Despite regular safety inspections and safety planning, numerous fatal accidents related to scaffold take place at construction sites. Current practices relying on human inspection are not only impractical but also ineffective due to dynamic construction activities. Furthermore, a scaffold typically consists of multiple bays and stories, which leads to complexity in its structural behaviors with various modes of failure. However, previous studies considered only a limited number of failure cases for a simple one-bay scaffold while exploring machine-learning (ML) approaches to predict safety conditions. Thus, the authors have proposed an approach to monitor a complicated scaffolding structure in real time. This study explored a method of classifying scaffolding failure cases and reliably predicting safety conditions based on strain data sets from scaffolding columns. Furthermore, the research team successfully enhanced the predicting accuracy of ML classification by the proposed self-multiplication method to increase the number of features such as strain data sets. Implementation of the proposed methodology is expected to enable the monitoring of a large, complex system at construction sites.

For effective pre-construction process, engineers create a 4D building information model (BIM). However, the rich information associated with the creation of the 4D BIM leads to manual effort. Past studies explored methods that automatically generate construction schedules using 3D building models. However, no method properly utilized relationships between building elements to robustly generate 4D BIMs that are structural stable during installation. This research presents an approach to the automated generation of structurally stable construction sequences using a 3D BIM. Focusing on steel erection, we create a framework integrating a 3D BIM and algorithms to create a 4D BIM with detailed steel erection sequences of individual elements. This research explores an approach to a variation of static indeterminacy for each installation process of steel elements. The principle of this approach is based on the relationships among the nodes and the connections among steel elements, information about which is available to those involved in the project. For validation, we test a prototype software program using a BIM for a real-world construction project. The results indicate that the prototype utilizing the static indeterminacy variation could generate a large number of random sequences and successfully transforms them into stable sequences. This study establishes the foundational step of generating constructible sequences using structural information in BIM which is found to be more robust than previous approaches, and results of this study can lead to follow up studies for full automation such as automated analysis and optimization of the sequences.

Structural integrity is essential for safety in infrastructure, as it can help prevent catastrophic failures and financial losses. The significance of vibration-based damage detection has grown substantially in fields such as civil and mechanical engineering. Concurrently, the advancements in computational capacities have facilitated the integration of machine learning into damage detection processes through post-processing algorithms. Nevertheless, these require extensive data from structure-affixed sensors, raising computational requirements. In an effort to address this challenge, we propose a novel approach utilizing a pre-trained convolutional neural network (CNN) based on images to identify and assess structural damage. This method involves employing wavelet transform and scalograms to convert numerical acceleration data into image data, preserving spatial and temporal information more effectively compared to conventional Fourier transform frequency analysis. Six acceleration data channels are collected from carefully chosen nodes on a mini bridge model and a corresponding finite element bridge model, to train the CNN. The efficiency of training is further enhanced by applying transfer machine learning through two pre-trained CNNs, namely Alexnet and Resnet. We evaluate our method using different damage scenarios, and both Alexnet and Resnet show prediction accuracies over 90%.

Traffic accidents caused by road icing are a serious global problem, and conventional de-icing methods like spraying chemicals have several limitations, including excessive manpower management, road damage, and environmental pollution. In this study, the carbon nanotubes reinforced de-icing coating for the road system with a self-heating function was developed as part of the development of a new system to prevent accidents caused by road icing. The electrical characteristics of the fabricated coating were analyzed, and the carbon nanotube coating heating performance experiment was conducted to measure the temperature increments by applying a voltage to the coating at a sub-zero temperature using an environmental chamber. In addition, the coating was installed on the road pavement and the applicability was investigated through a heating test in winter. As a result of the experiment, the coating made with the higher carbon nanotube concentration presented higher heating owing to its higher electrical conductivity. In addition, the coating showed sufficient heating performance, although the maximum temperature by Joule heating decreased for the entire coating at sub-zero temperatures. Finally, field tests demonstrated the potential of electrically conductive coatings for de-icing applications.

Passive wireless sensing systems, particularly passive antenna sensors, offer a viable alternative to traditional wired and active sensors for long-term structural health monitoring due to their simplicity, easy installation and maintenance, and ability to measure strain without an external power supply. Customization of antenna shape can also adapt to various structural geometries. However, sensor fabrication using chemical etching is expensive and time-consuming, which is unsuitable for limited-quantity production. To address this, this study explores the potential of extrusion-based additive manufacturing to produce cost-effective passive wireless antenna strain sensors. The study investigates polylactic acid’s mechanical and electromagnetic properties for substrate design and uses multi-physics simulation to estimate strain-sensing performance. The obtained results show similar strain-sensing performance to sensors produced through chemical etching, making the manufacturing process a promising alternative.