This paper aims to develop a novel prediction tool based on the machine learning framework to evaluate the compressive strength and effective porosity of pervious concrete material from its compositions. To address this difficult task, 14 data sources were collected from the literature to build a dataset of 164 samples. The dataset included seven mixture design features (e.g., aggregate-to-cement ratio, water-to-cement ratio, minimum coarse aggregate size, the presence of sand or silica fume, effective porosity, and the compressive strength). This dataset was trained and tested by the most relevant machine learning methods: the extreme gradient boosting method (XGB), the random forest regression method, and the support vector machine method. The Particle Swarm Optimization method was applied to tune the models’ hyperparameters. It was observed that the extreme gradient boosting method significantly outperformed the accuracy of the other methods. Relatively high R-squared values of 0.92 and 0.88 were obtained for the compressive strength and effective porosity predictions. Furthermore, to account for the role of compaction, the original database was refined to obtain a 36 samples subset that considered compaction energy. Based on our assessment of this subset, results yielded superior R-squared values up to 0.99 for compressive strength, and 0.97 for effective porosity, revealing the effectiveness and accuracy of this research.

The purpose of this study is to identify the differences in porosity and permeability coefficients when the mixing ratio of aggregates is different and to present the mixing ratio satisfying the strength requirement of compressive specified in a specification of Korea. Three mix ratios were suggested by considering various aggregate sizes and three cylinders were made for each ratio. The porosities of those cylinders were evaluated through the compression and water permeability test, measuring the weight of specimens in underwater and analysis of the pictured Computed Tomography (CT) image. Experiments have shown that it is best to mix 50% for 5–10 mm aggregates, 45% for 2–5 mm aggregates, and 5% for sand in terms of strength and permeability. In addition, as the proportion of fine aggregates increased, the porosity and permeability decreased. Moreover, the effectiveness of maintenance method was also examined in this study.

This study proposes a new approach to construct predictive formulas for the permeability of Pervious Concrete (PC), which depends on PC mixture and porosity. To achieve this, a dataset of 195 samples collected from different sources was used. In the dataset the permeability is dependent on porosity, aggregate-to-cement ratio (AC), maximum nominal sizes (MS) of coarse aggregate, and water-to-cement or binder ratios (WC). From the dataset and through applying simple regression techniques, several analytical functions based on the Kozeny-Carman model were constructed and evaluated for their effectiveness in implementing independent datasets and similar analytical functions. Furthermore, for the first time, the Genetic Programming-based Symbolic Regression method was adopted to construct hybrid models combined with the Kozeny-Carman analytical model. The equation of the hybrid model ensures both basic physical conditions and efficiency while being simple enough for engineering-level applications.