The Simulation And Prediction Of Two Passes Grind-hardening

The grind-hardening is a new production process utilizing grinding heat effects, coupling with mechanical effects to induce the surface hardening directly for no harded steel.With its unique characteristics and superiority,it has a broad application future in the metal material machining and modified properties of surface,thus bringing about satisfactory technical and economic effects and remarkable social ones. In the recent years,the applied fundamental studies and technology development of actively-controlling grinding-hardening are highly concerned by engineering experts and scholars abroad.Grind-hardening is not only a complex process coupling mechanical,dynamical and thermal effects,but a process containing distinct changes of microstructure and performances in the surface layer of the grinded workpiece.Basing on the experiments of the two passes grind-hardening,the two passes grind-hardening process has been simulated;The LM-BP neural network system of grind-hardening is developed.The main research contents and contributions of this paper as follows:(1) Two passes grind-hardening heat flux model is established and the phase transformation model is gived in this paper.The tempering parametersλis introduced into this paper.Using the additivity of the tempering parametersλand the relationship of fractions of phase transformation andλ,high temperature and fast temper of martensite phase transformation model is established.(2) DEFORM-2D finite element software is re-developd.And the temperature field,microstructure and micro-hardness of the two passes grind-hardening is numerical simulated and nonlinear analyzed by using this finite element software.The forming mechanism of two passes grind-hardening is revealed.Through comparing the simulation results with test results,the effectiveness of the numerical simulation in this study is verified.(3) Grind-hardening mechanism model and neural network model are combined to develope the LM-BP neural network predicton system for grind-harding.The system is developed by using Software C++ Builder 6.0 as front-end developing tool and Software SQL Server 2000 and MATLAB 6.5 as background tool.The L-M algorithm is used to train the BP network.Based on the COM technology,mixed programming of software C++ Builder and MATLAB is carried out.Using the ODBC and BDE technology,the connection of software C++ Builder and SQL Database is carried out.(4) The system is applied to predict the thickness of grind-hardening layer is predicted by using the grind-harding neural network predicton system.Predicted result is in concordance with the the experimental result.And the accuracy and reliability of neural network model is verified.This research paper is a significant part of the project of National Natural Science Foundation of China(No.50675088) and Jiangsu Natural Science Foundation (No.BK2006077).The research has significant theoretical value and practical meaning for analyzing the forming process,revealing the forming mechanism, enriching and developing the theories and facilitating the practical application of grind-hardening.