Prediction Of Piston Pump Leakage Based On Time Series Method

As the core power component of hydraulic system,piston pump has lots of advantages,such as high output pressure,small flow pulsation,high power etc.Therefore,it is widely used in aerospace,marine ships,engineering machinery,national defense and military industry and many other fields.However,China’s hydraulic industry is large and not advanced at present.The high-pressure piston pump depends on imports mainly and it is listed as one of 35 "bottle-neck" key basic components by the country at present.The reliability of piston pump is vital to the operation of mechanical equipment and its closely related to national security and economy.With the urgent demand of the market for the safety,long life and intelligence of hydraulic equipment,it is particularly important to improve the performing reliability of piston pump by carrying out the study on the performance degradation of piston pump.Based on this,this thesis regards the swashplate axial piston pump as the research object,and carries out the research from the following aspects:(1)By analyzing the physical mechanism of performance degradation caused by friction and wear of four key friction pairs inside the piston pump,the leakage quantity was determined as the external characteristic quantity reflecting the degradation,and the change relationship between the leakage quantity of the friction pair and its corresponding equivalent clearance was analyzed.The leakage of piston pump forms a random time series with the increase of operation time and the time series analysis method is selected to predict the leakage flow of the pump.(2)Based on the leakage data of piston pump,ARIMA model is established to predict the leakage data of single group of piston pump in different time periods and the leakage data of multiple groups of piston pump in the same time period.The analysis shows that ARIMA model fits the data with long-term trend well,the prediction is accurate,and the fitting result is poor when the data fluctuates greatly.By comparing different filtering methods and combining the characteristics of leakage data,HP filtering was used to decompose the leakage data into trend term and fluctuation term data,and the HP-ARIMA combined model was established for the trend data,and the HP filtering method significantly improved the prediction accuracy.(3)Combined with statistical correlation method to verify that the leakage data of piston pump has nonlinear and heteroscedasticity characteristics,the extended threshold autoregressive moving average SETARMA model and conditional heteroscedasticity model ARIMA-GARCH were established respectively,and the same comparison method was used to analyze the prediction results to verify its feasibility.The prediction results show that the nonlinear model is more accurate than the linear model.(4)Considering the leakage data of piston pump,the improved methods for time series model are proposed from different angles.For the leakage data of the single pump,the number of training set data is comprehensively selected,and the error weighted least square method was used to optimize the model parameters to reduce the influence of data errors.A rolling prediction method was proposed to improve the accuracy of multi-step prediction.Based on the similarity of the degradation process of several pumps,a weight prediction model was established by means of calculating the mean value of model parameters.Compared with the traditional time series model,the improved method can improve the accuracy of leakage prediction in different degrees.