Research On Data-driven Intelligent Fault Diagnosis Technology Of Marine Hydraulic Pump

With the continuous improvement of the degree of ship automation,the operating speed and power of various machines continue to increase,and the hydraulic system is widely used in marine equipment such as deck equipment and ship engine rooms.As an important part of the ship’s hydraulic system,the hydraulic pump is essential to ensure the normal operation of the ship’s hydraulic system.However,since the hydraulic pump has been working in a harsh and complex environment for a long time,the hydraulic pump might be damaged and malfunction.If the hydraulic pump fails and is not diagnosed and maintained in time,it will cause the entire hydraulic system to be paralyzed,thus affecting the entire The normal navigation of the ship.With the rapid development of big data and artificial intelligence technology and its wide application in various fields,the application of big data and artificial intelligence technology to the fault diagnosis of marine hydraulic pumps has a very important engineering application value.In this paper,hydraulic pump is considerated as the research object.First of all,the composition structure and characteristics of the hydraulic pump is analyzed,The working parameters of the hydraulic pump are clarified as well as the working mechanism of the hydraulic pump and its typical failure modes(such as deformation,wear,corrosion,and fatigue)are used to carry out the hydraulic pump failure simulation test design.Next,the principle of data-driven fault diagnosis of hydraulic pump is described,and the monitoring data of hydraulic pumps of the test platform are pre-processed by removing outliers,EEMD noise reduction and smoothing.The pre-processed monitoring data were subjected to time-domain feature extraction based on statistical analysis,frequency-domain feature extraction based on power spectrum,CEEMDAN-based time-frequency domain feature extraction,and AR model-based time series feature extraction to obtain a large number of get a feature set that can represent the fault characteristics of the hydraulic pump;on this basis,through feature selection based on Spearman coefficients and feature fusion technology based on principal component analysis,comprehensive fault characteristics that can indicate the fault state of the hydraulic pump can be obtained.Finally,a fault diagnosis model based on quantile regression neural network is established to predict the sample’s fault type through the probability level,and the particle swarm optimization algorithm is used to optimize the model parameters to improve the accuracy of fault diagnosis.The research results show that the method proposed in this paper can accurately identify the fault types of marine hydraulic pumps and has a good identification effect for different types of hydraulic pump faults.This has great application value for the realization of online monitoring,fault diagnosis and maintenance of marine hydraulic pump.