Study On Applying Al Techniques In Separation Process Synthesis

Process synthesis is one of key techniques in the process design and development. In recent two decades, it has been a popular research area in chemical engineering, but the combinatorial problem associated with process synthesis restrains the algorithm based on mathematic models to be widely used in practical applications.Separation process synthesis is the typical problem in the process synthesis and separation process is one of two key processes in chemical production flowsheets, therefore how it is designed affects the benifit of the whole production process directly. However, traditional experient methods have been used in design companys to synthesize separation process by hand but computer, which includs selecting separation method, mass separation agent, separation device, etc., sequencing the split points, optimizing the separation flowsheets, and so on.In this dissertation, the method to synthesize separation processes by the artificial intelligent techniques is explored, so as to make the problem be solved and make the foundation for realizing computerization and automization in the whole procedure of process design.Firstly, the problem of separation process synthesis is decomposed into several sub-problems, which are abstracted into two problem types: selection and synthesis, based on their specialities. Two kinds of knowledge structure models, heritachy and object—specificity relation models, related with selection are proposed here, so as to reduce the selecting space, simplify the solving procedure and improve the searching efficience. The models are available for avoiding the mistakes, such as redundancy, duplication and contratory, when using production system to acquire knowledge. The rule proposed to storing the selection rules can further improve the searching efficience for selection typo problem. The concept of self-answer is introduced to lower the demand to the user. The certification factors of the conclusion are calculated by fuzzy models in self-answer system.Secondly, the heuristic rules about sequencing split points are reviewed, the...