This study, proposes the use of a novel rubber-sand concrete (RSC) material, which comprises rubber particles, sand, and cement, as an aseismic material in practical engineering construction. The uniaxial compressive strength (UCS) of damping materials is an important factor that directly affects the seismic activity in underground structures. To predict the UCS of RSC, artificial intelligence model back propagation neural network (BPNN), which is optimized through four swarm intelligence optimization (SIO) algorithms: particle swarm optimization algorithm (PSO), fruit fly optimization algorithm (FOA), lion swarm optimization algorithm (LSO), and sparrow search algorithm (SSA), is used. The dataset for the prediction models was obtained from uniaxial compression tests in the RSC laboratory. The performances of the four hybrid intelligence models were evaluated using six performance indicators: the root mean square error (RMSE), correlation coefficient (R), determination coefficient (R²), mean absolute error (MAE), mean square error (MSE), and sum of square error (SSE).The prediction capability of these models was graded based on these indicators using a ranking system. The results show that the prediction ability of the LSO-BPNN hybrid model is better than that of the three other hybrid models, with RMSE of (1.0635, 1.2352), R of (0.9887, 0.9713), R² of (0.9776, 0.9165), MAE of (0.7257, 0.8243), MSE of (1.1352, 1.5256), SSE of (64.7074, 36.6151), and ranking score of (24, 24) in the training and testing phases, respectively. Therefore, the LSO-BPNN hybrid model is an efficient and accurate method for predicting the UCS of RSCs. Sensitivity analysis showed that rubber and sand were the most important elements that affected UCS prediction, followed by cement, with the lowest relative importance being RPZ. This study provides guidance for the extension and application of RSC materials to underground seismic engineering.

The combination of rice husk ash and common concrete both reduces carbon dioxide emission and solves the problem of agricultural waste disposal. However, the measurement of the compressive strength of rice husk ash concrete has become a new challenge. This paper proposes a novel hybrid artificial neural network model, optimized using a reptile search algorithm with circle mapping, to predict the compressive strength of RHA concrete. A total of 192 concrete data with 6 input parameters (age, cement, rice husk ash, super plasticizer, aggregate, and water) were utilized to train proposed model and compare its predictive performance with that of five other models. Four statistical indices were adopted to evaluate the predictive performance of all the developed models. The performance evaluation indicates that the proposed hybrid artificial neural network model achieved the most satisfactory prediction accuracy regarding R² (0.9709), VAF (97.0911%), RMSE (3.4489), and MAE (2.6451). The proposed model also had better predictive accuracy than that of previously developed models on the same data. The sensitivity results show that age is the most important parameter for predicting the compressive strength of RHA concrete.

The application of aseismic materials in foundation engineering structures is an inevitable trend and research hotspot of earthquake resistance, especially in tunnel engineering. In this study, the pelican optimization algorithm (POA) is improved using the Latin hypercube sampling (LHS) method and the Chaotic mapping (CM) method to optimize the random forest (RF) model for predicting the aseismic performance of a novel aseismic rubber-concrete material. Seventy uniaxial compression tests and seventy impact tests were conducted to quantify this aseismic material performance, i.e., strength and energy absorption properties and four other artificial intelligence models were generated to compare the predictive performance with the proposed hybrid RF models. The performance evaluation results showed that the LHSPOA-RF model has the best prediction performance among all the models for predicting the strength and energy absorption property of this novel aseismic concrete material in both the training and testing phases (R²: 0.9800 and 0.9108, VAF: 98.0005% and 91.0880%, RMSE: 0.7057 and 1.9128, MAE: 0.4461 and 0.7364; R²: 0.9857 and 0.9065, VAF: 98.5909% and 91.3652%, RMSE: 0.5781 and 1.8814, MAE: 0.4233 and 0.9913). In addition, the sensitive analysis results indicated that the rubber and cement are the most important parameters for predicting the strength and energy absorption properties, respectively. Accordingly, the improved POA-RF model not only is proven as an effective method to predict the strength and energy absorption properties of aseismic materials, but also this hybrid model provides a new idea for assessing other aseismic performances in the field of tunnel engineering.