针对滚动轴承故障分析的特点，构造了基于p-范数模糊推理神经网络，指出它可以对Sugeno-Takagi模型进行逼近，因而更便于学习，并克服了单纯前向神经网络训练中容易陷入局部极小及收敛速度较慢的缺点。该神经网络应用于滚动轴承的四类故障诊断，与实验结果符合很好，取得了良好的故障诊断效果。“,”Considering the special features of the fault analysis of rolling bearings, a p-norm fuzzy inference based neural network is presented． The network is capable of gradually approaching to the network of Sugeno-Takagi model in reference ［3］， and thus provides a learning convenience． The problem of being easily trapped in local minimum and slow in convergence as regards the general back-propagation networks are avoided． The application to the fault diagnosis of rolling bearings shows that the results are in good agreement with that of the experiments．