Advancing Air Balancing in Hvac Systems: Cfd Analysis of Non-Fully Developed Flows and Gpr-Based Prediction of Damper Degrees

Abstract

Air balancing, a critical process in HVAC systems, optimizes airflow distribution to enhance indoor environmental comfort. This research highlights the inadequacies of existing air balancing techniques, particularly under non-fully developed flow conditions within ducts. Traditional models, such as the Darcy-Weisbach formula, assume fully developed flows, overlooking the complex interactions present when local fittings coupled. To mitigate this, the study utilized computational fluid dynamics (CFD) to scrutinize resistance disparities in interconnected elbows, not accounted for in traditional models assuming fully developed flow scenarios. Consequently, we introduced an innovative local resistance calculation methodology, tailored for these specific flow conditions. Moreover, we established a refined pressure balance model, considering adjacent influences, to accurately determine the damper's pressure loss corresponding to the targeted airflow in each terminal zone. Due to inherent flow constraints, actual damper flow attributes can significantly differ from standard handbook specifications, causing discrepancies in damper degree adjustments derived using linear interpolation methods. Addressing this, we innovated a prediction technique for damper degree adjustments, employing Gaussian Process Regression (GPR). This approach harnesses the pressure differential across the damper and the airflow rate to forecast the requisite damper adjustments. Empirical validation attests to the method's precision, with relative errors post air balancing in the tested branches limited to 4.66%, 8.09%, 7.15%, and 7.05%, demonstrating the efficacy and reliability of the proposed improvements in air balancing strategies.

Keywords: Air balancing, Ventilation system, Non-fully developed flow, Resistance calculation, Gaussian process regression, machine learning