⚡ electrical engineering

Mechanical Performance and Strength Prediction of Rubberized Silica Fume Concrete: An Experimental and Machine Learning Investigation for Sustainable Construction

This study investigates the mechanical performance of rubberized silica fume concrete through experimental testing and machine learning, demonstrating that while rubber powder and silica fume enhance specific properties and existing design codes underestimate strength, advanced predictive models like Artificial Neural Networks offer superior accuracy for estimating compressive and flexural strength to support sustainable construction.

Original authors: M. Shariq, A. Fuzail Hashmi, M. Ayaz, M. Tanveer, M. Asif Ansari

Published 2026-06-25

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: M. Shariq, A. Fuzail Hashmi, M. Ayaz, M. Tanveer, M. Asif Ansari

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 the construction industry is like a giant bakery. For years, they've been baking "concrete cakes" using standard ingredients: cement, sand, and rocks. But the world is facing two big problems: a mountain of old tires piling up in landfills (which are toxic and hard to break down) and a need to use less cement to save the planet.

This research paper is like a team of bakers trying to solve both problems at once by inventing a new, eco-friendly recipe. They wanted to see what happens if you swap some of the sand for crushed rubber powder (from old tires) and replace some of the cement with silica fume (a fine industrial byproduct).

Here is a simple breakdown of what they did, what they found, and how they used computers to predict the results.

1. The Experiment: Mixing the "Eco-Cake"

The researchers created 16 different batches of concrete.

The Rubber: They replaced the sand with rubber powder in amounts ranging from 0% (plain concrete) all the way up to 35%.
The Secret Ingredient: For half of these batches, they added 10% silica fume to the cement mix.
The Test: They baked these mixes into blocks and waited 7, 28, and 56 days. Then, they crushed them to see how strong they were, bent them to see how much they could flex, and sent sound waves through them (like an ultrasound for a baby) to check their internal quality.

2. The Results: What Happened?

Think of concrete as a sponge. When you add rubber, it's like putting soft, squishy sponges inside a hard brick.

The Rubber Effect: As they added more rubber, the concrete got softer. It became less able to hold heavy weights (compressive strength dropped). The sound waves traveled slower through it, meaning the material became less dense and had more tiny air pockets.
- However, there was a sweet spot. When they added just a tiny bit of rubber (5%), the concrete actually became better at bending without breaking. It was like adding a little bit of rubber to a brittle cookie to make it slightly chewy instead of shattering.
The Silica Fume Effect: This was the game-changer. When they added the silica fume, it acted like a super-fine filler. It plugged up the tiny holes left by the rubber and glued everything together tighter.
- The Result: The silica fume boosted the strength of every mix, even the ones with lots of rubber. It turned the "soft" rubber-concrete back into something much stronger and more durable.

3. The Problem with Old Rules

Engineers usually use "rulebooks" (design codes) to guess how strong a new type of concrete will be. The researchers tested their new rubber-concrete against these old rulebooks.

The Finding: The rulebooks were wrong. They consistently guessed the strength would be much lower than it actually was. It's like a weatherman always predicting a blizzard when it's actually just a light snow. The old rules didn't account for the unique way rubber and silica fume work together.

4. The Solution: Teaching Computers to Predict

Since the old rulebooks failed, the researchers turned to Machine Learning (ML). Imagine training a smart robot to be a "strength predictor."

The Training: They fed the robot data from their 16 experiments. They showed it: "Here is the amount of rubber, here is the silica fume, and here is the strength we got."
The Contest: They tried six different types of "robot brains" (algorithms) to see which one learned best:
1. Artificial Neural Network (ANN): A brain-like system that learns patterns.
2. Regression Trees & Boosted Trees: Systems that make decisions like a flowchart.
3. Support Vector Machines & others: Mathematical models that draw lines to separate data.
The Winner: The Artificial Neural Network (ANN) was the champion. It was the best at understanding the complex, non-linear relationship between the ingredients and the strength. It made the fewest mistakes.

5. The Final Tool: A New Formula

While the ANN was the smartest, the researchers wanted something practical for engineers to use on a construction site without needing a supercomputer. So, they used the data to create a simple mathematical formula (a regression model).

This formula allows an engineer to plug in the age of the concrete, the amount of rubber, and the amount of silica fume, and it will accurately predict the strength.
They even used this formula to predict how well the concrete would bend (flexural strength) just by knowing how strong it was in a straight crush test.

Summary

This paper is a success story of recycling and smart math.

Recycling: They proved you can use waste tires in concrete, but you need to be careful with the amount.
Enhancement: Adding silica fume fixes the weakness caused by the rubber.
Prediction: Old engineering rules don't work for this new mix, but a new, simple formula (derived from smart computer learning) can predict the strength accurately.

The goal isn't just to make "green" concrete, but to make it predictable so builders can trust it and use it safely in real-world construction.

Technical Summary: Mechanical Performance and Strength Prediction of Rubberized Silica Fume Concrete

Problem Statement
The global accumulation of waste tires presents a significant environmental challenge, with traditional disposal methods like landfilling and incineration causing land degradation and toxic emissions. While the incorporation of recycled tire rubber into concrete (rubberized concrete) offers a sustainable disposal pathway, it typically results in a reduction of mechanical properties, particularly compressive strength, due to the weak bonding between hydrophobic rubber particles and the cement paste. Conversely, silica fume is known to enhance concrete strength through pozzolanic reactions, yet its synergistic effect when combined with varying percentages of rubber powder remains under-explored. Furthermore, existing empirical design codes (IS, Euro-code, ACI, NZS) consistently underestimate the modulus of rupture for rubberized concrete, highlighting a critical gap in predictive tools for this modified material.

Methodology
The study employed a dual approach combining experimental investigation with advanced data-driven modeling:

Experimental Program: Sixteen concrete mixes were prepared using 43-grade Ordinary Portland Cement (OPC). Rubber powder (0% to 35% by weight of fine aggregate) and silica fume (10% by weight of cement) were used as partial replacements. The mixes included variations with and without silica fume, maintaining specific water-to-cement ratios (0.44 for mixes without SF, 0.49 for mixes with SF).
Testing Protocol: Specimens were cured under submerged conditions at 27±2°C. Testing included Ultrasonic Pulse Velocity (UPV) at 28 days to assess quality, compressive strength at 7, 28, and 56 days, and flexural tensile strength at 28 days.
Predictive Modeling:
- Machine Learning (ML): Six algorithms—Artificial Neural Network (ANN), Regression Tree (RT), Boosted Tree (BT), Gaussian Process Regression (GPR), Support Vector Machine (SVM), and Linear Regression Model (LRM)—were trained on 80% of the dataset to predict compressive strength.
- Regression Analysis: Multivariable regression analysis was performed to develop analytical equations for age-dependent compressive and flexural strength.
- Validation: Experimental results were compared against empirical equations from major design codes. Sensitivity analysis was conducted on the best-performing ML model to determine the influence of input parameters.

Key Results

Mechanical Performance:
- Compressive Strength: Increasing rubber powder content led to a progressive decrease in compressive strength. The 35% rubber mix (without silica fume) experienced a 70% reduction compared to plain concrete. However, the addition of 10% silica fume significantly mitigated this loss, enhancing strength across all rubberized mixes.
- Flexural Tensile Strength: A unique trend was observed where adding up to 5% rubber powder increased flexural strength by 6.2% due to the elastic nature of rubber delaying fracture. Beyond 5%, strength declined. Silica fume consistently improved flexural strength, with the 10% SF mix showing an 8.7% increase over plain concrete.
- UPV: Pulse velocity decreased from 4.70 km/s to 3.75 km/s as rubber content increased from 0% to 35%, indicating a shift from "excellent" to "good" concrete quality, attributed to reduced density and increased porosity.
Design Code Comparison: All evaluated design codes (IS, Euro-code, ACI, NZS) significantly underestimated the experimental modulus of rupture, with underestimations ranging from 27% to 56%.
Model Performance:
- Among ML models, the Artificial Neural Network (ANN) demonstrated superior predictive accuracy for compressive strength, achieving a test $R^2$ of 0.9905 and the lowest RMSE (1.3793). Sensitivity analysis identified cement content as the most influential parameter, while rubber powder was the least influential.
- While ANN showed high accuracy, a multivariable regression model was selected for practical application due to its simplicity and explicit formulation. This model achieved an $R^2$ of 0.98 for compressive strength prediction.
- The regression model was successfully extended to predict flexural tensile strength based on compressive strength data, with all data points falling within a ±10% error margin.

Significance and Contributions
The paper claims to provide a comprehensive framework for the sustainable use of waste rubber and silica fume in construction. Its primary contributions include:

Synergistic Optimization: Demonstrating that combining rubber powder (up to 5%) with silica fume can enhance flexural properties while maintaining acceptable compressive strength, offering a viable path for sustainable construction.
Code Limitation Identification: Highlighting the inadequacy of current international design codes in predicting the behavior of rubberized silica fume concrete, thereby necessitating new, material-specific models.
Practical Predictive Tools: Developing and validating analytical models (Eq. 12 and Eq. 14) that allow engineers to accurately estimate age-dependent compressive and flexural tensile strengths without relying on complex computational tools. These models facilitate the integration of waste materials into real-world structural design.
Environmental Impact: Providing a technical basis for utilizing waste tires in concrete, thereby addressing waste management issues and promoting the recycling of industrial by-products.

The study concludes that while machine learning offers high-precision prediction, the derived analytical regression models provide a practical, accessible tool for practicing engineers to design structures incorporating rubberized silica fume concrete.