Research On Degradation And Life Evaluation Technology Of Hydraulic Pump Based On Data Drive

With the rapid development of the military industry and equipment manufacturing industry,military engineering machinery and equipment are constantly moving towards precision,integration and intelligence.In the military vehicle hydraulic system,the hydraulic pump is the core component that realizes the conversion between hydraulic energy and mechanical energy.Its performance and life span play a vital role in the safe and reliable operation of the entire hydraulic system.Due to the complex and diverse environment of military vehiclemounted hydraulic systems in the working process,they are susceptible to many factors such as impurities,moisture,dust,etc.Which accelerate the wear and failure of internal components of the hydraulic pump.In order to grasp the service capability of the hydraulic system in time,it is necessary to evaluate the degradation state and the remaining use life(RUL)of the hydraulic pump equipment to ensure the reliability and safety of the military on-board hydraulic system.Due to the complexity and variability of the wear failure mechanism and influencing factors of hydraulic pumps,the degradation process and failure mode will show a multi-factor coupling phenomenon,which is difficult to describe with accurate physical models.In addition,in military hydraulic systems,hydraulic pumps have the characteristics of high reliability,long life,and small samples.It is difficult to evaluate equipment degradation and remaining life using a life distribution model based on failure statistics.Therefore,from the perspective of mechanical performance degradation,this paper carries out a data-driven hydraulic pump degradation and life assessment technology research.The main contents are as follows:1)By analyzing the mechanism of hydraulic pump wear degradation,it is possible to better perceive and describe the complexity and variability of hydraulic pump wear failure,and to measure the impact of solid particle pollutants on the performance and life of hydraulic pump components.Taking plunger pumps and gear pumps as research examples,conducting pollution sensitivity tests,using logarithmic nonlinear regression and flow attenuation data under various test conditions,to solve the pollution sensitivity coefficients of pollutant particles in each size range,so as to construct a failure-based the life evaluation model of the mechanism and the shortcomings of the existing models are summarized.2)Research on the extraction and screening of hydraulic pump degradation characteristics to mine performance degradation information during equipment operation.On the basis of wavelet packet threshold denoising,multi-dimensional feature extraction is performed from the perspective of time domain,frequency domain and time-frequency domain,and then feature standardization,feature dimensionality reduction,feature selection and performance degradation features in the feature selection process.A series of studies were conducted on the construction of evaluation indicators.Taking the common gear pumps in hydraulic pumps as the research object,it can be seen that this method can effectively extract and screen the performance degradation characteristics of hydraulic pumps,and establish a foundation for the construction of the remaining life assessment model in the subsequent chapters.3)Through the use of fuzzy C-means(FCM)clustering algorithm,the division of hydraulic pump degradation stages and the formulation of corresponding failure standards are realized.Because the hidden markov model(HMM)has certain limitations in characterizing the actual degradation process of hydraulic pumps,it is necessary to use the hidden half markov model(HSMM)derived from it for modeling,and use forward-backward the algorithm,Viterbi algorithm,and Baum-Welch algorithm solve the basic problems of the model,improve the model to make it suitable for multi-characteristic sequences,and introduce the degradation factor into the model to modify the state transition matrix,thereby constructing a model based on FCM-HSMM’s hydraulic pump remaining life assessment model.On the basis of high-quality feature screening research examples,the accuracy of the remaining life assessment model built is verified by comparison with traditional methods.4)According to the construction and selection methods of time series samples under degradation characteristics,an integrated learning model based on support vector regression(SVR)is established from two perspectives to evaluate the remaining life of the hydraulic pump.The time series sample extraction technology based on sliding window is used to integrate the time change characteristics of the degraded series into the training sample set.The SVR,which is good at dealing with multi-feature learning problems,is selected as the base learner,and its basic principles and kernel function construction are elaborated and analyzed in detail.The proposed method is verified by combining high-quality feature screening research examples,and compared with the prediction results of the remaining life assessment model of hydraulic pump based on FCM-HSMM.It can be seen that the integrated learning method can effectively compensate for the prediction defects of each base learner,so that In the process of building the base learner.It does not have to be obsessed with the optimization of model parameters,and improves the tolerance and robustness of data set quality,model training and parameter optimization.