基于神经网络和遗传算法的薄壳件注塑成型工艺参数优化

建立基于神经网络和遗传算法并结合正交试验的薄壳件注塑成型工艺参数优化系统.正交试验法用来设计神经网络的训练样本,人工神经网络有效创建翘曲预测模型;遗传算法完成对影响薄壳塑件翘曲变形的工艺参数(模具温度、注射温度、注射压力、保压时间、保压压力和冷却时间等)的优化,并计算出其优化值.按该参数进行试验,效果良好,可以有效地减小薄壳塑件翘曲变形,其试验数值与计算数值基本相符,说明所提出的方法是可行的.     

An Improved Fourier Five-Sensor (IF5S) method for separating straightness and yawing errors of a linear slide based on multiple sensor parameter sets and least square regression technique

In this paper, an Improved Fourier Five-Sensor (IF5S) measurement method is proposed for separating the straightness and yawing motion errors as well as determining the profile of a linear slide. The previous F5S method [3] used the constant parameters initially to estimate the profile function based on three sensor equations for different angle ranges. The profile estimation and error separation are implemented via an iterative method which can only yield acceptably accurate results with tremendous computational efforts. Here, the improved F5S method applies the least square regression technique instead of the iterative method to estimate the profile functions by using three distinct sets of parameters and different fused sensor data according to the travel of the linear slide. Various errors can then be separated based on the calculated profile function. Simulation results confirm that the IF5S method provides better performance and effectiveness as compared to the previous F5S method.