Spot welding quality check using artificial intelligence / Jagadisa Rajarathnam

Spot welding is widely used in the automotive industry as the preferred method to weld the body parts. However, a bad quality spot weld can cause problems to the production line such as downtime and monetary losses. It can even cause fatal accidents if the defect body reaches the customer. The current methods of evaluating the quality of a spot weld is either too slow or it is not sufficiently accurate. These methods include, isolating a body to perform destructive test or by using ultrasonic test. This thesis studies on determining the spot weld quality based on a captured image of the spot point. In order to accurately determine the quality, a complex image processing algorithm is applied. The module is fed with pictures of spot welding with good quality and spot welding with bad quality. It then learns the attributes of these images and builds a database with corresponding figures for the two categories. These figures are then used to determine the quality of spot welds. To further enhance the system, artificial intelligence is introduced to it. The system uses the images to build its database and as the size of the database increases, it becomes more accurate. The model has an accuracy of 85% for positive image detection and 75% for negative image detection. It has a marking accuracy of 100% for detecting the spot welding regardless of positive or negative. Keywords: spot welding, image processing. artificial intelligence, neural network, k-nearest neighbo