Regression-based Approach for Concrete Compression Strength Estimation

Abstract

Concrete strength is a critical parameter in civil engineering because it governs the load-bearing capacity, durability, and long-term performance of buildings, bridges, and other infrastructure. Conventional determination of compressive strength relies on destructive laboratory testing of cured specimens, which is time-consuming, labor-intensive, and costly, and can only be performed after the concrete has hardened. In this project, we develop a machine–learning based concrete strength prediction system that uses mix design parameters such as cement content, water, aggregates, fly ash, superplasticizer, and curing age to estimate compressive strength rapidly and non-destructively. Using Python with libraries like pandas, NumPy, scikit-learn, matplotlib, and seaborn in the Google Colab environment, the dataset from the UCI Machine Learning Repository is preprocessed, explored, and modeled with several regression algorithms (linear regression, decision tree, random forest) to identify the most accurate predictor. By leveraging statistical insights and automated model selection, the system provides reliable strength estimates in seconds, reducing the need for repeated laboratory testing, cutting project costs, accelerating decision-making in construction planning, and contributing to safer, more efficient, and sustainable engineering practices.