Planning and flexible work assignment to operators in aeronautic assembly lines: a systemic approach for addressing ergonomic and economic risks – PER4MANCE

The contribution will focus on the development of new mathematical models and optimization methods in order to address efficiently hard problems of combinatorial optimization under high uncertainty conditions. The project will make use of methods from industrial engineering, operational research and artificial intelligence, but also from physical and cognitive ergonomics.

To improve the robustness of solutions, we have proposed a model for representing solutions for “compiled” scheduling problems in the form of a decision tree. Uncertainties on the data are represented by intervals of uncertainty, under the assumption that the decision maker has the possibility, at certain times during the execution of the scheduling, to ask questions about the uncertain data. This will provide the additional information helping to reduce the uncertainty. We have developed an algorithm to build a robust decision tree, whose nodes correspond to the questions asked by the decision maker. Each branch of this node corresponds to a possible answer and offers a new robust solution (in view of the information obtained) compatible with the scheduling carried out up to this node. Such a tree allows the decision maker to adapt the current solution online, while retaining certain robustness properties. Finally, we carried out experiments to measure the quality of the solutions provided by the tree according to the number of questions dealt with. The results show that we obtain the solutions which dominate in the sense of Pareto the solutions found by the reactive approach.

We also modeled the rescheduling problem. The proposed model takes into account the constraints of precedence and resources, the jobs of the operators, their skills and the constraints of the work areas. We have studied different optimization criteria that can be used in this context. We have developed a constraint programming model for the problem under consideration by implementing several proposed multi-objective functions. The performance of these models was tested and the impact of each criterion on the characteristics of the solutions found during the resolution was analyzed.

Experiments on the effectiveness of modeling physical strain

Experiments for the analysis of managers' stress while taking decsions under uncertainty

Validation of the general approach

Valorisation of results.

1. T. Portoleau, C. Artigues, R. Guillaume, Decision trees for robust scheduling, to appear in 17th International Workshop on Project Management and Scheduling (PMS 2020), Toulouse, France, 2020
2. T. Portoleau, C. Artigues, R. Guillaume. Robust Predictive-Reactive Scheduling: An Information-Based Decision Tree Model, International Conference on Information Processing and Management of Uncertainty in Knowledge-Based Systems, 479-492, 2020.
3. N. Berti, C. Artigues, O. Battaïa, R. Guillaume, D. Battini. Heuristic approaches for scheduling manufacturing tasks while taking into account accumulated human fatigue, IFAC-PapersOnLine 52 (13), 963-968, 2019.
4. T. Portoleau, C. Artigues, R. Guillaume. Arbres de décision robustes pour l'ordonnancement proactif/reactif sous incertitude, ROADEF 2020.
5. T. Portoleau, C. Artigues, R. Guillaume, H. Fargier, Contingency scheduling pour le RCPSP robuste : travaux préliminaires, dans Atelier Multi-Agents, Flexible, Temporel, Epistémique et Contingent (MAFTEC 6.5) des 14èmes Journées Francophones sur la Planification, la Décision et l'Apprentissage pour la conduite de systèmes (JFPDA 2019), Toulouse, France, 2019
6. T. Borreguero Sanchidrian, T. Portoleau, C. Artigues, A. Garcia Sanchez, M.Ortega Mier, P. Lopez. Multimode Time-Constrained Scheduling Problems with Generalized Temporal Constraints and Labor Skills, Working paper.
7. O. Battaïa, L. Sanmartin, C. Pralet. Dealing with disruptions in low-volume manufacturing: a constraint programming approach, Procedia CIRP 81, 1372-1375, 2019.

The project considers the case of the final assembly lines in aeronautical industry. Such lines are paced i.e. each aircraft must visit all workstations and change its position in a synchronized way between all workstations. Due to technical, economic and financial constraints, usually such lines are characterized by a relatively small number of workstations (<10) and a relatively large number of tasks assigned to each workstation (between 200 and 3500 tasks). Because of this large number of tasks per workstation, the workstation schedule is sensible to frequent uncertain events: delays in the supply of necessary parts, absence of a qualified operator, quality problems, etc. As a consequence, the workstation schedule is constantly challenged and should be frequently corrected in order to respect the takt time constraint. However, the corresponding planning is a hard combinatorial optimization problem. In practice, the optimization of the initial planning may take several working days. As a result, corrective actions are often carried out without global vision and their quality strongly depends on the experience of the manager.
In order to improve the current situation, project "PER4MANCE" aims to develop a novel systemic approach based on the idea to combine robustness, flexibility and resilience in the optimization models. Taking into account the available data, the uncertain events can be classified as frequent, less frequent but relatively probable and rare. The systemic approach will be based on different solution methods dedicated to manage the risks associated with these three categories. For the frequent risks, we propose to explore robust optimization models in order to guarantee the stability of the initial planning and minimize the impacts of the uncertain events. For the less frequent risks, flexible planning models will be studied based on intelligent storage of alternative solutions pre-calculated in advance for possible uncertain events. To deal with rare uncertain events, fast and efficient methods for re-planning will be developed in order to bring the system back to a stable state as quickly as possible.
In order to improve the working conditions for the operators of assembly lines, the project will also consider the problem of the physical load evaluation for assembly tasks in aeronautic context. Since the existing methods are not adapted for dynamic and flexible work reassignment, a new evaluation method will be developed on the basis of energetical consumption. The obtained information about the tasks strain will be used at all levels of task planning for a more equitable work assignment and for prevention of musculoskeletal disorders.
The expected positive impact of this novel systemic approach includes: improvement of the quality of the resource planning and the efficiency of the resource utilization, reduction of the delivery delays and improvement of social conditions for operators and managers by reducing working stress. From the scientific point of view, the contribution will focus on the development of new mathematical models and optimization methods in order to address efficiently hard problems of combinatorial optimization under high uncertainty conditions.