Manufacturing is an obvious domain for the adoption of advanced AI methods. Already, convolutional neural networks are used for image classification to improve quality control and reinforcement learning is used to increase efficiency on production lines. Industrial robots are natural recipients of AI-enabled upgrades. Improving already highly mechanized and automated processes that result in mass-produced objects does not typically present any ethical questions. But when the processes being replaced are manual craft processes which result in artisanal objects, the challenges are not only technical.
Craft processes typically involve highly dexterous and flexible work, that relies on a limited set of tools but on a wealth of human intuition and experience. Both the manual workers and consumers place great value on the human involvement in the process and on the perceived authenticity of the product.
This project will explore the technical challenge of developing an AI approach to flexible process automation, as well as the questions relating to replacing the human factor in these processes. In particular, the focus will be on flexible sheet metal forming processes such as spinning and the English Wheel. A responsible automations system should consistently deliver within the target engineering specifications and be interpretable by production engineers. These processes are ubiquitous in workshops and are used for a vast range of products, from musical instruments to car panels. Craft workers typically train for years before mastering these processes, and the resulting products command a premium over mass-produced items. At the same time, the flexibility of these processes means that no dedicated tooling needs to be created which reduces lead time, costs and environmental impact.
The project will employ prototype mechanized processes available within the Advanced Design & Manufacturing Centre to implement and assess AI toolpath generation approaches. Candidate methods for flexible process control include neural networks and Bayesian optimisation. The flexibility of the processes means several degrees of freedom must be controlled in parallel, typically by providing continuous tool trajectories. The output of the process is a deformed part whose geometry must match a desired outcome. Therefore, a dataset of parametric trials which combines input toolpaths and output geometries can inform an AI approach. Suitable optical sensing instrumentation will have to be incorporated into the existing hardware to facilitate data collection and real-time process monitoring.
In addition to addressing these technical challenges, the project will seek to engage with expert craft users of the manual processes to capture and to incorporate their knowledge. Beyond the technical control questions, it will aim to understand to what extent AI can substitute the human factor and will engage with craft experts to answer the following: How can craft experts be motivated to impart their knowledge to AI systems? When are craft products superior to their machine-made counterparts? How does quality assessment and assurance compare between craft processes and AI-powered industrial systems? How does the choice of AI approach and data sources facilitate or inhibit these objectives?
This project is associated with the UKRI Centre for Doctoral Training (CDT) in Accountable, Responsible and Transparent AI (ART-AI). We value people from different life experiences with a passion for research. The CDT's mission is to graduate diverse specialists with perspectives who can go out in the world and make a difference.
Applicants should hold, or expect to receive, a master's degree or first or upper-second bachelor's degree in a relevant subject.
Informal enquiries about the project should be directed to Dr Loukaides.
Formal applications should be accompanied by a research proposal and made via the University of Bath’s online application form. Enquiries about the application process should be sent to [Email Address Removed].
Start date: 2 October 2023.
Continue with Facebook
