. 2026 Aug 6;16:24317. doi: 10.1038/s41598-026-56105-4

Table 1.

Comparison of our differentials with the approaches of the related work.

	Dataset	Degradation index(DI)	Health index(HI)	Piece-wise	Filters	Models	Similarity based	Time series	Results
Our proposal	Azure AI guide for PdM solutions	Yes	Yes	Yes	Savitzky –Golay and Kalman Filter	RRF and DNN (CNN, LSTM)	Yes	Yes	Regression coefficient () and accuracy above 90%. We compare RMSE, and we allow failure prediction with considerable anticipation of five days or one hundred and twenty telemetry cycles.
Wang et al.⁶	PHM08 Challenge	No	Yes	No	No	Regression Models and Moving Average Method	Yes	No	The results of the article are guided by competition. The score of the result sent was among the best.
Zheng et al.²⁷	C-MAPSS, PHM08 Challenge and Milling	No	No	Yes	No	Multi-Layer Perceptron (MLP), Support Vector Regression (SVR), Relevance Vector Regression (RVR), CNN and LSTM	No	Yes	The results are based on the comparing of the RMSE. Two tables are presented comparing the models used for each dataset.
Dong et al.³³	C-MAPSS Jet Engine Failure	No	No	No	No	LSTM	Yes	Yes	Results based on RMSE and presentation of RUL graphs.
Bruneo and De Vita³²	C-MAPSS	No	No	No	No	LSTM	No	Yes	Was made comparisons with literature results and presented the RMSE by reducing it by about 35% and 54% approaches, with the RMSE score equal to 11.42 and a percentage error equal to 0.12.
Liang et al.¹⁰	C-MAPSS	Yes	No	No	No	Linear Regression, Spearman-coefficient-based model	Yes	-	Results based on MAE and presentation of RUL graphs comparing with two other methods with considerable improvement.
Cai et al.²⁵	C-MAPSS	No	No	No	Particle Filter	Rao-Blackwellized Particle Filter (RBPF)	Yes	No	The comparison is with the Benchmarks, and using the RMSE and score criteria, claim that improved prediction compared with other existing RUL prediction methods.
Shi et al.³⁷	C-MAPSS	No	Yes	Yes	Exponential Filter	DA-LSTM	Yes	Yes	A detailed proposal for a novel DA-LSTM lightweight model based on exponential smoothing and concerns a can achieve high prediction accuracy with fewer parameters.