A linguistic-based decision-support framework for enhancing production ramp-up effort

Production ramp-up is still known to be one of the most resource-intensive stages across the production lifecycle, particularly in terms of cost, time, and knowledge, despite advances in adaptable and reconfigurable manufacturing paradigms. Here, production systems are brought to full production by tuning system parameters to achieve the targeted key performance indicators. However, as the interactions between equipment and process are often not well understood for new production lines, ramp-up is heavily dependent on the experience and knowledge of the human operator. Currently, this knowledge and expertise are often not captured nor analysed for future reuse and, as a result, often lost. This thesis is focused on providing means to enable the capture and analysis of these data towards facilitating a faster and more efficient ramp-up process.

One of the key challenges this research addresses is the better integration of human operator knowledge into the ramp-up process. First, a model has been defined to formalise the ramp-up process and human input in the form of observations, assessments, and adjustments. Second, it has been investigated how these captured data can be further analysed using natural language processing. The results from this analysis are then used for the development of an overall decision-support framework for the ramp-up process for plug-and-produce assembly systems. The novelty of this framework lies in integrating linguistic-based approaches for the analysis of the data as well as the communication of the extracted knowledge. This has been realised through the incorporation of a conversational, personal assistant in the form of a chatbot.

The proposed decision-support framework has been tested and validated using different scenarios to ramp up a robotised workstation. It was shown that the decision-support mechanism can reduce the number of trials required to ramp up the system under consideration by over 28%.