Weld joint penetration determination is the key factor in welding process control area. Not only has it directly affected the weld joint mechanical properties, like fatigue for example. It also requires much of human intelligence, which either complex modeling or rich of welding experience. Therefore, weld penetration status identification has become the obstacle for intelligent welding system. In this dissertation, an innovative method has been proposed to detect the weld joint penetration status using machine-learning algorithms. A GTAW welding system is firstly built. Project a dot-structured laser pattern onto the weld pool surface during welding process, the reflected laser pattern is captured which contains all the information about the penetration status. An experienced welder is able to determine weld penetration status just based on the reflected laser pattern. However, it is difficult to characterize the images to extract key information that used to determine penetration status. To overcome the challenges in finding right features and accurately processing images to extract key features using conventional machine vision algorithms, we propose using convolutional neural network (CNN) to automatically extract key features and determine penetration status. Data-label pairs are needed to train a CNN. Therefore, an image acquiring system is designed to collect reflected laser pattern and the image of work-piece backside. Data augmentation is performed to enlarge the training data size, which resulting in 270,000 training data, 45,000 validation data and 45,000 test data. A six-layer convolutional neural network (CNN) has been designed and trained using a revised mini-batch gradient descent optimizer. Final test accuracy is 90.7% and using a voting mechanism based on three consequent images further improve the prediction accuracy.

The overall purpose of this research is to develop a real-time ultrasound based system for controlling robotic weld quality by monitoring the weld pool. The concept of real-time weld quality control is quite broad, and this work focuses on weld penetration depth monitoring and control with laser ultrasonics. The weld penetration depth is one of the most important geometric parameters that define the weld quality, hence remains a key control quantity. This research focuses on the implementation and optimization of the laser phased array generation unit and the development of signal analysis algorithms to extract the weld penetration depth information from the received ultrasonic signals. The system developed is based on using the phased array technique to generate ultrasound, and an Electro-Magnetic Acoustic Transducer (EMAT) as a receiver. The generated ultrasound propagates through the weld pool and is picked up by the EMAT. A transient FE model is built to predict the temperature distribution during welding. An analytical model is developed to understand the propagation of ultrasound during real-time welding and the curved rays are numerically traced. The cross-correlation technique has been applied to estimate the Time-of-Flight (ToF) of the ultrasound. The ToF is then correlated to the measured weld penetration depth. The analytical relationship between the ToF and penetration depth, obtained by a ray-tracing algorithm and geometric analysis, matches the experimental results. The real-time weld sensing technique developed is efficient and can readily be deployed for commercial applications. The successful completion of this research will remove the major obstacle to a fully automated robotic welding process. An on-line welding monitoring and control system will facilitate mass production characterized by consistency, high quality, and low costs. Such a system will increase the precision of the welding process, resulting in quality control of the weld beads. Moreover, in-process control will relieve human operators of tedious, repetitive, and hazardous welding tasks, thus reducing welding-related injures.