Risk management for industrial systems and services

The general scientific research problem of this axis is related to decision-making in uncertain contexts. More specifically, we are interested in defining strategies to identify as soon as possible the appearance of defects or process drifts and to define the actions to be implemented to prevent or correct them. The search for efficiency in terms of decision-making shows the limits of traditional approaches. The scientific obstacles to be removed focus on the proposal, on the one hand new technical and economic models for optimising decision-making processes under uncertainty, and on the other hand new strategies for detecting deviations taking into account less traditional and more complex hypotheses. The methodologies developed consist in defining and modelling the evolution of performance indicators as well as estimating the impact of a given decision on their future evolution. The proposed tools are based on the development of elaborate mathematical formalisms that are borrowed from statistics and probabilities combined with stochastic optimization methods.
 

The main areas of application are:

• monitoring of production and service processes,
• modelling and optimization of maintenance policies.

With regard to process monitoring, the focus is on statistical methods to detect, as quickly as possible, drift(s) from a nominal situation considered acceptable. The proposed methods make it possible to take into account various statistics (coefficient of variation, ratio of random variables, compositions) and various situations (continuous / discrete, independent / autocorrelated, univariate / multivariate, normal / non-normal), to take into account parameter estimation or to propose non-parametric approaches and to integrate costs related to actions and decisions if necessary (economic approach).
 

1. Modelling the technical and economic performance of maintenance and maximising the expected payback period of multi-component systems (modelling the degradation process and the imperfect efficiency of maintenance actions, stochastic optimisation),
2. Integrated planning of maintenance and production of goods (in connection with the team's second theme) by developing methods and tools to solve this joint problem (Heuristics based on Lagrangian relaxation, Dantzig-Wolfe decomposition and column generation, genetic algorithms, etc.).

Design, planning and scheduling of production and service systems

Some of the team's projects deal with scheduling and planning issues, which are widely encountered in the literature. The team is committed to carrying out various studies to identify and characterize new problems that emerge in current contexts whose concerns are evolving. A first part of the work, of a purely theoretical nature, can also be included in the team's transversal axis, particularly around complexity approaches: identification of sub-cases, proof of NP-completeness in particular. On the other hand, work is being carried out to remove the scientific obstacles to the solution of problems dedicated by the proposal of solution methods resulting from Operational Research approaches: tree resolution methods, mathematical modelling, column generation models, but also different meta-heuristics such as Limited Discrepancy Search, Large Neighborhood Search, Tabu Search...

The team's various contributions are mainly structured in the field of production line design (reconfigurable, collaborative environments, etc...), context workshop scheduling with parallel machines, batch processing, cross-dock...), project scheduling (taking into account competence, generalized precedence...), production planning (maintenance and joint production, multi-site planning, integration of financial risks, etc.) and human resources planning (appointment management, nurse rotation, etc.) and can find applications in both the goods production sectors (or find examples in metal parts production or the aeronautics industry) and in services (health system, project management for example).

Design and optimization of logistics and transport networks

The SLP team is working on the development of models and algorithms for transport optimization. The tools developed are used to support decision-making at a strategic level, mainly during network design (supply chain, distribution network / transport). A second set of contributions focuses on the optimization of vehicle tours, with applications at the operational level or for the simulation of systems under development. The contributions of the transport optimisation team have applications in passenger transport, freight distribution, transport within the supply chain and in services. The research aims either to improve the performance of algorithms for solving known problems or to solve new problems. The SLP transport team's research can be presented on four major levels :

• Design of logistics networks, supply chain and transport networks :
  • Integration of environmental aspects into supply chain design
  • Design of collaborative distribution network
  • Design of bus networks
• Integration of transport decisions with other organizational decisions :
  • Joint optimization of logistics hubs and vehicle routes locations
  • Robust optimization of production and gas inventory management in conjunction with distribution rounds
  • Integration of production planning and distribution
• Optimization of vehicle tours in the service sector (2) :
  • Technicians' tours
  • Health transport, specialized transport for people with reduced mobility
  • Waste collection

Cross-cutting and fundamental themes

A lot of work is being done between the team's axes. The most important of these are integrated production and maintenance planning, which involves the integration of stochastic phenomena into the production planning process.

In addition, independent or parallel theoretical contributions to the work carried out in the team's three historical areas are made.

A first part focuses on the design of meta-heuristics for solving optimization problems for an objective:

• Contribution to the solution of combinatorial optimization problems with ALNS metaheuristics: a large number of works are carried out with ALNS metaheuristics in vehicle tours and network design. They contribute to a more precise analysis of the key components of the method and to its evolution. In particular, we are studying the hybridization of this method with the PLNE.
• Contribution to the resolution of continuous optimization problems with the PSO metaheuristic.

A second section addresses current obstacles encountered in multi-objective optimization. It aims to propose new knowledge for solving large multi-objective optimization problems, either combinatorial or in mixed numbers. Efficient algorithms are produced to facilitate the treatment of multi-objective NP-difficult problems in order to solve them effectively. In particular, this concerns algorithms based on polyhedral approaches, integrating sharp planes, generalizing bi-objective methods to multi-objective situations and moving towards the implementation of branch-and-bound/cut multi-objective methods. Multi-objective meta-heuristics complete this section, particularly from the perspective of hybridization or mat-heuristics. More specifically, this work includes the following points:

• Study of the links between relaxation and scalarization methods to determine limiting sets in a context of multi-objective optimization.
• Cutting and branch-and-cut method in a context of multi-objective optimization.
• Dynamic methods for branch variable selection and active node selection for branch-and-bound and branch-and-cut multi-objective algorithms.
• Approached algorithms in multi-objective combinatorial optimization and articulation in a three-phase scheme.