Abstract. The structural integrity of masonry plays a vital role in ensuring the safety and longevity of the infrastructure. Traditional methods of crack detection are often destructive, labor-intensive and limited in scalability. This study presents a non-destructive automated framework for masonry crack classification and segmentation, combining both machine learning models (Random Forest and XGBoost) and deep learning architectures (Convolutional Neural Networks (CNN) and UNet). A custom image dataset of brick surfaces was collected using a mobile phone setup with controlled focal distance and lighting conditions. The images were annotated with binary crack masks using LabelMe, enabling supervised training and evaluation.

Handcrafted features such as edge density, texture metrics, and intensity statistics were extracted for ML models, whereas DL models learned spatial hierarchies directly from preprocessed images. Experimental results demonstrate that while Random Forest algorithm achieved a classification accuracy of 71%, CNN outperformed with 89% accuracy. For pixel-level segmentation, UNet trained on binary masks delivered a high Intersection over Union (IoU) of 0.81 and Dice coefficient of 0.89, effectively detecting crack boundaries and shapes. Comparative analysis revealed that lightweight models like RF/XGBoost are suitable for edge deployment, while UNet is preferable when spatial precision is critical.

The study also highlights the sensitivity of DL models to annotation quality and training strategies, such as the limitations of using Canny edge-based labels. Visual overlays of segmentation results affirm that UNet simulates human-level crack recognition more effectively than bounding box classifiers.

This study contributes to sustainable construction practices by supporting damage assessment and potential reuse decisions within a circular economy framework. By integrating machine learning and deep learning methods into a unified, real-world applicable pipeline, this research bridges the gap between theoretical model development and practical, scalable crack detection assessment for built environment monitoring.

More: Publication

Abstract. In this research, a ten-year predictive maintenance model for wooden (Glulam) beam is developed based on Life Cycle Analysis (LCA) using Python and a Random Forest Regressor model which forecasts moisture content and humidity for 10 years, identifying when maintenance is necessary by setting thresholds which pinpoints for two specific dates across the decade. The energy consumption and CO2 emissions including drying 1m³ of Glulam are calculated at 0.33 GJ and 122.16896 kg CO2e, respectively. A comparative analysis with concrete reveals that using Glulam instead of concrete for an equivalent volume could avoid 287.83 kg CO2e, equating to a 70.2% reduction in emissions. This study emphasizes the environmental benefits of utilizing Glulam over concrete, highlighting the role of predictive modeling in decision-making process and sustainable construction practices, thereby contributing to the broader field of circular economy in built environment. Through life cycle analysis based on the prediction model, the project determines the significant potential of Glulam beams in reducing the carbon footprint of construction materials. 

keywords: LCA, maintenance, Glulam, moisture content, humidity, temperature, prediction model, regression model, carbon emission.

Lead acid batteries comprise electrolytes, lead, lead alloy grids, lead paste, organics and plastics. These components include a significant amount of toxic, dangerous, combustible, and explosive compounds, hence posing possible risks for both human and aquatic life. The primary objective of this research is to study the hazards posed by the recycling of lead acid batteries, specifically the metal, liquid, and plastic wastes generated in the process. Twenty-one surface water samples were collected from the Bhairab, Moyur, and Rupsha rivers, with three taken at the disposal point and the remaining eighteen taken at 165-foot intervals both upstream and downstream. Distinct variations in the quality parameters of surface water samples were evident upstream, whereas downstream showed negligible fluctuations compared to the disposal points. The pH and electric conductivity (EC) were found to be 2.1, 2, and 2.3; and 3564, 3422, and 3503 µS/cm for Bhairab, Moyur, and Rupsha rivers, respectively, at the disposal points. The TDS, COD, and BOD values were significantly higher than the standard limit for all three samples at the disposal points. The lead concentrations at the disposal points of the Bhairab, Moyur, and Rupsha rivers were determined to be 0.732, 0.617, and 0.795 ppm, respectively. Disposal points show about 8 to 10 times higher lead toxicity than the far-located samples from both up and downstream. It has been demonstrated that a significant proportion of both male and female workers are dealing with hypertension-related conditions. The workers in the open market experienced a range of health conditions, including headaches (37%), colic pain (25%), nausea (14%), tremors (15%), anemia (6%), brain damage (1%), cancer (1%), and kidney damage (2%). Hence, it is imperative to enhance the recycling process of lead-acid batteries while simultaneously minimizing and closely monitoring the discharge of industrial effluents.

This study explores the design and planning of an independent tram track infrastructure, focusing on its effectiveness, cost, and sustainability throughout its life cycle. The infrastructure comprises network elements and relations, situated in a suburban area with uneven terrain. The key components include a tram track routing along a traffic dam, intersected by a footbridge, and extending over a steel bridge with a retaining wall abutment. The tram track, positioned on the dam, addresses ground leveling challenges, while the steel bridge spans over water bodies or deep terrain depressions, proving more economical than extended dam construction. The retaining wall supports the bridge on steep slopes. The footbridge serves as a vital link for pedestrians, cyclists, and potentially pipelines or animals, maintaining connectivity disrupted by the tram route, especially near stations. This paper aims to integrate these five systems into a cohesive framework, developing a combined ontology and parametric model by thoroughly examining their interrelations and connection points.
http://130.149.22.198/wp/?page_id=7769 

For this project traffic infrastructure has been researched. The civil systems individually chosen by the group members were two steel bridges and two subsea tunnels.

Traffic infrastructure consists of complex civil systems designed for a long lifetime. Tunnels and bridges compose important supporting systems within traffic infrastructure, which improve accessibility and transportation efficiency. These are however components that are subject to degradation due to both outer and inner factors. Among outer factors traffic loads must be addressed. Inner factors account for such as the main construction materials, which for the chosen systems are concrete and steel.

The chosen civil systems are naturally connected within traffic infrastructure. For the integration, the systems are considered as parts of a roadway. The systems are considered to be physically connected, and the connection is illustrated below. This simplifies the integration, as the outer factors leading to degradation that are related to traffic can be considered equal for the systems. The combined system considered is a tunnel followed by a bridge. The tunnel length is set to 1000 m, whereas the bridge length is set to 300 m.
http://130.149.22.198/wp/?page_id=7796

ABSTRACT

Due to the advantageous criteria like high strength, durability and availability in Bangladesh of jute and carbon fibre reinforced polymer (JFRP & CFRP), they are commonly used to enhance the column properties. In this study unidirectional and untreated jute and carbon fibre has reinforced the composites with the matrix material epoxy. Hand lay-up techniques were used to construct total 18 specimens having various shapes, constituent coarse aggregate and no. of layers of JFRP and CFRP. These columns were subjected to cyclic loading until it reaches the ultimate stress and fails. The data from Kaleida graph shows the effect clearly that confined columns are strengthened by unidirectional JFRP which has given an adequate improved capacity of resisting cyclic loading with the increment of no of layers of FRP. The procedure is made by the guidelines for the design and construction of externally bonded FRP systems for strengthening concrete columns (ACI 440.2R-08).