Applying Non-destructive Testing and Machine Learning to Ceramic Tile Quality Control

One of the requirements of the Industry 4.0 is the concern about the stability of its product development. Applying emerging technologies to improve product quality control is a significant step in achieving balance in the production process. The acoustic emission is a highly efficient technique used in industry to test products to detect structural failures. This type of method is well known as non-destructive testing. We assert that quality control of the products can be significantly improved by combining the acoustic emission with machine learning techniques. To achieve our claim, we built a dataset from ceramic tiles acoustic emission, and we model two different machine learning techniques, such as Support Vector Machine, with five distinct kernel functions, and k-Nearest Neighbors, to classify the quality of ceramic tiles. In this work, we use several ceramic tiles as the objects to be tested. To extract the features of the ceramic tiles sound, we have built a prototype to hit the ceramic tile to record the sounds. As the results shown, we obtained a high precision classification of the control quality of the ceramic tile, ranged from 95% to 99.9%, depending on the parameters set and the algorithm chosen.