Today, the use of artificial intelligence to design house plans has become increasingly popular. However, most of these artificial intelligence systems use machine learning and deep learning algorithms that are trained based on the pixel data set of residential plans and do not take into account structural features and construction limitations in their solutions. In this article, in order to improve these defects–the various functions of AI and the cost and time-consuming nature of traditional plan design systems–a rule-based artificial intelligence is introduced for designing residential plans based on a structural system named 'PeyDal'. This AI has been expanded in the analysis and production of plans according to the needs and preferences of the real user in the design and implementation of the plan.
This research is a combination of descriptive-analytical and experimental methods; at first, initial information was collected through the descriptive-analytical method, and then, using the experimental method, an algorithm based on the Peydal System was developed for the design and implementation of houses. Next, the research aims to identify the necessary data sets for classifying the spatial needs of users in different types of plans with different sizes, describe the structure of converting plans into constructive modules in the Peydal System, produce a plan-generating AI design pattern in the form of a reduction of acceptable and desirable results for the user and develop a plan-generating algorithm with the help of artificial intelligence.
The results of the study include the achievement of a digital design plan algorithm with high industrialization capability, the structurality of the algorithm's solutions due to matching with the Peydal Modular System, and the increase in response quality and speed by moving from continuous plan design to discrete design.