Thermal signal reconstruction and employment of K clustering method for inspection of additive manufactured polymer parts

Due to the increase in the application of additive manufactured components in the industry, developing fast and accurate methods for defect evaluation of these products has become vitally important. In this study, a PLA sample was inspected by thermography. Several artificial defects varying in size and depth were produced in the specimen. A projector with 2 KW in power was utilized to heat the sample for the 15s. The infrared camera recorded the sample’s temperature during the heating period and 30s after shutting down the source. Afterward, the best frame of raw data was selected. The contrast of defective and sound regions improved with applying the well-known Thermal Signal Reconstruction (TSR) image processing method to enhance the automatic detectability of defects. The contrast enhancement was studied quantitatively via adopting signal to noise ratio (SNR) parameter. According to the acquired results, TSR’s 1st derivative image had the highest average of SNR. This amount was approximately four times higher than that of the best frame of raw data. Ultimately, to identify the defects automatically, k-means clustering was adopted. By comparing the segmented images, it was proved that the adopted process was successful in improving automatic defect detection. While the proportion of detectable defects through segmented image concluded from the best frame of raw data was only 70 percent, the figure for segmented images concluded form 1st and 2nd derivative of TSR was substantially higher at 100 percent.