Porcelain Bushing Internal Defects Intelligent Detection Based on Transient Thermography

Porcelain Bushing are prone to internal defects and the existing testing methods are difficult to realize on-line detection conveniently. A transient thermography method was proposed for internal defects detection. In this method, the transient thermal response of the sample surface is excited by a constant heat source. The time-region thermograms during the cooling period are collected and used for analyzed. The samples with flat bottom hole defects were detected at a distance of 1m. The temperature curve and defect state of each pixel are extracted and used for support vector machine (SVM) training. The trained model is used for three additional detection conditions: changing distance (1m, 0.5m and 2m). F 1 -scores of these trials are 0.825, 0.794 and 0.688, respectively. The established and optimized model has certain generalization performance for porcelain bushing detection at different distances. Changing detection distance will degrade the performance of the model, especially increasing detection distance.