An Image Processing and Machine Learning Based Approach to Parametric Optimization and Defect Detection in Laser Additive Manufacturing

Abstract

In the era of Industry 4.0, economic competitiveness is ensured by incorporating new newlineand upcoming information technologies into the latest manufacturing techniques. newlineIndustry 4.0 consists of various key technologies such as the industrial internet of newlinethings, cloud computing, big data, autonomous robots, additive manufacturing or 3D newlineprinting, etc. Additive manufacturing or 3D printing is considered a game-changing newlinetechnology for the manufacturing world. Because it has the ability to manufacture newlinecomplex geometries, add delicate features to an existing product, remanufacture or newlinerepair or refurbish a defective and worn but usable product, and mass customization newlineof a product. However, the industrialization of additive manufacturing or 3D printing newlineis challenging due to several issues such as repeatability, defects, measurement of newlinefeatures, and surface roughness. To solve these issues researchers are trying to propose newlinevarious methodologies, for example, to solve the issue of defects and features newlinemeasurement researchers are proposing the usage of image processing and machine newlinelearning. newlineDeposition geometry features measurement and detecting defects in the additively newlinemanufactured components is the most challenging. In the past, researchers have used newlineimage processing and machine learning approaches to address these challenges. The newlineimage processing approach can be used to calculate deposition geometry parameters newlinesuch as width, height, and area. It can be used to quantify the total and effective newlinedeposition geometry features in additively manufactured components. Simultaneously, newlinein past, researchers have used the image processing approach to detect defects in the newlineadditively manufactured components. However, it has been observed that the image newlineprocessing approach had certain limitations such as categorizing the defects, prediction newlinedepending on the quality of images, etc. It has been found in past literature that the newlinemachine learning approach is most beneficial and highly accurate in detecting defects