Long before algorithms, simulations, and seismic codes existed, traditional Indian architecture had already mastered resilience. Today, at the Earthquake Engineering Research Centre (EERC) at IIIT Hyderabad, researchers are revisiting that wisdom through AI, drones, computer vision, and machine learning to understand how structures survive disasters. From studying heritage buildings and earthquake-resistant vernacular homes to developing AI-based crack detection systems and open-source structural assessment tools, the centre is building a bridge between ancient engineering intuition and modern computational science. In conversation with assistant professor Dr Jofin George, CE explores how technology and traditional knowledge are redefining structural safety and earthquake resilience in India.
Excerpts
What is the central problem the EERC is trying to solve today?
At its core, EERC is trying to answer one question that lies at the intersection of engineering and public responsibility: How do we ensure the safety of our built environment before the next earthquake strikes, not after?
India’s infrastructure reflects its cultural history, from ancient stone temples and colonial masonry arch bridges to post-independence concrete housing and rural masonry buildings. Many of these structures are located in moderate to high seismic zones and require proper evaluation of their earthquake vulnerability.
Our approach follows a dual philosophy. On one hand, we use drones, computer vision, and machine learning to make structural assessment faster and scalable. On the other, we study what ancient master builders understood intuitively and apply those lessons to modern disaster management and heritage conservation.
How does blending traditional building wisdom with modern computational tools change the way we understand structural safety?
This understanding changes structural safety from a simple load-and-resistance approach to one focused on energy dissipation mechanisms that help structures survive earthquakes.
Take Bhonga houses in Kutch, Gujarat. These cylindrical mud huts have no steel reinforcement, yet their circular geometry prevents stress concentration and crack initiation. It is an elegant physics-driven design refined through generations.
Another example is the Kath Kuni construction of Himachal Pradesh, where alternating layers of timber and stone interlock without mortar. The system resists earthquakes by dissipating energy effectively.
Computational tools now allow us to quantify what traditional knowledge achieved intuitively. We can create numerical and kinematic models, run limit analysis, and evaluate their performance probabilistically. The combination of traditional wisdom and modern tools offers a richer understanding of structural safety.
Why are heritage structures a key focus in your research on earthquake resilience?
Heritage structures have already survived centuries, often in high seismic zones, because of their robust geometries and structural forms. By studying them, we identify material and geometric properties that enhance resilience and apply those lessons to modern construction.
These structures are also culturally irreplaceable. Unlike modern buildings, a medieval temple cannot simply be reconstructed once lost.
Our colleague Dr P Pravin Kumar V Rao leads retrofitting research using textile/wire-reinforced mortar composites, natural fibres, and fibre reinforced polymer composites to strengthen heritage buildings without compromising their historical character. Such interventions aim to improve seismic resilience while preserving architectural integrity.
What makes arches such an interesting structural system to study through the ‘hanging chain’ principle?
An arch is one of the simplest structural forms derived from nature. In 1675, Robert Hooke observed that a freely hanging chain forms a catenary, where every link is under pure tension. When inverted, the same shape becomes a perfect arch carrying pure compression.
This allows materials strong in compression, such as brick, stone, and lime mortar, to create stable structures. If the line of thrust, or path of compressive forces, stays within the arch cross-section, the arch remains safe.
At EERC, PhD student Kanukuntla Rajkumar is studying 3D collapse modes in masonry arches. During IIIT Hyderabad’s annual R&D showcase, the principle was demonstrated using a voussoir model and hanging chain.
What inspired the development of an open-source tool for calculating thrust lines in arches?
The inspiration came from the gap between theory and practice. Although thrust-line analysis is centuries old, applying it to complex arch geometries usually requires computation beyond the reach of many practicing engineers and conservationists.
Students from the MTech CASE programme built an open-source GitHub tool to make theoretically grounded structural assessment accessible and peer-reviewable at no cost. Structural assessment of heritage infrastructure, we believe, is a public good.
How does this tool change the way engineers approach structural evaluation in the field?
The tool is still a pilot project, but it demonstrates that rigorous assessment methods can be made accessible without expensive software.
Engineers can input arch geometry and loading conditions and instantly visualise the thrust line instead of performing calculations manually. This allows rapid assessment of old arch bridges and gateways while showing how student-led research can produce practical engineering solutions.
In what ways does automation make structural assessment more accessible compared to traditional methods?
Automation changes how expert time is used. Instead of spending days on visual surveys, engineers can focus on interpretation and decision-making.
Traditional structural assessment is labour intensive and subjective. Automation supports a ‘scan-to-analysis’ pipeline where drone surveys capture precise 3D geometry and create digital twins that can be used for multiple simulations and evaluations.
How are drones transforming the way structural data is collected and analysed?
Drones provide safe and consistent access to locations that inspectors cannot easily reach, such as the intrados of arch bridges or upper façades of heritage structures.
At EERC, a MeitY-sponsored project on drone-enabled kinematic screening of masonry structures uses drone surveys to capture 3D geometry and convert digital models into kinematic analysis models. Deriving analysis models directly from point-cloud data reduces uncertainty significantly.
How do computer vision systems interpret structural damage in real-world conditions?
Real-world structural damage presents major challenges because surfaces may contain vegetation, weathering stains, repair patches, or distorted camera perspectives. Models trained only on clean datasets perform poorly in such environments.
To address this, our datasets deliberately include real-world variations so the system learns robustness. The geometric detection output is then transferred to a RAG engine for interpretation. Instead of one model handling both detection and reasoning, separate components perform specialised tasks. The result is a system designed to support engineers with evidence-based preliminary analysis.
What level of accuracy can AI-based crack detection systems achieve compared to manual inspection?
AI-based crack detection systems are not substitutes for manual inspection; they are tools that support field engineers.
The computer vision system developed by undergraduate students Aniket Gupta and Samarth Srikar for unreinforced masonry buildings can identify, map, and quantify cracks with millimetre-level precision. Entire façades can be analysed much faster than through manual inspection.
The real advantage lies in consistency. Human inspectors may disagree on crack characteristics or causes, whereas AI systems trained on standardised datasets provide more uniform analysis.
At EERC, crack data is linked to a retrieval-augmented generation engine grounded in FEMA 306 guidelines. This allows the system not only to detect cracks but also to infer probable causes and suggest recommendations, moving from observation to diagnosis.
What challenges come with integrating AI and robotics into traditional civil engineering workflows?
The biggest challenge is the multidisciplinary nature of combining AI, robotics, and civil engineering. Experts from all three domains must work together, though mathematics provides a common language.
Another challenge is that civil engineers dislike ‘black-box’ systems. Engineering decisions require transparent reasoning. This motivated our adoption of an RAG-based framework grounded in established standards like FEMA 306. We need physics-informed systems, not just statistically trained models.
There is also the challenge of trust and adoption. Engineering codes evolve slowly and incorporate lessons from real failures. AI tools must demonstrate not only effectiveness but also clearly understood limitations and failure modes before they are accepted widely.
What is the next big leap you are aiming for in earthquake engineering research?
My work is moving toward integrating drone surveys, computer vision diagnostics, and machine-learning-based collapse prediction into a unified pipeline deployable before earthquakes occur.
At the same time, Dr Sunitha Palissery studies the seismic performance of modern structural configurations and machine-learning approaches for classifying earthquake signals. Dr Pravin Kumar Venkat Rao is advancing minimally invasive retrofitting for heritage structures, while Dr Shubham Singhal focuses on precast, prestressed, and modular construction systems and their machine-learning applications.
Underlying all this is the vision of Prof Pradeep Kumar Ramancharla, who insists that research must reach communities through awareness tools, engagement, and accessible guidance for people building in seismic zones without formal engineering support.
The next leap is not a single technology, but the convergence of all these efforts into systems that genuinely reach the communities that need them most.
Across all these initiatives, the larger ambition remains preparedness rather than post-disaster reaction. The centre’s research attempts to democratise structural safety by creating affordable, open, and scalable tools that can assist engineers, conservationists, policymakers, and local communities alike. Whether through heritage retrofitting, drone-based surveys, AI-assisted diagnostics, or traditional construction insights, EERC’s work reflects a broader shift in engineering: resilience must be designed into structures long before the ground begins to shake beneath them.