Defect Detection Of Heavy Aluminum Ultrasonic Bonding Joint Based On Level-set And Random Forest Algorithm

As an effective bonding technology,ultrasonic wire welding doesn't require additional welding materials,high temperature environment,and no harmful gas.It is an energy-saving and environment-friendly bonding method,and becoming more and more popular in industrial manufacture.As an indispensable part of industrial automation,the continuous reform and innovation of industrial automation have put forward higher requirements for industrial bonding technology: high precision,strong robustness and good environmental adaptability.However,the traditional ultrasonic bonding technology needs manual adjust bonding force and bonding energy due to the wear of the splitter in the bonding process.Machine vision feedback has become an important research direction in the field of industrial automation because of its high precision and real-time performance.Using visual feedback to monitor the bonding quality of ultrasonic bonding process in real time provides a new idea for realizing the automation and high efficiency of bonding equipment.Aiming at the problem of improper bonding parameters caused by wear of splitter in bonding process,the industrial camera was adopted to take pictures of bonding joints which are classified to determine the type of defects,so as to adjust the control parameters.The algorithm based on the gray variance of pixel neighborhood is designed to locate bonding joint region efficiently.The redundant non-solder joint region from the localized solder joint image is removed by improved gray projection algorithm.On this basis,the defect area of solder joint is segmented by Level-Set algorithm to determine the solder joint defect area,and the linearly separable features is extracted by KPCA.Finally,random forest algorithm is used to classify the features to determine the defect category,which can be used as a reference for bonding parameter adjustment.Servo motor,ultrasonic transducer,ultrasonic driver,industrial camera,force sensor are used as experimental equipment to construct a complete ultrasonic bonding equipment.Based on the bonding equipment,the control software of the whole system is layout by Qt,which can realize the integrated control of motion subsystem,vision subsystem,ultrasonic subsystem and force control subsystem.Finally,the bonding joint location,redundancy removal,defect segmentation,feature extraction,defect classification algorithm mentioned in this paper are experimented step by step.The classification accuracy of the improved algorithm reached 91% through the complete experiments,which met the actual processing requirements.