The characteristics information of fault in the axis piston pump was investigated. It was of assistance in the identification and classification of faults. From all the faults of axis piston pump, two main kinds of faults were chosen, containing port-plate wear and slipper wear. Mass data of the two main faults characteristics is abstracted from original vibration signals of axis piston pump. The data is transformed by wavelet package and mathematics and selected by genetic algorithm-partial least squares (GA-PLS) in order to get the optimization fault characteristics set. Aiming at some problems of longer training time and weight adjustment, a new AdaBoost algorithm based on uniformly distributed weight and exponential loss function was proposed. The classification models were constructed with AdaBoost Ml and improved AdaBoost algorithm. It is proved that improved AdaBoost algorithm with the optimization set of few characteristics can get better classification results.%对轴向柱塞泵故障特征信息的研究有助于辅助完成轴向柱塞泵故障类型的鉴别和分类.从轴向柱塞泵的所有故障中,选出两种典型故障:缸体与配流盘磨损、柱塞滑履松动.从轴向柱塞泵原始振动信号中提取这两种故障特征的数据,经过小波包变换、数学变换以及遗传算法和偏最小二乘回归相结合(GA-PLS)特征选择后,确定最优的故障特征集.为了解决训练时间较长及权重调整过适应等问题,提出一种基于均匀分布权重和指数损失函数的改进型AdaBoost算法.分别使用AdaBoost M1,改进型AdaBoost构建分类模型比较其分类效果.结果表明:改进型AdaBoost使用仅含有少量的特征组成的最优特征集,可以得到较好的分类结果.
展开▼
