The application of lean principles traces back to the Toyota Production System (TPS), pioneered by Toyota Motor Corporation in Japan in the 1950s to address the inefficiencies of traditional mass production. Likewise, in off-site construction various building elements, including precast concrete panels, are manufactured in a controlled environment. Once produced, these components are transferred to the construction site for installation. Hence, the precast concrete manufacturing process encounters relatively close challenges to hurdles in mass production due to its factory-based production. In my poster, I am going to explore the applicability of integrating various Lean principles within the precast concrete manufacturing industry. The objective is to mitigate idle time and enhance efficiency and productivity. Through this exploration, I am going to illuminate pathways for optimization and innovation within the precast concrete manufacturing sector, aligning it with the lean philosophy of value creation and waste elimination.

The growth of the construction industry is estimated to increase by 4.4% by 2024. Due to the controlled environment and increased productivity, offsite construction has gained rapid interest from the construction sector. However, offsite workplace environment is dynamic, where hazardous situations may occur. These occurrences can be attributed to congested working environments and the utilization of construction equipment, such as robotic machines. To mitigate the risk of potential injuries arising from hazardous situations, it is essential to monitor the trajectories of workers and moving obstacles and predict their future paths accurately. Machine learning is a promising approach used in trajectory prediction. This project implemented a Transformer-based machine learning model to simulate workers and moving obstacles behavior separately and predict their future trajectories. The model takes 8 frames as input then predicts the future 12 frames. To assess the trajectory prediction accuracy of the trained models in real construction settings, tests were conducted using actual construction operations data. The model achieved promising results with a Mean Average Displacement (MAD) of 1.25 and Final Average Displacement (FAD) of 2.35.

The increasing demand for efficient and eco-friendly building practices has led to the development of new construction methods to address building issues concerning environmental impacts. However, alongside the incorporated benefits, they also introduce obstacles and challenges that impact the final product. Alternatives to decrease carbon dioxide emissions by reducing construction materials usage are strongly emerging in the industry due to environmental harm and its impact on climate change. In many developed countries such as Canada, different acts and measures are being taken to achieve net-zero emissions shortly, fostering a collaborative commitment across the industry to eliminate millions of tons of greenhouse emissions. This study examines construction methods regarding material usage, carbon footprint, and limitations identified during structural design and architectural modeling. This approach encompasses the analysis of concrete slabs and wall methods, in which embodied emissions are assessed in the manufacturing, transportation, material waste, and construction operations stages.

Wearable Cognitive Assistance Devices (WCADs) are gadgets you wear that help you out while completing demanding tasks. These devices use technologies such as artificial intelligence, machine learning, and wearable sensor devices to deliver immediate cognitive aid. They offer guidance and support by issuing alerts or notifications that are tailored to the wearer's cognitive condition and ongoing tasks. One of the common technologies used to serve such purpose is cognitive architecture. WCADs have been adopted in various industries which include healthcare, manufacturing, sport and fitness, and military and defense with various cognitive assistance functions. These sensors can provide real-time data and insights to both workers and management, helping to prevent accidents, enhance performance, optimize workflows, and improve decision-making. This research aims to explore the application of these technologies as a tool to enhance workers’ cognitive abilities and decision-making process through real-time feedback to proactively introduce remedial measures.

Driven by provisional project requirements, the heavy industrial construction supply chain (HICSC) is complex, dynamic, and uncertain. The behaviors of suppliers further complicate the HICSC, posing additional challenges in managing material inventory and project schedules for general contractors. This can negatively impact project duration and cost. Existing research in the construction supply chain has focused on logistic optimization, material storage and inventory, supplier selection, knowledge management, and coordination. None, however, has centered on identifying the critical materials and assessing the impact of supplier’s reliability on project duration within the HICSC. To fill the gap, this study developed a discrete event model of a typical material procurement process within HICSC from a general contractor’s perspective in the Simphony environment. This study illustrated how supplier reliability affects the material lead time and identified critical material based on the various scenarios extracted using both artificial and real-world historical project data sets.

The construction industry, characterized by its dynamic nature and inherent unpredictability, demands innovative management solutions to enhance predictability and efficiency. This research presents a comprehensive framework integrating multiple project control methods, chaos theory, data-driven analytics, and simulation-based decision support systems to address the complexities of modern construction projects. By employing a mix of Design Science Research (DSR) methodology, quantitative analyses, machine learning models, and simulation techniques, this study aims to develop and validate a multidimensional approach that significantly improves management effectiveness and outcome reliability. The research is structured into four pivotal studies: the first applies DSR, Monte Carlo simulation, and quantitative analysis to assess the integration of various control methods. The second study explores the integration of chaos theory into project management through entropy calculations and simulations. The third paper develops a machine learning model coupled with a warning dashboard for proactive productivity management. Finally, the fourth study investigates the efficacy of a simulation-based decision support system in optimizing construction project recovery plans.

Crane-related accidents on construction sites are often due to operators' errors from suboptimal cognitive states like distraction or overload. Traditional monitoring methods like questionnaires cannot capture these states dynamically during operations. This project aims to introduce a novel framework using mobile electroencephalogram (EEG) technology and a deep neural network to continuously and non-invasively monitor operators' cognitive states. The system will classify cognitive states into optimal and three suboptimal conditions: Mind Wandering, Effort Withdrawal, and Mind Blindness/Deafness. These sub optimal states are good indicators for cognitive Over/Under engagement of operators on the task, making the operation more susceptible to error and failure. After validating the system, it will be implemented on different construction sites and collect data from actual construction operation to gather insights on the operation from operator cognitive perspective. These new insights will be manualized for practical purposes in new projects crane location setup and further improvement of in-cabin user interface setup.

As the elderly population expands, the active mobility of older adults emerges as a critical issue for both individual well-being and societal prosperity. Despite the scholarly significance, the elderly’s mobility remains restricted due to a variety of stressful interactions with the built environment in their daily trips. Recent advancements in wearable biosensing technologies have demonstrated the potential to continuously detect environmental barriers by measuring stress in older adults’ daily trips. Subsequently, this study aims to (1) develop a wearable biosensor and geographical analysis-based framework to precisely locate older adults stress in their built environment, and (2) propose multifaceted intervention strategies, such as urban improvements to alleviate identified stressors, motor and cognitive training, and a mobile application that suggests routes minimizing stressful interactions. This integrated system promises to significantly advance urban mobility for older adults through less invasive and less laborious means, thereby facilitating healthy aging.

Safety management is crucial in construction, requiring robust protocols for project success. A well-designed safety management system integrates safety measures into all project aspects, from hazard identification to emergency response. However, traditional methods sometimes struggle to comprehensively address safety due to the dynamic nature of construction sites. Traditional methods often fail to address safety comprehensively due to the dynamic nature of construction sites. This study introduces an innovative causation model to enhance safety management by employing systems mapping methods. By understanding the complex relationships among various factors influencing safety outcomes, it aims to augment existing systems. The research focuses on three objectives: proposing an ecosystem perspective on safety leading indicators, analyzing contributors to safety incidents comprehensively, and developing causation models to anticipate safety levels on construction projects. Through these objectives, this research aims to improve decision-making and prevent incidents, ultimately enhancing overall safety management systems in construction projects.

Clash detection has been argued as one of the most beneficial BIM (Building Information Modelling) applications. However, the clash resolution process is still manually conducted and time-consuming, and BIM information is not fully utilized to facilitate automatic clash resolution. This research explores machine learning application through two main avenues: Categorizing clashes by image recognition and numerical data. By applying an Artificial Neural Network multilayer algorithm to different combinations of clashes a precision of over 80%was achieved. Image recognition was also employed using YOLO v9's supervised CNN algorithm. This research makes a valuable contribution to BIM and model coordination. Furthermore, the development of a predictive model for clash significance presents new possibilities for professionals in the industry to enhance the efficiency of model coordination meetings by considering the disciplines, elements, and volumes of the clashes.

Off-site construction (OSC) is considered in the Architecture, Engineering, and Construction (AEC) Industry an innovative engineering system, which provides several benefits in the execution of construction projects such as better quality, increase productivity, and safety, reduced labor intensiveness and construction time, and better sustainable performance. However, their use remains globally low. It is attributed to the lack of understanding of project management in OSC and methodologies to facilitate the adoption of OSC. To address this issue, the aim of the research is to propose an Off-site construction Maturity Model (OSC-MM) based on Lean construction management practices and tools that involve the main OSC project phases. This OSC-MM will allow measuring the readiness of large and medium-small organizations, assess and improve processes transversally for the execution of OSC projects, as well as serving as a prescriptive methodological tool to effectively adoption of OSC in the AEC industry.

Construction projects often face cost and time overruns, requiring adjustments to meet deadlines. Accelerating construction through schedule adjustments demands increased productivity and workflow efficiency, potentially raising costs if not managed properly. The crashing technique reduces activity durations and project duration, affecting time-cost trade-off decisions and CO2 emissions. This study aims to shorten a case study's construction project duration using stick-built and panelized construction methods. Stick-built relies on resource deployment and linear optimization for crashing activities at minimum costs and CO2 emissions. The panelized construction method involves offsite prefabrication to compress critical activity workload and transport it for onsite installation. The project manager seeks to accelerate construction activities, facing the choice between stick-built resource allocation or the panelized method. The study aims to determine which strategy, when implemented, results in the lowest cost and CO2 emissions.

Fire incidents are highly likely to occur on construction sites due to flammable materials, complex environments, and large workforces, potentially threatening the safety of workers, properties, and nearby buildings. Construction workers’ awareness and ability to identify fire risks and respond effectively are crucial for improving on-site fire safety. Emerging technologies, such as extended reality (i.e., virtual reality, augmented reality, and mixed reality) and bio-sensors, have shown promising effects in investigating and training to improve construction safety. However, their application in fire safety is still limited. Therefore, this study aims to discuss and propose a conceptual framework that integrates XR and bio-sensors to enhance construction workers' fire safety awareness and response capabilities. This framework includes a fire risk identification and response training model tailored for various fire scenarios, which can aid in developing future training prototypes.

We developed a new computer vision algorithm that can automatically detect and track construction vehicles in real-time using low-power computers. This technology addresses the safety risks and inefficiencies caused by traditional manual monitoring methods that are prone to human error. In the construction industry, site monitoring systems lack advanced artificial intelligence (AI) capabilities and rely on low-power processors that cannot efficiently run traditional AI models. Our solution uses a state-of-the-art AI model called YOLO v8s, optimized with Intel's OpenVino technology, to enable real-time vehicle detection and tracking with high accuracy, even under challenging construction site conditions. This innovative approach overcomes the limitations of traditional monitoring methods and brings the power of advanced computer vision technology to construction site management. With this new technology, construction companies can integrate advanced AI capabilities into their site operations, thereby improving safety and productivity without the need for expensive high-performance computing resources.

Game engine technology has been studied extensively lately in offsite construction (OSC) research due to its ability to develop virtual reality (VR) environments that can be used in the decision-making process prior to the actual project implementation. However, accessing these VR scenes typically requires the participant to be in a VR lab or at least possess VR-specialized hardware and software. Also, these models are typically accessed in isolation with no real-time connectivity with other stakeholders, limiting their collaboration efficiency and questioning their applicability in real life. Thus, this study proposes a web-based multi-player framework based on game engines and blockchain technologies to promote collaborative decision-making processes in OSC projects. The developed system allows users to access a cloud simulation-optimization (SO) model to evaluate several decisions based on identified key indicators. Two case scenarios are studied to ensure efficient connectivity among the OSC stakeholders and the security of the developed network.

Auditory warnings serve as a final defender to many types of accidents at construction sites. However, excessive construction noise hinders workers from noticing the warning sounds, thereby failing them to properly react to the warned risks. Existing adaptive systems, which primarily modulate sound pressure levels (SPL) according to the noise, are effective because the SPL of noise at construction sites often exceeds the legal limitation, leaving no room for the warning sound to be louder. To address this issue, our system adapts acoustic features (e.g., pitch, entropy) of warning sounds according to those of the ambient noise, thereby ensuring warning noticeability regardless of noise SPL. A computational model for acoustic adaptation is established by observing changes in warning noticeability through brain signals while varying the acoustic combinations of warnings and construction noise on participants, using a brain sensor and generative artificial intelligence. After extensive field evaluation, the system will be prototyped as a module attachable to heavy equipment, making its initial field use. The proposed system will significantly improve the effectiveness of auditory warnings, and ultimately construction safety.

In the field of construction and urban planning, adherence to building codes is crucial to ensure safety, efficiency, and sustainability of the projects. However, going through extensive regulatory documents, such as the National Building Code of Canada (NBCC), can be exhaustive. To tackle this challenge, this project will develop a conversational question answering system designed specifically for NBCC. Compared to traditional querying methods that require substantial time to navigate the large volume of text in regulatory documents, this research not only eases efforts but also enables the extraction of precise information more quickly. We are utilising advanced search algorithms and Large Language Models (LLMs) to achieve the desired task. By leveraging the capabilities of LLMs, which excel at understanding and generating human-like texts, we facilitate natural, user friendly and intuitive interactions with the NBCC.

Traditional construction site inspection involves repetitive in-person visits to ensure safety and progress of projects. In-person site inspection is time-consuming, costly and inconsistent especially for remote projects. By using robots to enable real-time remote visual inspection can provide inspectors with enhanced and readily available construction site information, improving safety, quality and progress tracking. Additionally, integrating visual inspection with AI can assist inspectors to detect issues and make informed decisions. This study proposes a new approach to inspect and monitor construction site in real-time using a wheeled robot and Virtual Reality (VR) headset. The proposed framework in this study uses a cloud-based solution to integrate a robot with live 360-degree video and a VR application for visualizing and remote navigating the construction site, as well as using computer vision to detect possible issues automatically. This approach improves collaboration, situational awareness, and reduces repeated site visits—all of which contribute to increased project efficiency.

This project proposes new methods for estimating and measuring productivity. In this regard, the appropriate detail level is clarified, determining the necessary data granularity for effective productivity estimation and measurement. Later, this data is linked with designed work packages. During the estimation process, the impacts of contextual and pre-determined factors are considered for more accurate results. During execution, these factors' impact can be quantified and recorded in a dataset for future use. Moreover, this research proposes an approach to integrate all the required data into one single platform and link them to the BIM model to improve the estimation and measurement accuracy and the storage of critical information. This new method is anticipated to help project managers to enhance productivity through improved decision-making. Another expected benefit is that construction companies can utilize the continuously updated productivity data stored in the database to enhance their project estimation in the planning phase.

The construction industry, characterized by its physically demanding tasks, faces significant ergonomic challenges that frequently lead to musculoskeletal disorders and reduced productivity among workers. Recent advancements in industrial exoskeletons offer promising avenues to alleviate physical strain and enhance worker performance. These wearable devices are designed to support repetitive motions and heavy lifting, and have shown positive effects in controlled laboratory studies. Despite these promising results, there has been a notable lack of in-field evaluations specifically within the construction industry. This gap underscores the need for comprehensive studies to assess the practicality and effectiveness of exoskeletons in real-world construction settings.

This research presents an innovative approach for detecting and evaluating defect propagation of structures using image matching and convolutional neural network (CNN). Employing the Scale-Invariant Feature Transform (SIFT) technique for image matching, this study determines similarity scores to identify the most relevant image pairs. Subsequently, Mask R-CNN is applied for defect identification and pixel-level segmentation in the selected top-score image pair. Post-processing algorithms are then customized to compute defect properties (width, length and area) for each image and facilitate a comparative analysis of defect growth over a certain time interval. By integrating SIFT for image matching and Mask R-CNN for defect identification and segmentation, this methodology offers a robust framework for accurately tracking defect evolution. The findings from this study hold significant potential for enhancing defect detection and evaluation processes across various domains, especially in industries reliant on precise monitoring and structural integrity management.

Light wood frame structure has become a predominant form of residential construction in North America due to its advantages of lightweight, fast construction, and environmental friendliness. As primary load-bearing structures, floor systems often require labor-intensive manual design and drafting processes, which are subjective and inefficient. With the rapid development of Artificial Intelligence (AI) technology, this study proposes a method using deep reinforcement learning algorithm to design floor systems. This method adopts material cost as the evaluation criterion and aims to provide optimal floor component (joists and built-up beams) selection for different plan dimensions. The method performed well as compared with existing datasets generated from manual design.

Tower and Mobile Cranes are some of the most commonly used heavy equipment in all construction sites, and any crane failures could lead to significant human and monetary losses. Moreover, rigging configuration determination is a critical task that requires the rigging crew to have significant experience and knowledge of various failure modes that can be encountered when performing lifting operations. However, despite the criticality of training riggers, there has yet to be a comprehensive tool used to train and guide inexperienced riggers, and hence, more practical tools are needed. This paper proposes a framework for using Virtual reality (VR) and simulation to train riggers to identify the optimal rigging configurations based on the lift type and the external conditions. Through 3D modeling, the critical components of the rigging system are modeled to accurately simulate the rigging system and their performance when faced with critical loading scenarios. The developed framework is expected to allow inexperienced riggers to identify critical failure modes and enhance construction operations' overall safety performance and productivity. Furthermore, several scenarios are assessed based on historical evidence for rigging configuration failures, and the efficiency of the training tool is assessed through real-life scenarios and tests.

In industrial modular construction sequencing, traditional methods often rely on labor-intensive manual planning, leading to inefficiencies and delays. To address these challenges, utilizing AI-based planning emerges as a promising solution. In this context, this research explores the application of Reinforcement Learning (RL) techniques within a 3D GridWorld environment. By examining RL algorithms, particularly Deep Q-Networks, the study aims to automate the generation of construction sequences for a basic predefined industrial module structure, while considering factors like scaffolding requirements. This initiative aims to bridge the gap between automated sequencing and industrial construction methodologies, thereby propelling advancements in industrial construction planning. It lays the groundwork for more streamlined and adaptable approaches in automatic sequencing for industrial modular construction.

Buildings' operational and embodied environments are responsible for a large portion of global greenhouse gas emissions. To reduce embodied carbon, eco-friendly materials and structural optimization are crucial. From a sustainability point of view, timber is listed on the top. Currently, mass timber building construction is experiencing a huge surge in North America and Europe. However, mass timber systems are generally more costly, compared to conventional systems built with other common structural materials. Structural design optimization can help in this regard and promote the construction of this sustainable alternative, to traditional systems. The current research is developing a method of optimizing glue-laminated timber (glulam) beam design through the use of reinforcement learning (RL). An RL agent, called the Proximal Policy Optimization algorithm, is trained to design the beam. It is found that the agent successfully designs the beam and minimizes the cost for a given beam length and loading condition.

This research leverages Quality Function Deployment (QFD) and generative design to optimize the layout design process, targeting the limitations of flexibility and speed inherent in traditional design methodologies. Specifically, it focuses on transforming complex client requirements into precise design specifications through the House of Quality (HoQ), the foundational tool of QFD. The study aims to develop a holistic framework that integrates these methodologies into architectural workflows, enhancing efficiency and adaptability. Using Dynamo, generation and evaluation algorithms are scripted and optimized via NSGA-II to create user-centric designs based on predefined objectives. The results are directly applicable to Building Information Modeling (BIM) systems, bridging the gap between conceptual development and practical application. This approach not only meets diverse client needs effectively but also advances the adaptability and efficiency of architectural design processes.

This study explores the correlation between the risks, time, and cost associated with the use of cherry pickers in construction projects. The idea to enhance safety measures to optimize project efficiency. Bayes Theorem has been utilized to evaluate the risks associated with cherry pickers based on historical fatality incident reports. The dynamics between risk, time, and cost before and after the implementation of safety improvements will be investigated. The aim is to strike a balance between mitigating risks effectively while keeping project timelines and budgets manageable. Empirical findings and graphical representations offer valuable insights into the integration of safety protocols with cherry-picker operations, facilitating informed decision-making for construction project management. The study contributes to the advancement of safety practices in construction and provides practical recommendations for balancing safety, time, and cost considerations.

The construction industry is a dynamic platform where every project is unique, time, and cost-bounded. That uniqueness makes it very difficult to have a fixed set of procedures to tackle disruptions. Construction resilience is the ability to withstand and recover from disruptive events that happen under high levels of uncertainty during the whole life cycle of a construction project. Different strategies and procedures are being adopted by the industry to mitigate and solve problems the industry faces, structural resilience, supply-chain resilience, and infrastructure management. In the fast-paced realm of industrialized construction, a profound understanding of disruptive events is indispensable for companies, especially off-site construction ones, striving to build a resilient and adaptive foundation for the manufacturing stage. This research aims to help the growing modular construction market by establishing a framework of decision-making processes for the manufacturing stage in controlled production environments.

This research evaluates an augmented reality (AR) task assistance system designed specifically for construction workers, focusing on its usability, human-centered design, and user satisfaction. The study highlights key system features, development, and initial tests in a simulated construction lab. A mixed-method approach was employed, utilizing eye tracking, questionnaires, and interviews to gather feedback pre- and post-iteration. Experimental findings demonstrate significant improvements in user satisfaction and system usability after a round of iterative enhancements. User feedback underscores the system’s potential in enhancing work safety and operational convenience, while noting areas for improvement like latency and low-light usability. The findings of this study lay an empirical, user-data-driven foundation for AR system interaction design and iteration, supporting its potential for broader application within the construction industry. Future research will focus on expanding system functionalities and conducting extensive tests across a more diverse range of work environments.

The high labor demand on construction projects, coupled with the limited and unpredictable availability, creates a dynamic condition, which have been dealt with by labor unions for a long time. As a supplement, non-unionized workers from open shops who are ready to work can be scheduled to meet the demands. Uber, a ride-hailing service provider, has effectively tackled an analogous but more complicated scheduling problem. Available drivers are driving in the city, waiting for calls, signing out whenever they choose. Orders can be placed anytime, and the trip information is unpredictable. Uber is designed to dynamically update drivers’ statuses, assign orders to drivers based on specific optimization rules. Inspired by the Uber approach, this paper aims to develop a framework to support union managers in setting various optimization objectives and rescheduling within a short timeframe. Three scenarios are explored to expand the uncertainties that union managers face.

This project introduces two innovative, cost-effective methods for estimating Ground Reaction Force (GRF), the force exerted by the ground in response to pressure. These methods, which use mobile devices and computer vision technology, provide accurate GRF measurements, rivaling those of more expensive equipment. The first method captures vertical movements at the waist using a smartphone’s accelerometer, offering GRF estimates during activities like walking. The second method uses a computer vision system called OpenPose to track key points around the waist, analyzing shifts in these points to compute GRF. Both techniques apply Newton’s second law of motion to refine movement data into precise GRF measurements. The results verify the accuracy of both methods in capturing the timing and intensity of GRF, making them affordable options for dynamic force monitoring. Additionally, this data can be used in Finite Element Analysis (FEA) to further analyze structural responses.

This research explores the application of Extended Reality (XR) technologies which combining virtual and augmented realities to enhance training in the construction industry. Focused on collaborative scenarios, this research project employs XR to simulate complex construction environments where personnel can train together, regardless of their physical location. A significant innovation of this research is the integration of Digital Twin technology, which creates a real-time, virtual replica of physical sites. This integration allows for dynamic, interactive training sessions that mirror actual working conditions, enabling workers to safely practice in a controlled, yet realistic setting. The study seeks to answer the question: How can XR and Digital Twin technologies be effectively combined to facilitate collaboration in construction training? By addressing this, the project aims to enhance collaborative practices, improve safety, and increase communication efficiency across construction teams, and contribute to new standard of training in the industry.

As steel manufacturing involves a range of processes, the complex variables and parameters can be hard to manage. There would be potential drawbacks concerning safety issues, waste management, inventory management, process optimisation, quality control, and supply chain management without proper project management. Hence, the objective of this research is to adopt lean principles and Digital Twin (DT) to improve the steel production facility’s daily project management. By following lean principles, we will take a deeper look at project management and identify the key constraints for steel manufacturing processes such as space, materials, labour, and equipment. Afterwards, we will create a digital copy of the factory, called a Digital Twin, to better understand and manage the challenges that can come up with while managing these constraints.