量子 BP 神经网络在发动机故障诊断中的应用

摘要

In order to solve the problem of slow convergence speed and low classification accuracy in common BP neural network (CBPN),a model of quantum BP neural network (QBPN)was pro-posed.Quantum computation was introduced into CBPN.The structure of QBPN contained input lay-er,hidden layer and output layer,where input and transfer function were represented by quantum bit, and the output was real value.Firstly,the real-valued training samples were transformed into quantum training samples,which was as the input.Then,with transfer function,the quantum weights were cal-culated,and the network parameters were updated to achieve the required result.Finally,the trained network was used for fault diagnosis,and the result was output in real value.The proposed method was applied in fault diagnosis of engine.The results indicate that,compared with CBPN,QBPN has great advantages of convergence speed,classification accuracy and executing time.%为了解决普通 BP 神经网络收敛速度慢、分类正确率低等问题，提出一种量子 BP 神经网络算法。该算法在普通 BP 神经网络中引入了量子算法，量子 BP 神经网络结构由输入层、隐含层和输出层组成，其中，量子神经元的输入和传递函数均由量子比特表示，输出结果为实数。首先，该算法将实数值训练样本变换为量子态训练样本，从而作为算法的输入。然后，通过传递函数，计算量子态权值并更新网络参数以达到训练效果。最后，利用训练好的网络进行故障诊断，并将结果以实数值输出。将该方法应用于发动机故障诊断，实验结果表明，与普通 BP 神经网络相比，量子 BP 神经网络算法在收敛速度、分类正确率和执行时间等方面具有明显的优势。