基于KPCA与MPSO–BP注射成型工艺参数优化

Study on Parameters Optimization of Injection Molding Based on KPCA and Modifi ed Particle Swarm Optimized Back Propagation Neural Network

针对注塑过程中工艺参数的优化选择问题,提出一种融合核主元分析方法 (KPCA)与改进粒子群算法优化BP神经网络的成型工艺参数优化方法。首先,对正交实验数据作为训练样本的工艺参数利用核主元分析方法进行降维、拨冗余,约减网络结构;其次,针对BP易陷入局部最优、收敛效率低的不足,改进粒子群算法中粒子速度与位置更新策略并优化BP算法的权值和阈值,从而构建了工艺参数预测模型。在此基础上,实现粒子群算法寻优最佳的注射成型工艺参数。结果表明,该方法能够更快、更好地获得注射成型中的工艺参数,且以此工艺参数进行实验,塑料件的翘曲变形量、收缩率均较小。

For optimization selection of process parameters in injection molding process,molding process parameters optimization method which fuses kernel principal component analysis （KPCA） method,modified particle swarm optimization （PSO） algorithm and BP neural network is proposed. Firstly,the kernel principal component analysis method is employed to reduce thedimensionality,get rid of the redundancy,do reduction of network structure. Subsequently,a prediction model about process parameters and the optimized objectives is established, because BP is easy to fall into local optimum and low convergence efficiency,particle swarm algorithm in the particle＇s velocity and location update strategy is improved and the weights and thresholds of BP algorithm are optimized. On this basis,the injection molding process parameters of the optimization of particle swarm optimization is realized. The experimental analysis shows that the war page and the shrinkage of the plastic parts both become smaller with the proposed method,which demonstrates this algorithm can provide a faster and better tool to optimize injection molding process parameters.