A simulation-optimization based heuristic for the online assignment of multi-skilled workers subjected to fatigue in manufacturing systems

- We aim to dynamically assign workers to machines so as to reduce the mean flowtime.
- The proposed approach takes the impact of fatigue into consideration.
- A bi-criteria decision making problem is stated and addressed using TOPSIS.
- Simulation-optimization is used offline to adapt the heuristic.
- Experimental results show the efficiency of our approach.

Manufacturing systems are often characterized by a stochastic and uncertain behavior in which frequent changes and unpredictable events may occur over time. Moreover, the customers' demands can sometime evolve drastically along time. In order to cope with such changes in the manufacturing system state, and to optimize given performance criteria, the assignment of multi-skilled workers to the machines in the system can be decided online, in a dynamic manner, whenever workers become available and need to be assigned. Indeed, the starting and completion times of jobs in such systems cannot be predicted, so that static optimization approaches turn out not to be relevant. Several studies, in the ergonomics literature, have outlined that the operators' performances often decline because of their fatigue in work. In particular, in manufacturing contexts, fatigue can increase the processing times of jobs. Several online heuristic have been published, but to the best of our knowledge, they do not cope with this consequence of fatigue. We propose to solve this dynamic multi-skilled workers assignment problem using a new methodology, which aims to provide an adaptable dynamic assignment heuristic, which is used online. Our approach takes the impact of fatigue into consideration, in order to minimize the mean flowtime of jobs in the system. We suggest computing more realistic task durations, in accordance with the worker's fatigue. The heuristic uses a multi-criteria analysis, in order to find a compromise that favors short processing times and avoids congestions. The Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) method is used to select the machine where to assign the worker. Since in our case no expertise is available, an offline adaptation process, based on simulation optimization, is used to identify the weights needed by TOPSIS, so as to better fit with the system specificities. A Job-Shop system is simulated to illustrate the proposed approach. The performance of the suggested heuristic is assessed and compared to two other workers assignment rules, which are widely used in the scientific literature because of their efficiency on the mean flowtime: SPT and LNQ. The comparisons are made under different conditions (staffing level, operators' flexibility). A sensitivity analysis is also performed to analyze the impact of the way how fatigue affects the task duration. Our experimental results show that our heuristic provides better results in every case studied. Several important research directions are finally pointed out.