一种基于多测量向量模型的机械振动信号联合稀疏重构方法

A joint sparse reconstruction method for mechanical vibration signals based on multi-measurement vector model

针对目前机械设备越来越智能化、高速化、集成化和复杂化,传统的压缩感知模型(单测量向量)获得的测量信息比较单一,需要在不同监测点分别进行测量获得多个信号数据,浪费时间,并且忽略同一机器不同监测点之间信号的相关性,为充分利用信号间和信号内的相关性,更大程度上减少冗余性和采样时间,提出一种基于多测量向量模型的机械振动信号联合稀疏重构方法。重点研究了重构方法的设计:基于粒子群算法,首先通过时间稀疏贝叶斯算法求解出初始解,然后结合贪婪算法的修剪技巧并加入自适应粒子激活机制进行位置更新寻找最优解,最后对振动信号进行精确重构。实验结果显示,该方法较其它方法而言能有效的恢复机械振动信号且重构误差相对较小。

Aiming at mechanical equipment becoming more and more intelligent,high-speed,integrated and complicated,The traditional compressed sensing model(single measurement vector)obtains single measurement information,it needs measuring at different monitoring points to obtain multiple signal data,it wastes time and ignores signal correlation between different monitoring points on the same machine.Here,to make full use of inter-signals and intra-signal correlations and further reduce redundancy and sampling time,a joint sparse reconstruction method for mechanical vibration signals based on the multi-measurement vector model was proposed.The design of the reconstruction method was studied emphatically.Firstly,based on the particle swarm optimization algorithm,the initial solution was solved with the time sparse Bayesian algorithm.Then,the pruning technique of the greedy algorithm combined with the adaptive particle activation mechanism was used to do position updating and search the optimal solution.Finally,the vibration signal was reconstructed accurately.Test results showed that compared with other methods,this method can effectively recover mechanical vibration signal and the reconstruction error is relatively smaller.