Application of Visual Recognition Based on BP Neural Network in Architectural Design Optimization.

Area of Science:

• Building Science
• Computational Engineering
• Artificial Intelligence in Architecture

Background:

• Optimizing architectural design parameters is crucial for enhancing building performance, particularly ventilation.
• Existing methods may lack efficiency in establishing complex relationships between design choices and performance outcomes.

Purpose of the Study:

• To propose and validate an architectural design optimization method based on BP neural networks.
• To establish a clear mapping between key architectural design parameters and building ventilation performance.
• To identify optimal design parameters for improved building ventilation.

Main Methods:

• Utilized a Backpropagation (BP) neural network for architectural design optimization.
• Employed uniform experimental design to generate initial building design samples.
• Trained the BP neural network on sample data to establish parameter-performance relationships.
• Calculated optimal design parameters based on the established mapping for building ventilation.

Main Results:

• Successfully established a mapping relationship between architectural design parameters (spacing coefficient, air outlet area, window sill height) and building ventilation performance.
• Identified optimal building design parameters through the BP neural network model.
• Demonstrated the effectiveness of the proposed method in optimizing building ventilation design.

Conclusions:

• The BP neural network-based method provides a valuable approach for architectural design optimization.
• The established mapping facilitates informed decision-making for improving building ventilation.
• The findings offer practical reference for optimizing architectural designs for better performance.