<html><head></head><body><div style="font-family: Verdana;font-size: 12.0px;"><div>Acquisition of constraint models in the context of agile manufacturing</div>
<div> </div>
<div>Within the framework of decision support tools for the configuration and monitoring of production workshops, the thesis will focus on the acquisition of models at different levels: from extraction and pattern tracking models on data flows to workshop resource optimisation models. In order to be transposable to different situations, these models will have to be parameterised. In order to converge faster towards realistic models, the acquisition process will sometimes require active learning in connection with a digital twin that will be used as an oracle in the acquisition process. Finally, these models will have to be interpretable in terms of business concepts and, for some, executable in solvers such as those related to constraint programming and/or linear programming.</div>
<div> </div>
<div>The scientific objective of the thesis is to design model acquisition approaches in the context of production workshops at two levels: on the one hand, models for monitoring and analysing production from data flow from the workshop's machines and sensors, and on the other hand, models for generating scheduling models based on data. These models should be interpreted in terms of business (e.g. machines, capacity, setup, settings, stability, cost). In order to obtain models that are more in line with reality, the acquisition process will take into account a targeted interaction with an expert to discriminate certain elements of the model and take into account his preferences, or a more important interaction with a digital twin. In order to be transposable to slight variations in the resources available on the shop floor, some models will have to be parameterised (e.g. by the number of machines, their capacity, setup time or runtime matrices). These models will be tested on at least two concrete cases from industrial cases in the energy or automotive industry, where these models will be used by standard optimisation libraries.</div>
<div> </div>
<div>Context: ASSISTANT H2020 project (Learning and robust decision support systems for agile manufacturing environments)</div>
<div> </div>
<div>Contact: Nicolas Beldiceanu (nicolas.beldiceanu@imt-atlantique.fr)</div></div></body></html>