基于支持向量机与遗传算法的故障模式识别及趋势预测方法研究

发布时间：2018-02-16 13:26

本文关键词： 支持向量机（SVM）遗传算法(GA) 模式识别趋势预测小波变换滚动轴承齿轮　出处：《北京化工大学》2012年硕士论文　论文类型：学位论文

【摘要】：本文开展了基于支持向量机（Support Vector Machine—SVM）与遗传算法(Genetic Algorithms—GA)的故障模式识别及趋势预测方法研究，利用支持向量机对滚动轴承典型故障进行了模式识别，同时应用预测模型对齿轮状态趋势进行预测，并利用遗传算法分别对支持向量机分类过程和趋势预测过程进行了优化分析，主要工作如下：（1）基于SVM可以解决小样本学习问题这一优点，提出利用SVM对滚动轴承在正常、内圈缺陷、外圈缺陷和滚动体缺陷条件下工作的四种状态信号进行识别分类,为了提高分类识别率，利用遗传算法具有优良空间搜索性能的特点，对分类过程中的两个重要核参数初始值进行优化，提出了基于GA算法的改进SVM识别方法，研究结果表明：核参数初始值经过GA优化后SVM分类识别率得到了明显提高，能较好地实现轴承典型故障类型的识别。（2）为解决低转速滚动轴承故障特征难以提取的问题，利用小波变换技术具有高低频分离、局部细化和时频域内特征提取等性能优点，提出基于小波变换技术的低转速滚动轴承故障特征提取方法，对低转速轴承正常、外圈缺陷、内圈缺陷和滚动体缺陷等四种状态下的振动信号进行诊断分析，并结合SVM对轴承典型故障进行了分类识别，由分析结果可知，利用小波变换与支持向量机技术相结合的方法处理低转速滚动轴承故障问题能够取得很好的效果。（3）为了预测齿轮状态趋势发展状况，，建立三阶函数方程预测模型对齿轮趋势发展进行模拟分析研究，利用GA良好的空间搜索性，提出基于GA的预测模型函数优化方法，将获得的新预测模型函数与通过线性拟合原理获得的二阶、三阶函数做了对比分析研究。研究结果显示：经过GA优化后获得的三阶函数方程预测模型能够实现齿轮故障趋势发展预测模拟。
[Abstract]:This paper carried out based on support vector machine (Support Vector Machine - SVM) and genetic algorithm (Genetic Algorithms GA) to study the fault pattern recognition and prediction method based on support vector machine for pattern recognition of typical faults of rolling bearings, and applied to the prediction of the gear state trend, and the support vector forecasting classification process and trend analysis process was optimized by genetic algorithm, the main work is as follows:
(1) the advantages of SVM can solve small sample learning problems based on the proposed for rolling bearings in normal, inner defects with SVM outer defects and four kinds of rolling state signals of defects under the working conditions of the recognition and classification, in order to improve the classification accuracy, algorithm has excellent performance characteristics using the genetic search space, the two important parameters in the process of classification of the initial nuclear value optimization, this paper proposes an improved SVM recognition method based on GA algorithm, the results show that: the kernel parameter initial value GA optimized SVM classification rate has been significantly improved, can achieve better recognition of typical bearing fault types.
(2) to solve the problem of low speed rolling bearing fault feature extraction difficult problem, using wavelet transform technology with high frequency separation, local refinement and time-frequency domain feature extraction performance advantages, the low speed of wavelet transform technique of rolling bearing fault feature extraction method based on the low speed bearing outer ring inner ring is normal, defects, defects and the rolling defects diagnosis and analysis of vibration signals under four conditions, combined with the SVM classification of typical bearing faults, according to the analysis results, combined with the method of using wavelet transform and support vector machine technology with low speed rolling bearing fault problem can achieve good results.
(3) in order to predict the trend of gear fault, the establishment of three order function equation prediction model was simulated and analyzed to study the development trend of gear, the use of GA good space search, optimization method is proposed to predict the model function based on GA function, a new prediction model will be obtained by linear fitting with the principle of order two, order three the function to do the comparative analysis. The research results show that: three order function equation obtained by optimized GA prediction model can simulate the development trend of gear fault prediction.

【学位授予单位】：北京化工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3

【引证文献】