Image processing and optimization based automation approach in the design of reinforced concrete buildings

This study aims to automate the structural system design process from architectural plans by integrating image processing and heuristic optimization methods. In the initial stage, digital architectural plans are analyzed using image processing techniques, enabling the structural axis system and potential column locations to be identified in an objective and repeatable manner. Based on this data, the placement of columns along the generated axis system and the selection of column and beam cross-sections are formulated as an optimization problem, considering both code requirements and cost criteria. Since the solution space is largely discrete and only a small subset corresponds to valid solutions, heuristic algorithms are employed instead of classical derivative-based methods. Within this scope, different optimization algorithms were tested, their performances compared, and the algorithm providing the best results was identified. The evaluation of each candidate solution involves algorithmic modeling, numerical analysis, and the computation of a customized objective function. As these operations can be conducted independently across solutions, the process carries high parallelism potential at the evaluation level, even though the optimization algorithm itself does not operate in parallel. Accordingly, a parallel computing framework was developed to enable simultaneous evaluations on multi-core processors. Structural analyses were performed using the OpenSEES platform due to its high computational speed, while evaluation procedures and penalty functions were implemented within the MATLAB environment. The developed system provides an end-to-end automation approach that, through the comparison of different algorithms, enables the attainment of an optimum structural system design with minimal user input.