基于GA-BP神经网络的发动机失火诊断

Engine Misfiring Fault Diagnosis Based on GA-BP Neutral Network

为了提高神经网络在发动机失火故障诊断中的准确率,提出了GA-BP神经网络算法。分析了发动机故障时的尾气变化情况,提出了发动机故障诊断规则;分析了BP神经网络原理,指出其训练速度慢、容易陷入局部极值问题;使用遗传算法对神经网络结构和参数进行优化,得到最优网络结构,将优化后的模型参数作为初始值再次进行BP算法优化;将此算法与自适应动量BP神经网络进行对比,GA-BP神经网络不仅缩短了训练时间,而且故障诊断准确率也大大提高。

To improve engine misfiring fault diagnosis accuracy, GA-BP Neutral Network Algorithm is proposed. Tail gas variation is analyzed in engine failure situation, and engine fault diagnosis rules are put forward. Principle of BP Neutral Network is analyzed, and the shortcomings of slow training speed and easy to fall into local extreme is discovered. Genetic Algorithm is used to optimize structure and parameters of Neutral Network, and the optimal parameters is optimized by BP algorithm once again. GA-BP neutral network and Adapting Momentum BP neutral network are used to diagnosing engine fault, training time of BP neutral network is much shorter than the other, and its fault diagnosis accuracy is much higher than the other.