⚡ electrical engineering

Cyber-Physical Seismic Resilience: Integrating Digital Twins and Machine Learning for Performance-Based Earthquake Engineering of CLT and Masonry Structures in High-Seismicity Regions — Lessons from the USA, Japan, and Proposals for Peru

This study proposes a transformative Performance-Based Earthquake Engineering framework that integrates Digital Twins and Machine Learning to enhance the seismic resilience of Cross-Laminated Timber and masonry structures in Peru by benchmarking against US and Japanese practices, introducing novel analytical tools for uncertainty reduction, and bridging the gap between construction-phase modeling and real-time operational monitoring.

Original authors: PAUL RICARDO PRUDENCIO GALVEZ

Published 2026-06-26

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: PAUL RICARDO PRUDENCIO GALVEZ

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are building a house in a place where the ground shakes violently, like Peru. For a long time, builders have followed a strict "recipe book" (prescriptive rules) to make sure the house doesn't fall down. But this paper suggests we need a smarter approach: a "performance-based" system that learns from the house's actual behavior, not just the recipe.

Here is the core idea of the paper, broken down into simple concepts and analogies:

1. The Problem: The "Recipe Book" vs. The "Live Pilot"

Peru sits on a very active tectonic plate, making earthquakes a constant threat. Currently, the country relies on older building rules that don't account for new, high-tech materials like Cross-Laminated Timber (CLT) (think of it as super-strong, glued-together wood panels) or how to best use confined masonry (brick walls with concrete frames).

While countries like the USA and Japan have moved on to using "Digital Twins" and "Machine Learning" to predict how buildings will survive earthquakes, Peru is still catching up. The paper argues that Peru needs to stop guessing and start using data to build safer, smarter structures.

2. The Solution: The "Digital Twin" (The Building's Shadow)

The paper introduces the concept of a Digital Twin.

The Analogy: Imagine a building has a perfect, invisible "shadow" floating in the cloud. This shadow is a computer model that looks exactly like the real building.
How it works: As the real building is being built and later as it sits there, tiny sensors (like a fitness tracker for a building) send data to the shadow. If the real building sways, the shadow sways too. If a screw loosens, the shadow knows immediately.
The "Construction Phase" Twist: The author proposes starting this "shadow" while the building is being manufactured. Sensors are embedded in the wood panels before they even leave the factory. This ensures the shadow is perfectly calibrated to the real building from Day One, rather than trying to guess what the building is like after it's finished.

3. The Brain: Machine Learning (The Super-Intelligent Assistant)

A Digital Twin is useless without a brain to interpret the data. That's where Machine Learning (ML) comes in.

The Analogy: Think of the ML as a super-smart assistant who has read millions of books on how buildings behave during earthquakes.
What it does: Instead of running slow, expensive computer simulations every time there is a tremor, this assistant uses a "Random Forest" (a type of AI) to instantly predict how much the building will sway. It's like having a weather forecaster who can tell you exactly how much rain will fall in your specific neighborhood, instantly, rather than waiting for a general forecast for the whole country.
The Result: The paper claims this AI can predict building movement with over 94% accuracy, which is a huge leap forward.

4. The New Rules: Updating the "Scorecard"

The paper proposes four specific new formulas to replace or upgrade the old ways of calculating risk:

Formula 1: The "Uncertainty Discount."
- Old way: We assume the worst-case scenario because we aren't sure exactly how the building will act, so we overestimate the cost of damage.
- New way: As the Digital Twin gathers more real data, the AI becomes more confident. The paper introduces a factor (called αML) that lowers the estimated cost of damage as we learn more. It's like getting a discount on your insurance premium because you've proven you are a safe driver.
Formula 2: The "Rocking Wall" Balance.
- For wooden (CLT) buildings that are designed to rock slightly during an earthquake (like a tree bending in the wind), the AI constantly updates the math to account for how the metal connectors are wearing out or how the tension in the cables is changing.
Formula 3: The "Vibration Health Check."
- Just as a doctor listens to your heartbeat, the Digital Twin listens to the building's natural vibration. If the pitch of the vibration changes, the AI knows the building is damaged, even if you can't see it with your eyes.
Formula 4: The "Fast-Forward" Predictor.
- Instead of waiting hours to run a complex earthquake simulation, the AI gives an instant answer on how bad the shaking will be for the building's stories.

5. The Roadmap for Peru

The paper doesn't just dream; it offers a step-by-step plan for Peru:

Years 1–2: Build a database. Test 50+ wooden connections and run thousands of computer simulations specifically for Peruvian cities like Lima and Cusco.
Years 2–3: Build a pilot project. Construct a 6-story wooden building in Lima with sensors embedded in the walls during construction to test the "Digital Twin" system.
Years 3–5: Expand the system to brick (masonry) buildings in other areas and create a free, open-source software platform for Peruvian engineers.
Year 5+: Update the national building codes (NTP E.030) to officially include these new, data-driven methods.

The Bottom Line

The paper concludes that by combining Digital Twins (the real-time shadow) and Machine Learning (the super-brain), Peru can reduce the uncertainty about earthquake damage by up to 35%. This means better decisions on where to invest money, safer buildings, and a move away from outdated rules toward a future where buildings "talk" to engineers to keep everyone safe.

Technical Summary: Cyber-Physical Seismic Resilience in Peru

Problem Statement

Peru faces a critical seismic vulnerability due to its location at the convergence of the Nazca and South American tectonic plates, evidenced by historical events such as the 2007 Pisco (Mw 8.0) and 1970 Ancash (Mw 7.9) earthquakes. Despite this high hazard, the Peruvian construction sector remains constrained by prescriptive design methodologies (NTP E.030/E.090) that lack Performance-Based Earthquake Engineering (PBEE) frameworks. While Cross-Laminated Timber (CLT) and confined masonry offer potential for seismic resilience, their implementation in Peru is hindered by a lack of real-time monitoring, experimental databases, and regulatory provisions. Furthermore, the transition to PBEE is computationally prohibitive due to the reliance on expensive non-linear time-history analyses (NLTHA), and there is a significant gap in integrating Digital Twins (DT) and Machine Learning (ML) to reduce epistemic uncertainty in loss estimation.

Methodology

The study employs a Systematic Literature Review following standardized engineering research synthesis guidelines. The process involved four phases:

Identification: A Boolean search in Scopus and Web of Science yielded 412 records using terms related to Digital Twins, Machine Learning, CLT, Masonry, and Seismic/PBEE.
Screening & Eligibility: After removing duplicates and reviewing titles/abstracts, 84 records were assessed.
Inclusion: 38 high-impact studies were selected for data extraction, focusing on DT/ML applications in structural engineering with a seismic focus on CLT or masonry.

The paper synthesizes these findings to benchmark global practices (USA, Japan) against the Peruvian reality and proposes a novel, modernized PBEE framework. This framework integrates a Construction-Phase Digital Twin (CP-DT) architecture that bridges Building Information Modeling (BIM) with operational Structural Health Monitoring (SHM).

Key Contributions and Proposed Framework

The paper introduces a modernized PBEE framework tailored for Peru, centered on the CP-DT and four specific technical innovations:

ML-Enhanced Expected Annual Loss (EAL):
The standard FEMA P-58 loss integral is augmented with an ML-derived uncertainty reduction factor, $\alpha_{ML}$ (where $\alpha_{ML} \in (0, 1]$ ). As the Digital Twin accumulates sensor data, $\alpha_{ML}$ is continuously calibrated by a Random Forest model, reducing epistemic uncertainty and lowering the estimated EAL from its inflated baseline.
- Formula: $EAL_{DT} = \int_0^\infty [\sum P(DS_j | EDP) \cdot C_j] \cdot \alpha_{ML} \cdot d\lambda(IM)$
Kinematic Formulation for CLT Rocking Walls:
A restoring moment formulation for CLT rocking walls with post-tensioning (PT) and energy dissipators is updated via the DT. The DT uses ML regression on strain gauge data to update long-term prestress loss ( $T_{pt,0}$ ) and fatigue-affected yield forces ( $F_{hd,y}$ ), ensuring the model reflects the "as-built" physical reality.
Frequency-Based Damage Index ( $DI_{CLT}$ ):
A global structural damage index is computed using ambient vibration data from the DT. It compares natural frequencies derived from Stochastic Subspace Identification (SSI) against theoretical BIM frequencies.
- Formula: $DI_{CLT} = 1 - \frac{1}{N} \sum_{i=1}^N (\frac{f^n_{DT,i}}{f^n_{BIM,i}})^2 \cdot w_i$
- Scale: 0 indicates undamaged; approaching 1 indicates severe degradation.
Random Forest Surrogate for Rapid Prediction:
To bypass the computational cost of full NLTHA, a Random Forest (RF) surrogate model maps seismic and structural parameters (e.g., spectral acceleration, period, mass ratio) to Maximum Inter-Story Drift Ratio (MIDR). The model is trained on a synthetic database of 5,000+ NLTHA records specific to Peruvian hazard conditions.

Results

Comparative Maturity: A comparative matrix highlights that while the USA and Japan possess advanced PBEE integration, real-time loss estimation, and extensive experimental databases, Peru currently relies on basic prescriptive codes with no PBEE or DT implementation in timber/masonry.
Model Performance: The proposed Random Forest surrogate achieved a coefficient of determination $R^2 > 0.94$ in cross-validation for MIDR prediction, demonstrating high fidelity in predicting structural demand.
Uncertainty Reduction: The integration of DT and ML is projected to reduce epistemic uncertainty in seismic loss estimation by up to 35%.
Implementation Roadmap: The paper outlines a four-phase roadmap for Peru:
- Phase 1 (Years 1–2): Establish a Peruvian CLT Seismic Database (50+ cyclic tests, 5,000+ NLTHA records).
- Phase 2 (Years 2–3): Deploy a CP-DT pilot on a 6-story CLT building in Lima with embedded MEMS sensors.
- Phase 3 (Years 3–5): Extend CP-DT to confined masonry and develop an open-source platform compatible with ETABS and OpenSees.
- Phase 4 (Year 5+): Inform regulatory revisions (NTP E.030/E.090) with performance-based provisions.

Significance and Claims

The paper asserts that the integration of Digital Twins and Machine Learning into PBEE for CLT and masonry is a practically achievable objective with immediate benefits for seismic risk reduction in Peru. Its primary significance lies in:

Bridging the Gap: Providing a quantified gap analysis and a concrete technology transfer strategy from mature markets (USA/Japan) to the Peruvian context.
Regulatory Modernization: Offering a technical basis for updating Peruvian regulations (NTP E.030/E.090) to include performance-based design and digital monitoring requirements.
Economic Efficiency: Demonstrating that reducing epistemic uncertainty through DT-ML integration leads to more accurate, risk-informed investment decisions, which is critical for a nation with high seismic vulnerability and limited resources.
Lifecycle Calibration: Establishing a CP-DT architecture that ensures the digital model is calibrated to physical reality from the manufacturing and assembly stages, rather than relying solely on post-construction assumptions.

The study concludes that this approach transforms the paradigm from prescriptive compliance to performance-based resilience, directly addressing the challenges of local materials, construction practices, and economic constraints in Peru.