Research On Crankshaft Optimal Rolling Parameter Using Neural Network

The research on engine crankshaft rolling craft, equipment and correlative technology has become one of the key technologies which are favorable to current auto industry development, especially to saloon car industry development. But by far, the research both here and abroad on crankshaft rolling craft hasn't enough, for example, it lacks systematic research on influences that crankshaft material, residual stress state and rolling craft parameters act on rolling layer and crankshaft fatigue intensity, besides, it lacks a suit of scientific and rational methods on the selection of crankshaft rolling parameters and the estimation of rolling effect. Therefore, the paper spreads wide and deep research aiming at the problem of engine crankshaft rolling craft parameters in the actual production.The main work in this paper includes: 1. It carried through relative systematic kinematics and dynamics analysis of engine crank connecting link system and got force analysis model of crankshaft on the concrete work state, and it provides essential datum for FEM analysis. 2.It made a static FEM analysis of concrete working state and a machining FEM analysis, and it made a concrete analysis for the results.3.Based on Matlab, it worked out the program of orthogonal test about every factor affects residual stress during rolling and it made orthogonal test.4.Based on the Matlab, it worked out artificial neural network(ANN)program and got relative steady network, and it was utilized for prediction of residual stress after rolling, at the same time, it got corresponding relation between rolling parameters and residual stress and it provides rolling craft and equipment credible references.Based on the theory of the optimal residual stress, the paper used firstly ANSYS to make the simulation aiming at concrete working state and got corresponding optimal residual stress. At the same time, it used DEFORM to make fillet rolling experiment and got many groups residual stress aiming at different rolling parameter. Then it used ANN to get corresponding relation between rolling parameter and residual stress. Finally, it got optimal rolling parameter aiming at above optimal residual stress using the ANN, and it provides actual productioncredible references.