DATA QUALITY AS A RELIABILITY MULTIPLIER: QUANTIFYING THE IMPACT OF STANDARDIZED WORK PROCESSES ON RELIABILITY MODEL ACCURACY

Abstract

In modern industrial systems, predictive reliability models are central to maintenance strategies, yet their accuracy is fundamentally constrained by the quality of input data. While significant research has focused on algorithmic improvements for fault detection and remaining useful life (RUL) estimation, the role of standardized work processes (SWP) as an enabling multiplier for data quality and, subsequently, model accuracy remains underexplored. This paper quantifies the causal relationship between adherence to SWP, resulting data integrity metrics (accuracy, completeness, consistency, timeliness), and the performance of a supervised machine learning-based reliability model. Using a controlled manufacturing case study over six months, we compare model accuracy across two regimes: conventional data logging (Phase 1) versus SWP-integrated data capture (Phase 2). Results demonstrate that implementing SWP improves overall data quality by 41.2%, which translates to a 28.6% increase in reliability model accuracy (from 67.2% to 86.4%) and a 35% reduction in false positive alarms. The proposed framework establishes data quality as a force multiplier for reliability predictions, offering a pragmatic pathway for industries to maximize returns from existing AI investments without additional algorithmic complexity.