基于人工神经网络的机床刚度建模

Machine tool stiffness modeling based on artificial neural network

将人工神经网络应用于机床刚度建模,针对基本BP算法收敛速度慢、易陷入局部极小点的不足,从步长和搜索方向两方面对基本BP算法进行了改进,并引入具有全局寻优能力的粒子群算法。通过统计误差计算的次数、设定多组初始权值及方差分析等方法,对几种优化算法在机床刚度建模中的应用效果进行了比较,最后以输出误差最小时的连接权值建立了机床刚度神经网络模型。

By applying the artificial neural network in modeling the machine tool stiffness,to eliminate the insufficiency that the basic BP algorithm is low in convergence rate and is easily sunk into the local minimum point,the improvement to the basic BP algorithm was conducted from aspects as step and search direction and by introducing the particle swarm optimization with global optimization capability.By means of methods as totaling the number of error calculation,setting multi group of initial weights and variance analysis,through comparing the effect on machine tool stiffness modeling by several kinds of optimization algorithms,the machine tool stiffness model was established by using connection weights when the output error is minimum.