Research

The problems we work on

Each problem below is stated the way an owner agency meets it. Under it: how we approach it, what our published work has found, where the limits are, and the papers to read.

Inspections cannot keep up with the number of ageing bridges

Joint work with Chulalongkorn University.

Visual inspection is slow, inconsistent between inspectors, and cannot reach every surface. Thailand's bridge stock is large and much of it is decades old.

Published work

Approach

UAV surveys that deliver high-resolution imagery of the structure, and nonlinear finite-element analysis that turns the observed damage into a structural assessment. AI damage detection is being developed in the NRCT projects.

What the work found

On an RC bridge in service for more than seventy years, UAV imagery located the flexural-shear cracks precisely enough to compare them with the crack patterns from nonlinear finite-element analysis. The study hypothesized that the cracks came from fatigue under repeated loading, and showed the potential of UAV surveys to cut inspection time and cost.

Limits

Image-based methods see surfaces. Internal defects, corrosion behind sound cover and prestressing losses still need non-destructive testing and sensors.

Papers

UAV-Enhanced Analysis of Fatigue-Induced Damage in RC Bridge Girders Using Nonlinear Finite Element Methods (IABSE Symposium, Tokyo 2025: Environmentally Friendly Technologies and Structures: Focusing on Sustainable Approaches, 2025)

The bridge on the drawings is not the bridge in the field

Joint work with Chulalongkorn University and the Asian Institute of Technology.

Assessments are usually based on design drawings. After decades of service, geometry, stiffness and behaviour have all drifted from what was drawn.

Published work

Approach

3D LiDAR scanning for as-built geometry, ultrasonic pulse velocity for in-place stiffness, accelerometers and dynamic strain gauges under traffic, and a digital twin built on the as-is data rather than the as-designed data.

What the work found

For a balanced cantilever bridge, the point-cloud model exposed discrepancies between as-designed and as-built conditions; modal frequencies, mode shapes and neutral-axis depth extracted from the measurements agreed well with the as-is twin, so dynamic strain and neutral-axis parameters can serve as sensitive indicators of local damage.

Limits

One structure and a baseline campaign. Long-term monitoring and repeated surveys are needed before change detection becomes routine.

Papers

Digital twin-based structural health monitoring and measurements of dynamic characteristics in balanced cantilever bridge (Resilient Cities and Structures, 2025)

Corrosion is not uniform, so an average understates the risk

Joint work with Waseda University and Lehigh University, and on individual papers with Tongji University, Kanazawa Institute of Technology and Central Nippon Expressway. Builds on the director's doctoral research at Waseda University.

Chloride-induced corrosion varies along and across every bar. In prestressed beams, capacity is governed by the worst local loss, which an average measurement never shows.

Published work

Approach

Corrosion measured along the bars and strands of RC and prestressed specimens; random-field models fitted to the measured spatial patterns; 3D nonlinear finite-element analysis of members with simulated corrosion.

What the work found

Transverse correlation of steel weight loss among multiple tensile bars was quantified experimentally and shown to change the failure probability of RC girders. In prestressed beams, capacity and ductility were governed by the maximum local weight loss of a single wire in the maximum-moment region, and prestress widened corrosion-induced cracks.

Limits

Accelerated corrosion in laboratory specimens. Field corrosion is slower and its spatial statistics may differ.

Papers

Effect of the interaction of corrosion pits among multiple tensile rebars on the reliability of RC structures: Experimental and numerical investigation (Structural Safety, 2021)
Effects of spatial corrosion distribution and prestressing levels on the structural performance of deteriorated PC beams (Construction and Building Materials, 2024)
Investigation of corrosion-induced cracks using corrosion products quantified by an X-ray technique and FE analysis of single- and multiple-rebar beams (Cement and Concrete Composites, 2024)
A novel combined experimental-machine learning approach to estimate the probabilistic capacity of RC beams with spatially correlated rebar corrosion in transverse and longitudinal directions (Engineering Structures, 2023)

How should an inspection change our confidence in a structure?

From the director's doctoral research at Waseda University, with colleagues at Lehigh University, Tokyo City University and Central Nippon Expressway.

Inspections produce local observations, such as crack widths at a few points. Reliability estimates need to absorb them without pretending they describe the whole member.

Published work

Approach

Random-field models of corrosion-related variables and a particle filter that updates them from spatially observed data, giving time-dependent reliability that reflects what was actually seen.

What the work found

In an example RC bridge girder updated from crack-width observations, the accuracy of the time-dependent reliability estimate improved as more locations were observed; the effect of the spacing between locations was also examined.

Limits

Illustrative example on one member type. Linking the framework to routine inspection records is part of the flagship program.

Papers

Random field-based reliability updating framework for existing RC structures incorporating the effect of spatial steel corrosion distribution (Structure and Infrastructure Engineering, 2021)

From one bridge to a network: which structure, which repair, in what order?

Budgets and engineers are limited. Decisions have to be made across hundreds of structures at once, on comparable data.

In progress

Approach

Bridge Intelligence: the full chain from survey to decision on a pilot bridge, linked to network-level prioritization, with data standards that owner agencies can adopt, under the NRCT digital twin project started in 2026.

What the work found

In progress. Results will be reported here as they are published.

Four research areas, implemented in phases

ARIES organizes its work in four research areas, implemented in phases. Two are active today and match the founding groups' expertise; two open as the center grows and colleagues from related fields join.

Deep Tech and Intelligent Systems

Active

Structural inspection with artificial intelligence, digital twins, unmanned aerial vehicles and LiDAR; autonomous inspection pipelines; computer vision for classifying structural damage.

Structural Systems

Active

Condition assessment and management of infrastructure systems from the component to the network level, covering bridges, buildings, dams and transport networks.

Resilience and Multi-Hazard Assessment

Next phase

Integrated risk assessment and structural reliability covering earthquake, deterioration, corrosion and climate-change impacts, together with life-cycle management of structures.

Advanced Materials and Durability

Next phase

Advanced materials such as graphene quantum dots and ultra-high performance concrete, computational deterioration modelling, and low-carbon concrete.

Research groups

Structural Assessment and Digital Twin

Structural assessment, digital twins and deterioration modelling of structures.

Lead: Asst. Prof. Dr. Supasit Srivaranun, Director

Geospatial and Field Operations

Unmanned aerial vehicles, LiDAR and field survey.

Lead: Asst. Prof. Kittipos Prapassorn

Further groups covering multi-hazard assessment and advanced materials are planned as the center expands.

Flagship program 2026 to 2028

Bridge Intelligence

Bridges are the proving ground for the first three years. The program builds a complete inspection-to-decision system: UAV and LiDAR survey, AI damage detection, a digital twin of each structure, condition assessment, and prioritization of maintenance across the whole network.

What is different: prioritization at the network level rather than bridge by bridge; an engineer in the loop at every quality-control step; and data standards that owner agencies can adopt directly, so results move from research into practice.

Survey

UAV and LiDAR

Image and point-cloud capture of the whole structure.

Detect

AI damage detection

Computer vision classifies cracks, spalling and corrosion.

Model

Digital twin

Geometry, damage and behaviour in one living model.

Assess

Condition assessment

Structural evaluation with human-in-the-loop quality control.

Decide

Network prioritization

Which bridge, which repair, in what order.

Designed for the agencies that own and maintain Thailand's bridges. It is at the research stage: the chain is being built and tested on a pilot bridge under the NRCT project.

Why this program, and where it leads →