本文给出了一种改进的神经元联想记忆模型及其在控制系统故障诊断中的应用。这种模型不要求输入的样本向量之间是正交的或近似正交的,而是对任意输入样本向量进行HADAMARD预处理,使得处理后的样本向量,与输出样本向量组成联想矩阵M,构成这种联想记忆模型。最后本文给出了一个简单的控制系统故障诊断的数值例子加以说明。 This paper presents an improved neuron associative memory model and its application in the fault diagnosis of control system. Instead of orthogonally or approximately orthogonally crossing the input sample vectors, this model performs a HADAMARD preprocessing on any input sample vector so that the processed sample vectors and the output sample vectors form a associative matrix M, This associative memory model. Finally, this paper gives a numerical example of fault diagnosis of a simple control system.