Multi-response optimization of sustainable ceramics produced from dredged sludge: An experimental study supported by machine learning

This study investigates the valorization of dredged sludges (DS) into sustainable ceramic materials within a circular economy framework. The effects of sintering temperature (900–1100 °C), soaking time, and pressure on the microstructure and mechanical properties were systematically analyzed. Response surface methodology (RSM) was applied to optimize the processing parameters. At the same time, machine learning models (multiple linear regression, polynomial regression, Random Forest, and XGBoost) were used as complementary tools to support data interpretation and prediction. The results show that temperature is the dominant parameter influencing densification, phase transformation, and compressive strength. A maximum compressive strength of 34.9 MPa was achieved under specific experimental conditions, whereas the RSM-optimized value reached 22.9 MPa, representing a compromise among multiple responses. Simpler models demonstrated more reliable and interpretable performance than more complex models due to the limited dataset size. The observed behavior was attributed to phase transformations and liquid-phase sintering, which promote densification and enhance mechanical properties. These findings highlight the potential of dredged sludges as alternative raw materials for ceramic production. However, further studies, including durability assessment and long-term performance evaluation, are required before practical applications can be fully validated.