Assessing the remaining useful life (RUL) of components is essential for decision-makers in various industries, including manufacturing, transportation, and aerospace. By providing a reliable estimation of how long a part can continue to operate effectively, RUL assessment can inform decisions regarding maintenance, repair, or replacement of the component. To develop a technique for estimating the RUL of a system that is both highly accurate and less time-consuming than current methods, firstly, using the principal component analysis (PCA) manner, features were extracted in a reduced format; then, a hybrid model, i.e., an ensemble model that combines linear, k-nearest neighbors (KNN), and Gaussian process regression (GPR) through weighted averages, was designed. The well-known C-MAPSS dataset from NASA was employed to evaluate the performance of the model. To optimize the model's performance, an optimization procedure, i.e., constrained nonlinear multivariable was applied to identify the tuned weight in the ensemble learning that minimized the root-mean-square error (RMSE) value. Findings indicated that the optimized hybrid model outperformed the linear, KNN, and GPR models separately, as well as many of the prior investigations dealing with the same dataset, as evidenced by its lower RMSE value and the execution time.
Ensemble Learning for Estimating Remaining Useful Life: Incorporating Linear, KNN, and Gaussian Process Regression
Rezazadeh N.;Perfetto D.;De Luca A.;Caputo F.
2024
Abstract
Assessing the remaining useful life (RUL) of components is essential for decision-makers in various industries, including manufacturing, transportation, and aerospace. By providing a reliable estimation of how long a part can continue to operate effectively, RUL assessment can inform decisions regarding maintenance, repair, or replacement of the component. To develop a technique for estimating the RUL of a system that is both highly accurate and less time-consuming than current methods, firstly, using the principal component analysis (PCA) manner, features were extracted in a reduced format; then, a hybrid model, i.e., an ensemble model that combines linear, k-nearest neighbors (KNN), and Gaussian process regression (GPR) through weighted averages, was designed. The well-known C-MAPSS dataset from NASA was employed to evaluate the performance of the model. To optimize the model's performance, an optimization procedure, i.e., constrained nonlinear multivariable was applied to identify the tuned weight in the ensemble learning that minimized the root-mean-square error (RMSE) value. Findings indicated that the optimized hybrid model outperformed the linear, KNN, and GPR models separately, as well as many of the prior investigations dealing with the same dataset, as evidenced by its lower RMSE value and the execution time.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.