Kernel ridge regression-based failure probability prediction method for ventilators
10.19745/j.1003-8868.2025090
- VernacularTitle:基于核岭回归的呼吸机故障概率预测方法研究
- Author:
Li-tian FAN
1
;
Zhu CHEN
1
;
Si-yuan XIE
1
;
Hao-jie LI
1
;
Qi-lin LIU
1
Author Information
1. 四川大学华西医院,成都 610041
- Publication Type:Journal Article
- Keywords:
kernel ridge regression;
ventilator;
failure probability prediction;
reliability analysis;
remaining useful life
- From:
Chinese Medical Equipment Journal
2025;46(5):73-77
- CountryChina
- Language:Chinese
-
Abstract:
Objective To propose a ventilator failure probability prediction method based on kernel ridge regression(KRR).Methods Firstly,the failure interval data of ventilators was collected and preprocessed to remove outliers.Secondly,the median rank method was used to estimate the failure probability.Finally,using the time data as the feature variable and the failure probability value as the target variable,a KRR model was established and trained by selecting the optimal kernel function and hyperparameter combination from radial basis kernel function,linear kernel function,polynomial kernel function,and S-type kernel function through grid search and cross-validation methods to predict ventilator failures.To verify the performance of the KRR model in predicting ventilator failure probability,it was compared with Weibull and its extended models.Results KRR achieved a coefficient of determination of 0.993 5,a mean squared error of 5.399 5×10-4,a root mean squared error of 0.023 2 and a mean absolute error of 0.018 3,outperforming Weibull and its extended models in prediction accuracy and error control.Conclusion The failure probability prediction method for ventilators based on KRR demonstrates exceptional performance in prediction accuracy and error control,and thus holds great potential for application.[Chinese Medical Equipment Journal,2025,46(5):73-77]
