Operator training dynamic simulations reviews

[mrtangent]

Dear all,

I am looking for some background information on dynamic simulators which are used for operator training. Partically the acceptance test for matches to the plant performance and how well the actual v plant responce has / can be modeled. I need to set up some acceptance criteria for a dynamic package however our experience with the dynamic modelling is rather limited.

 

[sshep]

We installed a dynamic simulator training system at one of our plants which had a history of operating problems involving chemical releases and bypasses of an incinerator. The immediate objective of the project was to train all operators on how to properly handle the various upsets which could lead to these releases. The longer term objective was to stay in operation in the face of regulatory agency scrutiny and fines. These objectives were successful. This was approximately 7 years ago so state of the art may have changed since then. Our model was developed by Aspen Technology Inc out of their Colorado office. Unfortunately we later sold off that plant so I am not sure if the dynamic simulator is still in service. I worked on the model development during that project, but not on the console communications.

The simulation should have a specific training objective- i.e. train staff of a new plant or to handle problem in an existing plant. To build a model to this detail in hopes it will give new operational insight is not economical in my opinion. If the model has a long term use then consider it a bonus. Ability to meet the model objectives will essentially set the FAT criteria.

I can offer these FAT criteria suggestions:

General
•Model should run at uniform speed (transients vs steady state) and in real time.
•Model should never fail at any legitimate operating condition or change.
•Model should start reliably from any preconfigured start point, intermediate save point, and current plant operation (if so designed).
•Model should only operate in real space. There can be no composition fractions less than zero or greater than one; no negative absolute pressures or other silly results of mathematical solutions.
•Model should either terminate with humanly understandable failure report, or continue to operate realistically (i.e. simulate relief to flare or overflow, etc) at all normal operating bounds. Note: Operators may take great pride in being able to "break" a training model, where as an engineering model does not have to run beyond the operating space.
•Model should alarm at all plant alarm points and shutdown in exactly the same manner as the plant.
•Controller tuning in the plant should be tested to see that it can be brought into the model as designed (hopefully automatically) and have the same response in the plant as the model. Do this by picking a few controllers which can be changed in the plant and check the open and closed loop responses in both the plant and model.

Unit Operations
•Trayed towers can give excellent model results if hold-up is modeled correctly- response times should not deviate by more than a few% from plant step test observations. Packed towers can be more difficult to model but should likewise not show large response time deviations.
•Rotating equipment dynamics should be realistic and include surge points, critical frequencies, etc.
•Reactors are often custom models and can be anything from empirical fits to rigorous first principles. For reactors you will have to choose your own criteria- conversion, duty, pressure drop, etc.
•Other unit operations are generally easy to model, but you may have special exchanger situations or other details which merit specific tests.

Miscellaneous
•Communications between the operator and trainer workstation should be reliable and work as intended.
•Weather related disturbances (if included) should be tested to see that they give a reasonable response relative to the plant.
•All other disturbance events which you have included in your project specification (i.e. starting equipment, instrument failure, exchanger fouling etc) should be tested and give a reasonable result.

This was my experience anyway. Best wishes, sshep

 

[brainstorming]

sshep gives excellent explanation and I can't say much on this except that last project I worked got Aspen Tech. dynamic simulator. It is working well and operations got training and they have been trained well for various plant upset conditions.
Things I remember are:
- Colours are important to show & reflect Shut down actions..etc
- Logics should be configured well

Regards

 

[mrtangent]

Thanks for the advice shepp and brainstorming

much appreicated.

 

[johngerhart]

2 suggestions :
1) look at NEXUs as a program for simulating and setting plant operating rules
2) Review the required resources , if you can't spare the people , don't bother with the project . Solid , dedicated time committment is an absolute requirement to succeed