IA-BP算法在谐振腔物料水分测量中的应用

Application of IA-BP algorithm in measuring moisture content

为提高微波谐振腔物料湿度测量精度,提出一种基于IA-BP优化算法的进化神经网络模型,对微波谐振腔湿度测量结果进行校正.模型首先利用IA算法能够保持解群分布多样性的特性进行全局搜索,从而得到最优解或次优解附近,然后根据BP算法基于梯度信息指导权值调整的性能进行局部搜索,进而避免在最优解或次优解附近震荡,并迅速收敛到最优值.结果表明该优化算法有效地避免传统BP算法易陷入局部极小,同时保持其高预测精度,且收敛速度快,具有寻优的全局性和精确性,提高了测量精度.预测湿度与实际湿度间的均方差为0.012 5,平均绝对误差为0.071 5,平均相对误差为0.118 6,确定系数为0.996 5.

Based on IA-BP optimal algorithm an evolutionary neural network model is presented to improve the accuracy. In the model, IA algorithm is first used for global search to obtain the region of optimal solution or suboptimal solution with IA＇ s ability of keeping diversity of population. And the BP local searching ability which avoids oscillating near the optimal solution or suboptimal solution and converges on the optimal solution speedy is considered. Experimental results showed that the IA-BP optimal algorithm makes the conventional BP algorithm avoid getting into infinitesimal locally effectively and has the merits of high prediction precision, rapid convergence, global superiority and accuracy for optimization. It improves the measurement precision with the mean squared error 0.012 5, the mean absolute error 0.071 5, the mean relative error 0. 118 6 and the certain coefficient 0. 996 5 between the predicted moisture content and the real value.