The Deformation Analyses And Expert System Researches On Crankshaft Rolling-aligning

As typical anisotropic parts, the crankshafts have stress concentrating in the places where the dimensions have been changed. A great deal of theoretical and practical researches have indicated, there exist fatigue destruction in the fillet transition places from cranks to principal axes and connecting rods. In order to improve crankshafts' fatigue intensity, the roll-strengthening technology is applied to crankshafts, however, the crankshafts have much low bending strength, they will be deformed after the rolling technology is applied, yet if the bending-aligning technology is applied to them once more, the press stress which has formed in roll-strengthening will be disappeared and lost roll-strengthening effects. The crankshafts can only be applied aligning-rolling technology, namely, certain places on the crankshafts are rolled using different rolling forces, until the needs of straightness are satisfied. Because of crankshafts complex shapes, it is inaccurate to calculate theoretically aligning forces and azimuths, the expert system of crankshaft fillet rolling-aligning need to be programmed by using the experience datum on various deformed cases of crankshaft fillet rolling, and the rolling-aligning of the crankshafts should be directed by this expert system. This paper is based on the research task of the intelligentized machining system on crankshaft fillet rolling and aligning. In this paper ,the technology parameters of crankshaft fillet roll-strengthening are studied; the distortion discipline of crankshaft fillet rolling is analyzed by using ANSYS engineering software; the neural network model of the crankshaft fillet rolling-aligning is built by using three layer back-propagation network; at the same time, the program design of the neural network expert system on the crankshaft fillet rolling-aligning is discussed. Concretely, several aspects as follows have been done:1 In this paper, it is studied how to determine the technology parameters such as rolling force, rolling loops and rolling velocity on crankshaft fillet rolling, and the effects of these parameters to the fatigue intensity of the crankshafts.2 The models of the connecting rods and principal axis neck are built, and the finite element analyses have been done to them by using ANSYS engineering software. As a result of the analysis ,the contour graphs of plastic deformation ,Von Mises stress and strain have been plotted after the crankshaft having been rolled. The distortion discipline of crankshaft fillet rolling have been found out, and thus can direct the work of crankshaft rolling-aligning and present references for crankshaft rolling- aligning expert system programming.3 The mathematics models and work mechanism of crankshaft rolling-aligning are discussed ,by combining the practice of crankshaft rolling-aligning, the principle of crankshaft aligning is thoroughly researched, such as how to determine the azimuths and forces of crankshaft rolling- aligning, so as to provide the theoretical basis for expert system building of crankshaft aligning ,and the three composing parts of the expert system, namely, productive mood rule, global database building and controlling strategy are analyzed in this paper.4 In view of the limitations of calculating theoretically and the complexity of crankshaft shape, the basis principle and frame of neural network expert system are presented by combining the development of the neural network in this paper. The neural network model of the crankshaft rolling-aligning is built by using three layer back-propagation training algorithm, it is discussed how to use reasonable method to choose the nodes of import layer, concealed layer, export layer. To the algorithm which have been used in this paper, several representative methods to improve study velocity of network, strengthen the abilities of the system's self-studying and self-adapting have been discussed. The problems of training incertitude, tediously long training time, the bad timeliness have been discussed, and the problems which the velocity of network...