DuPont is a big proponent of the digital twin, wherein all the elements and dynamics of physical plant assets can be replicated in software for process assessment and optimization and operator training.
“Using digital twins throughout the plant lifecycle is an idealized path for DuPont,” said Yuris Fuentes at the Siemens Automation Summit 2018. Fuentes is an engineering consultant in the Process Dynamics, Control, and Analysis Group in DuPont’s Engineering Technology Center. “DuPont has long been a proponent of virtual plant technology, which is why it created its own proprietary dynamic process simulator, known as DuPont™ TMODS. This simulation technology uses chemical engineering principles to simulate real-life operating chemical plants and can connect to an offline DCS for process configuration development, assessment and training.”
DuPont uses Siemens SIMIT simulation software in conjunction with its digital twin technology in three different ways to assess plant lifecycle issues.
The first of these uses what DuPont calls a “plain virtual plant— where we use dynamic simulation software with first-principle models, including all physical and thermodynamic properties, reaction kinetics and mass and energy balances,” said Fuentes. “You run this on the native HMI, not the DCS. It’s like a flight simulator for running a plant, with all regulatory control, sequence or batch logic being handled by the software.”
Another method used by DuPont is to connect the digital twin with the DCS so that the digital twin becomes a virtual plant, passing actual signals back and forth with the control system. “Now you are running on the DCS HMI and SIMIT just acts as a bridge to pass data from/to the DCS to the digital twin,” said Fuentes, noting that, in this scenario, DuPont connects its digital twin to SIMIT via OPC.
A third DuPont method involves using the digital twin to do “everything in a standalone fashion,” said Fuentes. “Here, we’re not directly connected to the plant DCS in any way. Instead, all the different aspects of the plant configuration, such as alarms and interlocks, regulatory control, sequence and batch logic are all simulated in the virtual plant” he said.
In this scenario, the DCS HMI is emulated within the simulation software with an acceptable degree of realism.
These three virtual plant alternatives are used by DuPont throughout a plant project lifecycle.
DuPont’s 6 Project Lifecycle Facets
The typical project lifecycle at DuPont includes six specific phases covering everything from the initial process conceptualization to the maintenance of ongoing operations. Virtual plant technology (i.e., dynamic simulation via a digital twin and SIMIT) has been utilized by DuPont in each of these areas. Fuentes noted that these six steps “show what is possible with the digital twin.”
Process Conceptualization. At this stage, dynamic simulation can be used to investigate process alternatives. This takes place during R&D or basic project data development. “The purpose is to augment steady-state economic analysis with operability analysis,” said Fuentes. “These analyses help ensure that the process will start up, shut down, maintain good operations, transition and reject disturbances.”
Tasks involved in this step include building a high-level dynamic model of the process for testing of process alternatives and augmenting basic data with equipment to ensure good operability while achieving desired quality and economics.
“This is where you find out if you can really operate at your conceptualized steady state,” said Fuentes. “And you can play around with it since it’s not connected to DCS at this point.”
Control Strategy Concept Development. During this stage, dynamic simulation is used to test alternatives and set a basic control strategy. This occurs before the control system vendor and system integrator are selected and before any DCS connection is made. The purpose here is to design effective batch and continuous control concepts that meet project objectives.
Tasks involved include increasing the fidelity of the dynamic model of the process; testing alternate control concepts while observing the impact on product quality, yield and energy use; demonstrating the simulation of the process to the project team using the best control approaches to get their buy-in; creating the functional requirements document using the simulator; and using the simulation as a supplement to the functional requirements document to help the system integrator understand the project requirements.
“Think of this step as a way to save time and money, as this model allows you to test simulations quickly because it can run faster than real time,” said Fuentes.
Development and Testing of DCS Control Configurations. Virtual plant technology can be used for the iterative development and testing of control configurations during production design. This is done when the system integrator is developing the DCS configuration. “At this point, you can catch any configuration problems—due to bad design or bad implementation—as early as possible to improve the quality and reduce the cost of developing the configuration,” said Fuentes.
Tasks involved in this phase include connecting the dynamic model to a DCS system using OPC; incrementally testing the DCS configuration as it is being developed; and turning over control of incremental portions of the process to the DCS to monitor process performance and operator graphics. The tasks done here with the Virtual Plant technology are “much like what you would do for a real plant startup,” Fuentes said.
“We like to work hand-in-hand with the DCS integrator team to test out configurations and catch problems early— before they are replicated to the entire configuration. We do this by making the simulation available at the vendor location where the configuration is being developed. This allows for iterative development and testing to be performed by both teams.”
Control System FAT (Factory Acceptance Testing). During the FAT, a DCS connected virtual plant (i.e., a dynamic model connected to the DCS via OPC) can be used to perform a more thorough checkout of the entire configuration running through important plant operations like startups, shutdowns, grade transitions and testing interlocks for continuous processes; and running multiple batches, testing various recipes and fault handling for batch processes.
“We’ve had great success catching configuration errors and identifying improvements prior to startup with this step,” said Fuentes. “By this point, you have the virtual plant set up to test everything, meaning that you can run sequences to make sure they do what they’re supposed to do. When you test a sequence, you not only see the immediate effect in the virtual plant, the model will tell you all that will happen as a result. For example, because you opened that valve here, you may not realize you’re overfilling a tank downstream. It’s not just immediate equipment feedback that you get from this step, but information about things beyond the device that you are you actually testing.”
Operator Training. The operator training typically occurs a few weeks before the “real plant startup,” and it can be done using a virtual plant connected to a DCS or with a standalone system. Fuentes said the goal here is to “minimize startup time and reduce incidents caused by poor operation.”
Specific tasks include training control room operators on the DCS, operator graphics and process operations; teaching operators how to handle both normal and abnormal situations; and running through the scenarios multiple times in this safe environment to reinforce the training.
“We’ve been doing operator training for initial startup at DuPont since the 1980s,” Fuentes noted. “You want to have operators get familiar with the process. You may not be training them on how to use the DCS at this point, but just having them understand how to run the process conceptually is a big benefit. That’s where standalone systems help. People may note that the faceplate are not be the same as in the DCS at this point—since it’s not DCS connected in standalone mode—but the processes are the same, so it helps operations personnel understand them. This is one of our main uses of virtual plant technology, because users learn how to run the plant before actually running it. You can even drive the plant to states you’d never be allowed to in the real world. This helps operators learn how to deal with potential worst-case scenarios.”
Ongoing Operations. After the plant is up and running, virtual plant technology can still be used for periodic retraining of operators and continuing assessment of control system modifications. As such, this step takes place throughout the life of the process with the purpose of maintaining operating discipline and reducing errors due to configuration or process changes.
Tasks involved include periodically retraining existing operators on normal and abnormal situations using the simulator; training all new operators using the simulator; and assessing all significant changes to both the control system or the control strategy by first implementing on the simulator.
“This is particularly valuable for testing changes to control or batch logic prior to running in the real plant,” said Fuentes. “Even after the plant is running, you can still use the model created for training. You can use it as a training tool and as a predictive tool (i.e., as a soft sensor) running in parallel with the real plant. We also have used the model in an advisory mode because it can run faster than the plant and provide predictions..”
According to Fuentes, Virtual Plant simulation technology becomes more cost-effective the more you use it. “If you’re only using it for operating training simulation, it’s easy to see it as a big investment you had to make. But the more you use it, if the model is good and used throughout the project as we do it, it is an easily justifiable expenditure.”
He added that DuPont is seeing this approach of incorporating simulation technology beyond operator training and into continuous plant development and improvement as “becoming more mainstream.”
