Application of multivariate joint model in life prediction of tunnel ventilation system
In this paper,we propose a data-driven multivariate joint model for predicting the Remaining Useful Life(RUL)of tunnel ventilation systems.By taking into account the variability in the monitoring process and the impact of environmental conditions,we conduct a joint analysis of multivariate monitoring data and failure data to identify shared random effects.This approach allows for the simultaneous estimation of parameters for the degradation rate model and the failure rate model,thereby reducing estimation bias and enhancing the accuracy of RUL prediction for tunnel ventilation systems.This methodology results in a significant improvement in the accuracy of dynamic RUL prediction.This study primarily utilizes a simulated fan bearing wear failure example to compare the prediction performance of our proposed method with the two-stage method and the unit joint model,using the average absolute error as the comparison metric,to assess the prediction accuracy of the model,the leave-one-out cross-validation method is employed.The bearing dataset is partitioned into a training set comprising 99 groups and a test set with 1 group.This process is repeated for each group within the dataset,rotating between designating a group as the test set and using the remaining groups as the training set.By systematically allowing each group of wind turbine bearing data to serve as the test set,we ensure a thorough evaluation of the model's performance and predictive capabilities.Upon analyzing the results of the computational study,it is evident that the prediction accuracy of the multivariate joint model surpasses that of the traditional two-stage method,highlighting its superior predictive performance.Furthermore,in the advanced stages of prediction,this model demonstrates higher accuracy in lifespan prediction compared to the unit joint model,showcasing a marginally improved predictive performance over the unit joint model.As the amount of available data increases,the prediction accuracy steadily improves,thereby confirming the efficacy of the model.The multivariate joint modeling approach adeptly identifies correlations between the evolution of monitoring states and occurrence of failure events.By minimizing estimation errors and enhancing accuracy,this method offers valuable decision support for ensuring the safe operation of subway tunnels.
万方数据
维普
