| Summary: | 碩士 === 大同大學 === 化學工程學系(所) === 92 === The simulated verifications and validations of the neural-network rate-function (NNRF) and the hybrid neural-network rate-function (HNNRF) approaches to modeling the batch reactor systems are provided. In chemical processes, some measurements may not be obtainable easily, and the designed NNRF model does not embed these state variables in the built dynamic model. To overcome this problem, the approximated physical model could be combined with the NNRF model to give the hybrid neural-network rate-function (HNNRF) model. The sequential pseudo-uniform design (SPUD) is used to locate desired but limited experiments to provide the NNRF and HNNRF models of the testing batch reactor systems with rich information. In this research, the NNRF model was applied to build the dynamic models of (a) the series-parallel reactions and (b) the submerged cultivation of monascus anka carried out in batch reactors; whereas, the HNNRF model was applied to a simulated polymethyl methacrylate (PMMA) solution polymerization system. The performance of the NNRF or HNNRF modeling approach is quite acceptable and can be applied to determine the optimal operating conditions of the processes.
|
|---|
