基于LMBP算法的液压油缸内泄漏故障诊断方法

针对液压油缸内泄漏故障诊断中提取时域参数过多以及各参数间相互交叉等问题，提出一种基于主成分分析（Principal Component Analysis，PCA）和改进的Levenberg—Marguard（LM）神经网络的诊断方法．首先采用Lu分解法对LM算法中逆矩阵的求解进行优化，以加快网络的收敛速度，然后提取压力信号的8个时域参数作为原始特征，采用PCA法对其进行降维和去相关，提取前2个主成分作为最终特征，输入到改进的LM网络中进行故障模式识别，并将诊断结果与LM算法和GA—BP算法进行仿真对比研究．研究结果表明：基于LMBP算法的故障诊断方法在减少识别误差和提高诊断速度等方面取得显著改善，是一种行之有效的液压油缸内泄漏故障诊断方法．

Internal leakage fault diagnosis approach of hydraulic cylinder using LMBP neural network

According to the fact that in fault diagnosis for internal leakage of hydraulic cylinder, excessive time-domain features are selected from pressure signal and the features interact, a fault diagnosis approach based on principal component analysis（PCA） and improved Levenberg-Marquardt （LM） neural network was proposed. LU decomposition was employed to improve the solution of contrary matrix in the LM algorithm simulated with MATLAB at first in order to increase the convergence speed of the network. And then eight time-domain parameters of pressure signal were selected as prime features, PCA was used to reduce the dimension and eliminate the correlativity, and the first two principal components were selected as the final features. At last, these final features were input into the improved LM neural network to identify faults. The simulation and comparison results of the proposed method, LM algorithm and GA-BP algorithm reveal that the proposed method can meet the recognition rate and increase diagnosis speed, so it is an effective fault diagnosis approach for internal leakage of hydraulic cylinder