Automatic model execution
Person accepts or rejects any operating advice
Real-time closed-loop models
As for open-loop models
Interface directly with plant control system and adjust the process automatically
Table 1. Real-time models are more capable but take more effort to make robust.
- advising on operating set points for individual equipment items;
- achieving a reconciled plant mass balance;
- determining product properties;
- analyzing energy usage;
- comparing actual versus design performance;
- responding to changing market conditions;
- meeting product specifications; and
- retaining and enhancing process knowledge.
Even though they may connect to real-time data systems, off-line models aren’t fully automated; a person normally initiates runs.
Models produced during the design phase usually require additional work before they can be used as off-line process models. After all, in design the simulation is created and run by an experienced engineer, who understands the constraints of the model and the range of valid conditions. If difficulties such as convergence failure occur, the design engineer knows how to overcome such problems.
For use in operations, the model must be tuned to match plant conditions and the particular calculations being executed. For instance, the plant setup may change from day to day — with different product grades being produced and individual units or controllers switched on/off. The off-line process models must account for these specifics.
In addition, because the simulations only are valid within a limited range of operating conditions this range must be strictly understood and enforced. The models will need to be made robust, so they always converge within the valid operating ranges. Model inputs (both those entered manually and those coming from real-time data systems) must be kept within these ranges; this often is done by running the models through a simpler custom interface, such as one based on Excel instead of their normal “engineering” user interface.
The sidebar provides other practical pointers for moving models from design to operations.
The next steps
If an off-line process model gets regular use in operations, it may be appropriate to convert it to a real-time open-loop model. The model execution then can be automated to occur, say, once per shift, every N minutes or when triggered by a process event. Such open-loop models also may write results back to the plant’s real-time data systems. However, the results of the model are always evaluated by a person, who ultimately accepts or rejects any advice or data.
|When moving models from design into operations, pay particular attention to the following points:|