Integrated planning and scheduling of engineer-to-order projects using a Lamarckian Layered Genetic Algorithm

This paper presents a new mathematical formulation for planning and scheduling activities of Engineer-To -Order (ETO) projects. It includes a new ETO strategy to reduce two principal impacts of the design uncertainty inherent in the ETO context: waste (of time and resources) and schedule instability. Our optimization approach is based on a two-level decision process to address, either sequentially or separately, the initial planning and the rescheduling stages. We also propose a hybrid Layered Genetic Algorithm combined with an adaptive Lamarckian learning process (LLGA). LLGA uses a new genetic representation (encoding format and decoding method) and a new cycle-avoidance procedure that guarantees solutions feasibility. LLGA is compared to the branch-and-cut procedure of CPLEX run on the proposed mathematical model on randomly generated instances with up to 340 operations. Our mathematical model shows a good performance for small and medium-sized instances, especially for the rescheduling stage. This performance deteriorates for larger instances (larger computing times and out-of-memory problems). However, the proposed heuristic is computationally stable and yields good-quality solutions in a reasonable computing time without requiring a large memory space. Our experiments also demonstrate the merits of our new ETO strategy in improving the robustness of the solutions.

Automated summary

This paper presents a new mathematical formulation for planning and scheduling activities of Engineer-To-Order (ETO) projects, along with a new ETO strategy to reduce the impacts of design uncertainty. The study proposes a hybrid Layered Genetic Algorithm combined with an adaptive Lamarckian learning process (LLGA) and compares it with the branch-and-cut procedure of CPLEX. The results show good performance of the proposed mathematical model for small and medium-sized instances.