Window-opening behaviour in China’s maternity hospitals: influencing factors analysis and prediction models

Abstract

Occupants’ window-opening behaviour has been proved that can directly influence the indoor environmental quality, thermal comfort, and human health. Despite studies on window-opening behaviour being consolidated in previous studies, the researches related to window-opening behaviour in residential and office buildings have been well conducted, while the investigations in healthcare facilities requires to be further explored. This research is the first study for investigating window behaviour in the maternity hospital, includes the analysis both in wards and offices, which could enhance researchers’ understanding of the window operation behaviour in maternity hospital through a significant amount of data, both from subjective and objective aspects.
Two aims and five objectives have been proposed corresponding to the five research questions, thus divides the work in this thesis into four aspects:
• Investigation of the window-opening behaviour pattern corresponding to different type of occupants in maternity hospital;
• Analysis of influence factors and differences in window-opening behaviours in different types of rooms, hospitals (in two different regions and climates) and seasons;
• Establishment of prediction model with ML algorithm for predicting window states in maternity hospital;
• Development of a weighted scoring system which uses field measured environmental data to determine the most suitable ML algorithm for accurately predicting window-opening state in maternity hospital.
To capture the occupants’ window behaviour in the Maternity hospital, the methods of real-time field measurement and questionnaire survey are used for data collecting in this thesis. To investigate the differences in window-opening behaviours in different regions and climates, the measurement has been conducted into two case study maternity hospitals in two cities: Ningbo and Beijing. Because of the unpredictability of human behaviour, the questionnaire method is used to collect people's subjective intention for summarising user behaviour habits and patterns and to supplement the actual measurements.
All the data would be filtered before analysis, followed by the correlation investigation of window states. Combining the two case studies in different maternity hospitals (Ningbo and Beijing), this thesis investigates the distinction caused by different target population (patients and doctors), climates and seasons. Besides that, measured parameters using chart analysis to define the potential influencing factors on window states and using MANOVA to investigate the level of influence of each factor on window-opening behaviour are presented. The ineffective factors would be screened out while the influential factors would be combined into different data sets according to their importance. After that, one commonly used and six advanced ML methods have been chosen to predict the window behaviour in this study, with the data coming from field measurement. After the establishment and validation of the ML models, a weighted scoring system with six criteria has been proposed to assess the candidate ML models with the same situation of dataset to provide the most suitable ML method for window-opening behaviour in maternity hospital.
This study is both meaningful in theoretical and practical aspects. In a theoretical perspective, firstly, this research fills the gap of the thermal comfort analysis on pregnant women in hospital wards. For researchers focusing on thermal comfort, the results could help them to understand the specific thermal requirement and thermal comfort of pregnant women in healthcare facilities, and provide a reference for improving indoor thermal environments for pregnant women, which in turn would benefit the physical health and mental comfort of the pregnant women and accompanying families.
Secondly, for researches focusing on window-opening behaviour, the results fill the gap of the analysis on window-opening behaviour of particular categories (pregnant women) in special health facilities (maternity hospitals). The results could provide researchers with knowledge of the indoor environment, thermal requirement as well as the window-opening behaviour in maternity hospitals, that enhance the understanding and relevant decision-making for healthy and comfortable indoor environment. Besides the above, through the analysis based on a significant amount of data, this study helps the researches to understand the regularities and drivers from both objective (field measurement results) and subjective aspects (questionnaire results) on occupants’ reactions with windows in health facilities. In the practical aspect, firstly, the research could help designers or operators to better understand the window-opening behaviour in healthcare facilities, and to provide a reference for future architectural and intelligent design of hospital buildings. The designers could get the reference for intelligent control of the air conditioning or ventilation or smart window systems, etc. This also could help operators improve the energy conservation strategy by the intelligent control for the usage of energy-using equipment (cooling, heating or ventilation systems) which would be affected by window states.
Besides, this research assessed different machine learning models and their accuracy and their ease of use for predicting window operation behaviour. Hence, it provided a weighted scoring system that can select a more suitable data-driven method that can potentially be further incorporated into building energy simulation tools. The weighted scoring system can not only used in the analysis of the window opening behaviour in maternity hospitals, it could be applied in window-opening behaviour area in different situations. With the change of the six criteria and weighting this system can be extended to the areas that need to use ML methods, it provides a reference for the researchers who need to adopt ML algorithms to select the most suitable ML methods for their researches.

Date of Award	Jul 2022
Original language	English
Awarding Institution	University of Nottingham
Supervisor	Wu Deng (Supervisor), Isaac Yu Fat Lun (Supervisor), Ali Cheshmehzangi (Supervisor) & Yupeng Wu (Supervisor)