Artificial Intelligence Enabled Predictive Safety Analytics for Proactive Hazard Identification in Industrial Manufacturing Environments

Abstract

This study quantitatively evaluated the effectiveness of an artificial intelligence–enabled predictive safety analytics system for proactive hazard identification in an industrial manufacturing environment. A quasi-experimental design combining interrupted time-series analysis, difference-in-differences estimation, hierarchical regression, and multilevel modeling was employed to assess whether AI-generated predictive risk scores contributed to measurable improvements in both leading and lagging safety outcomes. The analytic dataset included 2,184 line–shift observations and 312 workforce respondents across intervention and comparison production units. Descriptive results indicated that AI-generated risk scores decreased from a pre-implementation mean of 0.52 (SD = 0.16) to 0.41 (SD = 0.15) post-implementation (t = 8.74, p < .001). Exposure duration in high-risk zones declined from 81.2 to 69.3 minutes per shift (p < .001), and anomaly frequency decreased from 3.91 to 2.96 events per shift (p < .001). Near-miss frequency was reduced from 2.94 to 2.21 per 200,000 exposure hours (p = .002), while recordable incident rates declined from 3.81 to 3.05 (p = .015). Hierarchical regression analysis demonstrated that AI-generated risk scores significantly increased explained variance in leading safety indicators from R² = .055 to R² = .269. Multilevel modeling further showed that exposure duration (β = .29, p < .001) and anomaly frequency (β = .34, p < .001) significantly predicted near-miss outcomes. Interrupted time-series analysis revealed an immediate post-intervention incident reduction of −0.68 incidents (p = .004), with a sustained downward trend thereafter. Reliability testing confirmed strong internal consistency across safety constructs (Cronbach’s alpha range = .81–.89). Collectively, these findings provided statistically robust evidence that AI-enabled predictive safety analytics significantly enhanced proactive hazard detection and contributed to measurable reductions in industrial safety incidents within a complex manufacturing setting.