Executive Summary : | Development of relatively capital inexpensive, reliable, fast, reproducible, and accurate non-destructive technique has tremendous potential for quality assessment of welded components. In theproposed investigation, an intelligent defect detection protocol would be devised to assess the quality of welded assembly by amalgamating artificial neural network (ANN) with infrared thermography (IRT). The activity will involve the application of thermal imaging to portray various weld defects, which will be subsequently classified using ANN technique. The planned activity will offer a robust package, enabling the real-time identification of weld defects. Thus, the method has widespread application in quality judgment of critical welding components and may reduce dependency on a skilled workforce. The study will comprise of four components,(i) weld defect detection using lock-in thermography (LIT) i.e. one of the commonly used, active IRT techniques, (ii) Development of ANN based trained system for defect identification, (iii) Defect classification on the basis of geometry-location-nature, and (iv) Providing input for optimization of welding parameters to obtain a sound assembly with desired properties. The sensitivity / resolution of the technique would be initially established through detection of known defects or flaws in reference samples. Subsequently, this information would be used to identify unknown defects within the welds. Further, the feasibility of laser thermography (LT)to detect micrometric defects in weldments would be investigated. The basic approach in the proposed setup would be to derive a set of quality indicators for different types of weld defects. These quality indicators would be useful in locating and identifying defects during in-situ weld inspection and would aid in optimizing the welding parameters to obtain a joint of adequate efficiency. |