基于零件摩擦学性能的磨削参数优化

采用遗传算法和神经网络相结合，以表面形貌评价参数（表面算数平均偏差、表面支承指数、核心区液体滞留指数和谷底区液体滞留指数）为输入层，以磨削参数（砂轮转速、工件速度、横向进给量和背吃刀量）为输出层，建立多输入多输出的优化预测模型；针对不同使用需求的特定表面形貌结构，利用此模型预测流体润滑条件下相应的磨削工艺参数.验证实验采用与样本实验不同的机床、砂轮和工件尺寸，结果显示，预测值与实验值的最大偏差为12.87%，充分证明了该模型的优化精确性、可靠性和普适性；该模型可有效提高工艺方案的设计效率.
Abstract: An optimization model with multiple-input and multiple-output was achieved using genetic algorithm method and neural network method. Surface roughness parameters (arithmetic mean deviation, surface bearing index, core fluid retention index and valley fluid retention index) were taken as input layer factors, and the output layer factors were grinding parameters (wheel linear speed, workpiece linear speed, grinding depth and longitudinal feed rate). Moreover, this model was applied to predict grinding parameters for special surface topography aiming at different friction performances in hydrodynamic lubrication. Grinder, grinding wheel and workpiece size different from sample experiment were applied in the verification test. Results show that the maximum error between predicted values and experiments is only 12.87%, which implies the good accuracy, reliability and universality of this optimization model. This optimization model can efficiently improve the design efficiency of process plan.