Digital Twin and AI-Enabled Industrial IOT Framework for Predictive Maintenance and Secure Wireless Monitoring in Southeast Texas Refineries

Abstract

Petroleum refining in Southeast Texas requires equipment diagnostics and industrial communications, yet digital twin integration, inconsistent sensor information, wireless interruptions, and cybersecurity vulnerabilities constrain predictive maintenance and visibility. This study examines the associations of digital capabilities with perceived predictive maintenance effectiveness and secure wireless monitoring in refineries. Grounded in General Systems Theory, it adopts a quantitative, explanatory, cross-sectional, embedded multiple-case design covering ten refinery cases in the Beaumont-Port Arthur and Greater Houston-Galveston Bay areas. The study specifies 312 professional responses from maintenance and reliability engineers, instrumentation and automation specialists, operations personnel, and operational technology cybersecurity and IIoT practitioners, selected through purposive sampling. A structured five-point Likert questionnaire measures digital twin integration maturity, AI-enabled diagnostic and prognostic capability, IIoT sensor-data quality, wireless network reliability, and operational technology cybersecurity control maturity as predictors of two perceived effectiveness outcomes. The analysis incorporates descriptive statistics, Cronbach’s alpha, construct-validity assessments, Pearson correlations, and separate five-predictor multiple regressions to evaluate seven hypotheses at a .05 significance threshold. Reported mean scores are 3.40 for predictive maintenance effectiveness and 3.39 for secure wireless monitoring effectiveness, while internal consistency coefficients range from α = .897 to .924. Correlations between hypothesized predictors and maintenance effectiveness range from r = .394 to .486; those for secure monitoring range from r = .396 to .545. The maintenance model explains 61.8% of outcome variance (R² = .618; adjusted R² = .611), with AI diagnostics showing its strongest hypothesized standardized association (β = .344). The secure monitoring model explains 58.9% (R² = .589; adjusted R² = .582), led by wireless reliability (β = .419) and cybersecurity maturity (β = .408). All seven reported relationships meet p < .001. The framework prioritizes coordinated sensing, analytics, resilient communications, and cybersecurity for refinery technology evaluation, without establishing causal effects or measured reductions in failures, downtime, or attacks.