Files
Abstract
Facility layout planning is an intriguing and challenging problem that can be addressed from several perspectives, namely from an operational excellence point of view or in consideration of the well-being of occupants and their relative experiences within the facility. These two stances are known to be conflicting in nature since an improvement for one outlook is likely to have a negative impact on the other. At the same time, it might be necessary to address certain requirements and conditions when designing new or renovating existing facilities that can either have a positive or negative effect from an efficiency and/or human factors standpoint, thus revealing the underlying complexity of the facility layout problem (FLP). These inherent challenges make it difficult to apply exact methods for optimizing the layout of a facility, resulting in practitioners to resort to other techniques instead where an optimal design is failed to be guaranteed. There are three avenues that drive this dissertation research that are influenced by the aforementioned issues, including (1) coping with computational complexity and intractability, (2) consideration of infectious diseases in relevance to facility layout planning, and (3) increasing the applicability of exact methods for layout design practitioners (i.e., architects). For the first research avenue, a special variant of FLP known as the double-row layout problem is considered, where departments are placed along two rows that are separated by a central corridor. By modifying an already existing formulation in the literature and introducing additional symmetry-breaking constraints, it was found that solution times and optimality gaps were reduced for the most occurrences across 100 randomly generated problem instances of varying size (with respect to the number of departments). This is made possible by incorporating the minimum clearance requirements between departments in the pairwise distance constraints rather than the non-overlapping constraints. Doing so results in the number of binary variables to be reduced by 25% to 50%, respectively, compared to the existing model, thus lessening the overall computational complexity. These findings are key for the other topics in this dissertation. The second research avenue is in the context of layout design problems during pandemic-induced circumstances, where social distancing and reduced capacity constraints are enforced to reduce the spread of infection. A restaurant layout problem is the primary focus for this avenue in the dissertation (although it can easily be extended to other facility types) to assist restaurant owners in maximizing the number of seats that can be placed inside of the facility when operating under pandemic conditions. Similar to the first research avenue, social distancing requirements are embedded in the pairwise distance constraints. Note that optimizing the layout solely based on the number of seats in the restaurant (aka "naïve approach") may yield inferior performance in practice for a variety of metrics, such as generated revenue, table utilization, rejection rates, etc. (i.e., poor table utilization, parties not being seated due to an unsatisfactory table arrangement). To circumvent this issue, the naïve model is transformed into a two-stage stochastic program with recourse that incorporates scenarios generated from the probability distribution of variously sized parties to maximize the expected revenue. The assortment and arrangement of seats are determined during the first stage, and the assignment of parties to seats occurs in the second stage. When solving this problem with social distancing and reduced capacity constraints enforced, it was found that the stochastic program produced more revenue for up to 35.71% additional occurrences compared to the naïve approach, as well as improved rejection rates (in 10.5% additional occurrences) and table utilization under a simulated restaurant environment. The final research avenue is intended to assist architects in the layout design process by generating layout alternatives in consideration of adjacency specifications, the flow of occupants within the facility interior, and building code requirements that are set forth by relevant governing bodies, such as the International Building Code. Existing studies only address a subset of the adjacency specifications, while ignoring proximity and separation requirements that can be enforced to specify maximum and minimum tolerable separation distances between relevant pairs of departments, respectively. In addition, the placement of facility/department accessways have been ignored thus far in the literature, which are subject to the building code requirements that architects must abide by. As a result of these shortcomings, a two-phase optimization framework is proposed for generating block layouts and aisle network configurations for further refining the quality of potential layout design alternatives early in the pre-design phase. The proposed optimization model has a multi-objective function, which is represented as a weighted sum of several objectives. Multiple layouts are generated from various combinations of weights, and are evaluated by an architect to estimate their preferences. The resulting estimated preference can then be analyzed in an effort to find the optimal weights.