ARTIFICIAL INTELLIGENCE BASED QUALITY ASSURANCE OF SURFACE FINISH OF PARTS IN ASSEMBLY LINE

Abstract

Surface finish of machine parts is of considerable importance as it is included in the design and manufacturing of products to meet customers’ satisfaction. Above all it may be a major requirement for lapping, close and tight contact to prevent leakages or to minimize stress concentration. In light of the above a system was designed and the algorithm written in python programming language to determine the surface finish of parts in an assembly line. To realize this a dataset of 1800 images of metal surfaces was obtained. The defects were classified into six possible groups: patches, scratches, pitted, rolled and inclusion with 300 images of metal surfaces for each defect.. 1620 images were used in training set and 180 images in the test set. The images were trained and augmented so as to extract the textural features from the images. The images yielded to training. Training accuracy and error were obtained which validates the performance of the dataset during training. The model was also validated by testing it with a different dataset, and its performance established. The accuracy of the system was obtained by dividing the number of correct predictions by the total number of predictions made. The accuracy of the system was 98%, showing the efficiency of the system. It was also established that Artificial Intelligence based method of surface assurance, was more efficient than the contact process – which involve use of profilometer. The accuracy shows the level of conformance of the surface finish test of the parts during assembly line with the laid down specifications during product design. This system is recommended for commercialization and application in industries.