Integrated control of workshop management and product quality: application to the ACTA furniture company.

Summary This CIFRE thesis is part of a collaboration between Acta-Mobilier, a manufacturer of high-end lacquered facades, and the Centre de Recherche en Automatique Nancy. The idea is to take advantage of the concept of Product Controlled System in an industrial environment disturbed by numerous production loops and by a non-negligible rate of rework (non-quality) generating losses of parts, non-respect of deadlines, unstable workloads, etc... the impossible link between the product and an infotronic identifier making traceability more difficult. Work on scheduling and its optimization is hampered by these disturbances in the production chain, which make schedules untenable. Priority treatment of defective parts ensures a service rate that remains remarkable in relation to the percentage of parts to be repaired. But this also leads to losses of parts that prevent the complete delivery of the order. The scientific problem is based on the management of flows in a context of production disrupted by rework and the control of quality by evaluating its impact on the bottleneck. The issue of quality control has been addressed using neural networks capable of predicting the appearance of the defect to which they are dedicated according to production and environmental parameters. This anticipation allows to propose an alternative program to use or to postpone the task planning. The adaptation of the prediction model to the drifts of the physical model to the behavior considered as nervous is carried out "on line" with the help of control charts that allow to detect the drift and its starting date. In spite of this simplification of the flows, the piloting remains complex because of the normal loops of production and the residual non qualities. There are different states of saturation of the system for which the most adapted control rule is not always the same. This analysis is presented in the form of a two-dimensional map, each axis of which presents a key indicator of the non-quality rate and/or flow disturbance. Even if, contrary to algorithms, the best adapted control rule will not always be highlighted, this cartography presents other advantages such as the simplification of the control, the possibility for all users to have the important information on the state of the workshop at a glance, or the need for homogenization on the whole production unit. In this context, the intelligent container offers interesting perspectives with the will to trace a group of products having the same manufacturing range rather than products one by one, to share information such as its delivery date, its degree of urgency, to know which paths they have to take in the workshop and which are the possible alternatives or to communicate with the machines and the other systems of which the one of forecast of the quality and to retain information in the course of the manufacture of the products. The proposed system is therefore interactive and the container is at the heart of the decision. It signals its presence to the scheduling system only if the quality conditions are met, thus simplifying its work and allowing a simple traditional linear programming algorithm to perform this task, which is particularly complicated at first glance. On the other hand, it is the scheduler's responsibility to ensure the steering rule to be used and to request the corresponding information from the available batches. The contribution of this thesis is therefore a methodology for simplifying complex problems by dividing the tasks between different actor subsystems applied to the case of a company manufacturing high-end lacquered kitchen fronts.