再制造自动变速箱故障诊断方法研究

发布时间：2018-02-24 01:50

本文关键词： 再制造自动变速箱小波分析神经网络故障诊断　出处：《浙江工业大学》2012年硕士论文　论文类型：学位论文

【摘要】：随着我国汽车产业的飞速发展及自动变速箱装车率的不断提高，废旧汽车自动变速箱的报废量也在迅速增长。再制造凭借其节能节材等优势，成为各种处理废旧自动变速箱方式中的首选。由于再制造生产过程使用的原材料来自废旧自动变速箱，它的质量波动性是客观存在的。故本文设计了再制造自动箱故障诊断系统，，通过工厂采集数据、实验台数据分别对时域信号参数指标、傅立叶变换、小波变换、神经网络等分析方法在再制造变速箱振动信号分析与故障诊断中的应用进行具体分析。在分析国内外汽车再制造研究现状及故障诊断方法的基础上，本文的研究内容主要包括以下几个部分： 1.对汽车自动变速箱再制造的工艺过程进行调研完善，具体分析了变速箱的工作原理及损坏原因。针对再制造自动变速箱的正常工作状态以及换挡冲击、打滑、变矩器锁止力矩不足等典型故障信号，设计试验方案，运用时、频域分析方法对故障特征进行提取，研究相应的时域与频域特征。 2.基于小波分析和小波包分析的基本理论，采用基于小波包分解的信号特征向量提取方法，研究小波分析理论进行再制造自动变速箱典型故障特征量提取的特点。它能够对再制造自动变速箱的故障频率成分进行分析，运用小波分析的时频局部化特性能够捕捉到振动信号中的短时冲击和瞬变特征。 3.本文在分析故障诊断系统的基本诊断过程及故障诊断方法特点的基础上，结合神经网络的发展和特点，通过提取裕度指标、歪度指标、峭度指标、峰值指标及脉冲指标五个时域特征参数，并将其作为故障征兆向量，利用MATLAB建立基于振动信号的故障诊断神经网络并进行了验证。 4.针对BP神经网络容易陷入极小点、收敛速度慢的问题，本文采用了基于数值优化的改进算法，并在网络结构设计中，采用二次多项式逼近法确定隐含层节点数。基于上述改进和优化算法所建立起来的神经网络，能够对再制造自动变速箱的工作状况作出快速、准确的判断。
[Abstract]:With the rapid development of automobile industry and the increasing loading rate of automatic gearbox in our country, the scrap quantity of auto gearbox is also increasing rapidly. It has become the first choice in all kinds of disposal waste automatic gearbox. Since the raw materials used in the remanufacturing process come from the waste automatic transmission, its quality fluctuation is objective. Therefore, a fault diagnosis system for the remanufacturing automatic transmission is designed in this paper, and the data are collected through the factory. The application of time-domain signal parameters, Fourier transform, wavelet transform and neural network in vibration signal analysis and fault diagnosis of remanufacturing gearbox is analyzed in detail. Based on the analysis of domestic and foreign automotive remanufacturing research status and fault diagnosis methods, the research content of this paper mainly includes the following parts:. 1. Investigate and perfect the process of remanufacturing automatic gearbox, analyze the working principle and damage reason of gearbox, aim at the normal working state of remanufacturing automatic gearbox and shift impact, skid, Some typical fault signals, such as insufficient locking torque of torque converter, are designed and tested. The fault features are extracted by frequency domain analysis method, and the corresponding time domain and frequency domain features are studied. 2. Based on the basic theory of wavelet analysis and wavelet packet analysis, the signal feature vector extraction method based on wavelet packet decomposition is adopted. Wavelet analysis theory is studied to extract the typical fault characteristics of remanufacturing automatic gearbox, which can analyze the fault frequency components of the remanufacturing automatic gearbox. Using the time-frequency localization characteristic of wavelet analysis, the short-time shock and transient characteristics of vibration signal can be captured. 3. On the basis of analyzing the basic diagnosis process of the fault diagnosis system and the characteristics of the fault diagnosis method, combined with the development and characteristics of the neural network, this paper extracts the margin index, the crooked index and the kurtosis index. The five time domain characteristic parameters of peak value index and pulse index are used as fault symptom vector and the fault diagnosis neural network based on vibration signal is established by MATLAB and verified. 4. Aiming at the problem that BP neural network is easy to fall into minimum point and slow convergence speed, an improved algorithm based on numerical optimization is adopted in this paper, and in the design of network structure, The quadratic polynomial approximation method is used to determine the number of hidden layer nodes. The neural network based on the above improved and optimized algorithm can quickly and accurately judge the working condition of the remanufacturing automatic gearbox.
【学位授予单位】：浙江工业大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3

【相似文献】