Robotics and Computer Vision for Automated Inspection of Substation and Treatment-Facility Electrical Infrastructure

Abstract

This study addresses a persistent problem in safety critical electrical infrastructure inspection: robotics and computer vision pilots often show promising fault detection, but many organizations fail to operationalize these systems into compliant, auditable, and CMMS linked inspection workflows, limiting scalable adoption in substations and treatment facility electrical environments. The purpose of the study was to quantify, using real enterprise cases, how field capability and organizational enabling conditions influence perceived inspection effectiveness and adoption intention for robotics plus computer vision inspection supported by cloud and enterprise integration. A quantitative, cross sectional, case-based design was applied across two enterprise contexts, with N = 214 valid survey responses (substation cases n = 112, 52.3%; treatment facility cases n = 102, 47.7%) from maintenance engineers, inspection technicians, safety officers, and supervisors. Key variables included Robot Mobility and Coverage (MC), Vision Reliability and Evidence Quality (VR), Environmental Robustness (ER), Integration Readiness (IR), Safety and Procedure Compatibility (SP), Inspection Effectiveness (IE), Safety Improvement Perception (SIP), and Adoption Intention (AI); construct reliability was strong (Cronbach’s alpha = 0.82 to 0.90). The analysis plan used descriptive statistics, internal consistency testing, Pearson correlation, and multiple regression models, with IE predicted by MC, VR, and ER, and AI predicted by IR, SP, and IE. Headline findings showed SIP was highest (M = 4.22, SD = 0.58) and VR was also high (M = 4.10, SD = 0.63), while IR was lowest (M = 3.61, SD = 0.71), indicating that workflow and systems integration remain the main adoption bottleneck. Correlation results indicated IE was most strongly associated with VR (r = 0.61, p < .001), and AI was strongly related to IR (r = 0.56, p < .001) and IE (r = 0.58, p < .001). Regression results showed VR was the strongest predictor of IE (β = 0.39, p < .001), with MC (β = 0.21, p = .002) and ER (β = 0.18, p = .006) jointly explaining R² = 0.49. Adoption intention was driven primarily by IR (β = 0.31, p < .001), followed by IE (β = 0.28, p < .001) and SP (β = 0.19, p = .004), explaining R² = 0.54. Case level implications indicate that organizations should treat robotics plus computer vision inspection as an enterprise pipeline rather than a device purchase: prioritize high criticality assets where weighted IE was highest (4.31 in substations; 4.12 in treatment facilities), and accelerate adoption by improving CMMS integration readiness (M = 3.34) and formalizing procedural permissions for routine robot routes (M = 3.29) alongside evidence traceability and governance.