基于BP神经网络的复合材料注塑工艺优化

Optimization of Composite Injection Molding Process Based on BP Neural Network

针对纤维复合材料注塑件提出的高模量、低翘曲、高效率等要求,利用加权平均法兼容以上优点,通过层次分析法比较出制件力学性能、表面质量和生产效率的相对重要程度,构造判断矩阵得到不同权值,再利用BP神经网络算法对正交试验所得数据进行模拟,模拟所得数据误差在10%以内,最后,为提高数据的精度,通过缩短数据之间的步长,重新进行算法模拟得,到最佳工艺参数组合,并经过Moldflow反馈分析,证明了该方法的准确性。

The weighted average method is used in light of the requirements of high modulus,low warpage and high efficiency from the fiber composite injection molded parts.The relative importance of mechanical properties,surface quality and production efficiency is compared based on the analytic hierarchy process.Different weights are acquired by constructing the judgment matrix.The data obtained from the orthogonal experiment simulation is then simulated by means of the BP neural network algorithm,with the simulated data error set within 10%.Finally,the accuracy of the data is improved by shortening the step size between the data,and the optimal combination of process parameters is obtained by algorithm simulation.The accuracy of the method is proved by Moldflow feedback analysis.