An integrated hybrid machine learning and statistical analysis towards strength prediction of sugarcane bagasse ash-based sustainable concrete

Abstract

Concrete compressive strength prediction with bagasse ash incorporation as an additive in the concrete could be more accurately predicted via advanced machine learning and statistical validation methods within this study. It is acknowledged that the conventional models cannot capture the accurate and statistically robust complex interactions of variables involved in concrete mixtures, mainly when alternative materials are incorporated. In the present investigation, ANNs in the form of MLP model and SVR model are employed for compressive strength prediction and ANOVA analysis is performed to establish the statistical significance of the content of bagasse ash on concrete strength. The MLP model reached a good high accuracy of 95% on the training set and 93% on validation, while the MAE was 2.5 MPa, MSE was 1.2 MPa² with the SVR model reaching 92% accuracy on the training set and 90% on validation. ANOVA results showed strong effects of bagasse ash on compressive strength at F-statistic = 10.2, p = 0.001. The results showed that these data led to a significant improvement in predictive reliability relative to methods of conventional approaches. The outcome of such integration is the enhancement in the accuracy of concrete strength predictions, as well as support for the sustainable use of industrial by-products, prospective practical application of bagasse ash in the construction industry process.