有限元模拟的H型件自适应锻造专家系统

有限元模拟的Ｈ型件自适应锻造专家系统刘庆斌，季忠，刘马宝，吴诗　用传统人工智能方法建立的锻造专家系统智能水平低，获取新知识的能力差．用人工神经网络（ＡＮＮ）建立的锻造专家系统可以直接采用简单的运算来完成推理，且可以自组织、自学习，为知识获取和推理机制...

A Neural Network Expert System for H-shaped Forgings

Shivouri in 1994 said in a-n important survey paper1]: "O f late, application of artificial neural networks to process design have become popular." But he falled to give any clue aboutdetails about such application. This research summary is a brief progress report on our neural network expert syStem.We have achieved pre1Aninary success after partial completion of our proposed expertsyStem. As said before, it is based on neural network principle. In addition, it requires information obtained with FEM (finite element method) simulation. Our proposed expert system can be described as an open adaptive expert system. It is planned to possess the following desirable features: automatic process planning for H- shaped forgings I optimization of technological process ; prediction of defects ; providing information for die designer andCAD/CAE/CAM system. At this stage, it looks promising that production cycle can be significantly shortened and forging quality can be considerably improved. Fig. 1 is a flow chartof our proposed expert system. From Fig. 1, it can be seen that our proposed expert system is planned to possess the following technical features: (1) simulation information is extracted with FEM and organized by neural networks for the generation of knowledge base ;(2) the system has the function of self- learning; it is convenient for acquiring knowledge base ; (3) each rule is expressed digitallyl so it is convenient to organize and manage theknowledge base; therefore, it is easy to complete preform and final forging design as shown in Fig. 2; (4) our syStem is also different from existing expert system in that numericalcomputation instead of logic is used i (5) the syStem is made up of a BP network group; soit can realize automatic forging process planning, optimization of technological process, prediction .of defects, etc.