提出一种BP神经网络在线自学习方式及用该方式建立自适应非线性滤波器的方法,并提出了基于该BP滤波器对机器人系统进行在线实时状态跟踪、预报和突变性故障检测的方法。为了快速、正确地检测出故障,设计了一复合故障检测策略。对所提方式、方法及策略进行了仿真验证。 An online self-learning method based on BP neural network and a method to build an adaptive nonlinear filter using this method are proposed. The method of tracking, forecasting and catastrophic fault detection of robot system based on the BP filter is put forward. In order to quickly and correctly detect the fault, a composite fault detection strategy is designed. The proposed methods, methods and strategies were verified by simulation.