基于免疫危险理论的液压泵故障诊断方法研究

发布时间：2018-02-09 07:56

本文关键词： 液压泵故障诊断免疫危险理论特征选择包络解调小波簇　出处：《燕山大学》2012年硕士论文　论文类型：学位论文

【摘要】：液压泵一直被液压工程人员形象比喻为液压系统的心脏，它的健康状况将直接影响整个液压生产制造设备的正常工作。液压泵出现故障，轻者致使液压生产制造设备部分功能缺失，重则将造成严重的、灾难性的安全生产事故。因此研究液压泵的状态监测和故障诊断技术就显得尤为重要。目前，故障诊断技术正朝着自动化、智能化方向发展。受生物免疫危险理论模型的启发，本文提出基于免疫危险理论的特征选择算法实现了对具有众多信息的原始高维特征向量的降维。本文还应用免疫危险理论原理开发了具有学习、聚类、记忆特性的故障诊断算法，并将这一算法应用于液压泵的故障诊断中。免疫危险理论应用于故障诊断领域的案例并不是很多，目前国内外学者更多将这一理论应用于信息安全、机器学习、数据挖掘等领域。本文基于生物免疫系统危险理论模型识别机制，应用Matlab软件分别开发了具有能够降维高维数据的特征选择算法和故障诊断算法。特征选择算法将具有众多信息的高维特征向量降为低维特征向量，大大减少了后续故障诊断的时间。故障诊断算法将学习样本看作为抗原，，并通过抗体（随机检测器）对抗原（学习样本）的学习形成记忆抗体种群（成熟检测器），记忆抗体种群（成熟检测器）将识别抗原（测试样本）的再次侵袭。为验证本文算法的有效性，本文以实验室材料实验机的轴向柱塞液压泵作为诊断对象。应用加速度传感器和NI数据采集卡采集液压泵端盖振动信号，运用细化谱分析技术分析与确定液压泵各状态原始采集振动信号的共振频带范围；采用基于小波簇的包络解调方法对确定的共振段信号进行包络解调；将解调所得的包络信号进行2层小波包分解与重构，提取每一子带重构信号的时域、频域和时频域信息作原始特征向量；选择目标函数（各状态样本类间散度矩阵的迹和样本类内散度矩阵的迹的比值）作为特征选择后特征子集的分类性能评判函数，应用本文提出的基于免疫危险理论的特征选择算法，选择出了目标函数值最大时所对应的特征向量；最后，采用基于免疫危险理论的故障诊断算法对特征选择后的学习样本（抗原）进行学习，并生成最终的各状态成熟检测器（记忆抗体群）以便完成对测试样本（抗原）的状态监测和故障诊断。通过Matlab软件的程序仿真，验证了基于免疫危险理论液压泵
[Abstract]:The hydraulic pump has always been likened to the heart of the hydraulic system by hydraulic engineers. Its health will directly affect the normal operation of the whole hydraulic production and manufacturing equipment. Some of the functions of hydraulic production and manufacturing equipment are missing and heavy will cause serious and catastrophic accidents in production safety. Therefore, it is very important to study the condition monitoring and fault diagnosis technology of hydraulic pump. At present, it is very important to study the condition monitoring and fault diagnosis technology of hydraulic pump. Fault diagnosis technology is developing towards automation and intelligence. Inspired by the biological immune hazard theory model, In this paper, a feature selection algorithm based on immune hazard theory is proposed to reduce the dimension of the original high dimensional feature vector with a lot of information. The algorithm of fault diagnosis based on memory characteristic is applied to the fault diagnosis of hydraulic pump. There are not many cases in which immune hazard theory is applied in the field of fault diagnosis. At present, many scholars at home and abroad apply this theory to information security and machine learning. Data mining and other fields. This paper based on the biological immune system hazard theory model recognition mechanism, The feature selection algorithm and fault diagnosis algorithm are developed by using Matlab software. The feature selection algorithm reduces the high dimensional feature vector with a lot of information to the low dimensional feature vector. It greatly reduces the time of subsequent fault diagnosis. The fault diagnosis algorithm treats the learning samples as antigens. Furthermore, the memory antibody population (maturation detector) and memory antibody population (maturation detector) will recognize the re-invasion of antigen (test sample) through the learning of antigen (learning sample) by antibody (random detector). In order to verify the validity of this algorithm, the axial plunger hydraulic pump of the laboratory material experiment machine is used as the diagnostic object. The vibration signals of the end cover of the hydraulic pump are collected by using the accelerometer and NI data acquisition card. The resonance frequency band range of the original vibration signal collected by hydraulic pump is analyzed and determined by the technique of thinning spectrum analysis, and the envelope demodulation method based on wavelet cluster is used to demodulate the signal in the determined resonance section. The envelope signal obtained by demodulation is decomposed and reconstructed by two-layer wavelet packet, and the time domain, frequency domain and time-frequency domain information of each sub-band reconstruction signal is extracted as the original eigenvector. The objective function (the ratio of the trace of the scatter matrix between each state sample class and the trace of the divergence matrix within the sample class) is selected as the classification performance evaluation function of the feature subset after feature selection. Using the feature selection algorithm based on immune hazard theory proposed in this paper, the feature vectors corresponding to the maximum value of the objective function are selected. A fault diagnosis algorithm based on immune hazard theory is used to study the learning samples (antigens) after feature selection. The final state maturation detector (memory antibody group) was generated in order to complete the state monitoring and fault diagnosis of the test sample (antigen). The simulation of Matlab software proved that the hydraulic pump based on immune hazard theory was based on the immune hazard theory.
【学位授予单位】：燕山大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH137.51;TH165.3

【参考文献】