A new IoT-ANN framework provides civil engineers with a real-time, non-destructive method for monitoring concrete strength, reducing waste, accelerating timelines, and building smarter from the outset.
Study: Prediction of early age compressive strength of concrete using machine learning. Image Credit: fongbeerredhot/Shutterstock.com
In a recent study published in Scientific Reports, researchers presented an innovative approach to predicting the early-age compressive strength of concrete by integrating Artificial Neural Networks (ANNs) with the Internet of Things (IoT). This framework aims to overcome the limitations of conventional destructive testing methods, providing a faster, more accurate, and sustainable system for evaluating concrete performance in civil engineering applications.
The findings highlight the significance of the IoT-ANN framework for real-time monitoring and improved prediction of concrete strength during early curing stages.
Addressing Challenges in Concrete Strength Assessment
Concrete remains one of the most widely used construction materials due to its strength and durability. However, accurately estimating its compressive strength, particularly at early ages, has been one of the main challenges the industry has faced. Conventional assessment methods typically rely on destructive testing, which is labor-intensive, time-consuming, and produces significant waste.
Recent advances in structural health monitoring have introduced IoT-based systems that collect real-time hydration data. Combined with machine learning techniques like ANNs, these systems are now capable of modeling the complex relationships between temperature, curing conditions, and strength gain. In this study, the researchers took advantage of that synergy to create a predictive framework for early-age concrete strength.
While the ANN models were trained on compressive strength (CS) data collected from concrete cubes, their performance was primarily validated using flexural strength (FS) measurements from concrete prisms.
How the System Works
The core of the framework involved embedding LM35 temperature sensors directly into concrete cubes. These sensors were waterproofed and placed near the center of each cube, following ASTM guidelines. They recorded internal hydration temperatures and transmitted data in real time via an Arduino and ESP8266 Wi-Fi module to the ThingSpeak cloud platform.
From the raw temperature data, the team calculated the temperature-time factor (TTF) using the maturity method, a well-established approach for correlating temperature history with concrete strength gain.
They applied this system across five concrete grades (M20 to M40), casting three sets of cubes for each, and gathering data from 420 samples over a 28-day curing period. Input variables included material proportions (cement, water, aggregates) and TTF. Two ANN models were then trained:
- NN-GD (Gradient Descent)
- NN-LM (Levenberg-Marquardt)
Each model used 80 % of the data for training and 20 % for testing. Interestingly, while the models were trained using compressive strength data, their accuracy was validated against flexural strength results, offering an extra layer of performance evaluation.
Key Findings: Performance of IoT-ANN Integration
The integration of IoT sensors and ANN models proved to be highly effective. The collected temperature data aligned well with strength development patterns, confirming the validity of the maturity method in real-time applications.
Both ANN models predicted early-age strength with solid accuracy, but NN-LM performed better overall, producing lower root-mean-square errors. Across all concrete grades:
- The maturity method produced prediction errors ranging from 0.02 to 3.90 MPa
- ANN predictions fell within a -0.69 to -4.80 MPa range (still close to the experimental values)
Statistical analysis reinforced these findings. M30 concrete had the lowest mean absolute percentage error (MAPE), while M35 showed the most consistent results, with a prediction-to-strength ratio (PSR) closest to 1.0. These metrics confirmed that the ANN models, paired with real-time IoT data, could support practical decision-making on-site, such as when to remove formwork or proceed with structural loading.
Practical Applications for the Construction Industry
This framework has direct, practical implications for the construction industry. By monitoring concrete hydration as it happens, engineers can make informed decisions earlier in the process without relying on destructive tests or waiting for lab results.
Benefits include:
- Faster project timelines, thanks to early decision-making
- Reduced material waste, as fewer test specimens are needed
- Improved sustainability, with lower energy and material usage
- Scalability, as the approach can be adapted to different grades and environments
While the study was conducted under controlled lab conditions, it lays the groundwork for real-world implementation. The researchers noted that field trials are a necessary next step to validate performance across diverse settings and mix designs.
Conclusion and Future Directions
The results clearly showed that combining IoT sensing with machine learning enhanced concrete quality control and helped streamline construction workflows. However, several areas still require further exploration to validate and extend the system's capabilities.
Future work should focus on testing the framework in real-world job site conditions, where variables such as environmental fluctuations and inconsistent curing practices could affect performance.
Expanding the study to include a broader range of concrete mixes and curing environments would also help determine how generalizable the model is across different applications.
In addition, integrating this system with Building Information Modeling (BIM) or digital twin platforms could enable dynamic updates and closer alignment with construction schedules. Finally, exploring hybrid ANN models may further improve predictive accuracy and adaptability, especially in more complex or variable field conditions.
By advancing real-time, non-destructive testing into a more digitally integrated domain, this work provides a foundation for smarter, more sustainable construction practices that respond quickly to on-site conditions.
Journal Reference
Palanisamy, A.K., &. et al. Prediction of early age compressive strength of concrete using machine learning. Sci Rep. DOI: 10.1038/s41598-025-29233-6, https://www.nature.com/articles/s41598-025-29233-6
Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.