DESIGN AND DEVELOPMENT OF A MOBILE ROBOTIC PLATFORM WITH FABRICATION OF A 4-DOF MANIPULATOR FOR CONDITION MONITORING AND WELD SEAM INSPECTION IN REACTOR PRESSURE VESSEL HEADS

Abstract

Manual weld inspection in hazardous and confined industrial environments is time-consuming, inconsistent, and exposes human inspectors to significant safety risks. This study presents the design and development of a robotic weld inspection system integrated with a multi-sensor condition monitoring framework for automated detection of weld defects in complex confined structures such as Reactor Pressure Vessel (RPV) heads. The proposed system combines a mobile robotic platform with a 4-DOF manipulator, ultrasonic sensing, temperature and pressure monitoring, and real-time data visualization. The core contribution of this work lies in the integration of robotic inspection hardware with data-driven analysis implemented in R. Ultrasonic signals were processed using statistical threshold-based anomaly detection for defect identification, while environmental sensor data were analyzed to validate operational stability. A Shiny-based dashboard was developed to enable real-time multi-sensor visualization and interactive monitoring of inspection data. The physical prototype was also fabricated. Additionally, a confusion-matrix-based evaluation framework was used to assess the performance of the defect detection model. Experimental results demonstrated that the proposed system achieved high classification performance with an accuracy of 0.99, precision of 0.989, recall of 1.000, and an F1-score of 0.995. Correlation analysis between ultrasonic signals and environmental parameters revealed negligible relationships, with correlation coefficients of −0.0115 (temperature) and 0.0773 (pressure), confirming that defect detection is primarily driven by structural variations in weld seams rather than external environmental conditions. The results validate the robustness and reliability of the proposed robotic inspection framework. The integration of robotics with R-based statistical computing provides a scalable, interpretable, and cost-effective solution for industrial weld inspection. The proposed system significantly improves safety, efficiency, and reliability compared to conventional manual inspection methods and demonstrates strong potential for deployment in industrial condition monitoring applications.