基于WPT-DDT-SVM的混凝土泵车液压泵故障诊断研究

发布时间：2018-03-15 21:17

本文选题：故障诊断　切入点：小波包　出处：《湖南大学》2011年硕士论文　论文类型：学位论文

【摘要】：液压泵在现代工程机械等众多机械设备中被广泛使用,统计表明,在所有工程机械设备的故障中,液压泵的故障比重约占30-40%。实时在线监测液压泵运行状态、早期预测预警液压泵故障,对于减少液压泵的损坏率、降低工程机械作业现场停工期具有十分重要的意义。本论文以此为背景,提出一套系统的从信号降噪到故障分类的混凝土泵车液压泵故障诊断方法,重点研究故障状态分类改进算法。针对小波包分析存在的混频问题,并结合故障信号信号弱、非平稳、存在突变信号等特点,本文采用频带调整的小波包单子带重构技术进行特征分析。实验证明该方法能有效消除混频现象。针对特征向量中包含有的不相关和冗余信息,并结合在线故障诊断系统高实时性要求,采用距离区分技术对特征信息进行筛选。实验证明该方法能有效减低计算复杂度,提高分类精度。重点针对已有在线故障诊断方法在大数据量、强噪声下分类速度较低、分类精度不理想等问题,并结合液压泵故障类别数大、工作环境恶劣的特点,提出一种适应于混凝土泵车液压泵在线诊断的状态识别算法——容错度自适应支持向量机。该方法主要从四个方面对分类速度做了改进：(1)引入容错度因子进行模型训练；(2)优先选择能将某一类单独分离出来的二分类器；(3)在满足(2)的基础上选择平均支持向量机少的分类器；(4)引入增量学习算法对参数进行自适应调整,提高多故障诊断中对新故障类别和新故障数据的适应性,保证系统的分类精度。实验证明该方法在明显提高分类速度的同时保证了较高的分类精度。
[Abstract]:Hydraulic pump is widely used in modern construction machinery and many other mechanical equipments. Statistics show that the proportion of hydraulic pump fault is about 30-400.Real time on-line monitoring hydraulic pump running state. The early prediction and warning of hydraulic pump failure is of great significance to reduce the damage rate of hydraulic pump and reduce the shutdown period of construction machinery operation site. This paper presents a systematic fault diagnosis method for hydraulic pump of concrete pump vehicle from signal denoising to fault classification, with emphasis on the improved algorithm of fault state classification. Aiming at the mixing problem of wavelet packet analysis, and combining the characteristics of fault signal, such as weak, non-stationary and abrupt signal, etc. In this paper, the feature analysis is carried out by using the frequency band adjusted wavelet packet single subband reconstruction technique, and the experiment results show that this method can effectively eliminate the mixing phenomenon. In view of the irrelevant and redundant information contained in the feature vector and the high real-time requirement of the on-line fault diagnosis system, the distance discrimination technique is used to screen the feature information. The experimental results show that the proposed method can effectively reduce the computational complexity. Improve the classification accuracy. Aiming at the problems of existing on-line fault diagnosis methods, such as large amount of data, low classification speed under strong noise, low classification accuracy and so on, combined with the characteristics of large number of fault types and poor working environment of hydraulic pump, This paper presents a state recognition algorithm for on-line diagnosis of hydraulic pumps of concrete pump vehicles, called fault-tolerant adaptive support vector machine. This method mainly improves the classification speed from four aspects: 1) introduces the fault-tolerant factor to model. On the basis of the selection of two classifiers which can separate a certain class separately and satisfy the requirement of 2), we select a classifier with fewer average support vector machines (K4)) and introduce an incremental learning algorithm to adjust the parameters adaptively. The adaptability of new fault types and new fault data in multi-fault diagnosis is improved, and the classification accuracy of the system is guaranteed. The experimental results show that this method can obviously improve the classification speed and ensure higher classification accuracy.
【学位授予单位】：湖南大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：TU646;TH137.51

【相似文献】