Research On Prediction Method Of Similarity Remaining Life Of Mechanical Equipment Based On Data

The use and development of mechanical equipment reduce the intensity of labor,improve the production efficiency and the speed of social and economic development.With the diversification of equipment service content and the accuracy of service requirements,the structure of mechanical equipment tends to be complex,and the components are more precise,showing the characteristics of high reliability and long life.Once the equipment has unexpected failure,it will also bring huge losses.Therefore,how to improve the degradation state of the key components of the equipment and the accurate prediction of the remaining life has become the core problem to ensure the safe operation of the equipment and reasonable maintenance arrangement.Driven by the fourth industrial revolution,the underlying sensing technology comprehensively detects the operation status of mechanical equipment,collects and records the data generated in the service process.The extensive construction and use of industrial big data platform provides a large number of life cycle data of similar equipment operation degradation.In the case of complex structure and failure mechanism of mechanical equipment,based on the data resources provided by big data environment,using data-driven method to predict the remaining life of mechanical equipment can achieve excellent results.In the datadriven method,due to the similarity,the residual life prediction method has the advantages of no need to establish degradation model,no need to set failure threshold,strong robustness and so on,which is suitable for predicting the residual life of mechanical equipment in the big data environment with sufficient reference samples.However,the current similarity remaining life prediction methods consider only a few factors in similarity calculation,and the method does not make full use of the historical data of service samples.These deficiencies will affect the final prediction results of the remaining life of mechanical equipment.In view of these deficiencies,this paper makes the following research.Firstly,in view of the problem that the similarity measurement in the traditional similarity remaining life prediction method is a single consideration,the paper improves the remaining life prediction method of the similarity of mechanical equipment from the perspective of the similarity calculation method.At the overall level of similarity calculation,the method of combining dynamic time warping and translation similarity calculation is used to calculate similarity,and the time factor is introduced in the specific calculation to improve the calculation method of translation similarity.Through the improvement of the similarity calculation method,the accuracy of the prediction result of the remaining life prediction method of the similarity of mechanical equipment is improved.Secondly,this paper studies the limitations of the mechanical equipment similarity remaining life prediction method,which is not good for the early and medium-term equipment prediction and the insufficient utilization of the historical data of service samples,and puts forward an improved mechanical equipment similarity remaining life prediction method based on recurrent neural network.The long short term memory neural network is used to mine the historical data of mechanical equipment,predict the subsequent changes of the mechanical equipment degradation index,and then use the predictive sequence to increase the length of the mechanical equipment degradation index sequence before and mid-term service,and finally use the combined sequence to perform similarity rest.Life prediction to improve the accuracy of the remaining life prediction of the early and mid-term service samples.Finally,the data-driven mechanical equipment similarity residual life prediction method is applied in the case of turbofan engine residual life prediction,and the realization process of the research content is described in detail,and the rationality and effectiveness of the proposed method is proved through the result analysis.