A model agnostic eXplainable AI based fuzzy framework for sensor constrained Aerospace maintenance applications
Bharadwaj Dogga, Anoop Sathyan, Kelly Cohen
TL;DR
This work tackles sensor-cost constraints in aerospace maintenance by proposing a model-agnostic, explainable AI framework that integrates SHAP, UMAP, and fuzzy c-means to prioritize sensors for prognostics. A regression neural network predicts remaining useful life (RUL) on the NASA C-MAPSS FD001 dataset, while SHAP explains feature importance and fuzzy clustering captures RUL bins under sensor constraints; SHAP-informed feature selection enables clustering with $70\%$ less data without sacrificing decision quality, as demonstrated across multiple data cases. The results show that SHAP-guided reductions improve cluster validity metrics (e.g., MI, Homogeneity, V-measure) and enable comparable maintenance decisions with far fewer sensors, signaling practical benefits for prognostics and health management (PHM) in aerospace. Overall, the framework offers a scalable, model-agnostic path to sensor prioritization and cost-effective maintenance planning in aging jet engines and similar PHM contexts, with potential applicability to finance and healthcare as well.
Abstract
Machine Learning methods have extensively evolved to support industrial big data methods and their corresponding need in gas turbine maintenance and prognostics. However, most unsupervised methods need extensively labeled data to perform predictions across many dimensions. The cutting edge of small and medium applications do not necessarily maintain operational sensors and data acquisition with rising costs and diminishing profits. We propose a framework to make sensor maintenance priority decisions using a combination of SHAP, UMAP, Fuzzy C-means clustering. An aerospace jet engine dataset is used as a case study.