基于GABP神经网络的超声辅助AWJ冲蚀深度预测及参数优化

Ultrasonic Assisted AWJ Erosion Depth Prediction and Parameter Optimization Based on GABP Neural Network

多孔HAP生物陶瓷材料硬度低、脆性大,传统生物陶瓷零件切削加工方法容易产生细小裂纹、应力集中等缺陷,微细磨料水射流加工技术可有效解决这些难题。本文采用超声辅助微细磨料水射流对多孔的HAP生物陶瓷块进行加工试验。基于GABP神经网络方法,利用获得试验数据样本来训练和检测GABP神经网络,建立了加工参数如系统压力、靶距、振幅等的微细磨料水射流冲蚀深度预测模型,预测误差和为0.007044,利用遗传算法进行参数寻优,较传统BP神经网络误差和降低了61.506%,大大提高了预测精度,实现了不同参数组合下冲蚀深度的预测。该预测和优化结果表明,当采用系统压力为25MPa,靶距为7.576mm,振幅为13.883μm时,可以获得最大冲蚀深度,其值为3.296mm。

Due to the low hardness and high brittleness of porous HAP bioceramics,traditional cutting and processing methods of bioceramics are prone to produce fine cracks,stress concentration and other defects.Micro abrasive waterjet processing technology can effectively solve these problems.In this paper,ultrasonic assisted micro abrasive waterjet is used to process the porous HAP bioceramics.GABP neural network based method,using the test data obtained samples to train and test GABP neural network,set up the processing parameters such as system pressure,stand off distance,amplitude of micro abrasive waterjet erosion depth prediction model,the prediction error is 0.007044,and the use of genetic algorithm for parameter optimization,and reduce 61.506%than traditional BP neural network error,greatly improve the prediction accuracy,implements the erosion depth under different parameter combination forecast.The prediction and optimization results show that the maximum erosion depth of 3.296mm can be obtained when the system pressure is 25MPa,the target distance is 7.576mm,and the amplitude is 13.883μm.