Research On The Healthy Life Prediction Technology Of Typical Parts Of Electromechanical Equipment Based On Data-driven

With the continuous development and advancement of industrial information technology,electromechanical equipment is increasingly developing towards precision,complexity and intelligence.While creating convenience for human production and life,it also brings new challenge to equipment fault prediction and health management.Once the mechanical equipment fails,it will affect the healthy operation of the entire electromechanical system,resulting in serious production events and even more serious loss of life and property.Bearings are one of the key components of electromechanical equipment,most mechanical equipment cannot be healthy and stable due to fault,and there is a direct or indirect relationship with bearing damage.Therefore,this topic uses the current research methods of mainstream life and health prediction at home and abroad: based on the data-driven method,taking the bearing as an example to study the health life prediction technology of electromechanical equipment,the main research contents are as follows:Firstly,with the help of all the vibration signals collected throughout the life cycle of the bearing,we analyze and compare the changes of different time domain features and frequency domain features during the whole life cycle of the bearing.Finally,we select four characteristics of root mean square,crest factor,kurtosis index,and FSPS as the degradation indicators to predict the remaining useful life of bearing.Then,in view of the shortcomings of Mean-DNN and other methods,we propose a deep association neural network(ASDNN)prediction model that combines K-nearest neighbor regression prediction model and fully connected deep neural network prediction model,and is compared with four typical methods to illustrate its superiority and feasibility by IEEE PHM The 2012 data set.Secondly,we propose a rolling bearing residual life prediction method based on the multivariate ASCNN prediction model,and use the evaluation indicators to evaluate the model prediction performance.We use the public dataset for experimental verification,and compare the results with the ASCNN model using the original degradation feature index to illustrate the effectiveness of the prediction method.In view of the shortcomings of this paper,we summarize and forecast in the fifth chapter of the article,work out the next step of the research work,and hope that we can continue to learn and improve.