基于优化稀疏编码学习的机械设备电子故障检测研究

Mechanical and electronic fault detection based on optimized sparse

针对现有机械设备电子故障检测方法非线性逼近性能差的问题,提出了基于优化稀疏编码学习的检测算法研究。采用了稀疏表达的方式来识别机械设备电子故障信号,提高了检测算法全局寻优的能力,避免陷入局部最优解;通过提升过完备字典模型内部原子结构与故障信号的匹配度的方式,获取了精度更高的稀疏解;促使稀疏逼近后重构信号的周期性与原始信号保持一致,并引入了特征自学习方案;最后采用分段的方式提取了各段信号的稀疏表征,改善了对原始故障信号的控制与检测性能。研究结果表明:提出检测算法在信号故障特征提取方面与原始信号周期性峰值匹配度更高,重构信号的控制误差较低,在稀疏度值超过100时的时间消耗相对于现有检测方法具有更明显的优势。

Aiming at the problem of poor nonlinear approximation performance of existing mechanical equipment electronic fault signal detection methods,a detection algorithm based on optimal sparse coding learning was proposed.The sparse expression was used to identify the electronic fault signal of mechanical equipment,which improved the global optimization ability of the detection algorithm and avoided getting into the local optimal solution.The sparse solution with higher accuracy was obtained by improving the matching degree between the atomic structure and fault signal in the over-complete dictionary model.The periodicity of the reconstructed signal after sparse approximation was consistent with the original signal,the feature self-learning scheme was introduced.Finally the sparse representation of each segment of the signal was extracted in a segmented way,and the control and detection performance of the original fault signal was improved.The results show that the proposed detection algorithm has a higher matching degree with the original signal periodic peak in signal fault feature extraction,a lower control error in reconstructed signal,and a better time consumption when the sparsity value exceeds 100 than the existing methods.