This white paper discusses the application of model fidelity to dynamic process simulators used for control system software acceptance testing or operator training.

We are often asked to help users define the simulation complexity or fidelity required to test automation system application software or train operators. This is probably one of the most confusing tasks when specifying or applying process simulators. Wrong decisions at this point of the project can confuse the user into spending much more money than needed and may result in a simulation system that is inoperable.

This is an important subject and warrants a careful review to help sort out the real needs of the application versus the claims of vendors or industry consultants. Because MYNAH solutions are built for automation software acceptance testing and operator training, this discussion will focus on those applications and only touch on other applications like process design.

Simulation Fidelity or Complexity

Fidelity is the common industry term used to describe the rigorousness of process models and the level of complexity. Many times a simulation application will be categorized as low, medium, or high fidelity. A description of each category is as follows:

• Low Fidelity - simple IO signal modeling, device tiebacks, value initialization. The model will require user intervention to respond to automation system actions.
• Medium Fidelity - mass balance model, heat balance model. Process streams are modeled as composed of a single component. The model will run automatically and respond to automation system actions and process changes.
• High Fidelity - complete mass balance, rigorous heat balance, reaction kinetics. Process streams are modeled with individual component thermodynamic properties.  The model will run automatically and respond to automation system actions and process changes in a very similar manner to the designed process.

In addition, the term First Principle simulation is also used interchangeably with High Fidelity. This is a misuse of the term. First Principles modeling is correctly defined as a simulation based upon laws of conservation of mass, energy, or motion using physical and chemical characteristics of the process.

Empirical modeling refers to the use of realistic limits to the model using actual or assumed process correlations or data. In actuality, every properly implemented medium or high fidelity process model will use some degree of First Principle and Empirical modeling.

A better term than simulation Fidelity is simulation Complexity. The needs of the application will dictate the best application of model fidelity. A model for a distillation tower will require a higher degree of complexity to function correctly than one for a tank farm. However, more complex models cost more to engineer and maintain. More complex models may also be less flexible when the model is applied to different uses.

Dynamics - Essential for Testing and Training

Simulators that are developed for process automation system software acceptance testing and operator training have the additional need for good dynamics. There are two categories of process models:

• Steady State models are generally used for plant and process design. A steady state model does not explicitly simulate transitions between process states. Inputs to the model affect the outputs from the model without time delays or lags.
• Dynamic models account for transitions between process states. Outputs to the model are affected by the inputs to the model along with the time delays and lag characteristics of the model.

Steady state models are not usable for automation system software acceptance testing or operator training. These applications require Dynamic process models.

Applying Simulation Complexity

There are two factors that dictate the level of Complexity of simulation that should be used: the application and the nature of the process.

The application refers to the part of the automation system application software being tested or the objectives of the training session. Time is of the essence in most cases. The simulation must be ready when the project dictates it is needed.

The lower the level of Complexity of the simulation, the quicker it can be developed and updated. A general guideline for simulation Complexity based upon the automation system level is:

• Control Modules - Tieback Simulation, Automated Test Scripts
• Equipment Modules - Tieback Simulation, Limited Dynamics
• Sequence, Batch - Mass Balance, Temperature and Pressure Dynamics
• MES, Advanced Control Applications - Mass Balance, Heat Balance, Kinetic Models

The nature of the process also will dictate the level of Simulation Complexity required. The less complicated the process, the lower the level of Complexity that will meet the need. Mass-balance, heat-balance models will be adequate for simulation of standalone and many batch processes. More complex models and simulations will be required for continuous, highly-integrated process units common to refineries and petrochemical plants.

Within a process unit, some unit operations may require a different level of complexity than others. A tank farm can be simulated with a simple mass balance, while a complex distillation tower will need a rigorous, dynamic, mass-balance and heat balance.

The term Selective Fidelity has been used to describe a cost effective balance of high and medium complexity models, selectively applied to the unit operations of the process. This approach provides the best balance of simulation performance, cost, and flexibility.

The guidelines above apply to simulation systems for Software Acceptance Testing and Operator Training. The goal of the simulation is to provide accurate input signals to the control system so that the system can be tested and effective operator training can be realized.

In contrast, the goal of steady-state models for process design is to provide accurate models of the process so that the design of the process can be analyzed and tested. Applying the requirements for steady-state process design to models for testing and training will force the user into a complicated, expensive, inflexible solution.

Other Factors to Consider

Simulation fidelity discussions tend to get most of the attention when specifying simulation systems; there are other factors that deserve equal or greater attention.

The integrity of the off-line process automation system deserves attention. In order to achieve successful testing and training of the automation system, the control system must run in native, standard mode without modification to the configuration. This requires the use of simulation systems that support Non-Intrusive Interfaces. Non-intrusive simulation interfaces are external to the automation system and interact with the IO level of the system without modifying the control system configuration.

Simulation system Flexibility is another key requirement. During testing, the simulation system must be capable of being run in several modes. These may include fully simulated mode or a semi-manual mode. In a semi-manual mode, the user needs to be able to introduce process signals or run automated scripts without fighting the simulation. The simulation system must be easily modified. Changes to the control system that will affect the simulation system must be incorporated into the simulation system quickly and easily.

Operator Training Simulators should allow the development of complex or simple training Scenarios. In order to preserve the integrity and the maintainability of the process models, these scenarios should be external to the simulation models. The simulation engine must also be flexible enough to allow a wide variety of training system architectures including concurrent or parallel sessions.