自由曲面加工误差预测——基于模拟退火算法优化的BP神经网络算法

Machining errors prediction of free-form surfaces:BP neural network algorithm optimized by simulated annealing algorithm

针对三坐标测量机测量效率低的问题,建立了自由曲面加工误差预测模型。采用基于模拟退火算法优化的BP神经网络算法对自由曲面上若干个点的加工误差进行预测,结合模拟退火算法的概率突跳特性,在解空间中随机寻找目标函数的全局最优解,从而改进BP神经网络算法。为进一步提高算法的预测精度,采用加工误差分解的方法剔除点集中的奇异点。用三坐标测量机对自由曲面上若干个点进行测量并获得加工误差,将预测结果与试验结果进行对比验证。结果表明,平均绝对误差指标达到了1.70μm,且最大绝对误差为7.12μm,说明该优化算法具有较好的预测性能。

A model of machining errors prediction of free-form surfaces was established aiming at low measuring efficiency of CMM.Firstly,BP neural network algorithm optimized by simulated annealing algorithm was used to predict the machining errors of points on the free-form surface;combined with the probability jump characteristics of simulated annealing algorithm,the global optimal solution of the objective function was randomly found in the solution space to improve the BP neural network algorithm.Secondly,the machining error decomposition method was used to eliminate singular points to further improve the prediction accuracy of the algorithm.Thirdly,several points on the free-form surfaces were measured by CMM to obtain the machining errors.Finally,the predicted results were compared with the actual results in an experiment.The experimental results show that the algorithm has reliable predictive performance of prediction with the average absolute error index 1.70μm and the maximum absolute error 7.12μm.