Full Session with Abstracts
The main actual problem in modular design is that is often divided between the design of the module unit itself, and the design of the modular building. The first ones try to optimize values such as module size, story drift or connection type with important parameters such as layout design or transportation costs left out of the study. On the other hand, design methodologies focusing on the modular building as a single entity, mainly focus on the layout of such building, where the main characteristics of the modular unit are considered to be given data and not further optimized. From this, it is noted that these design methodologies seem not to work together, which makes it difficult to believe their result might correspond to a true optimal solution: an optimal stand-alone module design might, in some cases, be the best solution for a certain design, when an optimal layout design might not be the best considering the costs involved in the stand-alone module used. Hence, despite the obvious relationship between the module characteristics and the final building design, it is difficult to understand both with a single design method.
In this paper, a new design method is presented to solve the division between modular units and modular building design, and a multi-population genetic algorithm is introduced to solve this trade-off cost optimization problem, considering all the variables and risks in the construction process. First, the layout problem is translated into matrix form for an easier encoding and the constraints on the problem are presented. Then, all the costs and risk considered are introduced and justified, trying to maintain as much parallelism to a real situation as possible. After, the general scheme for the genetic algorithm is presented step by step, explaining further the different operators in it, such as crossover, mutation, or migration. Finally, parameter results using different values of initial population and sub-populations is carried out and a study case featuring the typical layout evolution is highlighted. The obtained results show that the proposed GA is successful in finding the optimal layout, hence minimum cost, for a given module size architectural constraints. This method also enables designers to introduce conditions or preferences depending on the particular problem being analyzed, which makes helpful in assisting during the early-stages. It can also be used to systematically compare building costs of different design alternatives considering the characteristics of the modules used as well as the building ones.