Laser beam welding quality monitoring system based in high-speed (10 kHz) uncooled MWIR imaging sensors

The combination of flexibility, productivity, precision and zero-defect manufacturing in future laser-based equipment are a major challenge that faces this enabling technology. New sensors for online monitoring and real-time control of laserbased processes are necessary for improving products quality and increasing manufacture yields. New approaches to fully automate processes towards zero-defect manufacturing demand smarter heads where lasers, optics, actuators, sensors and electronics will be integrated in a unique compact and affordable device. Many defects arising in laser-based manufacturing processes come from instabilities in the dynamics of the laser process. Temperature and heat dynamics are key parameters to be monitored. Low cost infrared imagers with high-speed of response will constitute the next generation of sensors to be implemented in future monitoring and control systems for laser-based processes, capable to provide simultaneous information about heat dynamics and spatial distribution. This work describes the result of using an innovative low-cost high-speed infrared imager based on the first quantum infrared imager monolithically integrated with Si-CMOS ROIC of the market. The sensor is able to provide low resolution images at frame rates up to 10 KHz in uncooled operation at the same cost as traditional infrared spot detectors. In order to demonstrate the capabilities of the new sensor technology, a low-cost camera was assembled on a standard production laser welding head, allowing to register melting pool images at frame rates of 10 kHz. In addition, a specific software was developed for defect detection and classification. Multiple laser welding processes were recorded with the aim to study the performance of the system and its application to the real-time monitoring of laser welding processes. During the experiments, different types of defects were produced and monitored. The classifier was fed with the experimental images obtained. Self-learning strategies were implemented with very promising results, demonstrating the feasibility of using low-cost high-speed infrared imagers in advancing towards a real-time / in-line zero-defect production systems.