A Conceptual Design Framework for Integrating Smart PPE Monitoring with Proposed Predictive Analytics in Smart Manufacturing Safety Management

Manufacturing safety is being digitalized through wearable sensors, connected-worker platforms, and artificial intelligence (AI), but the evidence does not yet justify treating a conceptual architecture as a validated predictive safety system.This study therefore proposes a conceptual design framework for integrating smart personal protective equipment (PPE) monitoring with future predictive analytics in smart manufacturing.We conducted a structured integrative synthesis of peerreviewed literature published from 2018 through September 2026 in Scopus, IEEE Xplore, and ScienceDirect, supplemented by targeted searches of occupational safety agencies, standards bodies, and primary manufacturer documentation.Search concepts combined smart PPE or workplace wearables with occupational safety, manufacturing, ergonomics, human-robot collaboration, IoT, anomaly detection, and machine learning.Sources were assembled by theoretical sampling from anchor reviews and their citation networks, and selection stopped at conceptual saturation of the framework's design space rather than at exhaustive coverage of the field.We coded eligible sources by industrial sector, hazard, sensing modality, connectivity, analytical method, validation maturity, and governance coverage against a published codebook.We organized recurrent capture-transmit-analyzeact functions into four technical layers.We linked them to a cross-cutting governance and assurance plane covering accountability, worker participation, purpose limitation, data access and retention, human decision authority, and algorithmic oversight.The resulting conceptual outputs distinguish calibrated rule-based alerts, statistical anomaly detection, supervised hazard classification, and prognostic modeling; specify candidate inputs, ground-truth labels, and validation metrics; and define controls for sensor drift, missing data, latency, false alarms, model drift, uncertainty, and alert fatigue.Numeric operational thresholds are intentionally excluded because no sensor system, dataset, simulation, or field trial was used.The contribution is therefore a traceable design framework and staged research agenda, not evidence of predictive performance or safety improvement.Future work must calibrate sensors, prospectively validate models across worker subgroups, evaluate human factors and privacy, and demonstrate that the system remains advisory relative to independently engineered safety controls.