A recent review article published in Case Studies in Construction Materials provided a comprehensive analysis of 296 deep learning (DL)--based studies in Construction Engineering and Management (CEM) from 2014 to 2024 identified six key application areas where DL has been effectively utilized. These include equipment management, schedule management, structural health monitoring, site safety management, workforce management, and intelligent design.
Study: Classification and application of deep learning in construction engineering and management – A systematic literature review and future innovations. Image Credit: SWKStock/Shuterstock.com
Advantages of DL in CEM
Deep learning (DL) is based on extracting sophisticated feature representations from data through multiple layers of neural networks. These networks consist of interconnected neurons that process input data using weights and activation functions. In parallel, Construction Engineering and Management (CEM) is a multidisciplinary field primarily focused on planning, construction operations, and maintenance (O&M).
While DL has demonstrated transformative impacts across academic and practical domains, its adoption in CEM is still in its early stages due to the field's complexity and inherent uncertainties. However, recent computational advancements have significantly enhanced DL's ability to address complex challenges in CEM, unlocking numerous benefits in visualization, automation, optimization, and prediction.
DL excels at converting complex construction data into clear visual representations, enabling engineers to understand project progress better. This improved clarity reduces errors, minimizes delays in information dissemination, and enhances decision-making efficiency and accuracy.
Moreover, DL efficiently integrates and analyzes large volumes of accumulated data to uncover latent insights. These insights can be seamlessly incorporated into project management software to automate data analysis and support informed decision-making.
With its powerful data processing capabilities, DL facilitates real-time analysis of construction data and schedules, enabling autonomous, optimal, and timely adjustments to timelines and resource allocation.
By analyzing historical data patterns and trends, DL also provides accurate predictions of future conditions. These predictive capabilities can anticipate project schedules, budget overruns, safety risks, and other potential challenges, making DL an invaluable tool for CEM.
Challenges of DL Applications in CEM
While DL provides intelligent solutions for CEM, its practical applications in CEM are still limited due to certain challenges. The primary challenges are related to the DL algorithms themselves: the black box challenge and the cybersecurity threat.
The black box challenge refers to the difficult understanding and articulation of the internal mechanisms of a DL model. Initially, machine learning developers prioritized model performance and application efficacy over clarifying the model’s decision-making processes.
This becomes a major hurdle in practical applications like CEM, as decision-makers are required to comprehend the model's logic and rationale to ensure its accuracy and safety.
The cybersecurity challenges related to the use of DL models stem from their inherent security vulnerabilities. DL models require continuous learning and innovation to remain effective in performing specific tasks. This makes DL a potential source of malware threats.
The application of DL in CEM is further restricted due to dataset dependencies, as DL requires a large volume of data for training in the recognition of numerous features and patterns.
Generally, the unlabeled data required for most CEM tasks is sufficient or can be easily collected, but obtaining high-quality, accurately labeled data is difficult due to limited availability and high collection costs. Moreover, privacy concerns arise while procuring some datasets. Therefore, most CEM datasets are limited, inhibiting the DL model from reaching its full potential.
Application of DL in CEM is generally expensive due to high costs related to hardware equipment, data collection and processing, algorithm deployment, and ongoing operations and maintenance. Additionally, a lack of collaboration and integration, standards and specifications, and technicians with diverse skill sets challenge the large-scale application of DL in CEM.
Conclusion and Future Prospects
Overall, the researchers comprehensively explored the potential of DL in CEM and the associated challenges. Compared to other digital technologies like building information modeling (BIM) and machine learning, the use of DL in CEM is in the early stages. Accordingly, the researchers suggest various potential future applications of DL in CEM.
DL-based applications are mostly limited to specific targets, such as personnel behavior and equipment status, in current construction site management. Meanwhile, an effective synergy across the global construction industry is lacking.
In such a scenario, DL-based knowledge graphs can comprehensively capture construction site information and improve the overall connection of the sites, facilitating multimodal construction site management.
Virtual and augmented reality (VR/AR) technologies can significantly enhance visualization and collaborative efficacy for CEM through DL-driven real-time rendering and interaction with BIM. Additionally, DL can enable intelligent design with data sharing through current data exchange standards such as Industry Foundation Class.
Furthermore, incorporating large language models (LLMs) for text data analysis can enhance the decision-making process. By training LLMs on various CEM-related texts to learn the domain-specific language patterns, terminology, and semantic relationships, data analysis becomes more accurate.
Thus, the findings of this study can serve as a fundamental reference for construction engineers and researchers, enabling them to thoroughly explore other applications of DL in CEM.
Journal Reference
Li, Q. et al. (2024). Classification and Application of Deep Learning in Construction Engineering and Management – A Systematic Literature Review and Future Innovations. Case Studies in Construction Materials, e04051. doi: 10.1016/j.cscm.2024.e04051. https://www.sciencedirect.com/science/article/pii/S2214509524012038
Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.