New Approaches for Automated Intelligent Quality Inspection System Integration of 3D Vision Inspection, Computational Intelligence, Data Mining and RFID Technology

Abstract

Product quality is one of the critical issues for a company to be competitive in the global market. Quality inspection is an essential step in quality control process, to examine the semi or final products for the following production process, prevent the interruption of the manufacturing process and reduce the economic loss. Defected final products will lower the loyalty of customers and competitiveness in the market; while defected semi products will affect the following production process even damage the manufacturing devices, thereby leading to the economic loss for the company. With the requirements of inspection speed and accuracy, as well as of special production conditions, sensor based quality inspection has been well developed, which also enables the automated quality inspection combining with data analysis and artificial intelligence techniques. Automated, smart quality inspection is developed to meet the need for accurate, fast and objective quality inspection. Artificial intelligence or data mining techniques are generally applied to achieve the computer aided decision support or decision making. This thesis presents a framework of automated intelligent quality inspection system, integrating the structured light based 3D vision system, computational intelligence/data mining approaches and RFID technology. The structured light system (SLS) is applied as the data acquisition system in the framework, which obtains the point cloud of the inspected product. The quality inspection system consists of two subsystems. The first one is matching based quality inspection involving the point cloud processing, 3D point cloud template matching and applications of computational intelligence during these processes, such as the 𝑘-d tree searching, particle swarm optimization for the automated matching process; while the other is data mining based quality inspection involving the feature extraction, data mining modelling and classifier evaluation. Typical data mining methods such as decision tree, artificial neural networks (ANN) and support vector machine (SVM) are all studied for the data mining based inspection case. Moreover, as the integration system of the framework, RFID system is introduced in this thesis and the combination with the quality inspection system is also studied. The outcome of the thesis can be applied in production line in manufacturing industry and combining with the WPM or MES system.