Process simulation and modelling

What is the difference between them?

Process simulation is a subset of process modelling

The 'process lifecycle'

A process can be considered to have a lifecycle, from the time it is conceived (conceptual design), through the various R&D and engineering phases to commissioning and operation and finally decommissioning

A powerful general-purpose process modelling tool such as gPROMS has a role to play in every one of the stages, not just simulation of the process end engineering designs.

Process simulation and process modelling refer to different things.

In the gPROMS context, process modelling involves building a mathematical model of the process (or of a product, for that matter) by describing its fundamental physical and chemical relationships – without specifying how they are to be solved.

Process simulation is merely one of the activities that you can perform with that process model.

Process simulation is often an exercise in 'molecule accounting', and is often performed by relatively inexperienced engineers.

Constructing a high-fidelity process model, on the other hand, requires deep modelling and process expertise, and is usually performed by an experienced specialist – sometimes working in conjunction with R&D personnel.

What does this mean in practice?

Much process simulation is carried out using "off-the-shelf" black-box models that provide little competitive advantage, or purely steady-state models that do not capture the complexity of process operation. Most process simulators on the market provide little in the way of custom modelling or data-fitting capabilities.

By contrast, PSE's gPROMS is a true process modelling tool such that allows you to create and apply first-principles models that – particularly when used in the context of a model-based engineering approach – provide high-fidelity predictive information for a unit or process. This information can be applied in many different ways to generate value.

What can you do with a true process model that you can't with a simulation?

Consider, for example, a detailed model of a fluidised bed reactor and its surrounding flowsheet.

With this model you can of course perform steady-state and dynamic simulation runs to see what happens if feed conditions are varied. What is more, you can do that with a high-fidelity custom model that closely reflects your actual process rather than a ‘generic’ process.

However with the model you can also:

estimate parameters using gPROMS's model-based data analysis and validation techniques on that model, against experimental data. This can enhance predictive accuracy significantly, and provides information that can be used in formal risk analysis.
design experiments to refine the parameter estimations and reduce the risk associated with measurement inaccuracy.
perform optimisations – dynamic or steady-state – on the model, to directly calculate optimal trajectories or values or design variables or operating conditions rather than undertaking lengthy trial-and-error invetigations.
generate linearised models for use in control design applications or Model-based Predictive Control (MPC), gain scheduling or any other activity that requires linear models.
because this is a model and not a simulation, simulate 'backwards' to find out what feed or unit values give rise to the desired product qualities, at no additional cost in terms of execution time or complexity of model.

Only with a process model will you be able to perform all the activities required to model across the process lifecycle, from conceptual design and laboratory experimentation through detailed engineering design to operation.