数理計画法を用いた液相拡散接合エキスパートシステムの構築

- 液相拡散接合プロセス最適化システムの開発（第１報）-

抄録

The computer-aided optimisation system ("Bayesian Expert System") for transient liquid phase bonding process which integrated the alloy design of insert metal and optimisation of processing parameters has been systematised utilising the radial basis function neural network with Bayesian inference. Bayesian Expert System consisted of three main systems, i.e. the supporting system for planning the experimentation, the supporting system for alloy design of insert metal and the supporting system for optimisation of processing parameters. The latter two systems involved the formulation and its verification module with Bayesian RBF network, the search module of extremum by the genetic algorithm, the multi-objective optimisation module by the satisficing trade-off method and visualisation tool for outputting the obtained results. The Bayesian Expert System realised the more accurate formulation of evaluation factors based on the measured data with scatters and resulted in the feasibleness of becoming the advanced optimisation of TLP-bonding process.