Siemens Evolution of a digital twin with an ethylene plant as an example User manual

White Paper

Edition 01/2020

Chemicals

Evolution of a digital twin with an ethylene plant as an example

Concept and implementation

siemens.com/chemicals

Content Introduction

The article describes a concept for the development and

1 First steps with a digital twin 3

1. 1 Models and simulation in the lifecycle of a plant 3

1. 2 Landscape of the models 4

1. 3 Vision 6

2 Example application - Steam cracker 5 3 Design and evolution of a digital twin 9 4 Use of the digital twin 13

4. 1 Engineer ing 13

4. 2 Plant operation 14

4. 3 Re-engineering of the digital twin 18

5 Summary 19 6 References 20 7 Authors 21

integrated use of a digital twin over the entire lifecycle of a process plant. Various aspects of a digital twin are defined and described. Possible benefits of a digital twin in successive phases of its lifecycle are discussed in detail. The concept is realized in form of a demonstrator using the example of a steam cracker for ethylene production

[36]

Evolution of a digital twin

1 First steps with a digital twin

The term „digital twin“ itself evokes a wide range of associations. With human twins, we think of common inherited traits, similar characteristics and characters, and the often astonishing parallels in their ways of life. Although there is still a large discrepancy between the interest in digital twins (> 500 million Google links) and the number of real applications, the term „digital twin“ is more than just a buzzword in the process industry. There are indeed many different concepts, but also initial approaches to concrete implementations.

At first glance, the large number of different types of digital twins appears to be confusing. Depending on the viewpoint of the observer, typical terms such as

◾ Product digital twin ◾ Automation digital twin ◾ Production digital twin ◾ 3D digital twin ◾ Asset digital twin ◾ Process digital twin ◾ etc.

can be found in literature, lectures and conferences.

A digital twin of a process plant as an integrated concept covers three core points: The digital twin of the product, the digital twin of the production plant and the digital modeling of the performance of the product and production.

The functional scope of a digital twin essentially depends on its purpose. In the process industry, this can be everything from the safety analysis, product simulation or the optimization of the production process, right up to economic benefit formulation.

Parts of an integrated digital twin are among others, planning data from the design and engineering phase, plant data from the operating phase, and descriptions of the plant behavior in the form of models. The individual simulation models that belong to the digital twin are specifically tailored to the planned use and satisfy the respective requirements for accuracy in this regard.

Like the real system, the digital twin develops across the plant life cycle and integrates the currently available data and knowledge bases in a step-by-step, integrated way. It not only describes the system‘s behavior, but solutions for the real system are also derived from it [1].

The individual components of a digital twin are largely stateof-the-art already today. New perspectives come from the approach of integrating the individual models and software tools into an integrated, semantically coupled system, via the various hierarchical levels of a plant and via the various phases in the lifecycle of a plant.

1. 1 Models and simulation in the lifecycle of a plant

Each simulation can be considered as a virtual experiment with the goal of better understanding a system [2]. The system characteristics are modeled in a sufficiently accurate mathematical representation and calculated using common computer programs. The creation of a simulation model is thus always purpose-oriented and context-specific, i.e. it serves to answer one or more special questions. To this end, a simulation model can, for example, describe the physical, chemical, energetic and/or IT behavior of a system over time [3]. Simulations are more or less frequently used nowadays in all phases of the plant’s lifecycle and can be compiled into the following four groups:

◾ Simulation for virtual commissioning ◾ Virtual commissioning simulation ◾ Training simulation (OTS: operator training system) ◾ Simulation during operation

These four use cases of simulations are shown in Figure 1 over the lifecycle of a process plant.

Evolution of a digital twin

Plant lifecycle

Conceptual design

Figure 1: Use cases of simulations in the lifecycle of a process plant

Basic planning

Basic Design

Detailed design

Virtual commissioning and simulation-based

Setup Commissioning Operation

engineering

Design simulation [4, 5]: Use of a steady-state process simulation for the plant engineering and design. The result is represented by energy and mass balances and mass balance, the Process Flow Diagram (PFD) and data sheets for the individual units and devices. Sometimes dynamic process simulations are already used in this phase. This enables modeling of the transient behavior between operating points of the process for example, for a better design of start-up and shut-down behavior.

Virtual commissioning and simulation-based engineering [6–10]: After finishing the plant equipment design, the

automation system design will be performed. For a safe and efficient operation of the plant, the distributed control system play a key role. Therefore a correct functioning of the system is essential. The use of simulation support in this phase by signal and function testing of the engineered process control system against virtual plant models. This simulation models represent the behavior of all devices that communicate with the automation system. The configuration of the automation program that will later be used in real operations should be the one that is tested. To this end, it will either be run on the real hardware (a programmable logic controller (PLC)) as so-called hardware-in-the-loop configuration, or on an emulated hardware, as so-called software-in-the-loop configuration [11].

Since 2013, the GMA expert committee 6.11 has been dedicated to working out VDI/VDE guideline 3693 [12] on the topics of virtual commissioning. Test configurations, test methods and model types which are used in the context of virtual commissioning are introduced in sheet 1 of this guideline.

Operator training [13, 14]: The goal of a training simulation is to prepare operating personnel risk-free, efficient and realistic for their future tasks. This encompasses both working with the process control system and with the process itself.

Maintenance and modernisation

Operator Training

Operation-associated decision

support and optimization

Depending on the intended application spectrum the simulation component requirements for user interface, model accuracy, model details and validity differ greatly.

Operation-related decision support and optimization [13, 15]: The use of simulations in the operating phase is very wide. That can vary from soft sensor for monitoring and control applications up to model predictive controllers. The operator can receive support for his future decisions, by examining various production scenarios before active intervention in the process.

1. 2 Landscape of the models

The mathematical models available for the digital twin are highly diverse. Depending on the functional requirements in the intended life-cycle phase application, the degree of model accuracy can range from moderate for examination procedure controls and control strategies up to exact replica of process dynamics required for tuning control parameters. In Table 1, different components of a digital twin are displayed as lines, with the respective purpose in the lifecycle being categorized in columns.

In Table 1 the functional requirements of a digital twin, broken down by requirements for different plant components (lines) and different units or phases in the life cycle of the plant (columns). The first lines make statements about the respective model requirements. The following abbreviations are used: CPM: Control Performance Monitoring, RLT: Remaining Life Time prediction, CM: Condition Monitoring, XR: VR/AR S upport, MPC: Model-Predictive Control, RTO: Real Time Optimization, EKF: Extended Kalman Filter or soft sensor, PLC: Progammable Logic Controller; OS: Operator System; ES: Engineering System; EDD: Electronic Device Description; SFC: Sequential Function Chart; APF: Advance Process Functions

Evolution of a digital twin

Category Real twin Digital twin Software

Model types Examples

Model requirements

Units Reactor,

Components Valves, pumps,

Automation technology

Documents Process flow

Table 1: Functional requirements of a digital twin over plant lifecycle

cracker, column, etc.

m o t o r s

Sensors Device models SIMIT, EDD X X X X CM

PLC hardware Emulation SIMIT-VC X X X

PLC software Copy of the PLC

PLC user interface

diagram (PFD)

P&ID Object-oriented P&ID COMOS P&ID X X X

Data sheets EDD, @eclass X X X X

Recipes, procedures

Signal lists Link between parts

Layout planning

Measured data Measured data

Model accuracy ++ ++ + – + ○ + ++ – +

Model details ++ ++ + – + – + + – ○

Scope + + ++ – ++ ○ ++ ++ ○ +

Steadystate model

Dynamic model (coarse, wet run)

Dynamic model simplified and linearized

Dynamic model (precise)

Dynamic model (expanded by wear)

FEM (flow, thermodynamics, CFD)

Material flow models Preactor X

Characteristic curve, data, fields

s oftware

Copy of the PLC interface

Object-oriented PFD COMOS Feed

Emulated PLS SIMATIC Batch,

of the digital twin

3D model COMOS

a rchive

programs

gPROMS, AspenPlus, Pro-Il

SIMIT, Matlab

Matlab, PID tuner, CPM X X X MPC CPM

gPROMS, ACM

gPROMS

Star-CCM+

SIMIT, Excel, ValveApp,

ES project

Virtual OS

APF, SFC

COMOS, HW-Config, SIMIT

Walkinside

OSI PI, InfoPlus.21, Historian

Planning Commissio-

Basic engineering, concept

Detailed engineering,

operation modes

Automation concept

XX RTO

X X X EKF

XX XX X CM

XXX

XX X

XXXXX X X

XXX

XXXR

ning

Configuration

Virtual commissioning

Operation Main -

Real commissioning

Training

Optimization, APC

Production planning

XX RTL

X X X CPM

tenance

Maintenance planning

Evolution of a digital twin

Reading example: A dynamic simplified linearized model of a unit (reactor, cracker) can be used in the plant life cycle for the following tasks: Planning of the automation concept (PID controller design), virtual commissioning, real commissioning (using or updating the model for PID tuning), optimization in the operating phase (MPC), and control performance monitoring (CPM). For this application the model will be transferred into different software tools respectively updated, for instance Matlab, PID-Tuner, MPC-Configurator, CPM. An exemplary use of the model within the framework of an APC application (e.g. MPC) sets greater requirements to the model accuracy and details than an application in maintenance planning (CPM). The exact dynamic model is realized in gPROMS and can be used for planning of procedural alternatives (recipes) within detail engineering, for planning automation concepts (for example loop-paring: allocating of actuators to control loops), and in operation phase for training purpose (operator training) and model based soft sensors (EKF).

The virtualized representation of a device, system or even an entire plant requires the description of the real behavior using models. For simulation and modeling software, there is a comprehensive offering on the market, often specialized for devices, machines, instruments and plants, such as pumps, distillation towers, polymerization reactors or steam crackers. Yet, the integrated use of the models in a digital twin leads to a new level in usability of the model, thanks to the high degree of connectedness.

In information technology, for the modeling of complex systems a distinction is made between static type level (class definition) and dynamic instance level (objects). For example, a pump is a device with static properties defined by type, which generates a reproducible pressure or flow rate in a value-added process. Instance-specific values are dynamically assigned to the properties at each individual real pump. Comprehensive domain know-how from process and automation technology as well as the corresponding software programs are required for creating models.

1. 3 Vision

Considering the listed advantages, the question arises why simulation are not integrated and by default used over the entire plant lifecycle today. In addition, today‘s use cases are often isolated from one another, i.e. models and experiences are seldom reused. The literature already contains suggestions on how simulations can be used in a more integrated way. Bausa and Dünnebier [4] are investigating, for example, how mathematical models from the design phase can be reused for optimization of the later operating phase.

The role that simulation plays in the product and production lifecycle was examined in the section on discrete production in [16]. Of course, industrial practices are still far away from the integrated use of simulations. How the integrated use of simulations in the lifecycle of processing plants has been implemented to date and how it can be implemented more comprehensively, was worked out in detail in [17]. The following image of the future is depicted in [18,19]:

„In the future, simulations will be systematically used and will be an integral part of the normal engineering and operating processes over the entire life cycle of process plants. The basis for the engineering and operation of a plant will be a virtual depiction of the plant. Decisions will be evaluated and made based on the virtual plant. New plants will first be planned and developed virtually, and even in existing plants, no changes will be made before a preceding check in the virtual plant. Once developed, models will be reused and refined over the course of the lifecycle. This will be supported by available exchange and co-simulation standards. The configuration of the simulation models is done modularly to allow reuse and an efficient layout. Simulation models (modules) can be connected to each other in the sense of „plug-andsimulate“. Models are provided by manufacturers of the real components as standard feature to use them for the layout of the overall virtual plant. The process of creating models is easy and the first models for the virtual plant can be derived from existing planning data (especially for existing plants for later creation of the virtual plant). This is possible on the basis of integrated information and data management between planning, simulation and operating data. A connection between real and virtual plants allows continuous optimization of the virtual depiction and support in answering operations-associated questions. Thanks to the parallel operation of the virtual and real plants, continuous optimization relative to factors such as costs, time, energy consumption and resource consumption of the real plant is possible. A highly fluctuating demand can best be answered by the virtual plant, because predictive simulation calculations can always ensure the ideal operating state. In addition, the engineering and operating know-how is always up-to-date and can be called up by all participants in the form of the models and data of the virtual plant. Finally, simulation is part of the training at institutes of technology and is a widely accepted technology and method.“

This article takes on this image of the future and puts it in the context of the discussions about the „digital twins“, which have been going on for some years. In addition, the concrete example of an ethylene plant shows how far this vision can be implemented already today.

Evolution of a digital twin

2 Example application - Steam cracker

A „steam cracker“ (cracking furnace) is given in this article as an example of a practical application of the general concepts.

Steam cracking is a method in the petrochemical industry in which long-chained hydrocarbons (naphtha, but also ethane, propane and butane) are converted by means of thermal cracking in the presence of water vapor into short-chained hydrocarbons such as ethylene, propylene and butane. A cracking furnace is one of the most complicated units in petrochemical plants. It is used to manufacture intermediate products, which are mainly processed to become plastics (such as polyethylene), paints, solvents, or insecticides.

Exhaust Gas

Hydrocarbon (educt)

The cracking furnace is a tubular reactor with several separate coils in which the mixture is heated up to temperatures of approx. 840 °C. The long-chain molecules are thermally cracked within fractions of seconds.

In Figure 2 a simplified steam cracking furnace with one tube coil is shown. COT („Coil Outlet Temperature“) and TMT („Tube Metal Temperature“) describe temperatures that are relevant for the control concept described later in the digital twin use chapter.

NOx CO O2

Dilution Steam

Air

Fuel Gas

Figure 2: Schematic representation of a steam cracking furnace with tube coils

TMT

TI TI

COT

Cracked Gas (product)

+ 15 hidden pages