Optimising Brazing Processes through Learning-Based Visual Servoing and Collaborative Robotics

Fei, Haolin and Kennedy, Andrew and Wang, Ziwei (2024) Optimising Brazing Processes through Learning-Based Visual Servoing and Collaborative Robotics. PhD thesis, Lancaster University.

Text (2023haolinphd)
2023haolinphd.pdf - Published Version
Download (3MB)

Abstract

In manufacturing safety-critical components, brazing stands out for its ability to form strong, cost-efficient joints between dissimilar materials. Despite its importance, the brazing process often falls short in efficiency and precision due to its reliance on manual labour. Simultaneously, full automation, while enhancing certain operational aspects, lacks the adaptability and decision-making prowess inherent to human operators. This thesis addresses these challenges within the brazing process by advocating for a synergistic integration of robotics and artificial intelligence (AI) in a human-robot collaboration (HRC) framework. It uniquely combines human expertise with advanced machine capabilities, aiming to refine brazing operations beyond the reach of solely human or automated endeavours. Central to the thesis is the development of a category-agnostic object localisation strategy. This technique enables robots to recognise and position brazing filler metal (BFM) across a diverse array of joint configurations without prior specific knowledge of the objects. By leveraging AIdriven insights, this approach significantly enhances operational precision and adaptability, illustrating its utility in complex assembly tasks where traditional methods fall short. Building on this foundation, a learning-based visual servoing method is introduced. This innovative approach allows robots to dynamically adjust their actions in real-time based on visual feedback, navigating complicated environments and performing tasks with heightened accuracy. Such capability is crucial for ensuring the consistent placement of BFM under varying conditions, demonstrating a marked improvement in the process’s reliability and efficiency. Finally, an intuitive human-robot collaboration framework is proposed. This model is designed to seamlessly integrate the strengths of both humans and robots, facilitating a partnership that leverages the precision of automation and the judgement of human operators. Through examples such as collaborative adjustment of brazing parameters in response to realtime observations, the framework underscores the importance of human insight in augmenting robotic capabilities. This approach not only advances the brazing process by mitigating the reliance on skilled labour and enhancing safety standards but also lays a foundation for applications beyond brazing, highlighting the transformative potential of integrating human and robotic expertise in industrial processes.