Font Size: a A A

Based On Bp Neural Network Rack Welding Dosage Of The Optimization Study

Posted on:2006-09-22Degree:MasterType:Thesis
Country:ChinaCandidate:Y XiaoFull Text:PDF
GTID:2191360155465085Subject:Mechanical design and theory
Abstract/Summary:PDF Full Text Request
In the process of engineering design, people always hope that their design project's capability is the best of all under all kinds of restrictions. Optimize of jointing quantity is a significative work which can reduce cost, predigest technice . Jointing has all kinds of weld mode and infection factors of jointing process are very much which make the problem of optimize of jointing quantity is very complexity and traditional optimize method can not solve. On the base of reading a large amount of reference documentations and the characteristics of the problem , auther research this optimize problem with nerve net.Is's difficult to establish a exact expression between the length of welding line and maximal stress of welding frame when the welding frame is in the condition of working . With the non-linear mapped function of nerve net ,the author build a nerve net , realizing the mapped connection between the length of welding line and maximal stress . The author trained the nerve net with the data which from the finite element analysis . Aiming at the problem of the maximal stress point's jump in structural analysis and the disadvantage of nerve net which has several input and several output ,the author improve on the net. Besides, the author discuss the problems of hiding layer quantity, the cell number of hiding layer and the training of net. With the building of BP nerve net, realizing the replace of finite element analysis with nerve net and using this nerve net in the forecast the maximal stress when the welding frame in the working conditions. This method conquer the malpractice which is lacking feasibility in realism of the optimize which is base on finite element analysis , because it takes much time in the finite element analysis.The autor unite the BP nerve net with the covering complex method , using the training stylebooks from the structural analysis of UG II to train the nerve net. With the use of trained nerve net in the covering complex method, realizing the optimize of the jointing quantity of the punch-rivet machine frame, and checkout the optimize result with the structural analysis function of UG n. All these prove the method of this paper is feasible.
Keywords/Search Tags:BP nerve net, finite element analysis, covering complex method, optimize design
PDF Full Text Request
Related items