In-process sensing of weld penetration depth using non-contact laser ultrasound system

Welding is one of the main methods used to join structural members. Of the many types of welding processes, Gas Metal Arc Welding (GMAW) is one of the most widely used. One of the largest challenges involved in production of welds is ensuring the quality of the weld. Automatic control of the welding process requires non-contact, non-destructive sensors that can operate in the presence of high temperatures and electrical noise found in the welding environment. Laser generation and Electromagnetic acoustic transducer (EMAT) reception of ultrasound were found to satisfy these conditions. Previous research worked towards development of ultrasonic time of flight based weld penetration depth measurement techniques. One such technique, the Rayleigh Generation Longitudinal to Shear Time of Flight (RGLS ToF) technique was developed recently.;The objective of this research was to compensate for the changes in time of flight exhibited when the RGLS ultrasonic technique is used during welding, thereby reducing penetration depth measurement error. A numerical model based on generalized ray theory was developed to determine the dominant frequency and means through which Rayleigh waves reach the bottom surface of a plate. The model results were validated experimentally. The underlying assumptions made when developing the technique were investigated and shown to be incorrect. In addition, the RGLS wave was not present in received ultrasonic signals.;Since the RGLS wave was not present in the received ultrasonic signals, an alternative wave path was selected. An automated weld inspection system was developed to permit inspection of welds during and after welding. Using the inspection system, experiments were performed to identify the wave paths that reach the EMAT. Of the candidate wave paths, the longitudinal diffracted longitudinal to shear (LdLS) wave was found to be the strongest candidate that is not subject to interference by other waves. A theoretical model was developed to determine the optimal placement of the laser generation location and EMAT. The technique was then validated experimentally and found to perform well.;The inspection system was then used to inspect samples during and after welding. The times of flight of the LdLS wave were measured under both conditions and shown to be larger during welding because of the decreased wave speeds at elevated temperature. The root mean square (RMS) difference in wave speed was shown to decrease as the sensing system was placed further away from the welding torch. The differences in time of flight due to the temperature field during welding were large enough to produce negative penetration depth measurements and errors as large as 5 mm. In order to improve the in-process penetration depth measurement, two neuro-fuzzy penetration depth prediction models were developed. The first model compensates for the temperature induced error by producing an estimate of the room temperature time of flight based on the in-process time of flight and the time history of the wire feed rate. This model can be trained without performing any destructive measurements. The performance of the model was very good. The RMS difference in the estimated time of flight and the post-welding time of flight was reduced by a minimum of 68% and as much as 88%. The estimated time of flight was used to measure the penetration depth. The root mean square error (RMSE) of the penetration depth measurement was comparable to that obtained using the LdLS technique after welding. The second model predicts penetration depth directly by using destructively obtained measurements in the training process. This model performs significantly better than the offline LdLS technique and the time of flight error compensation model. By using the penetration depth prediction models, the accuracy of laser ultrasonic techniques has been drastically increased. The in-process weld penetration depth measurement techniques developed in this research are effective and is suitable for application towards real-time weld quality monitoring and control. Real time weld quality control has the potential to drastically reduce costs, material waste, and human injury and increase throughput of manufacture of welded structures.