-8f89-2a1a59bf471f", "startTime": 431425, "endTime": 441386, "dropped": false, "measurementSessionId": 1, "sensorId": 1, "createdAt": "2021-05-20T08:59:24.000Z", "updatedAt": "2021-05-20T08:59:24.000Z"

}, {

"id": 225, "uuid": "caaadf34-4a68-42d1-90a7-a2e22b79f80e", "startTime": 442240, "endTime": 452230, "dropped": false, "measurementSessionId": 1, "sensorId": 1, "createdAt": "2021-05-20T08:59:24.000Z", "updatedAt": "2021-05-20T08:59:24.000Z"

}, {

"id": 229, "uuid": "61d39719-57d5-4f05-8b10-e74a3e2245d4", "startTime": 453083, "endTime": 462934, "dropped": false, "measurementSessionId": 1, "sensorId": 1, "createdAt": "2021-05-20T08:59:24.000Z", "updatedAt": "2021-05-20T08:59:24.000Z"

}, {

"id": 233, "uuid": "191e34a6-0e0d-4e30-9a1b-4a0c9b732148", "startTime": 463787, "endTime": 473778, "dropped": false, "measurementSessionId": 1, "sensorId": 1, "createdAt": "2021-05-20T08:59:24.000Z", "updatedAt": "2021-05-20T08:59:24.000Z"

}, {

"id": 237, "uuid": "8e285103-6d5e-4b1d-b536-31c3de9d5b66", "startTime": 474632, "endTime": 484484, "dropped": false,

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

"measurementSessionId": 1, "sensorId": 1, "createdAt": "2021-05-20T08:59:24.000Z", "updatedAt": "2021-05-20T08:59:24.000Z"

} ], "dataColumns": [ {

"name": "Data Point ID", "unit": "", "format": "integer", "key": "data_point_id"

}, {

"name": "Resistance Gassensor", "unit": "Ohms", "format": "float", "key": "resistance_gassensor"

}, {

"name": "Temperature", "unit": "DegreesCelsius", "format": "float", "key": "temperature"

}, {

"name": "Pressure", "unit": "Hectopascals", "format": "float", "key": "pressure"

}, {

"name": "Relative Humidity", "unit": "Percent", "format": "float", "key": "relative_humidity"

}, {

"name": "Time Since PowerOn", "unit": "Milliseconds", "format": "integer", "key": "timestamp_since_poweron"

}, {

"name": "Real time clock", "unit": "Unix Timestamp: seconds since Jan 01 1970. (UTC); 0 = missing", "format": "integer", "key": "real_time_clock"

}, {

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

"name": "Error Code", "unit": "", "format": "integer", "key": "error_code"

}, {

"name": "Cycle Step Index", "unit": "", "format": "integer", "key": "cycle_step_index"

}, {

"name": "Cycle ID", "unit": "", "format": "integer", "key": "cycle_id"

} ], "specimenDataPoints": [ [2201,38254.632812,38.803349,983.621399,20.660652,431425,1615022588,0,0,221], [2202,1483520.5,38.364086,983.688721,20.657352,431705,1615022588,0,1,221], [2203,1388474.625,37.36187,983.655579,20.794451,433104,1615022590,0,2,221], [2204,1225246.75,36.993301,983.63562,21.063087,437313,1615022594,0,3,221], [2205,171294.75,37.652184,983.66449,21.170946,438013,1615022595,0,4,221], [2206,167594.109375,37.884438,983.679016,21.192097,438713,1615022595,0,5,221], [2207,163892.453125,37.967743,983.635315,21.155724,439413,1615022596,0,6,221], [2208,30146.019531,38.407005,983.650452,21.11739,440113,1615022597,0,7,221], [2209,27309.580078,38.59634,983.617065,21.088247,440813,1615022597,0,8,221], [2210,24917.267578,38.697319,983.634521,20.989874,441386,1615022598,0,9,221], [2241,22448.263672,38.752861,983.624512,20.851637,442240,1615022599,0,0,225], [2242,583309.625,38.293404,983.699219,20.752279,442531,1615022599,0,1,225], [2243,543380.1875,37.26342,983.652344,20.541948,443943,1615022600,0,2,225], [2244,473307.15625,36.957958,983.648743,20.68368,448017,1615022605,0,3,225], [2245,80935.820312,37.609268,983.647949,20.769163,448718,1615022605,0,4,225], [2246,79725.945312,37.828896,983.652527,20.804911,449417,1615022606,0,5,225], [2247,78551.703125,37.927353,983.671753,20.806301,450118,1615022607,0,6,225], [2248,20990.488281,38.359039,983.63562,20.809202,450818,1615022607,0,7,225], [2249,21541.568359,38.558472,983.617615,20.770164,451518,1615022608,0,8,225], [2250,22122.365234,38.641781,983.617249,20.681274,452230,1615022609,0,9,225], [2281,22686.990234,38.712467,983.644836,20.586311,453083,1615022610,0,0,229], [2282,615384.625,38.265633,983.713257,20.30512,453374,1615022610,0,1,229], [2283,557127.3125,37.238174,983.68457,20.352257,454648,1615022611,0,2,229], [2284,498539.4375,36.937767,983.645142,20.609156,458861,1615022615,0,3,229], [2285,84377.0625,37.594124,983.659485,20.710518,459561,1615022616,0,4,229], [2286,82580.648438,37.821327,983.678955,20.772936,460261,1615022617,0,5,229], [2287,81476.765625,37.899582,983.665833,20.767719,460961,1615022617,0,6,229], [2288,20765.736328,38.338844,983.657532,20.786839,461662,1615022618,0,7,229], [2289,21326.224609,38.543327,983.627319,20.769142,462362,1615022619,0,8,229], [2290,21940.349609,38.634209,983.640198,20.70701,462934,1615022619,0,9,229], [2321,22832.679688,38.709942,983.635437,20.622875,463787,1615022620,0,0,233], [2322,610432.1875,38.242912,983.704041,20.544394,464079,1615022621,0,1,233],

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec BME AI-Studio Manual

[2323,564031.9375,37.207878,983.680054,20.412897,465492,1615022622,0,2,233], [2324,495643.75,36.927666,983.649597,20.671202,469565,1615022626,0,3,233], [2325,84544.25,37.581501,983.654846,20.814461,470265,1615022627,0,4,233], [2326,83797.054688,37.808704,983.679871,20.87174,470966,1615022628,0,5,233], [2327,82341.585938,37.89201,983.675903,20.85638,471665,1615022628,0,6,233]

]

}

BST-BME688-AN001-00 Date 2021-09-07 Version 1.6.0

distribution, as well as in the event of applications for industrial property rights.

Bosch Sensortec GmbH

Gerhard-Kindler-Straße 9 72770 Reutlingen / Germany

contact@bosch-sensortec.com www.bosch-sensortec.com

Modiﬁcations reserved Preliminary - speciﬁcations subject to change without notice Document number: BST-BME688-AN001-00 Revision 1.6.0

Seeed Studio Grove Air Quality Sensor BME AI-Studio Manual

Specifications and Main Features

Frequently Asked Questions

User Manual