摘要
在经典的循环神经网络中,时间序列的前后关系往往被忽略,通常无法获得长期的整体记忆;同时,其权值的传递和更新主要依靠梯度下降法实现,均导致在旋转机械剩余使用寿命预测中表现出较低的预测精度和较高的计算成本。鉴于此,提出一种量子基因链编码双向神经网络(QGCCBNN)的旋转机械剩余使用寿命预测方法。在QGCCBNN中,设计了量子双向传递机制,建立时间序列的前后关系,根据输出层的反馈对网络的权值参数进行重新调整,从而实现输入信息与网络整体记忆之间更高的一致性,使其具有更好的非线性逼近能力。此外,为了提高全局优化能力和收敛速度,构建量子基因链编码代替梯度下降法传输和更新数据,采用量子比特概率幅实数编码,并通过相位选择矩阵将损失函数最小值对应的余弦和正弦量子比特概率幅和当前时刻的量子比特概率幅进行比较,以实现对网络权值参数方向性并行更新。由于QGCCBNN在非线性逼近能力和收敛速度方面的优势,所提出的QGCCBNN旋转机械剩余使用寿命预测方法可以获得更高的预测精度和计算效率。所提方法对双列滚子轴承的剩余使用寿命的预测值为6.33 h(实际值为7.17 h),预测误差仅为0.84 h,预测相对误差仅为11.72%,该实验结果证明了所提方法的有效性。
In the classical recurrent neural networks,the pre-relationship and post-relationship are usually neglected.The long-term overall memory is generally inaccessible.Meanwhile,the weights are transferred and updated mainly by the gradient descent method,which leads to low prediction accuracy and high computation cost in the residual useful life(RUL)prediction of rotating machinery(RM).In view of this,a quantum gene chain coding bidirectional neural network(QGCCBNN)is proposed to predict RUL of RM.In QGCCBNN,the quantum bidirectional transmission mechanism is designed to establish the pre-relationship and post-relationship of time series for readjusting the weight parameters according to the feedback from the output layer.Therefore,the higher consistency between the input information and the overall memory of the network can be realized.Meanwhile,the QGCCBNN has better nonlinear approximation ability.In addition,to improve the global optimization ability and convergence speed,the quantum gene chain coding instead of gradient descent method is established to transmit and update data.The qubit probability amplitude real number coding is utilized.The cosine and sinusoidal qubit probability amplitudes corresponding to the minimum loss function are compared with those of the current time by the phase selection matrix for the directional parallel updating of the weight parameters.On this basis,a new RUL prediction method for RM is proposed.The high prediction accuracy and desirable efficiency can be achieved due to the advantages of QGCCBNN in terms of nonlinear approximation ability and convergence speed.RUL prediction result of a double-row roller bearing by the proposed method is 6.33 h(the actual RUL is 7.17 h).The prediction error is only 0.84 h,and the relative prediction error is only 11.72%.The effectiveness of the proposed method is verified.
作者
程阳洋
李锋
汤宝平
田大庆
Cheng Yangyang;Li Feng;Tang Baoping;Tian Daqing(School of Mechanical Engineering,Sichuan University,Chengdu 610065,China;State Key Laboratory of Mechanical Transmissions,Chongqing University,Chongqing 400044,China)
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2020年第7期164-174,共11页
Chinese Journal of Scientific Instrument
基金
机械传动国家重点实验室开放基金(SKLMT-KFKT-201718)
中国博士后科学基金第60批面上资助项目(2016M602685)
四川大学泸州市人民政府战略合作项目(2018CDLZ-30)资助
关键词
量子计算
量子双向传递机制
量子基因链编码
剩余使用寿命预测
旋转机械
quantum computing
quantum bidirectional transmission mechanism
quantum gene chain coding
residual useful life prediction
rotating machinery
