多尺度移不变稀疏编码及其在机械故障诊断中的应用

Multi Scale Shift Invariant Sparse Coding for Robust Machinery Diagnosis

提出了一种融合多尺度信息的高效移不变稀疏编码算法,并用于机械故障诊断.将移不变稀疏编码作为分类器应用于故障诊断,直接对振动信号进行训练和识别,不需要提取特征和预处理.为进一步提升效果,将不同尺度的移不变稀疏编码分类器融合在一起.经实验验证,即使在训练样本和测试样本负载不同的情况下,文中方案仍然能够以较高的准确率识别出轴承的故障位置和程度.与其他方法相比,文中方法的准确率、鲁棒性更高,具有一定的工程应用价值.

An efficient shift invariant sparse coding （SISC） algorithm which combined multi scale information was proposed for machinery fault diagnosis. The SISC was applied as a classifier for fault diagnosis, and it could directly train and recognize the vibration signals without feature extraction or pre-processing. What＇s more, multi scale SISC classifiers were combined for better performance. Validated by experiments, although the loads of training samples and testing samples were not the same, this scheme could still precisely determine the fault location as well as the severity of fault for bearings. Compared with other methods, the proposed algorithm shows high accuracy , strong robustness and a certain value for engineering application.