CFD Data Reduction for Indoor Environments: Use of Statistical (Linear Regression) and Neural (Back-propagation) models

1.0 ABSTRACT Dynamic data visualization in real-time involves rapid simulations of computerized models. Although computing power has increased exponentially in the last few decades, detailed simulations using software introduce time-delay that defeat the notion of realtime data visualization of the generated results. Approximate techniques can pave way for efficient generation of continuous flow of data, thereby enabling real-time visualization. In this project, statistical (linear regression model) and neural (backpropagation) methodologies are used to learn and approximate post-processed CFD data to ensure rapid data generation for dynamic visualization. An indoor space is simulated using CFD software under various design conditions. Post-processed CFD data is used as training data to learn the AI system developed. The learnt system is simulated to generate rapid CFD data under various conditions to allow real-time visualization through AR technologies, in real-space. While, under normal conditions, CFD simulations cannot generate results that could be visualized in real-time, the integration of AI techniques enables efficient CFD data reduction for real-time visualization.