Model-based reasoning and control of process plants: Dissertation

In addition to stabilizing feedback control, safe and economic operation of industrial process plants requires discrete-event type logic control, for example automatic control sequences, interlocking and protections. A lot of complex routine reasoning is involved in the design and verification & validation (V&V) of such automatics. Similar reasoning is required in planning and scheduling plant operations, action planning and fault diagnosis. Much of the required reasoning is so straightforward that it could be accomplished by a computer if only there were plant models which allow versatile mechanised problem solving. Such plant models and related inference algorithms are the main subject of this report. Applying model-based reasoning in the design and V\&V of plant automatics and in action planning during plant operation is discussed. A prototype tool called ISIR (Interval SImulation and Reasoning) for model-based reasoning is presented. The approach is based on the principles of QSIM, an algorithm for qualitative simulation, and it is implemented in constraint logic programming language CLP(R). ISIR is applied to the verification and synthesis of a simple discrete-event control strategy of a continuous process, to a power plant feed-water system and to various standard examples from the qualitative simulation literature. The emphasis of the report is on presenting the principles of the approach, and on demonstrating its potential applications. Because constraint logic programming and the programming techniques developed in this work may be utilized as such, programming examples and most of the ISIR-algorithm are discussed in detail. The results are well-defined principles, a prototype implementation, some demonstrations and ideas for the specification of a general-purpose tool.