基于正交试验和神经网络的轴系主轴承润滑特性优化

Optimization of Crankshaft-Bearing Lubricating Characteristics Based on Orthogonal Experiment and Neural Network

选取影响发动机轴系主轴承润滑特性的13个主要参数,每个参数取3个不同的值,按照正交试验的要求共得到27组不同的组合,运用仿真计算的方法得到此27种组合下的最小油膜厚度、最大油膜压力和摩擦损失功率.采用极差分析的方法确定了影响最小油膜厚度、最大油膜压力和摩擦损失功率因素的主次关系,利用BP神经网络理论建立了轴系主轴承润滑特性的神经网络模型并进行了训练和验证,然后利用该模型对影响润滑特性的主要参数进行了优化.结果表明,运用正交试验和神经网络相结合的方法进行轴系主轴承润滑特性的优化设计,减少了工作量并能得到满足精度要求的结果,对主轴承的优化设计有一定的指导意义.

13 major parameters affecting lubrication characteristics of engine main bearing were selected, and three different values for each parameter were taken. In accordance with the requirements of orthogonal experiment, 27 groups of different combinations were obtained. The minimum film thickness, maximum film pressure and friction loss power of these 27 combinations are obtained by using simulation method. The primary and secondary power relations of factors that influence the minimum film thickness, maximum film pressure and friction loss power are determined by using range analysis method. The neural network prediction model of main bearing lubrication character is built by BP neural network theory and makes its training and certification. Main parameters that affect the lubricating properties are optimized by using this model. Results show that workload is reduced and accuracy requirement of the results is satisfied by the combination of orthogonal experiment and neural network methods to optimize the design of main bearing lubricating properties. This has some guidance for the optimal design of the main bearing.