Abstract: Remanufactured crankshaft’s balance is one of the important factors affecting the performance of the remanufactured engine. The connotation of uncertainty for crankshaft main journal runout are studied, diameter, cylindrical and crankshaft runout, bending and other factors. Considering the distance between the probability distribution of each factor and the ideal value, a quantitative uncertainty measure model is built. Based on the above study, remanufactured crankshaft’s balance control method based on improved BP neural network (BPNN) is proposed. Network model adjusts weights and thresholds of input layer to the hidden layer by gradient descent, and alters weights and thresholds of hidden layer to the output layer by support vector machines. The method can effectively improve the pass rate of the remanufacturing crankshaft’ balance, and reduce the scrap rate of crankshafts. The application example demonstrates the feasibility and effectiveness.

Key words: balance, crankshaft, remanufactured, uncertainty