Modelling issues associated with the synthesis of process operations on a plant-wide scale are discussed, and a hierarchical, distributed, object-oriented modelling framework described. The modelling approach is generic, and is demonstrated to be suitable for computer-automation of the synthesis of operating procedures. In Part II, provably correct and complete nonlinear planning algorithms based on these models are presented. It is shown that domain-independent planning methodologies using the functional operators required for the synthesis of operating procedures are computationally intractable. Consequently, domain-specific knowledge is exploited within the modelling structure presented in this paper in order to define a tractable methodology for nonlinear planning of process operations. In this paper, we have discussed some of the issues involved in automating the synthesis of plant-wide operating procedures for continuous chemical process. In particular, we have focussed on the question of how to effectively model, on the computer, knowledge about the operation of chemical plants. Previous attempts at modelling have either relied on the restrictive operator- based modelling paradigm, or have employed extremely simplified or situation-specific models. A modelling technique based on a functional operator structure was also considered. This representation accounted for the fact that the outcome of applying an operator is dependent upon the state of the system before execution, and was shown to reduce to the conditional action scheme investigated by researchers in artificial intelligence. Planning with conditional operators involves a plan generation step which is NP-hard, and is therefore considered to be computationally intractable. This justifies the adoption of a domain-specific approach to the planning of process operations, and explains why no algorithmic solution has yet been developed. Realizing the limitations of the operator model, a hierarchical, distributed, object-oriented modelling structure was described. The models are expressive enough to allow for automatic generation of interunit constraints, which is accomplished through propagation of known values through a network of relations representing the physical laws governing the behavior of a chemical plant and constraints induced by practical engineering considerations. The modelling framework is sufficiently general so as to allow the use of both quantitative information as well as qualitative knowledge, such as order-of-magnitude estimates. Examples of the construction and use of the hierarchical modelling structure were presented, and implementational issues discussed. The nonlinear planning methodology that makes use of the modelling structures described here is presented in Part II of this series.
ASJC Scopus subject areas
- Chemical Engineering(all)
- Computer Science Applications