Study on the influence of lead ions on soil fissure development and intelligent prediction model under dry–wet cycle conditions

This study investigates the fissure development law of soil contaminated with varying degrees of lead ions under the influence of dry–wet cycles. The effects of different lead ion concentrations on the development of fissures in silty clay, red clay, and expansive soil were systematically investigated through dry–wet cycle tests, and parameters such as fissure area and fractal dimension were precisely quantified and analyzed using image processing techniques. The study’s results indicate that the presence of lead ions significantly promotes the development of fissure areas in three soils. Based on this, four fractal dimension prediction models, namely, bidirectional long short-term memory network (BiLSTM), gated recurrent unit (GRU), extreme gradient boosting (XGBoost), and regression by relevance vector machine (RVM), are constructed, among which the RVM model exhibits optimal performance. To further improve the prediction accuracy, the sparrow search algorithm (SSA) and particle swarm optimization algorithm (PSO) are introduced to optimize the model parameters, and it is found that the SSA- RVM model performs the best, and its mean squared error (MSE), root mean squared error (RMSE), and mean absolute percentage error (MAPE) metrics are reduced by 3.55%, 6.98%, and 16.79% compared with that of the RVM model, respectively. This study supports the optimization of ecological remediation techniques for contaminated soils and the risk assessment of heavy metal-contaminated sites.