AI-Assisted NDT Analytics for Early Crack Detection in Energy-Infrastructure Pressure Vessels Using Advanced Manufacturing Data

Abstract

Early detection of cracks in energy-infrastructure pressure vessels remains a critical technical challenge because small, irregular, or noise-affected defects can be difficult to distinguish from benign structural responses, while manufacturing and inspection data are often fragmented across different systems. This study aimed to evaluate how artificial intelligence-assisted non-destructive testing analytics and advanced manufacturing information contribute to Early Crack Detection Performance in pressure-vessel integrity management. A quantitative, cross-sectional, case-study-based research design was employed across energy-infrastructure organizational cases involving pressure-vessel fabrication, welding, NDT inspection, maintenance, reliability assessment, quality assurance, and asset-integrity activities. The sample comprised NDT engineers and technicians, mechanical and integrity engineers, welding and manufacturing engineers, quality and reliability personnel, maintenance professionals, and AI/data analytics specialists. Of 320 questionnaires distributed, 301 were returned and 290 valid responses were retained, producing a usable response rate of 90.6%. The key explanatory variables were AI-Enabled NDT Data Analytics, Advanced Manufacturing Data Integration, Intelligent Defect Pattern Recognition, and Predictive Crack Risk Analytics, while Early Crack Detection Performance represented the dependent variable. Data were analyzed using descriptive statistics, Cronbach's alpha, KMO and Bartlett testing, Pearson correlation, multiple regression, ANOVA, and regression diagnostics. Early Crack Detection Performance achieved the highest mean, M = 4.17, SD = 0.54, followed by AI-Enabled NDT Data Analytics, M = 4.13, SD = 0.56, and Intelligent Defect Pattern Recognition, M = 4.08, SD = 0.58. All predictors were significantly associated with Early Crack Detection Performance, with correlations ranging from r = .64 to r = .72, p < .001. The combined model explained 68.7% of performance variance, R = .829, R² = .687, adjusted R² = .683, F(4, 285) = 156.42, p < .001. AI-Enabled NDT Data Analytics was the strongest predictor, β = .30, followed by Intelligent Defect Pattern Recognition, β = .27, Predictive Crack Risk Analytics, β = .23, and Advanced Manufacturing Data Integration, β = .20, all p < .001. The findings imply that integrating AI-based inspection analytics, manufacturing data, intelligent defect recognition, and predictive risk assessment can substantially strengthen the sensitivity, consistency, reliability, and timeliness of early crack detection in critical pressure-vessel infrastructure.