TL;DR
This paper introduces a novel approach to optimize factory layouts in semiconductor manufacturing by allowing "shared clearances" between machines. This innovative strategy significantly reduces both the physical footprint of the factory and the cost associated with material handling, achieving savings of 7-10%.
What Problem Does This Paper Address?
In manufacturing, especially in high-tech industries like semiconductor production, efficient factory layout is crucial. Machines require a certain amount of space around them for operation, maintenance, and to prevent interference (minimum clearance). Additionally, space is needed for human access and material staging (additional clearance). Traditionally, these clearances are treated independently, leading to potentially excessive spacing. This paper addresses the inefficiency of this approach by proposing a "double-row layout problem with shared clearances" (EDRLP-SC). The core problem is that if two machines are adjacent, the additional clearance needed for one machine might not be simultaneously required by the other, allowing for this space to be shared. This sharing can lead to substantial savings in both the overall factory area and the distances materials need to travel, thus reducing handling costs.
Methods
The authors tackle the EDRLP-SC by first developing a Mixed Integer Linear Programming (MILP) formulation to precisely define the problem and its objectives: minimizing both material flow cost and layout area. To solve this complex, multi-objective problem, they propose a hybrid approach combining:
- Multi-Objective Tabu Search (MTS): This meta-heuristic is used to efficiently search for optimal sequences of machines within the two rows. It explores different arrangements of machines, aiming to find a set of non-dominated solutions (solutions where no objective can be improved without worsening another).
- Heuristic Rules (HR): These rules are designed to quickly determine how additional clearances can be shared between adjacent machines. They aim to exploit the possibility of sharing to minimize the overall space required.
- Mathematical Programming (MP): For the best machine sequences identified by MTS, MP is used to find the optimal exact machine locations and determine the final additional clearances. This step refines the solutions found by the heuristic search.
This hybrid approach is designed to overcome the computational challenges of finding optimal solutions for this complex layout problem.
Datasets
The paper evaluates its proposed algorithm on problem instances with 10, 20, 30, and 50 machines. These instances are generated with varying percentages of machines requiring additional clearances on both sides (parameter bi). The specific parameters for these instances are available for download as an electronic companion to the paper. No pre-existing benchmark datasets for this specific EDRLP-SC problem are mentioned; the authors appear to have generated their own test cases based on the problem definition.
Key Insights and Ideas
- The Power of Sharing: The most significant insight is that allowing adjacent machines to share additional clearances can lead to substantial reductions in both factory area and material handling costs. This challenges the traditional approach of treating all clearances as independent.
- Exploiting Adjacency: The effectiveness of sharing is directly tied to the adjacency of machines. The more machines that require additional clearances, the greater the potential for savings through sharing.
- Hybrid Approach Effectiveness: The combination of MTS for sequence optimization and HR for clearance sharing, followed by MP for precise location and clearance determination, proves to be a highly effective strategy for solving this complex multi-objective problem.
- MTS-HR Outperforms: The proposed MTS-HR algorithm consistently finds better Pareto-optimal solutions (in terms of quality and distribution) compared to standard multi-objective algorithms like NSGA-II and even an exact solver (CPLEX) for larger problem instances.
Key Results
- Significant Cost and Area Reduction: The study demonstrates that sharing clearances can reduce average material handling costs by 6.58% to 10.83% and average area consumption by 6.02% to 9.63%.
- Improved Solution Quality: The MTS-HR algorithm consistently finds high-quality, non-dominated solutions that effectively cover the Pareto front, offering a good trade-off between cost and area.
- Scalability: The MTS-HR approach shows good performance across various problem sizes, from 10 to 50 machines, outperforming other methods, especially for larger instances where exact methods become intractable.
- Computational Efficiency: While CPLEX struggles with larger instances, MTS-HR provides good solutions within reasonable computational times.
Limitations and Future Work
- Limited to Two Rows: The current formulation and solution approach are limited to a double-row layout. Future work could extend this to more complex multi-row layouts.
- Need for Lower Bounds: The authors suggest that establishing tighter lower bounds for the EDRLP-SC would be beneficial for better quantifying the effectiveness of their heuristic and for developing other solution approaches.
- Constrained Material Flow: The paper mentions that incorporating more constrained material flow scenarios, such as one-way material flow, could be a valuable extension.
- Continuous Location Formulation: For larger production environments, exploring continuous location formulations might be more suitable.
Summary updated