基于鱼群BP神经网络的板料成形分块压边力优化

采用人工鱼群算法与BP神经网络相结合的方法建立了分块压边力与成形质量的映射关系。首先以分块压边力为设计变量，通过基于最大最小原则的拉丁超立方取样设计方法抽取了BP神经网络的训练样本，并将通过仿真软件获得的成形质量指标作为BP神经网络的训练输出;其次通过人工鱼群算法优化的BP神经网络建立了分块压边力与成形质量的映射关系;然后采用粒子群算法对该映射函数关系式进行优化，得到最优分块压边力;最后将该最优分块压边力成形效果与整体压边力成形效果进行对比，结果表明成形效果大大改善。研究表明，采用该方法可以快速计算最优分块压边力，克服了分块压边力计算困难的缺点。

Optimization of Several Segmental Binder Force in Sheet Forming Based on BP Neural Network and Fish Swarm Algorithm

Relationship of several segmental binder force and forming quality was bulit, which was based on BP neural network and fish swarm algorithm. The first approach was gotten training samples for BP neural network by Latin hypercube designs based on maximum and minimum principle. Then simulation software was used to get forming quality and used as training outputs for BP neural network. Fish swarm algorithm was used to optimize BP neural network, which built relationship of several segmental binder force and forming quality.PSO was adopted to obtain optimal result of function relation. Compared with whole binder force,forming quality is improved greatly.The research fruits indicate that the method can calculate optimal several segmental binder force quickly. It can make up computational difficulties of several segmental binder force.