The integration of industrial and biogenic waste materials in soil stabilization provides an environmentally sustainable alternative to conventional binders. This study evaluates the influence of mussel shell powder (MSP) on both untreated, cemented, and recycled soils, where the recycled soil was initially stabilized with calcium carbide residue, cured for one year, ground into powder, and then re-treated with MSP. Unconfined compression, ultrasonic pulse velocity, and direct shear tests were conducted to assess the strength, compaction, and shear behavior of MSP-stabilized pure, recycled, and cemented soils. The results indicate that MSP addition reduced plasticity and improved soil workability. In recycled soils, 5% MSP provided optimal strength enhancement, while in cemented soils, 20% MSP was required for significant strength gains due to its role in secondary cementation. Freeze-thaw tests demonstrated that MSP-treated soils exhibited up to a 40% reduction in strength loss compared to untreated samples, improving durability in cold climates. The ultrasonic pulse velocity measurements showed strong correlations with unconfined compressive strength, confirming its potential as a nondestructive assessment method for stabilized soils. These findings highlight the potential of MSP as a sustainable stabilizer for improving soil mechanical properties, durability, and resistance to freeze-thaw cycles.

This research harnessed the potential of artificial neural networks (ANNs) to anticipate the characteristics of bricks derived from recycled soil. The study encompassed the production of bricks employing varying proportions of recycled soil, spanning from 0 to 50% with incremental steps of 10%. Subsequently, these bricks underwent exposure to both controlled and uncontrolled temperature conditions. Post-production, a curing process was initiated, followed by subjecting the bricks to comprehensive testing to evaluate their water absorption and compressive strength, a week after curing. Two distinct ANN models were accurately constructed and employed to predict the attributes of bricks post-burning under controlled and uncontrolled temperature settings. To gauge the accuracy and efficacy, the trained ANN model were assessed by analysing statistically, examining training graphs, and applying k-fold cross-validation techniques. The results showcased the capability of the ANN models in generating precise forecasts for water absorption and compressive strength values. Impressively, the ANN model exhibited high regression values of 0.99621 for bricks subjected to controlled temperatures and 0.99874 for those exposed to uncontrolled temperatures, underscoring the robustness and accuracy of the predictions.