This project focuses on studying the seismic behavior and failure mechanism of a new construction method that is based on combining TiABs with UHPC, namely ultra-high performance concrete structures reinforced with titanium bars (TARUHPC), to build bridges with better durability. To this end, various experiments and computer simulations are carried out to describe the interaction between these materials, structural response, and identification of failure modes of structural elements with TARUHPC.  Then, various detailed finite element models will be developed to characterize the bond strength between UHPC and TiABs, and to determine the seismic capacity of a bridge pier made of TARUHPC. In addition,  Machine Learning are used to predict the seismic response of these structures.  The results of the project will be synthesized in guidelines for the design and development of durable bridges with TARUHPC. This project hopes to validate a novel construction technique for durable (useful life of more than 100 years) and resilient infrastructure, which represents a milestone of innovation in the country's construction sector.

• PI: Luis Bedriñana

• Partners: Collaborative project with Idaho State University

• Funding from: UTEC Seed Fund RFP 2023

• Keywords: UHPC, Titanium alloy bars, seismic performance, bond strength, material properties, finite element, numerical simulations, Machine Learning.

This project proposes an integration of drone path automation with state-of-the-art AI algorithms for the automatic evaluation and report of surface damages on RC bridges. The main objective of this research is to propose a scalable, low-cost, system for the inspection and assessment of superficial damage in RC bridges using UAVs. To this end, the project will focus on two main technical challenges: (1) the automatic damage segmentation and properties retrieval of concrete structures using 2D images, and (2) the flight path optimization of UAVs for data collection of full-size bridges. The resulting system can automate the inspection process of bridges, reducing time, effort, and costs. Consequently, this project promises to revolutionize the way we maintain bridges, ensuring the safety and durability of our civil infrastructure.

• PI: Luis Bedriñana

• Partners: University of Alberta

• Funding from: University of Alberta - UTEC Faculty Grant

• Keywords: concrete structures; bridges; unmanned aerial vehicle; damage detection; machine learning; convolutional neural networks; deep learning; semantic segmentation; path automation.

This project uses various approaches to characterize the tensile behavior of concrete reinforced with steel and polypropylene fibers with partial replacement of coarse aggregate by using recycled sanitary ceramic. To investigate the differences in the tensile behavior, tests (e.g., 3-point bending) of different concrete mixtures will be carried out. With the experimental results, different simulations (finite element) will be carried out to delve into the failure mechanisms and residual capacity of the different concretes mixes, establishing recommendations for design variables. Similarly, different data-based predictive models are proposed for the tensile capacity of recycled concrete. Given that the amount of data with the studied materials in the project is limited, Transfer Learning techniques will be used to learn from the behavior of other concretes with greater data available (e.g. normal steel fiber reinforced concrete) and apply this knowledge to predictions in concretes with recycled aggregates and reinforced with fibers. This multi-focus strategy will not only improve the understanding of the mechanics of sustainable concrete materials, but can be useful in proposing practical tools for structural applications. The results of the project will contribute to the state of the art in sustainable concrete materials and will serve as a basis to produce more sustainable concrete with better mechanical performance. Likewise, practical tools will be proposed for the design of structures with these materials.

• PI: Luis Bedriñana / Carlos Benedetty

• Funding from: UTEC Seed Fund RFP 2024

• Keywords: concrete structures; concrete; steel fibers; polypropylene fibers, ceramic sanitary waste; recycled aggregates; sustainability; FE analysis; inverse analysis; transfer learning.

Heritage Building Information Modeling (HBIM) is currently a widely used strategy for conserving and managing heritage buildings. One key aspect of this approach is the scan-to-BIM procedure, in which a significant amount of geometrical data must be collected from a heritage building to be processed for 3D reconstruction. Manual annotation of elements from 3D point clouds is a time-consuming, complex, subjective, and expert knowledge-dependent process, especially in heritage buildings involving irregular geometries and juxtaposed classes. Then, different methods are needed for the annotation of point clouds for HBIM projects. This project discusses a framework for the automation of semantic segmentation of 3D point clouds for heritage buildings. Deep Learning models are trained to predict structural and non-structural classes in 3D point clouds from real heritage data. In addition, the segmentation and annotation time were compared between our methodology and the traditional protocols in HBIM projects, indicating the potential of the proposed framework to improve productivity in heritage projects. 

• PI: Diana Castillo

• Funding from: Thoma Fundation

• Keywords: Deep Learning; HBIM; Semantic Segmentation; Scan to BIM; heritage buildings; 3D point cloud.