JAKEY BLUEDean ChuStéphane Dauzère-Pérès2025-05-172025-05-172024-12-09https://www.scopus.com/record/display.uri?eid=2-s2.0-85219171558&origin=recordpagehttps://scholars.lib.ntu.edu.tw/handle/123456789/729395The conventional approach in Total Productive Maintenance (TPM) relies on Overall Equipment Effectiveness (OEE) as a key metric for assessing manufacturing productivity, despite its limitations as a lagging indicator due to reliance on historical data. Predictive Overall Equipment Effectiveness (POEE) addresses this limitation by utilizing Virtual Metrology (VM) to deliver real-time quality insights for process optimization. However, predicting long-term equipment conditions and OEE remains a challenge due to the need for real-time equipment data. To address this, a two-phased Predictive Equipment Health with Hidden Markov Models (PEHMM) is proposed. The offline phase uses historical manufacturing data to build behavioral models that capture equipment dynamics through Hidden Markov Models (HMMs). The online phase leverages these models, integrating production plans and optionally real-time sensor data, to predict short- or long-term equipment health and its impact on production yield. Validation using datasets like NASA IMS bearings shows promising results, but current applications are limited to diagnostics.advance process controlequipment prognosticsfault detection and classificationGaussian mixture modelhidden Markov model[SDGs]SDG8[SDGs]SDG9[SDGs]SDG12conference paper10.1109/ISSM64832.2024.10874870