Researchers have developed an interpretable deep learning framework that enables the early estimation of concrete compressive strength using real-time IoT sensor data. By leveraging multivariate inputs such as internal temperature, humidity, and environmental variables, the new model provides a reliable alternative to traditional destructive testing methods, which typically require a 7- to 28-day curing period before structural performance can be confirmed. This advancement offers a practical decision-support tool for construction management, facilitating improved quality control and optimized project scheduling.
Concrete compressive strength remains a critical indicator of structural integrity and safety in construction projects. Conventional assessment relies on physical destructive testing, an approach that is inherently time-consuming and limits the pace of onsite decision-making. The research team addressed these limitations by modeling the early-age hydration process as a temporal learning problem, integrating mix design parameters alongside real-time monitoring of curing conditions to capture the complex variables governing strength development.
The study utilizes a Bidirectional Long Short-Term Memory (Bi-LSTM) network, a specialized deep learning architecture capable of analyzing forward and backward temporal dependencies. By incorporating an attention mechanism, the model is able to identify the most influential time intervals within the curing process.
To address the challenge of model transparency, the researchers integrated SHapley Additive exPlanations (SHAP) alongside attention-based analysis. This methodology quantifies feature importance, revealing the critical curing stages that govern the final strength of the concrete.
Validation of the framework was performed using field data gathered from actual construction sites. Experimental results indicate that the approach achieves high predictive accuracy, with a mean absolute error (MAE) of 1.52 MPa and an R2 value of 0.94. These metrics demonstrate the framework’s ability to maintain precision under realistic environmental conditions, showing superior performance when compared against standard benchmark models such as Gated Recurrent Units (GRU) and Temporal Convolutional Networks (TCN).
The findings indicate that the proposed framework can reliably estimate both 7-day and 28-day compressive strength within the first 24 to 48 hours of curing. The integration of real-time sensing and explainable artificial intelligence represents a significant shift toward smart construction systems, effectively reducing the industry’s reliance on time-intensive destructive testing protocols.
Subject of Research: Deep learning-based predictive modeling for structural concrete strength analysis
Article Title: Explainable temporal deep learning for early prediction of concrete compressive strength using IoT sensor data
Article References: Shoaib, M., Khan, A., & Mohmand, M. I. (2026). Explainable temporal deep learning for early prediction of concrete compressive strength using IoT sensor data. Discover Applied Sciences. https://doi.org/10.1007/s42452-026-09712-9
Image Credits: AI Generated
DOI: 10.1007/s42452-026-09712-9
Keywords: Concrete Compressive Strength, Deep Learning, IoT Sensors, Bi-LSTM, Explainable AI, SHAP, Hydration Dynamics, Structural Performance, Construction Technology, Real-time Monitoring, Predictive Modeling, Machine Learning, Temporal Analysis, Material Science, Quality Control
News Source: Blake Davidson. (November 17, 2026). IoT-Integrated Deep Learning Enables Early Concrete Compressive Strength Prediction. Scienmag.



