SPECIAL REPORT: Dynamic simulation useful for reviewing plant control, design

Aug. 13, 2007
While plants continually become more complex, chemical engineers, with the advent of computers and commercial off-the-shelf simulation software, can now successfully redesign processes and control systems much faster and better than in the past.

While plants continually become more complex, chemical engineers, with the advent of computers and commercial off-the-shelf simulation software, can now successfully redesign processes and control systems much faster and better than in the past.

Technology used to optimize material flows and increase plant throughput, maximize equipment efficiency, review resource use, and enhance planning and logistics, is now mature. Furthermore, access to experimental data, integrated solutions and “what-if” scenarios is easier and more valuable than ever before.

This article explores some of the process modeling techniques used with emphasis on dynamic simulation.

Chemical engineers can perform difficult calculations in a short time and use tools fit for purpose, i.e., steady-state tools for basic design and dynamic modeling for reviewing controllability and distributed control system (DCS) checkout. For a project-driven organization, however, and in which reliability and accuracy of results are paramount, the following questions constantly arise:

  • Are the thermodynamics and the reactor kinetics correct?
  • Are the data correct? Do we understand the control scheme?
  • Are there any smart methods to perform the allocated tasks in less time?

Understanding unit operations, questioning the model-obtained results, and asking questions to the more experienced engineers are the strongest requirements for success in using dynamic simulations for process control and design.

Process modeling

Using first principle, rigorous models allows engineers to simulate any process in great detail and analyze scenarios and situations that, in the real plant, would be difficult to monitor or are dangerous. For plants that are being designed, one must:

  • Demonstrate overall controllability, safe operability, reliability, and the adequacy of process for all start-up, operating, transient, and shutdown conditions.
  • Meet regulatory demands for emissions and depressurizing.
  • Test and accept the plant’s DCS and emergency shutdown schemes.
  • Develop operating procedures and train the operators to maximize production and quickly react to abnormal situations.

Conversely, for existing plants, engineers are typically required to:

  • Maximize production and debottleneck specific plant areas.
  • Verify that a planned DCS update will be smooth with minimal process disruptions and ensure that new automations and sequences will operate as intended.
  • Improve control strategies and implement advanced control.
  • Financially optimize the costs to the added-value balance.

In this article, we first review the process modeling requirements for some of these obstacles. This article also examines the differences between standard steady-state process simulation and dynamic modeling.

Finally, the article also reviews special requirements of modeling for dynamic studies or a detailed checkout of the DCS logic and control loops.

Steady-state, dynamic modeling

Steady-state modeling means that the modeled process is solved only for a specific set of operating conditions. This is like a snapshot of the unit operation. Any change in the plant conditions requires solving the model again. After converging, the model should predict where the process will settle.

On the other hand, dynamic modeling provides information about the unit operation over time. All variables are solved at each time step and at any specific time that the process conditions are monitored. Compared to the steady-state “snapshot” equivalent, dynamic modeling is more of a movie than a single picture.

Using one or the other technique really depends on the operator’s requirements. For process design, a steady-state model of the unit is typically sufficient. When unit controllability is in question or the process response to transients must be investigated, however, a dynamic model is needed.

The main difference in dynamic and steady-state modeling has to do with the level of detail required. Steady-state modeling uses process specifications and focuses on process feasibility.

For example, if the temperature at the tube side of a heat exchanger is equal to a certain temperature, a minimum temperature approach is assumed. In dynamic simulation, the heat transfer coefficient of the exchanger should be estimated first from existing data (equipment datasheet, heat and material balance, or equivalent). Then the flows are estimated based on pressure drops and resistances across the heat exchanger. Only then, the temperature at the exchanger exit can be calculated.

Furthermore, one should understand the tricks used. For example, in steady-state modeling, pumps are rarely modeled; flow is possible from low to high pressure; distillation column reboilers and condensers are typically integrated within the first and last column equilibrium stage. In dynamic modeling, resistances across valves and piping are important; pressure and flow boundaries are typically used; and exact pump and compressor curves should be used.

Engineering studies

When increased confidence in the unit design is required, a detailed engineering study is recommended. This is typically the case for expensive equipment exhibiting controllability difficulties during major process upsets, like start-ups or shutdowns (normal and emergency). Model results are evaluated, issues uncovered, engineering solutions proposed, and finally, the unit design is modified.

The nature of engineering studies is such that modeling should be as detailed and exact as possible. In addition to the process and instrumentation diagrams and process flow diagrams, detailed datasheets are required for all equipment in scope.

Moreover, isometric diagrams are needed so that the actual pipe volume and resistances can be evaluated and taken into account. Lastly, the entire process-control philosophy must be incorporated so that the model-unit responds to disturbances in exactly the same way as the real unit.

The modeling engineer should be experienced and should have a clear understanding of the unit operation and critical parameters before starting the modeling. The engineer therefore focuses on the important modeling aspects and will be able to interpret the observed process behavior during testing. Typically, a thorough understanding of electric motors, steam and gas turbines, and relative inertia calculations is required.

DCS check, training simulators

A major undertaking for a new plant is development of the control system. It is the online, real-time software integrated with the necessary hardware that will ensure that the plant can be properly operated.1 2

Engineers test the DCS operation by actually triggering one by one all inputs and verifying that the output is behaving correctly.

Click here to enlarge image

Fig. 1 shows typical study results for a motor-driven compressor that is tripped. Before delivering this result to the client, the modeling engineer must verify that the results obtained by the model have physical meaning. The engineer must ask “is the reverse flow observed real or a model artifact?” or “are the calculated flows reasonable?”

This is a lengthy and costly endeavor. There is no easy way to follow though the DCS code, however, especially when complex sequences, alarms, and shutdown sequences are involved.

The preferred solution is therefore integration with a dynamic process model. System snapshots can therefore be saved easily and recalled when required, while process and DCS modifications can be readily made and evaluated.

Major differences between the dynamic models suitable for DCS testing and those suitable for training simulators have to do with the available graphics. This is due to the fact that engineers are performing the DCS testing, and operators will be using a training simulator. In the typical case, engineers do not have a great need for fancy graphics with buttons and emulated hand-switches or bypass signals. Many variables will be triggered directly from the model.

Trends in process modeling

The past few years have seen the rise of integrated modeling environments that provide the ability to model and optimize the plant (engineering suites, point solutions) or the whole enterprise (supply chain, demand planning, etc.). These products have matured; they are not “engineering tools” any more, but commercial off-the-shelf products, ready to be used.

Currently:

  • The main driving force behind major product development is model reuse and its ability to be used for the plant’s entire life.
  • Interactions of the major engineering suites with the “environment” become standardized.
  • Upward product compatibility is maintained.
  • Connectivity with external or client-delivered models and applications is achievable using a variety of tools (Dll, OPC, MS Excel, VBA, Fortran, C++, C#) is provided.
  • Windows-based systems take up the space previously reserved exclusively by UNIX.
  • Rival products exploit standard connectivity. Using the proper API calls or the proper model export and import facilities (text, xml) and some programming or scripting, one may use models under a competing product suite.
  • Modeling is becoming a commodity. Only web-based technologies supported by major groups seem to be those that will survive in the longer term.

References

  1. “Distributed control system,” Wikipedia, http://en.wikipedia.org/wiki/Distributed_control_system.
  2. “OPC dictionary and glossary of terms,” MatrikonOPC, http://www.matrikonopc.com/resources/dictionary.asp?range=a

The authors

Sissy Psarrou ([email protected]) is a senior engineer for Hyperion Systems Engineering, Athens. She has 8 years’ experience in the detailed modeling of refining, chemical, and polymer systems for engineering studies and operator training simulators. Psarrou holds a PhD in Chemical Engineering from the National Technical University of Athens. She is a member of the Technical Chamber of Greece.

Yiannis Bessiris ([email protected]) is a business development manager for Hyperion Systems Engineering, Nicosia, Cyprus. He has also served as a process engineer for Procter & Gamble, Brussels. Bessiris has 7 years’ experience in the development of large scale and full scope high-fidelity process simulations for the oil, gas, and power industries. He holds a MSc in chemical engineering from the University of Patras, Greece.

Ivor Phillips ([email protected]) is a technology manager for Hyperion Systems Engineering, Nicosia, Cyprus. With 15 years’ experience in the process simulation industry, he has also served as senior consultant and lead engineer at Fantoft UK Ltd. and ABB Simcon Inc. He holds a BSc in chemical engineering from Loughborough University, UK.

Vassilis Harismiadis ([email protected]) is a business development manager for Hyperion Systems Engineering, Athens. He has 9 years’ experience in the oil and gas industry with particular emphasis on the use of latest technology to improve plant effectiveness. His special expertise is in the area of operator training using virtual plant technology. Harismiadis holds a PhD from the National Technical University of Athens in the thermodynamic modeling of complex systems; most of the work was performed at the Shell Research Technical Center, Amsterdam. He is a member of the Technical Chamber of Greece.