Towards an Autonomous Robot-based Laser Cladding Repair Process: A Framework for Damage Detection, Localization and Path Planning

Abstract

With piling scientific evidence and growing public concerns about climate change and depletion of natural resources, policymakers are being forced to implement stringent environmental regulations. One such sector under scrutiny for the concerning pace at which it is consuming natural resources is the manufacturing sector. Repair and remanufacturing are deemed sustainable approaches due to their capability of restoring value in a damaged component and bringing it to like-new condition. However, in contrast to a manufacturing process benefiting from an automated environment, the automation level for repair and remanufacturing processes remains low. Moreover, the traditional repair process is tedious and time-consuming. This is mainly due to the stochastic return of used parts, making this process difficult to automate. With the aim of moving the repair industry towards autonomy, this study proposes a novel repair framework. The developed methodology presents a vision-based Robotic Laser Cladding Repair Cell (RLCRC) that has three features: (a) an intelligent inspection system that uses a deep learning model to automatically detect the damaged region in an image; (b) employing computer vision-based calibration techniques for converting damaged region in pixels to spatial coordinates and extracting the damaged volume; (c) generating a tool-path for depositing material to repair the worn component. In this research, the repair of fixed bends is selected as the case study. Fixed bends are cylindrical components used in directional drilling and are present in copious amounts in the oil and gas sector. The proposed RLCRC employs visual sensors (camera and time-of-flight sensor) to provide automatic and time-efficient damage detection and localization. At first, the performance of different deep learning models utilizing varying datasets is compared to obtain a model best suited for being implemented in the RLCRC. Captured images are analyzed by the selected model for the presence of damage. If damage is found, the model classifies and encloses the region of interest in a bounding box. Then, an algorithm is developed that leverages the pinhole camera calibration technique to localize the damage location spatially. By sending this location to a Time-of-Flight (ToF) sensor, the three-dimensional point cloud data containing the damage volume is acquired. Finally, a simplified tool-path generation method is explored that leverages off the polar coordinate system for depositing material in the damaged cavity to repair the component. Supported by case studies, the results obtained herein validate the efficacy of the proposed framework. Thereby enabling automatic damage detection and damaged volume extraction for worn fixed bends. Following the suggested framework, a time reduction of more than 63% is reported.