基于 PID 调节的多层前馈网及其应用研究

Neural Network Based on PID Adjustment and Its Application

介绍了多层前馈网的基本算法及其在镦粗压力计算中的应用．针对其基本学习算法（反向误差传播算法，ＢＰ算法）存在的收敛速度慢和易陷入局部最小值的问题，提出了基于ＰＩＤ调节的ＢＰ算法，由此建立起镦粗压力计算的神经网络知识库，经过测试集考核，证明该模型算法正确，能有效地提高模型的判别精度，是一种有效的和有前途的算法．

Some multilayer network's fundamental algorithms are introduced. Facing their disadvantages about convergent velocity's slowness and easily falling into the local minimums, a new BP algorithm based PID adjustment is presented. Its basic features are discussed and a chaotic sequence to display it is given. At last, this algorithm is used to set up a knowledge library calculating upset stress. The results of the test set show that it is an efficient and promising algorithm.