GENERAL INTRODUCTION
Research problem
Climate change and global warming, along with resource and energy shortages, pose significant challenges for humanity, particularly in the construction industry Rapid population growth and urbanization, especially in developing countries like Vietnam, are driving the demand for housing As construction projects increase in scale and height to accommodate residential and commercial needs, the industry becomes one of the largest consumers of energy and resources, contributing significantly to greenhouse gas emissions.
Traditionally, the emphasis in design, materials, and construction methods has been on aesthetics, functionality, and cost, often neglecting energy efficiency and environmental impact This oversight results in excessive energy consumption and increased CO2 emissions throughout both the construction and operational phases of buildings.
Buildings are projected to represent approximately 41% of global energy savings by 2035, as highlighted by the International Energy Agency Implementing effective energy efficiency measures can lead to substantial reductions in long-term operational costs, lower CO2 emissions, and provide economic advantages, ultimately enhancing environmental quality and overall quality of life.
Investors, project managers, and designers face the critical challenge of developing design solutions that balance functionality, aesthetics, and cost while optimizing energy efficiency and reducing CO2 emissions This approach is essential for achieving sustainable development in the construction industry for the long-term future.
Objectives of the topic
The objective of this study is to:
Developing a predictive model for energy consumption in buildings involves integrating Building Information Modeling (BIM), Building Energy Modeling (BEM), and Machine Learning This approach not only promotes sustainable construction practices but also highlights the significance of various influencing factors on energy efficiency.
- Developing a method for choosing design and construction options by Choosing By Advantages (CBA) method.
Scope of study
The scope of this study includes:
- Type of project: housing, office, apartment projects
- Object of study: BIM-based BEM in energy consumption simulation, the factors affecting the design selection decision related to sustainable construction and
”Choosing by Advantages” method, Machine Learning
- Object of survey: Experts in Sustainable construction field from Investors, Project managers, Contractors, Engineers.
Scientific and practical significances
This article assists design consultants and stakeholders in evaluating how building design features impact energy consumption By analyzing these influences, it enables the selection of optimal solutions to enhance energy efficiency and improve overall energy usage in buildings.
Propose a model to predict energy consumption based on design characteristics of buildings
The research results will become a reference base for future studies on energy efficient design solutions for buildings in Vietnam.
THEORETICAL BASIC AND RELATED RESEARCH
Definitions and concepts
In 1987, the Brundtland Report, released by the United Nations World Commission on Environment and Development, provided a widely accepted definition of sustainable development: it is the type of development that fulfills the needs of the present while ensuring that future generations can also meet their own needs.
Sustainable development focuses on fostering long-term growth that is equitable and environmentally responsible, addressing critical challenges like poverty, inequality, climate change, biodiversity loss, and resource depletion It aims to create a prosperous world for all while respecting the planet's carrying capacity Achieving this requires a comprehensive approach to decision-making that considers the interconnectedness of economic, social, and environmental factors.
The construction industry significantly impacts sustainable development, as highlighted during the First International Conference of CIB TG 16 on Sustainable Construction in 1994, where Professor Charles J Kibert defined sustainable construction as "the creation and responsible management of a healthy built environment based on resource-efficient and ecological principles." Unlike traditional construction, which prioritizes performance, quality, and cost, sustainable construction focuses on resource depletion, environmental degradation, and the promotion of a healthy environment This approach is guided by key principles established at the conference.
- Use renewable or recyclable resources (Renew/Recycle)
- Protect the natural environment (Protect Nature)
- Create a healthy, non-toxic environment (Non-Toxics)
Pursuing quality in the built environment is essential, as rigorously defined in the 1999 Agenda 21 on Sustainable Construction, published by the International Council for Research and Innovation in Building and Construction (CIB).
To address biases in the initial report caused by a predominance of contributors from developed countries, the council published a revised edition of the sustainable construction agenda for developing nations in 2001.
An energy efficient building offers an appropriate environment for habitation with minimal energy consumption and wastage of energy, thereby maximizing energy conservation [3]
An energy-efficient building provides residents with a comfortable living environment while minimizing resource use and energy consumption This approach encompasses the entire lifecycle of the building, including construction, operation, maintenance, and demolition Ultimately, an energy-efficient structure ensures optimal operational performance and thermal comfort for its occupants.
An energy-efficient building achieves a harmonious balance in energy use by integrating energy-efficient equipment, renewable energy sources, and passive solar design techniques.
In the context of building representation, 2D and 3D are commonly used terms; however, Building Information Modeling (BIM) encompasses additional dimensions that enhance project management and execution These dimensions include 1D, 2D, and 3D graphics, as well as 4D for timeline and scheduling, 5D for cost analysis, 6D focusing on sustainability and energy efficiency, and 7D related to facilities management, among others.
6D BIM enhances a building's geometric model by integrating sustainability data, which facilitates early energy usage estimations and promotes reduced consumption throughout the building's life cycle This approach enables the exploration of energy-efficient alternatives and the integration of renewable energy sources To develop this model, specialized software, referred to as a Building Energy Model (BEM), is necessary.
Building Information Modeling (BIM) is a powerful tool for optimizing envelope design and construction, especially in glass selection It allows architects and designers to assess solar exposure and thermal performance while identifying potential conflicts between envelope components and other systems A study by Chu highlights that BIM modeling enhances energy-saving solutions for building facades, enabling the analysis of energy performance and the selection of glass types that maximize efficiency Furthermore, Ghiassi and Zhang demonstrate that BIM simulation tools can evaluate various glass types, coatings, and thicknesses, ultimately leading to the selection of the most energy-efficient materials.
Utilizing Building Information Modeling (BIM) in envelope design enables informed material selection, particularly for glass and other envelope components, ultimately resulting in more efficient and sustainable building designs.
2.1.4 Multiple-criteria decision-making (MCDM) methods:
Businesses face numerous decisions daily, such as hiring staff, selecting technologies, and organizing operations It is reasonable to conclude that different decisions require distinct decision-making processes Roy (1974) identified and categorized these various decision-making processes, providing valuable insights into how businesses can approach their choices effectively.
- Describing: a description of each possibility and its key effects
- Sorting: dividing up all of the options into groups or categories
- Ranking: creating a ranking of all acceptable options
- Choosing: selecting the best option among all the alternatives (or a combination of them)
There are also other types of decisions described later by Belton and Stewart (2002):
- Selecting a Portfolio: to select a subset of options from a wider collection of available options objectives and desires disclosed by the decision-making process
Several approaches can support decision-making processes, particularly in the realm of Multi-Criteria Decision Making (MCDM) Belton and Stewart developed a useful taxonomy of MCDM techniques in 2022, while Arroyo introduced Jim Suhr's "Choosing by Advantages" (CBA) as a new category within this framework.
(2014) Four categories can be used to group MCDM techniques:
1 Goal-programming and multi-objective optimization methods (linear optimization)
2 Value-based methods (e.g., Analytic Hierarchy Process (AHP) and Weighting Rating and Calculating (WRC))
4 Choosing by advantages (e.g., CBA Tabular Method)
The literature on Multi-Criteria Decision-Making (MCDM) methods primarily discusses three key techniques, while a fourth approach, predominantly found in the lean community, is often overlooked in publications related to MCDM and operations research decision-making.
In the Architecture, Engineering, and Construction (AEC) industry, value-based methodologies, especially the Analytic Hierarchy Process (AHP), are highly favored and extensively documented While AHP is widely used, goal-programming and outranking methods are less prevalent in academic literature Additionally, Cost-Benefit Analysis (CBA) is primarily utilized within the lean community, distinguishing itself by emphasizing the comparison of alternatives through their benefits.
2.1.5 Choosing by Advantages (CBA) methods:
Relative research
Table 2.1: Summary of some previous relative research
No Author Topic Methodology Advantages Disadvantages
Building Information Modeling-Based Building Energy Modeling:
Investigation of Interoperability and Simulation Results
- Testing energy performance using the BIM-based BEM model
Proposed BIM-based BEM to solve design sustainable construction
Not propose a method to select optimal design options
Design Optimization of Energy Efficient
Residential Buildings in Mediterranean Region
- Determining cost-optimal efficiency packages by dynamic simulation software DesignBuilder and a building energy optimization software
Consider energy performance and cost of design option
Not consider other factors that impact design selection
Sustainability and Energy Efficiency: BIM 6D Study of the BIM Methodology Applied to Hospital Buildings Value of Interior Lighting and Daylight in
- Using Revit and its plugins to simulate energy consumption
Applied BIM-based BEM to solve design sustainable construction
Not propose a method to select optimal design options
No Author Topic Methodology Advantages Disadvantages
Multi-objective optimization of building energy consumption and thermal comfort based on integrated BIM framework with machine learning- NSGA II
- Simulating building energy in combination with BIM, Machine learning (ML), multi-objective optimization, and visual programming
Applied BIM-based BEM and Machine Learning to solve design sustainable construction
Not propose a method to select optimal design options
Building information modeling-based building design optimization for sustainability
- Integrating functions of modeling, simulation, analysis of thermal and lighting performance, and database in BIM in the sustainable building design process to estimate annual energy demand
- Using a PSO-based optimization process to find potential design solutions
- Finding optimum design scheme by pareto-optimal solutions
Applied BIM-based BEM and
Optimization algorithm to find the optimal design solution
Not consider other factors that impact design selection
No Author Topic Methodology Advantages Disadvantages
Deciding a sustainable alternative by 'choosing by advantages' in the ACE industry
US to gather real application examples and gain understanding of decision- making practices in green building design
- Application of two methods (AHP and CBA) to select sustainable alternatives in the AEC industry
- Comparing two methods AHP and CBA
Recommended that CBA should be incorporated in the lean construction body of knowledge
Not mention how to indicate the level of importance of factors affecting the decision
Comparing AHP and CBA as Decision Methods to Resolve the Choosing Problem in Detailed Design
- Literature study to identify the documented MCDM practices in the AEC industry
- Studying in - depth applications of AHP and CBA in the AEC industry
- Comparing and contrasting AHP and CBA in a case study using hypothetical
Found that CBA was superior to AHP in the context chosen in research
Not mention how to indicate the level of importance of factors affecting the decision
No Author Topic Methodology Advantages Disadvantages preferences with data from a real project
Comparison of Weighting- Rating Calculating, Best Value, and Choosing by Advantages for Bidder Selection
- Conducting a literature search comparing WRC, BVS, and CBA
- Building a case to compare the methods in the context of bidder selection based on the tendering procedure of the real project
Found that CBA provided additional benefits for helping public clients to differentiating between bidders
Not mention how to indicate the level of importance of factors affecting the decision
Forecasting Time-Series Energy Data in Buildings Using an Additive
Artificial Intelligence Model for Improving Energy Efficiency
- Building artificial neural network to predict energy consumption
Propose a model to predict energy consumption in buildings by artificial neural network
Not propose a method to select optimal design options as well as consider other factors that impact design selection
Proposing a hybrid metaheuristic optimization algorithm and machine learning model for energy use forecast in non- residential buildings
- SAMFOR model combines the support vector regression (SVR) and the firefly algorithm (FA) with the appropriate seasonal auto- regression moving average (SARIMA) model
Propose a combined machine learning model to predict electricity usage data in buildings
Not propose a method to select optimal design options as well as consider other factors that impact design selection
Combining BIM and CBA to select Design-Construction solution toward sustainable construction
Identifying factors affecting to decision of selecting Design-Construction solution through previous studies, expert s opinion
Collecting data about level of influence of factors through questionnaire surveys
Statistical processing by SPSS, MS Excel
Evaluating the level of importance of factors
Creating 3D models with the characteristics of a typical office by DesignBuilder software
Identifying primary simulation variables and their ranges of value
Simulation processing by DesignBuilder software
Energy consumption datasets with different values of variables
Creating RF model to predict energy consumption
Defining design alternatives having the optimal energy consumption
Using CBA to select the best design alternative
METHODOLOGY
Research procedure
The project aims to integrate Building Information Modeling (BIM) with Cost-Benefit Analysis (CBA) to identify sustainable design and construction solutions To effectively apply the CBA method for selecting optimal design alternatives, it is essential to identify potential design options and evaluate the key factors influencing the selection process The research involved two parallel processes to achieve these objectives.
The initial step involves identifying and ranking the factors influencing the selection of a design option This process starts with gathering insights from existing research and consulting industry experts Subsequently, a survey is conducted to evaluate the significance of these factors Following a comprehensive summary and statistical analysis using SPSS, the author establishes a ranking of the factors based on their importance.
The author identifies key factors influencing decision-making while exploring potential design alternatives Utilizing Designbuilder software, a 3D model featuring various design variables will be developed The energy simulation process in Designbuilder will yield options that optimize energy consumption based on selected objective functions A dataset containing diverse results from varying design variables will be employed to train a predictive model using the Random Forest algorithm This model will facilitate predictions with new design parameters, offering optimal energy consumption options These optimal alternatives will serve as viable design choices when applying the Cost-Benefit Analysis (CBA) method for selection.
Questionnaires are a widely used research tool for gathering primary data from a large number of respondents efficiently Utilizing survey panels enables quick data collection aligned with specific research objectives and target demographics.
The effectiveness of research results heavily relies on the careful design of the questionnaire It is essential to construct the questionnaire in a scientific and clear manner, ensuring that research objectives are well-defined Striking a balance between academic rigor and readability is crucial to maintain objectivity and enhance the reliability of the collected data.
The Questionnaire is designed based on documents and synthesis of previous studies,
Designing questionnaire following previous studies
Processing survey to collect data
The data collection procedure involves consulting experts in green building design and energy-saving measures, followed by conducting a comprehensive survey to gather responses To ensure the reliability and objectivity of the research, the questionnaire must be carefully designed.
Survey participants will evaluate the significance of various factors influencing their choice of design options for sustainable construction on a scale from 0 to 100 To facilitate this assessment, the author recommends using a five-level Likert scale to help participants visualize and score the importance of each factor effectively.
The survey included architects, engineers, and knowledgeable individuals involved in Green Buildings or energy-efficient structures, encompassing roles such as Investors, Design Consultants, Project Management Consultants, Cost and Contract Consultants, and Contractors Choosing the right participants is crucial, as it directly influences the accuracy and validity of the survey results.
Determining the appropriate sample size is crucial for enhancing the accuracy of research results Bollen suggests that the sample should be at least five times the number of observed variables Similarly, Hoang Trong and Chu Nguyen Mong Ngoc recommend that the selected sample size should be a minimum of four to five times the number of observed variables.
Therefore, with the number of solutions, equivalent to the initial observation variable of
12, the estimated sample size will be about 48-60 samples
The purpose of selecting an appropriate sampling method is to obtain a sample methods commonly used: (1) Non-Probabilistic Sampling and (2) Probabilistic Sampling
This study employs a non-probabilistic sampling method, specifically utilizing convenience sampling This approach not only saves time and costs but is also ideal for exploratory research and hypothesis testing.
The author utilized Google Forms to create a questionnaire, which was distributed to participants via email, Zalo, and Facebook Due to constraints in time and budget, most respondents were selected from the author's acquaintances and professional contacts, rather than a broader range of companies and individuals in the construction sector of Ho Chi Minh City While this convenience sampling method may not provide the same reliability as probability sampling techniques, it is deemed acceptable for the purposes of this thesis.
In order to increase the reliability of the data collection results, the author conducts screening to remove invalid answer sheets, specifically as follows:
- The answers are graded according to a fixed rule, or choose only one answer
- The answer choices are missing, or choose more than one answer
Table 3.1: Statistics on the amount of data collected
3.3.1 Testing the reliability of the scale
Testing the reliability of the questionnaire scale is crucial for evaluating the correlation among the observed variables of the main factor This study employs Cronbach's Alpha coefficient to assess the scale's reliability effectively.
The mathematical formula of the Cronbach’s Alpha coefficient [26]:
In this formula: ρ is the average correlation coefficient between the observed variables and N is the number of factors
Table 3.2: Meaning of Cronbach’s Alpha coefficient values
Cronbach’s Alpha Reliability Level α≤0.9 Excellent 0.8≤α 0.05 The mean values of the groups are not different
Sig Welch < 0.05 The mean values of the groups are different
Sig Welch > 0.05 The mean values of the groups are not different
Figure 3.3: One-way ANOVA analyzing process
Data Analysis Tools
3.3.1 Testing the reliability of the scale
Testing the scale's reliability is crucial for evaluating the consistency of the questionnaire, as it assesses the correlation among the observed variables of the main factor This study employs Cronbach's Alpha coefficient to measure the scale's reliability effectively.
The mathematical formula of the Cronbach’s Alpha coefficient [26]:
In this formula: ρ is the average correlation coefficient between the observed variables and N is the number of factors
Table 3.2: Meaning of Cronbach’s Alpha coefficient values
Cronbach’s Alpha Reliability Level α≤0.9 Excellent 0.8≤α 0.05 The mean values of the groups are not different
Sig Welch < 0.05 The mean values of the groups are different
Sig Welch > 0.05 The mean values of the groups are not different
Figure 3.3: One-way ANOVA analyzing process
Energy simulation software: DesignBuilder
DesignBuilder is an advanced building energy simulation software utilizing the open-source EnergyPlus compute kernel, designed to assist architects and engineers in managing energy, carbon, lighting, and environmental impact assessments With its high simulation performance and user-friendly interface, DesignBuilder offers significant advantages, such as the ability to easily select optimal energy solutions by comparing calculation results, enhancing designs to align with investor requirements, and efficiently handling simulations of large and complex buildings with high accuracy Additionally, it supports importing models from popular construction software like BIM and AutoCAD, and presents simulation results in various formats, including graphs and reports tailored for LEED and LOTUS certifications.
DesignBuilder utilizes a 3-D modeling core and integrated modules to deliver advanced analysis for specific design objectives The 3-D modeling core facilitates fast and easy construction visualization, while tools like solar visualization aid in assessing building shading EnergyPlus simulations provide insights into energy efficiency and thermal comfort, and the Radiance algorithm measures illuminance and natural lighting ratios Its user-friendly interface simplifies HVAC system design, allowing for cost-effective construction that considers carbon emissions By comparing design options against project goals, users can select the most optimal solutions Additionally, DesignBuilder analyzes results and generates reports aligned with green building evaluation criteria, simulating air movement to assess pressure, wind speed, temperature, and thermal comfort limits.
DesignBuilder is a highly regarded building energy simulation software in academia and energy research, known for its accuracy and user-friendly interface These features significantly reduce training time for users, allowing trainers to concentrate on critical design aspects and simulations Consequently, this thesis utilizes DesignBuilder to generate a dataset on energy consumption for office buildings.
Random Forest Algorithm
Random Forest (RF), introduced by Breiman in 2001, is a powerful supervised learning algorithm for classification and regression tasks Its notable advantages include the ability to handle datasets with numerous attributes, a rapid learning process, and high prediction accuracy, contributing to its growing popularity in recent years.
Random Forest (RF) is a proven classification algorithm that utilizes decision trees enhanced by Bagging and Bootstrapping methods During the RF learning process, input values are selected randomly or combined at each node while constructing each decision tree.
In the process of extracting a sample set from a training dataset, approximately two-thirds of the elements are utilized for computation, leaving one-third excluded from this sample The excluded elements serve a crucial role in estimating the error associated with the results generated by the composite trees in the Random Forest (RF) model.
3.5.2 The process of building a Random Forest model
RF modeling process includes 3 main steps [31]:
In Random Forest (RF) modeling, two key parameters are essential: ntree, which represents the number of trees in the forest, and mtry, the number of randomly selected attributes at each node for tree growth Breiman suggests setting ntree to 500 and mtry to the square root of the total number of attributes (√M) Additionally, the input dataset is typically split into two segments, allocating 70% for training purposes and 30% for testing to ensure effective model evaluation.
3.5.3 The advantages of Random Forest Algorithm
Random Forest algorithm is considered superior to some other prediction methods due to its several advantages:
Step 1: From the initial input data set, use Boostrap technique (returned random sampling) to generate sub-dataset S = {S 1 , S 2 , , S n }
Step 2: In each data set S i construct a data tree h i Instead of using all the candidate variables to choose the best split point, at each RF node randomly select a subset space of M attributes from the original M attributes (M M) Besides, the decision tree in the RF model is an unbranched decision tree.
Decision tree 1 Decision tree 2 Decision tree n
Step 3:The RF model makes predictions by polling the results of the decision trees.
The Random Forest model excels in managing complex, high-dimensional data by integrating predictions from multiple decision trees This approach minimizes overfitting and effectively captures diverse patterns within the dataset, making it a powerful algorithm compared to others.
Random Forest demonstrates remarkable robustness against noise and outliers within datasets by averaging predictions from multiple decision trees This averaging process minimizes the influence of individual noisy instances, enhancing the model's reliability when dealing with real-world data that frequently contains imperfections.
Random Forest is adept at managing large datasets characterized by high dimensionality, making it an ideal choice for big data applications Its ability to parallelize the training process and distribute data across multiple processors enhances its efficiency in handling extensive datasets.
Random Forest offers an estimation of feature importance, revealing the relative contribution of each feature to predictions This insight is crucial for feature selection, enhancing the understanding of various variables, and providing valuable perspectives on the problem domain.
Random Forest is effective in managing imbalanced datasets by balancing the influence of minority classes It achieves this by either randomly under-sampling the majority class or oversampling the minority class, which helps mitigate bias towards the majority class.
Random Forest effectively minimizes overfitting through its ensemble approach and inherent randomness in training By utilizing random subsets of data and features for constructing individual trees, it decreases the correlation among them This results in more generalized predictions and enhances performance on unseen data.
Random Forest is user-friendly and straightforward to implement, requiring minimal data preprocessing and feature scaling With its default parameter settings that perform effectively in various scenarios, it significantly reduces the necessity for extensive hyperparameter tuning.
Random Forest, although less interpretable than a single decision tree, offers valuable insights into feature importance, variable relationships, and decision-making processes, aiding in the understanding of the model's behavior.
- Versatility: Random Forest can be used for both regression and classification tasks, making it a versatile algorithm that can be applied to a wide range of predictive modeling problems
3.5.4 Evaluate the accuracy of the RF model
The accuracy of a predictive model is assessed by randomly selecting a set of result values and comparing them to the corresponding test values; the closer the predicted values are to the actual test values, the higher the accuracy of the prediction.
Besides, the study also uses the index Nash Sutcliffe Efficiency (NSE) to evaluate the predictive performance of the model:
∑ (𝑦 𝑛 1 𝑡 −𝑦̅) 2 n is the sample size y t is the value selected for evaluation x t is the predicted value ȳ is the mean of y t in the sample
The closer the NSE value is to 1, the more accurate the model's prediction performance is
Combining physical and statistical models using Machine Learning techniques as shown below:
In which: Training data is used for RF model, test data is used to check the accuracy of the predictive model after being trained
3.7 Software used in the study
In the thesis, research support software has been used for specific purposes as shown in the table below:
Creating 3D models with the characteristics of a typical office by
Identifying primary simulation variables and their ranges of value
Simulation processing by DesignBuilder software
Energy consumption datasets with different values of variables
Creating RF model to predict energy consumption
Defining design alternatives having the optimal energy consumption
Splitting Training data and Testing data
Figure 3.5: Procedure of building a model to predict energy consumption
Table 3.3: List of software used in the study
No Name of software Objective
1 IBM SPSS Statistics 22 Inferential statistics
Creating RF model to predict the energy consumption
This study analyzes the factors influencing material selection in sustainable construction by reviewing previous research, specialized technical documents, and consulting with experienced experts in green building and energy-efficient design The experts, with over a decade of experience in design, project management, and contracting, are affiliated with reputable companies in the construction sector The research identifies 12 key factors that affect design option decisions, focusing on both material characteristics and energy consumption, categorized into two distinct groups to align with the goal of promoting sustainable construction practices.
Table 4.1: Factors affecting to decision of choosing design-construction options
A Factors relating to energy consumption
A2 Total annual CO2 emission Expert
B Factors relating to designing features
B1 Sound insulation capacity Personal experience
Have joined Have not joined
B6 Impact to method statement and construction schedule Personal experience
B7 Impact to operation and maintenance Expert
B8 Availability in market Personal experience
Details of the Questionnaire are performed in Appendix 1
Details of the Survey results are performed in Appendix 2
1.4 Analyzing the characteristics of the study sample
The suitability of the survey participants
Table 4.2: Percentage of participants who have ever joined in Green Building projects or energy efficient buildings
No Response Quantity Percentage Graph
Software used in the study
In the thesis, research support software has been used for specific purposes as shown in the table below:
Creating 3D models with the characteristics of a typical office by
Identifying primary simulation variables and their ranges of value
Simulation processing by DesignBuilder software
Energy consumption datasets with different values of variables
Creating RF model to predict energy consumption
Defining design alternatives having the optimal energy consumption
Splitting Training data and Testing data
Figure 3.5: Procedure of building a model to predict energy consumption
Table 3.3: List of software used in the study
No Name of software Objective
1 IBM SPSS Statistics 22 Inferential statistics
Creating RF model to predict the energy consumption
This study analyzes the factors influencing material selection in sustainable construction by reviewing previous research, technical documents, and consulting with experts in green building and energy-efficient design The consulted experts, who have over a decade of experience in design, project management, and contracting, are affiliated with reputable construction firms The research identifies 12 key factors that affect design option decisions, focusing on both material characteristics and energy consumption, categorized into two distinct groups to support sustainable construction practices.
Table 4.1: Factors affecting to decision of choosing design-construction options
A Factors relating to energy consumption
A2 Total annual CO2 emission Expert
B Factors relating to designing features
B1 Sound insulation capacity Personal experience
Have joined Have not joined
B6 Impact to method statement and construction schedule Personal experience
B7 Impact to operation and maintenance Expert
B8 Availability in market Personal experience
Details of the Questionnaire are performed in Appendix 1
Details of the Survey results are performed in Appendix 2
1.4 Analyzing the characteristics of the study sample
The suitability of the survey participants
Table 4.2: Percentage of participants who have ever joined in Green Building projects or energy efficient buildings
No Response Quantity Percentage Graph
A significant 82.7% of survey participants have engaged in the implementation of green or energy-efficient buildings, while 17.3% have not participated but have actively researched and updated their knowledge on these sustainable building practices.
Table 4.3: Years of experience of the survey participants
No Response Quantity Percentage Graph
Ninety percent of respondents have over five years of experience in the construction industry, indicating that the majority possess significant expertise, which enhances the reliability of their responses.
Expertise of the survey participants
Table 4.4: Expertise of the survey participants
No Response Quantity Percentage Graph
Project ManagerDesignerSupervisorQuantity Surveyor
Comment: The data shows that the expertise of the survey subjects is very diverse, which contributes to making the research results more multidimensional and objective
The study reveals that architects, engineers, and project managers constitute 48% of the participants, indicating a significant representation of professionals with extensive expertise in the research topic.
Roles of the survey participants
Table 4.5: Roles of the survey participants
No Response Quantity Percentage Graph
Statistics indicate that survey participants have varied roles, with Client and Contractor groups representing a significant portion, thereby enhancing the multidimensionality and objectivity of the response outcomes.
Client Project Manager Design Consultant Contractor
Reliability of the scale is analyzed by IBM SPSS Statistics 22
Table 4.6: 1 st results of reliability testing
Scale Variance if Item Deleted
Cronbach’s Alpha if Item Deleted
A Factors relating to energy consumption
B Factors relating to designing features
The initial test revealed that variable B1, which measures sound insulation capacity, was excluded due to a total correlation coefficient of -0.052, falling below the acceptable threshold of 0.3 Following the removal of variable B1, the second analysis yielded new results.
Table 4.7: 2 nd results of reliability testing
Scale Variance if Item Deleted
Cronbach’s Alpha if Item Deleted
A Factors relating to energy consumption
B Factors relating to designing features
Scale Variance if Item Deleted
Cronbach’s Alpha if Item Deleted
The test results show that:
- Corrected Item-Total Correlations are suitable (All ≥ 0.3)
- Cronbach’s Alpha within 0,70 ≤ α ≤ 0.9 shows that the scale is good, meets the requirements of reliability
Following the assessment of the scale's reliability, the factors will be ranked according to the average influence values derived from the survey responses These results will then be normalized to a 100-point scale for application in the Cost-Benefit Analysis (CBA).
Table 4.8: Ranking factors through Mean values
Rank Variable Description Mean Std
8 B6 Impact to method statement and construction schedule 52.10 13.767 55
10 B7 Impact to operation and maintenance 47.85 14.499 50
Table 4.9: Results of One-sample T-Test
One-Sample T-Test Test value = 50 Variable t df
The variables A1, A2, A3, B2, B3, B4, B5, and B9 exhibit significance values (2-tailed) of 0.05 or less, leading to the rejection of the null hypothesis (H0) and acceptance of the alternative hypothesis (H1) This indicates that the average influence of all survey participants is significantly affected by factors other than the mean influence value of 50.
Variables B6, B7, and B8 exhibit significance values (2-tailed) of at least α = 0.05 Consequently, accepting the null hypothesis (H0) implies that the average influence of all survey participants on the factors is equivalent to 50, which represents the mean value of the average influence.
Mean difference analysis for the experience of the respondents
Table 4.10: Mean difference analysis for the experience of the respondents
Test of Homogeneity of Variances
B6 Impact to method statement and construction schedule 2.855 2 49 0.067
B7 Impact to operation and maintenance 00.993 2 49 0.378
Robust Tests of Equality of Means
B6 Impact to method statement and construction schedule 1.700 2 10.486 0.229
B7 Impact to operation and maintenance 2.680 2 10.816 0.113
Robust Tests of Equality of Means
The number of working years of the survey respondents did not affect the assessment of the influence of the factors: A1, A2, A3, B2, B3, B5, B6, B8, B9
The number of working years of the survey respondents affected the assessment of the influence of the factors: B4, B7
Mean difference analysis for the expertise of the respondents
Table 4.11: Mean difference analysis for the expertise of the respondents
Test of Homogeneity of Variances
B6 Impact to method statement and construction schedule 3.523 3 48 0.022
B7 Impact to operation and maintenance 6.751 3 48 0.001
Robust Tests of Equality of Means
A1 Total annual electricity usage 2.691 3 18.354 0.076 A2 Total annual CO2 emission 0.459 3 16.675 0.714 A3 Total annual discomfort hours 8 3 21.849 0.001
Robust Tests of Equality of Means
B6 Impact to method statement and construction schedule 1.369 3 16.745 0.287
B7 Impact to operation and maintenance 2.806 3 10.375 0.092
The expertise of the survey respondents did not affect the assessment of the influence of the factors: A1, A2, A3, B2, B3, B4, B6, B7, B8
The expertise of the survey respondents affected the assessment of the influence of the factors: B5, B9
Mean difference analysis for the role of the respondents
Table 4.12: Mean difference analysis for the role of the respondents
Test of Homogeneity of Variances
B6 Impact to method statement and construction schedule 3.326 4 47 0.018
B7 Impact to operation and maintenance 4.454 4 47 0.004
Robust Tests of Equality of Means
A1 Total annual electricity usage 0.958 4 19.163 0.453 A2 Total annual CO2 emission 0.436 4 20.016 0.781 A3 Total annual discomfort hours 1.54 4 18.294 0.232
B6 Impact to method statement and construction schedule 2.319 4 21.017 0.091
B7 Impact to operation and maintenance 3.578 4 20.742 0.023
The role of the survey respondents did not affect the assessment of the influence of the factors: A1, A2, A3, B2, B3, B4, B5, B6, B8, B9
The role of the survey respondents affected the assessment of the influence of the factors: B7
5.1 Simulation of building energy consumption using DesignBuilder
5.1.1 Procedure of building an energy simulation model
DesignBuilder software allows for the creation of simulation models either directly within the platform or by utilizing information models from architecture, structural, and MEP disciplines By incorporating geolocation parameters and selecting appropriate weather stations for data analysis, users can effectively build energy simulation models The energy model can then be exported as a GBxml file for import into DesignBuilder Designers can choose the parameters for calculation, assign input values, and run simulations to obtain the desired results.
Energy model in GBxml format
Figure 5.1: Procedure of building an energy simulation model in DesignBuilder
A typical office area 144 m2, height from floor to roof 3.8m, weather data at Tan Son Hoa station, Ho Chi Minh City
5.1.3 Defining design variables and input data to DesignBuilder
Based on existing studies and the ANSI/ASHRAE/IES Standard for Energy Efficiency in Buildings (excluding low-rise residential structures) as well as the LEED standard, the author has identified and outlined the variable parameters utilized in the simulation model, as detailed in the table below.
Item Variables Symbol Unit Reference
1 Ratio of Window on Wall WWR % [18], [33],
3 Heat transfer rate (U-value) of external Wall
4 Heat transfer rate (U-value) of
5 Solar Heat Gain Coefficient of
6 Heat transfer rate (U-value) of Roof R-U W/(m 2 K) [18], [33],
8 Cooling coefficient of performance CoP [33], [34],
U-Value measures the heat transfer through the central portion of glazing, excluding edge effects, and indicates the steady-state heat transfer rate per temperature difference between the environments on either side It is expressed in US Standard units as Btu/hr·ft²·°F and in SI/Metric units as W/m²·K.
The Solar Heat Gain Coefficient (SHGC) measures the total solar direct transmittance and the secondary heat transfer of glazing into a building This secondary factor accounts for heat transfer through convection and longwave infrared radiation from the absorbed solar energy by the glazing.
The coefficient of performance (COP) of a heat pump, refrigerator, or air conditioning system measures the efficiency of these systems by comparing the useful heating or cooling they provide to the energy input required for their operation.
Input design variables to DesignBuilder:
Based on the author's experience, the values of the design variables mentioned above are determined in detail as follows:
Item Variables Symbol Options Structure Description
- Outer layer: mortar 25mm thk
Item Variables Symbol Options Structure Description value) of external Wall
- Middle layer: 2 layers of burned clay brick – 180mm thk
- Inner layer: mortar 15mm thk
- Outer layer: mortar 15mm thk
- Middle layer: 1 layer of Autoclaved Aerated Concrete 200mm thk
- Inner layer: mortar 15mm thk
- Outer layer: mortar 25mm thk
- Middle layer: 1 layer of Aggregate concrete block 190mm thk
- Inner layer: mortar 15mm thk
4 Heat transfer rate (U- value) of
Laminated Solar control glass, 13.52mm - U = 5.1
- Solar control glass Tempered 6mm thk
- Clear glass Tempered 6mm thk
- Full tempered glass with Low E coating 6mm thk
- Clear glass Tempered 6mm thk
- Clear glass Tempered 6mm thk
- Clear glass Tempered 6mm thk
SHGC Laminated Solar control glass, 13.52mm - SHGC 0.45
Laminated Low E glass, 13.52mm - SHGC = 0.43
Laminated Clear glass, 13.52mm – SHGC = 0.82
6 Heat transfer rate (U- value) of
Concrete roof insulated by Rockwool – U 0.339
Item Variables Symbol Options Structure Description
Concrete roof insulated by XPS form – U = 0.282
Figure 5.3: Design variables setting in DesignBuilder
Define objectives and additional outputs
DesignBuilder software runs simulations based on input variables information and gives optimal results for objective functions The software allows the input of up to
2 objective functions The remaining output information if necessary, can be set in Additional Output
Based on the survey results on the importance of design elements towards sustainable construction, the author chooses two objective functions as follows:
Remaining factor – CO2 emission will be included in Additional output
Figure 5.4: Objectives and Outputs setting
After inputting the design parameter values, the model is executed to produce a comprehensive dataset of results for each set of parameters, ultimately identifying the optimal parameter sets for the objective functions.
Table 5.3: Simulation results in DesignBuilder
Algorithm by Python Programming Language
5.2.1 Procedure of creating an energy prediction model