⚡ electrical engineering

A predictive machine-learning and topology-optimization framework for the seismic retrofit of vulnerable housing with UHPC overlays and FRP nanocomposites: transposing the United States experience to Peru

This paper proposes an integrated framework that combines machine learning and topology optimization with UHPC and FRP-nanocomposite overlays to optimize the seismic retrofit of Peru's vulnerable informal housing, while adapting U.S. and Japanese best practices and standards to the local Peruvian context.

Original authors: PAUL RICARDO PRUDENCIO GALVEZ

Published 2026-07-06

📖 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 a house in Peru as a delicate, handmade clay pot. It's built by hand, often without a master plan, and it sits in a region where the ground shakes violently like a giant, unpredictable drum. When that drum beats hard (an earthquake), this pot is likely to crack or shatter.

This paper proposes a high-tech "armor" to save these pots, but instead of just wrapping them in thick, heavy blankets, it uses a smart, computer-designed suit made of super-materials. Here is how the author breaks it down:

1. The Problem: The "Leaky Roof" of Safety

In Peru, about 7 out of 10 homes are built informally. They aren't designed by engineers. If a massive earthquake hits (which scientists say is statistically likely), over a third of these homes could be destroyed.

The Analogy: It's like trying to stop a flood with a sieve. The current safety nets (building codes) exist, but they haven't caught up with the reality of how these homes are actually built.

2. The Solution: A "Smart Suit" for Houses

The author suggests a new way to fix these homes using two special ingredients:

UHPC (Ultra-High-Performance Concrete): Think of this as "liquid steel." It's a concrete mix so strong and dense that it can be poured in very thin layers but holds immense power.
FRP Nanocomposites: Imagine wrapping the house in a high-tech, carbon-fiber tape (like the kind used on race cars or tennis rackets) that is reinforced with tiny, invisible particles (nanotechnology) to make it even tougher.

3. The "Brain" of the Operation: Two Super-Tools

The paper doesn't just say "put this stuff on the wall." It uses two advanced computer tools to figure out exactly where and how to put it:

Tool A: The Crystal Ball (Machine Learning)
- What it does: Instead of guessing how strong the new concrete will be, the computer looks at thousands of past experiments (like a student studying for a test) to predict the strength with 93% to 99% accuracy.
- The Analogy: It's like a weather app that doesn't just guess if it will rain, but predicts the exact amount of rain based on historical data, so you know exactly how big an umbrella you need.
Tool B: The Sculptor (Topology Optimization)
- What it does: This tool acts like a digital sculptor. It takes a block of material and "eats away" everything that isn't needed, leaving only the most efficient shape to hold the weight.
- The Analogy: Imagine a stone mason carving a statue. They don't just chip away randomly; they remove stone from the middle to leave a strong, hollow arch that uses the least amount of stone to hold the most weight. This tool does that for the concrete overlay, creating diagonal "struts" (like the bones in a bird's wing) that guide the earthquake forces safely through the wall.

4. The Plan: Borrowing from the Best

The author looks at how other countries handle this and creates a "recipe book" for Peru:

The United States: They have a massive program (GRRP) that spends billions of dollars to fix homes, using computers to decide which houses are most at risk.
Japan: They have strict, detailed rules (like a master chef's recipe) for how to apply these high-tech materials, ensuring the job is done perfectly.
Peru: Currently, Peru has the "ingredients" (laws and some funding) but lacks the "recipe" for using these specific high-tech materials on informal homes.

The Paper's Goal: It takes the US "brain" (using computers to prioritize and predict) and the Japanese "hands" (strict construction steps) and adapts them for the Peruvian context.

5. The Result: A Lighter, Smarter Fix

By using this computer-designed approach, the paper claims you don't need to cover the entire wall in expensive material.

The Analogy: Instead of wrapping a broken arm in a cast that covers the whole body, you put a cast only on the broken bone and the joints that need support.
The Benefit: This saves money (using up to 70% less material in some cases) while keeping the house just as safe.

What the Paper Doesn't Claim

It is important to note what this paper is not doing:

It is not a report on a house that was actually built and tested in the real world yet. It is a "blueprint" and a mathematical proof.
It is not saying this will happen tomorrow. It admits that local materials need to be tested first and that the construction steps need to be tried in a lab before they can be used on every street in Peru.
It is not a medical study; it is strictly about engineering and math.

In Summary:
This paper is a proposal to stop guessing about earthquake safety. It suggests using a "smart computer" to design a custom, lightweight, super-strong armor for Peru's vulnerable homes, borrowing the best safety rules from the US and Japan, and adapting them to fit the local reality. It's about moving from "hope for the best" to "calculated safety."

Technical Summary: A Predictive ML and Topology-Optimization Framework for Seismic Retrofit in Peru

Problem Statement
Approximately 70% of dwellings in Peru are constructed informally without engineering design, predominantly using confined masonry with compressive strengths below the minimums required by the NTE E.070 standard. Located in a high-seismicity zone where a magnitude $M_w \approx 8.0$ event is statistically expected, these structures face severe vulnerability; under an intensity-IX scenario, over 32% could suffer Grade-4 damage. While the United States has advanced retrofit programs (e.g., GRRP) utilizing machine learning (ML) and machine-learning-assisted prioritization, and Japan has codified Ultra-High-Performance Concrete (UHPC) and Fiber-Reinforced Polymer (FRP) guidelines (JBDPA-2001), Peru lacks a nationwide retrofit program specifically for vulnerable housing and an explicit normative pathway for UHPC and FRP-nanocomposites. Furthermore, existing empirical design equations for these materials often under-predict capacity compared to ML models trained on large datasets, and topology optimization (TO) remains largely absent from masonry retrofit design.

Methodology
The study employs a transdisciplinary technical-synthesis design comprising three pillars:

Comparative Analysis: An eleven-row quantitative matrix was developed to contrast seismic retrofit practices in the United States, Japan, and Peru. This analysis covers regulatory frameworks, target housing typologies, vulnerability indices, dominant technologies, and the application of ML algorithms (specifically Random Forest, XGBoost, and Gradient Boosting Regression Trees).
Topology Optimization (TO): A two-dimensional plane-stress problem was formulated for a typical 2.40 m $\times$ 2.40 m confined-masonry wall. Using the Solid Isotropic Material with Penalization (SIMP) method, the study minimized strain energy subject to a 30% volume constraint on the UHPC overlay. The domain was discretized into a 96 $\times$ 96 mesh, with a 50 mm passive zone around tie-columns. The optimization utilized an optimality-criteria method with a sensitivity-filter radius of 1.5 element widths.
Predictive Modeling and Structural Equations: The framework integrates ML predictions for material strength with four specific structural engineering equations:
- SIMP Penalization: To drive density values toward binary states (0 or 1).
- ACI 440.2R: For calculating FRP contribution to nominal shear capacity.
- Mander-Priestley-Park Model: For determining the confined concrete strength ( $f'_{cc}$ ) of UHPC-jacketed columns.
- NTE E.030-2018: For computing the pseudo-acceleration spectrum ( $S_a$ ) for seismic design in Lima.

Key Results

Comparative Matrix: The analysis reveals that while the U.S. leads in financing scale and Japan in codified technological depth (particularly UHPC use), Peru lags in ML adoption, UHPC codification, and the use of nanotechnology-modified FRP. The predictive accuracy gap (ML $R^2$ values) is identified as the most actionable area for immediate improvement.
Optimization Outcome: The SIMP-optimized UHPC overlay converged in approximately 47 iterations. The resulting geometry features two diagonal compression struts connecting base corners to upper tie-columns, with widened flanges above and below openings. This layout aligns with force-flow paths and reduces material volume without compromising stiffness.
Predictive Performance: Referenced ML models demonstrate high accuracy: Random Forest models achieved a Pearson correlation coefficient ( $R$ ) of 0.94 for UHPC compressive strength, and Gradient Boosting Regression Trees achieved an $R^2$ of 0.987 for FRP-confined concrete.
Construction Adaptation: The paper proposes a construction process flow adapting Japanese JBDPA-2001 guidelines to the Peruvian context. Key adaptations include replacing steam curing with traditional moist curing, limiting overlay thickness to 40 mm, and using epoxy dowels for CFRP anchorage, consistent with NTE E.070-2020.

Significance and Claims
The paper claims its primary novelty lies in the first-time coupling of data-driven ML prediction with generative topology optimization specifically for the design of retrofit layers in informal Peruvian housing. By integrating these advanced computational methods with established structural equations, the framework offers a quantitative basis for sizing retrofit interventions that target critical stress paths rather than applying full-wall jacketing.

The study asserts that this approach:

Transfers Technology: It formalizes a roadmap for transposing U.S. and Japanese experiences (financing models and construction rigor) to the Peruvian normative and workmanship context.
Optimizes Resources: It enables a targeted-reinforcement strategy that reduces the volume of high-cost materials (UHPC and FRP nanocomposites) while maintaining confinement capacity at critical zones.
Provides a Replicable Framework: It codifies a construction-process flow and a set of four modern design equations that can serve as a reference for future regulatory updates and subsidy programs, such as the "Bono de Proteccion de Vivienda Vulnerable."

The authors maintain a modest stance, acknowledging that the framework has not yet been field-validated and that the ML models rely on laboratory data. They identify future work as necessary to validate the system through instrumented pilot testing, extend the optimization to 3D domains with nonlinear masonry behavior, and evaluate the specific incorporation of UHPC/FRP lines into Peruvian subsidy legislation.