论文部分内容阅读
测后诊断速度和诊断精度是模拟电路故障诊断性能的主要衡量指标。文中将神经网络的自学习和分类技术应用于非线性电路直流故障诊断，把反向传播（ＢＰ）网络训练成一部能诊断软、硬单故障的故障字典。考虑元件参数容差对诊断的影响，提出了优选训练样本的具体方法。此外，重新定义了ＢＰ网络的输出误差函数，使网络在训练时有较大的自由度。ＢＰ网络高度并行的信息处理能力决定了这种新型故障字典的诊断速度非常快。仿真实验结果表明，神经网络方法的综合性能要优于传统的故障字典法。
Post-test diagnostic speed and diagnostic accuracy of the analog circuit fault diagnosis performance of the main indicators. In this paper, neural network self-learning and classification technology is applied to DC fault diagnosis of non-linear circuits, and the BP network is trained into a fault dictionary that can diagnose soft and hard single faults. Considering the influence of component parameter tolerance on diagnosis, a specific method of training samples is proposed. In addition, the output error function of BP network is redefined to make the network have more freedom in training. The BP network’s highly parallel information processing capabilities determine the diagnostic speed of this new fault dictionary very fast. Simulation results show that the comprehensive performance of the neural network method is superior to the traditional fault dictionary method.