基于判别稀疏编码的液压泵故障诊断

为解决液压泵故障信号特征难以提取的问题,提出了一种基于判别稀疏编码的液压泵故障诊断新方法。在稀疏编码框架中引入Fisher判别准则,通过对训练样本进行字典学习,获取具有判别性的字典与稀疏系数,使用不同故障类别字典对测试样本进行稀疏表示,利用全局分类方法综合重构误差与系数偏差两方面参数,对液压泵故障信号进行识别。实验结果表明,对于不同状态下的液压泵振动信号,该方法可自适应地完成各类子字典的学习与模式识别过程,与传统方法相比,在液压泵故障诊断中具有更高的准确率和较好的稳定性。

Fault diagnosis of hydraulic pump based on discriminantive sparse coding

In view of the difficulty in extracting the fault signal features of the hydraulic pump,a new method for hydraulic pump fault diagnosis based on discriminative sparse coding was proposed. The Fisher discriminative criterion was involved in the sparse coding to get discriminative dictionary and sparse coefficient through the process of dictionary learning. Different sub-dictionaries were used to represent the test sample by sparse representation. Then the reconstruction error and coefficient deviation were used to identify the fault types of hydraulic pump by the global classification method. According to the vibration signals of hydraulic pump in different conditions, the method can complete the process of dictionary learning and pattern recognition adaptively. Compared with the traditional method,the method for hydraulic pump fault diagnosis has higher efficiency and stable performance.