Optimizing compressive strength prediction of pervious concrete using artificial neural network

• DOI: 10.1088/2631-8695/adb129
• Publication Date: 2025-01-31T22:59:21Z
• Authors: Sathushka Heshan ...; Asoharasa Janarth; Joseph Dharmar; Perinparasa Vinojan; Navaratnarajah Sa...; Daniel Niruban Su...
• Keywords: Pervious concrete

Abstract The prediction of compressive strength is crucial, as it influenced by various mix parameters such aggregate size, aggregate-to-cement ratio, and compaction. Accurate forecasting ensures optimized designs, enhancing both performance material efficiency in construction projects. novelty this study lies integrating machine learning techniques to predict the pervious concrete, incorporating these key improve predictive accuracy facilitate more precise sustainable design choices. For experimental study, 600 samples were prepared with varying ratios (3.0–5.0), compaction (0–60 blows from standard proctor rammer), size (4.75–25 mm) monitored for porosity strength. A modified Ryshkewitch model assessed alongside evaluations optimization. effect parameter variability on investigate uncertainty propagation. Key uncertainties are highlighted sensitivity analysis, output distributions produced Monte Carlo simulations, reducing essential practical applications, guarantees that forecasts remain constant across a range materials environmental circumstances. In addition, neural network models analyzed accuracy. Incorporating enhanced R 2 empirical 0.63 0.78 0.92, respectively, while was comparable observations. Aggregate size-based improved than 0.95 all cases, insisting dominant impact models. research concludes designs not only but also promote sustainability waste durability concrete. These findings provide valuable insights efficient environmentally friendly concrete urban infrastructure