| In modern industrial production processes,individual machine components have their own functions,thus ensuring the normal operation of the machine.Rolling bearings are the basic essential components of the mechanical industry and are known as "industrial joints".As a traditional confidential mechanical component,rolling bearings can be seen everywhere in modern industry,but rolling bearings are also very prone to damage,if they suddenly fail,and may cause equipment to stop running,serious consequences will be immeasurable.Moreover,by using the vibration signal,we can trace the developing history of rolling bearing,recognize its deterioration status,and predict its residual life.So as to timely and effectively implement equipment maintenance measures,from timely maintenance to situational maintenance transformation,not only can the equipment be operated safely,but also a great deal of human and material resources can be saved,so as to avoid unnecessary losses.Consequently,it is of great importance to analyze the original vibration signal of rolling bearings.The related feature information is extracted to predict the residual life of the rolling bearings.In this thesis,the life cycle vibration signal of rolling bearing is researchedand a new method is put forward to recognize the degradation status and residual lifetime of the rolling bearings,and a new method to deep learning on the basic that bidirectional long and short-term memory networks is put forth.In the course of the life cycle test,it is found that the rolling bearing vibration signal is erratic and non-linear.First of all,the pristine vibration signals of rolling bearings are parsed and processed by using Variational mode decomposition(VMD)signal to improve the signal to noise,so as to effectively extract the degradation characteristics.The fundamental principle and procedures of VMD are introduced,and the adaptive optimization of the modal components K and the penalty factor alpha in the VMD algorithm are adaptively optimized by combining the particle swarm optimization algorithm.Combined with experimental data,the feasibility and validity of VMD algorithm are validated.After extracting the time-domain,frequency-domain and time-frequency domain degradation features of the original vibration signal,the monotonicity,tendency,Pearson correlation coefficient and Spearman grade correlation coefficient are used to screen the features,construct the degradation feature parameter set,the starting point of bearing degradation is determined,and the foundation is laid for predicting and modeling the remaining service life of the remaining bearing.Then,exploiting the advantages of bidirectional long short-term memorynetwork in time series prediction,a network model on Bi LSTM is set up to anticipate the remaining service life of rolling bearings.Finally,by comparing it to onventional deep learning algorithm,the advantage of the remaining service life prediction model of rolling bearings make comparison with the signal analysis method and bidirectional long short-term memory network used in this paper is proved. |
