Component-Based Machine Learning for Multi-Element Aggregation and Interaction - Indoor Climate Prediction

Aug 5, 2026

Accurately and efficiently predicting indoor airflow, mass transport, and temperature distribution is crucial for designing environments that balance comfort and energy efficiency. Although computational fluid dynamics (CFD) can deliver these predictions, its high computational cost limits practical use during iterative building design. Data-driven surrogate models offer fast alternatives, yet reusability and generalization remain constrained. This work proposes component-based machine learning (CBML) that aggregates two learned prediction components to better capture spatial interactions for indoor climate prediction.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages10

Reading Time25 min

AccessFree to download

Abstract

Keywords

Introduction

Component-Based Machine Learning for Multi-Element Aggregation and Interaction: Indoor Climate Prediction Shaofan Wang1 , Philipp Geyer1 1 Leibniz Universität Hannover, Hannover, Germany ABSTRACT: Accurately and efficiently predicting airflow, mass transport, and temperature distribution in indoor environments is essential for the design phase, supporting exploration and decision-making with respect to indoor comfort and energy efficiency. However, the high computational cost of computational fluid dynamics (CFD) simulations remains a significant challenge and barrier for application of such methods. The data-driven models render a high-potential fast alternative to replace the CFD simulation to predict the indoor environment. However, generalization andreusability are still limited. Therefore, we propose component-based machine learning (CBML) for spatial flow prediction with better generalization ability than the current machine learning methods. In this paper, we tackle the aggregation of two data-driven prediction components; the success of this aggregation forms a fundamental method of CBML for CFD. The CBML surrogate model for this purpose includes three sub-models, a convolutional autoencoder with residual network (CAER) , a multilayer perceptron (MLP) , and a convolutional neural network (CNN) . Testcase data represent a 2D rectangular room equipped with two inlets on the left and right wall; prediction of the aggregation of two separate predictions of flow caused of left inlet and right inlet forms training and test object in this paper. The CAER serves an order reducer for reducing the dimensionality of spatial data and extracting the features of components. The MLP works as a predictor to map the boundary conditions with component features. Last but not least, the CNN is regarded as an aggregator to discover how two components affect others. Comparison of predictions with CFD simulation results shows a maximum absolute error of less than 0.09 m/s for 95% of the flow field in real-time prediction. Besides, the compressed features of single-inlet and multi-inlet are mapped into latent space by t-SNE, illustrating the correlations between features of different components. These outputs demonstrate that the CBML approach is able to predict the two aggregated flow fields, thus, effectively captures complex feature interactions among multiple components, and the latent features contain useful fluid information. KEYWORDS: Component-based machine learning (CBML) , Aggregation and interaction, Reduced order model (ROM) , CFD simulation, Indoor environment prediction 1. INTRODUCTION The attention on indoor hygiene and comfort has explosively increased since the COVID-19 pandemic resulting in an urgent demand for the rapid prediction of indoor environment (Lin et al. , 2024) . Besides, the design of the green building and passive house also requires indoor velocity and temperature distribution for energy-efficient ventilation (Cao et al. , 2014) . In order to achieve these design objectives , integrating indoor environment fast prediction into the design processes of buildings and their systems is important (Zhou et al. , 2020) . CFD is able to predict these spatial and temporal parameters , such as velocity and temperature distribution (Cetin, Avciand Aydin, 2020) but requires significant modelling and computational effort making them unsuitable for integration into the building design phase with its iterative decision-making. Furthermore, the CFD model is based on Navier-Stokes’s equations and numerical scheme , which is not friendly to architecture designers and building engineers , as they do not have professional knowledge of fluid mechanics and can easily lead to wrong model Settings. For this aim, some simplified CFD models are investigated , such as coarse-grid (Wang and Zhai, 2012) and fast fluid dynamics (Han et al. , 2021) . Besides, Recent studies have demonstrated the effectiveness of GPU-a

FAQs

Why is predicting indoor airflow and temperature important?

It supports indoor comfort and energy-efficient ventilation during building design by enabling rapid exploration and decision-making.

What challenge prevents direct use of CFD in design workflows?

CFD simulations require significant modeling effort and computational resources, making them hard to integrate into iterative architectural and engineering processes.

How does the proposed CBML approach improve prediction generalization and interaction modeling?

It aggregates two data-driven components using a CBML surrogate with sub-models (CAER, MLP, CNN) so the system captures complex interactions among multiple components and performs real-time spatial flow prediction.

Aug 5, 2026

Upload by River Wang