A Data-Driven Artificial Intelligence Framework for Predictive Maintenance in Smart Manufacturing

Authors

  • Yuchen Zhao School of Mechanical Engineering,Shanghai Jiao Tong University
  • Shuai Chen School of Control Science and Engineering,Shandong University
  • Robert L. Mitchell Department of Mechanical and Aerospace Engineering,University of California, Irvine
  • Daniel P. Foster Department of Electrical and Computer Engineering,Purdue University

DOI:

https://doi.org/10.9999/ijair.v1i1.1

Keywords:

Predictive Maintenance Artificial Intelligence Smart Manufacturing Data-Driven Modeling Industrial AI

Abstract

Predictive maintenance  (PdM) is a cornerstone capability for smart manufacturing, where maintenance actions are increasingly triggered by data and optimized against operational con- straints rather than by fixed schedules.  Despite rapid progress in sensing, industrial connectivity, and learning algorithms, practitioners still face persistent gaps:  (i) heterogeneous and imperfect data streams; (ii) distribution shift across assets, sites, and operating regimes; (iii) uncertainty in remaining useful life (RUL) estimation and failure risk forecasting; and (iv) the translation from  model  outputs  to  actionable  maintenance  decisions  under  cost,  safety,  and  availability requirements.

This paper proposes a data-driven artificial intelligence framework that unifies (1) an indus- trial data layer for acquisition, synchronization, and feature governance;  (2) a modeling layer that  supports  both  sequence-to-RUL  regression  and  time-to-event  (survival)  prediction  with calibrated  uncertainty;  and  (3)  a  decision  layer  that  maps  forecasts  to  maintenance  policies through cost-aware optimization.  We provide core mathematical formulations, pseudocode for the training and deployment pipeline, and implementation-ready design choices for edge–cloud execution in modern smart factories.  The manuscript is written to be reusable as an engineer- ing reference:  assumptions are stated explicitly,  interfaces between modules are defined,  and evaluation protocols are aligned with common public benchmarks and industrial constraints.

Downloads

Published

2026-01-30 — Updated on 2026-01-30

Versions

How to Cite

Zhao, Y., Chen, S., Mitchell, R. L., & Foster, D. P. (2026). A Data-Driven Artificial Intelligence Framework for Predictive Maintenance in Smart Manufacturing. International Journal of Artificial Intelligence Research, 1(1). https://doi.org/10.9999/ijair.v1i1.1