基于神经网络的汽车故障诊断.doc

《基于神经网络的汽车故障诊断.doc》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。

1、基于神经网络的汽车故障诊断1.问题描述这里，我选取了汽车变速箱作为本次故障诊断作为研究对象。在汽车变速箱的故障诊断中，是以不同振动信号下齿轮啮合次数作为故障诊断的参数。这里我们以啮合次数和故障的对应关系通过神经网络的方法进行训练：正常运行状态时啮合次数约为0-70万次、磨损运行状态时啮合次数为70-420万次、故障运行状态(断齿)时啮合次数为420万次以上。为此，通过汽车齿轮箱的振动信号经小波包分解，在各频带能量序列作为诊断的依据。2.神经网络设计学习样本与试验中，样本数量的选择应尽可能多，以包含尽可能多的故障类型。现选正常运行状

2、态的特征向量5个，因整个磨损状态所经历的运行时间比较长，故磨损运行状态的特征向量样本多一些，定为11个，断齿状态的特征向量样本5个。因此决策属性有三类，“0”表示正常运行状态，“1”表示磨损运行状态，“2”表示断齿状态。所用数据集见表1：表1学习样本U频率范围(Hz)D0~188(a)188~375(b)375~563(c)563~750(d)750~38(e)938~1125(f)1125~1313(g)1313~2500(h)161.02289.59371.756150.8058.51620.30533.27751.46302

3、85.11891.61668.181165.73411.59221.79035.27654.1540355.12495.44967.286159.6078.49321.54332.75453.4010485.376113.49966.016169.41010.65322.28339.52756.4220577.922101.10569.232157.2879.73720.66434.84353.8070672.1360168.31041.234109.1143.86718.65919.92827.08117147.527146.7

4、8459.772105.21217.20519.01938.90732.8931895.492197.01593.170182.34411.65127.73646.66959.43419103.998167.36273.005177.89412.80629.12243.57551.336110120.500168.11069.048140.13913.11425.82738.80243.172111163.397152.31379.554163.81123.12630.75646.08056.873112123.182166.60

5、067.123115.26720.87719.25042.20341.98011371.945211.17172.588137.73022.67122.11243.31545.104114164.045194.62968.603127.54116.12722.28641.45430.387115172.501308.04994.592177.90432.56936.06447.80352.628116100.973272.33980.250174.41422.56143.80350.03347.686117128.261294.4

6、31108.704255.53627.97735.22059.05571.603218133.521356.309111.461230.05029.39243.08856.88376.239219142.032338.22693.684287.42524.96938.71655.25673.801220171.641270.335103.851240.08331.78939.00454.77375.723221216.891371.829117.591285.67335.07739.50467.19179.9912此外还设计了三组

7、样本数据（见表2）用于测试，以验证神经网络的训练效果。表2测试样本U频率范围(Hz)D0~188(a)188~375(b)375~563(c)563~750(d)750~38(e)938~1125(f)1125~1313(g)1313~2500(h)168.80095.34667.424159.7668.11421.06435.15350.91902133.003290.54081.755169.59124.24336.09653.15842.85913261.994299.094132.719251.22236.56935.87

8、667.68790.5492将这些数据进行归一化处理，以方便训练和检验：表3学习样本（归一化）频率范围(Hz)UD0~188(a)188~375(b)375~563(c)563~750(d)750~38(e)938~1125(f)1125~1313