Research On Fault Diagnosis For Car Dumper Hydraulic System Based On Multiple Models PCA

The hydraulic control system occupies an important position in the manufacturing of high power equipment.The hydraulic system fault caused by oil pollution directly affects the efficiency and production cost of equipment,and brings inestimable loss to the enterprise.Therefore,fault prediction and fault diagnosis of hydraulic control system are carried out to ensure the smooth and efficient work of hydraulic control equipment.Car dumper is a typical hydraulic control equipment,which carries the task of bulk material unloading in railway transportation.Its production process has the characteristics of repeatability.Its high utilization rate and bad working environment result in a relatively high fault rate of hydraulic system.Therefore,this paper takes the car dumper hydraulic system as research object,uses multiple models static and dynamic multivariate statistical analysis method to extract the characteristic relation of its monitoring data.Fault monitoring is based on the high order statistics constructed by the extracted eigenvector.The results of study have important theoretical reference and practical significance to state monitoring and fault diagnosis of hydraulic system with repetitive working process.The simulation model of car dumper hydraulic system was set up,the gravity model of remaining coal in the unloading process of the car dumper is established based on the structural mechanics of coarse aggregate.This model was used as the load of simulation of hydraulic system.On the basis of correct simulation model,the fault characteristic curves of the system which are difficult to get in production were obtained by simulation.The characteristic curves were used as the reference data for fault identification.Due to the monitoring data of car damper has the characteristics of batch,multimode and changing work condition,the multiple models static principle component analysis(PCA)was used to extract the static relation between variables.A recursive adaptive algorithm was used to update the static feature to reduce the effect of changing work condition on the result of fault monitoring.Fault identification based on multiple PCA variable reconstruction.A directional fault subspace was proposed to describe the variation trend of variable in the fault state.Through the actual monitoring data,it is proved the method has high fault recognition rate and identification accuracy.The influence of time series relation of data on the feature vector of PCA and the control limit of squared prediction error(SPE)were analyzed.According to the definition of dynamic PCA feature extraction in papers,the dynamic PCA was divided into elimination time series relation dynamic PCA and augmented matrix dynamic PCA.The difference and connection of elimination time series relation dynamic PCA,augmented matrix dynamic PCA and static PCA were illustrated by an example of two dimensional variables.Due to the time series relation of car dumper hydraulic system monitoring data is nonstationary,the multiple models dynamic PCA was proposed.The nonstationary time series relations were represented by multiple models.The time series relations were extracted by dynamic PCA.The two types of multiple models dynamic PCA were used to extract the characteristic relation of the monitoring data.The monitoring results showed the fault recognition rate based on augmented matrix dynamic PCA was higher than static PCA and elimination time series dynamic PCA.The software and hardware of characteristic parameters monitoring of hydraulic system were designed.Fault diagnosis program for car dumper hydraulic system was developed based on PCA statistical characteristics.The system has passed field test,which provided convenience for car dumper maintenance personnel.