【作者】 朱昌堆；

【导师】 袁洪芳；

【作者基本信息】 北京化工大学 ， 控制科学与工程， 2012， 硕士

【摘要】 本课题来源于国家自然科学基金项目《机械故障无线传感网络监测与智能诊断方法研究》（项目编号：51075023）。声发射检测技术用于故障诊断领域是当前研究的热点问题之一，声发射信号由于处于高频范围内，不易受到各种低频振动噪声干扰，具有良好的应用前景。但是，声发射检测技术在工业生产应用中，如何进行有效的信噪分离，如何选择特征参数以提高状态识别的准确率，如何选择适用于不同要求的状态识别方法是目前存在的三个主要问题。本文针对声发射监测系统中存在的主要问题开展研究，在信噪分离上提出了以小波变换为核心，以强制消噪法和小波阈值消噪法相结合的信号预处理方法，并将信息熵的概念引入小波基函数的选择过程中，选择出与滚动轴承声发射信号较为吻合的小波基函数，提高了消噪的效果；研究了基于现代信号处理方法的声发射信号特征提取方法，提出了以互信息理论和距离测度相结合的特征参数评价因子，选择敏感度高冗余度低的状态特征参数；研究了基于参数有效性的状态识别方法，提出特征参数与状态类别的相关性和参数之间距离测度作为有效性系数改进模糊识别方法，提高了识别的准确率。将改进的模糊聚类法与包络谱分析法作为监测单元简易状态识别方法，而将蚁群聚类算法和粒子群聚类算法作为上位机精确识别方法，经过测试，这种监测单元简易识别与上位机精确识别结合的识别模式，提高了滚动轴承状态识别的效率。更多还原

【Abstract】 This project comes from the Nation Natural Science Funds project theresearch of mechanical fault wireless sensor network monitoring andintelligent diagnosis method, project number:51075023. Acoustic emissiontesting technology for fault diagnosis field is one of the hot research issues.Because it is in high frequency range, is hardly disturbed by vibration noisein low frequency. It has a good application prospect. But when the acousticemission test technology is applied in industrial production, it has theproblems: how to get effectively signal-to–noise separation, how to selectthe characteristic parameters in order to improve the accuracy of staterecognition, how to select the suitable recognition method for differentrequirements.The project research and design a complete set of rolling bearingacoustic emission monitor, which is based on the hardware design ofacoustic emission wireless monitor unit. This paper proposes the signalpretreatment method, which combines the forced de-noising method andwavelet threshold de-noising method based on wavelet transform; and the entropy introduced into selection wavelet function, it can help choosing thewavelet function fit for acoustic emission signal of rolling bearing. Thismethod can effectively separated signal-to-noise. This paper proposes thenew feature selection method based on combine mutual information theoryand distance measurement. The characteristic parameter subset based on thenew method can get higher recognition accuracy compared with thetraditional method. The fuzzy recognition is improved by the coefficients,which is the relationship between the parameters and state, also the distancemeasurement between parameters. The improved fuzzy recognition methodgets good recognition result. The envelope spectrum method and improvedfuzzy are used in recognitionsimple recognition of monitor unit, and the antcolony clustering recognition and particles swarm clustering recognitionalgorithm are used in accurate recognition of PC. The model combinessimple recognition of monitoring unit with accurately recognition of PC,improving the efficiency of the rolling bearing state recognition.更多还原