Summary
Predictive maintenance uses data to forecast when machines might fail so repairs happen before breakdowns. Usually, features that describe sensor data are chosen once and kept fixed, no matter how far ahead predictions are made. The authors show a new method that learns features based on quantiles — values dividing data distributions — tailored to different forecast time horizons. This approach improves accuracy, especially when predicting failures minutes, hours, or days ahead. They also find that features designed for short-term prediction do not work well for long-term forecasts unless adjusted, suggesting feature extraction should adapt to the prediction timeline.
predictive maintenancefeature extractionquantilesmachine learningmulti-horizon forecastingMLP-QRNNsensor dataF1-scoretemporal classificationdistribution-aware features
Authors
David J Poland, Daniele Ravi, Na Helian
Abstract
In data-driven predictive maintenance (PdM), feature extraction is usually treated as fixed preprocessing: a descriptor set is chosen once and reused while the downstream model or forecasting horizon changes. This paper isolates the representation-learning stage and presents a quantile-led feature-extraction framework based on a dual-stage MLP-QRNN hierarchy. QRNN1 learns a broad ten-quantile conditional distribution for each sensor channel, while skip-connected QRNN2 refines a retained mid-tail quantile set into compact, channel-resolved, distribution-aware features. A fixed thirteen-pipeline ablation spans 1-hour, 70-hour, and 30-day regimes across 72 machines in 9 industrial facilities, with the downstream temporal classifier held fixed within each regime. Increasing the retained mid-tail set from two to four quantiles improves 30- and 60-minute F1-score, reaching 75.92% and 72.44% with attention enabled. The results also show that representations do not transfer reliably beyond their design horizon unless feature capacity, temporal embedding, activation strategy, and sensor breadth are scaled with the forecasting task. The unmodified short-horizon extractor falls to 42.90% F1 at 70 hours, whereas horizon-conditioned extractors reach 60.38% at 70 hours and 79.97% at 30 days. The framework therefore supports treating PdM feature extraction as a horizon-dependent representational stage rather than fixed preprocessing.