ANN-based compactability integrated contours for target compressive strength design of pervious concrete

Target-oriented pervious concrete design remains challenging because its key properties are governed by complex interactions among mixture composition, aggregate characteristics, and compaction. Conventional compaction measures may not reflect the actual densification achieved. This study therefore introduces compactability as a response-based parameter within an artificial neural network–assisted mix-design framework. An experimental programme comprising 864 specimens was conducted using four aggregate size ranges, eight aggregate-to-cement ratios, and nine compaction levels. Compactability was quantified from the increase in wet density within a constant mould volume, and artificial neural network models were developed to predict the performances. Model reliability was assessed using training and validation datasets, error-based statistical measures, and 10-fold cross-validation. The models produced consistently strong predictions, with cross-validated coefficients of determination generally exceeding 0.93 and limited variation among the folds, indicating stable generalization without substantial overfitting. The results showed that increasing compactability generally increased wet density and reduced measured porosity, whereas its effect on compressive strength depended on aggregate size and paste availability. Smaller aggregate ranges were more sensitive to changes in compactability and aggregate-to-cement ratio, while the broadly graded 12–25 mm aggregate range promoted denser particle packing and a more gradual response. The principal innovation is the use of compactability as a transferable link between applied compaction, fresh-state densification, pore structure, and mechanical performance. Integrating the predicted strength, porosity, and wet-density contours enables mixtures to be selected from overlapping performance regions, providing a practical tool for target-oriented mix proportioning, reduced trial batching, and construction quality control.