本文關(guān)鍵詞：高速列車轉(zhuǎn)向架故障診斷智能決策方法研究　出處：《西南交通大學(xué)》2017年碩士論文　論文類型：學(xué)位論文

  更多相關(guān)文章： 高速列車 轉(zhuǎn)向架 粒子群優(yōu)化算法 支持向量機(jī) 故障診斷 特征提取 分級策略

【摘要】：在高速列車長期服役過程中,列車轉(zhuǎn)向架關(guān)鍵部件的性能蛻化與故障對列車的安全運(yùn)行造成嚴(yán)重威脅。列車運(yùn)行過程中通過在轉(zhuǎn)向架不同位置安裝各種類型的傳感器對轉(zhuǎn)向架進(jìn)行評估,對高速列車的安全運(yùn)營有重要意義。本文通過對監(jiān)測數(shù)據(jù)特征提取的分析,建立特征提取知識庫,構(gòu)建了故障診斷決策模型。由于支持向量機(jī)參數(shù)對其性能影響較大,通過改進(jìn)的粒子群優(yōu)化算法優(yōu)化支持向量機(jī)參數(shù)。對列車轉(zhuǎn)向架的原車、2種位置的空氣彈簧故障、4種位置的橫向減振器故障以及8種位置的抗蛇行減振器故障這15種工況,給出了基于分級策略的診斷框架。具體的研究工作如下:1、利用已有的多種列車轉(zhuǎn)向架振動信號特征提取方法,建立高速列車轉(zhuǎn)向架振動信號的特征提取知識庫,構(gòu)建了高速列車轉(zhuǎn)向架故障診斷決策模型,并對診斷決策模型進(jìn)行了數(shù)學(xué)描述。2、在故障診斷模型中,通過改進(jìn)的粒子群優(yōu)化算法選擇支持向量機(jī)的懲罰因子以及核函數(shù)參數(shù)。針對粒子群優(yōu)化算法易陷入局部最優(yōu)的問題,給出改進(jìn)方法,首先,速度更新公式乘以收縮因子,其次,慣性權(quán)重采用高斯函數(shù)遞減策略。利用公開數(shù)據(jù)進(jìn)行測試,結(jié)果表明使用改進(jìn)后的粒子群優(yōu)化算法對支持向量機(jī)參數(shù)優(yōu)化能夠提高分類的正確率。3、基于分級策略,給出了列車轉(zhuǎn)向架故障分級診斷框架,并根據(jù)實際情況確立了分級策略下高速列車轉(zhuǎn)向架故障診斷的分級順序,診斷順序依次為原車、空氣彈簧失效、橫向減振器失效與抗蛇行減振器失效。利用對振動信號離散傅里葉變換的幅值作為特征進(jìn)行原車的識別;利用振動信號的奇異譜熵、功率譜熵、小波能譜熵和小波空間特征譜熵對空氣彈簧失效、橫向減振器失效和抗蛇行減振器失效進(jìn)行故障識別;利用振動信號的奇異譜熵、功率譜熵、小波能譜熵和小波空間特征譜熵對空氣彈簧失效進(jìn)行故障定位。利用改進(jìn)后的粒子群優(yōu)化算法對支持向量機(jī)參數(shù)進(jìn)行優(yōu)化。結(jié)果表明分類結(jié)果的正確率較高,與已有的研究相比較正確率有較大提高。
[Abstract]:In the high-speed train during long term, pose a serious threat to performance degeneration and fault plane of train key components to safe operation of the train. The train running through the steering sensor aircraft of various types of installation in different positions to assess the bogie, is of great significance to the safe operation of high-speed train. This paper through the analysis on the extraction the characteristic of monitoring data, establish feature extraction knowledge base, constructs the decision model of fault diagnosis. Because the parameters of SVM has great influence on its performance, the machine to the volume by the improved particle swarm optimization algorithm to optimize the support parameters of train bogie. The original car, 2 position of the air spring fault, lateral damper fault 4 position and yaw damper fault 8 position of the 15 conditions, gives the diagnosis framework based on classifying strategy. The specific research work are as follows: 1, using the existing Frame vibration signal characteristic extraction method for multi train steering, the establishment of high-speed train steering characteristics of frame vibration signal extraction of knowledge base, the construction of high-speed train bogie decision model of fault diagnosis, and the diagnostic decision model and the mathematical description of the.2 in the model of fault diagnosis, the improved particle swarm optimization algorithm selection of support vector machine the penalty factor and kernel parameter. To solve the problem of particle swarm optimization algorithm is easy to fall into local optimum, the improvement method is given firstly, velocity updating formula multiplied by the contraction factor, secondly, the inertia weight decreasing strategy. Gauss function was tested by using public data, results show that using the improved particle swarm optimization algorithm for parameter optimization of support vector machine to improve the classification accuracy of.3, based on the hierarchical strategy, gives the train bogie hierarchical fault diagnosis framework, and according to the actual situation. The high-speed train bogie grading classification strategy under sequential fault diagnosis, the diagnosis is in the order of the original car, air spring failure, failure and yawdamper lateral damper. The amplitude of the vibration signal of the discrete Fourier transform as the feature of the original car identification; using the singular spectrum entropy of vibration signal and power spectrum entropy, wavelet energy spectrum entropy and wavelet space feature entropy on the failure of the air spring, failure fault recognition of lateral damper failure and anti hunting damper; vibration signal by using the singular spectrum entropy, power spectrum entropy, wavelet energy spectrum entropy and wavelet space feature entropy for fault location of air spring failure by particle. Swarm optimization algorithm is improved to optimize the parameters of support vector machine. The results show that the classification accuracy rate is higher, compared with the existing research on accuracy is greatly improved.

【學(xué)位授予單位】：西南交通大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：U279.323;TP18

【參考文獻(xiàn)】

相關(guān)期刊論文 前10條

1 ;國家發(fā)改委正式公布《中長期鐵路網(wǎng)規(guī)劃》[J];城市規(guī)劃通訊;2016年14期

2 馮曉琳;寧芊;雷印杰;陳思羽;;基于改進(jìn)型人工魚群算法的支持向量機(jī)參數(shù)優(yōu)化[J];計算機(jī)測量與控制;2016年05期

3 曲健;陳紅巖;劉文貞;李志彬;張兵;應(yīng)亞宏;;基于改進(jìn)網(wǎng)格搜索法的支持向量機(jī)在氣體定量分析中的應(yīng)用[J];傳感技術(shù)學(xué)報;2015年05期

4 魏峻;;基于和聲搜索算法的支持向量機(jī)參數(shù)優(yōu)化[J];河南科學(xué);2014年07期

5 莫贊;劉希良;謝海濤;;基于混合梯度下降算法的支持向量機(jī)參數(shù)優(yōu)化[J];網(wǎng)絡(luò)新媒體技術(shù);2014年02期

6 戴上平;宋永東;;基于遺傳算法與粒子群算法的支持向量機(jī)參數(shù)選擇[J];計算機(jī)工程與科學(xué);2012年10期

7 張迅;王平;邢建春;楊啟亮;;基于高斯函數(shù)遞減慣性權(quán)重的粒子群優(yōu)化算法[J];計算機(jī)應(yīng)用研究;2012年10期

8 于明;艾月喬;;基于人工蜂群算法的支持向量機(jī)參數(shù)優(yōu)化及應(yīng)用[J];光電子.激光;2012年02期

9 張煜東;吳樂南;王水花;;專家系統(tǒng)發(fā)展綜述[J];計算機(jī)工程與應(yīng)用;2010年19期

10 張衛(wèi)華;王伯銘;;中國高速列車的創(chuàng)新發(fā)展[J];機(jī)車電傳動;2010年01期

，

本文編號：1406517