Summary
Making sense of deep learning models that predict how long a machine part will last is tricky because the usual heatmaps highlighting important sensors don't always behave reliably. The authors trained three popular types of models and tested three popular explanation methods to see how well these methods reveal real influences on predictions when sensors are changed or even slightly noisy. They found that two methods, Integrated Gradients and occlusion, were better at showing true cause-and-effect on some models, while attention maps were easier to produce but less accurate. Their work helps engineers know which explanation tools to trust when assessing sensor data for maintenance decisions.
What this means in practice
- •For maintenance engineers: Choose reliable heatmap methods like Integrated Gradients or occlusion to interpret sensor data when predicting equipment lifespan from C-MAPSS-style models.
- •For industrial ai teams: Validate explanation tools for predictive maintenance models by assessing faithfulness and stability to improve trust in sensor-based predictions.
Tested on one dataset.
Authors
Manh Hien Nguyen, Ngoc Thanh Nguyen, Isabella Mendoza Cortes, Tam Khuat, Thanh Pham, Nhat Quang Tran, Ushik Shrestha Khwakhali, Loan Do
Abstract
Deep remaining-useful-life (RUL) models on NASA C-MAPSS are now routine, and so are heatmaps that colour sensors and timesteps. A heatmap that looks mechanical is not the same as an explanation an engineer can act on. We train three standard architectures - a 1D CNN, an LSTM, and a small Transformer encoder - on the official FD001 and FD003 splits with the piecewise RUL cap of 125 cycles and the official PHM08 asymmetric score. We then attach three attribution maps (Integrated Gradients, occlusion, last-layer attention) and evaluate them with the checks the XAI-for-PdM literature still under-reports: deletion/insertion faithfulness, Spearman stability under sensor-scale noise, agreement across training seeds, and cosine consistency inside RUL bins. Prediction error is a prerequisite, not the claim. The headline is which explanation method moves the RUL output when its top cells are removed, and which map survives a 5% input perturbation. Integrated Gradients and occlusion are similarly faithful on the LSTM; Transformer attention is cheap and temporally smooth but weakly faithful. All three maps are almost unchanged under 5% input noise, yet IG/occlusion agree only moderately across two LSTM seeds - stability to sensor jitter is not the same as stability to retraining. A secondary tabular check on the AI4I 2020 failure dataset shows the same deletion pattern for tree importances. We recommend occlusion or IG for any C-MAPSS-style report that will be read by a maintenance engineer, and we treat raw attention weights as a visualisation only.