Study on dynamic resilient modulus prediction model of subgrade fine-grained soil based on physical property parameters

Abstract

Abstract This study took a representative subgrade fine-grained soil of a highway as the research object, prepared specimens with different water content and compaction degrees, and investigated the effects of water content, dry density, plasticity index, dynamic bias stress and circumferential pressure on the dynamic resilient modulus M r by dynamic triaxial test. Subsequently, using the data from this study and previous research, a correlation analysis was conducted on the factors affecting the resilient modulus, identifying the factors that contribute significantly to the resilient modulus. The relationship between k i parameters and key physical property parameters was acquired by utilizing artificial neural network and genetic algorithm. Finally, a prediction model considering physical property parameters was established, and the prediction results were compared with those of the general model in standard. The study results show that M r decreases nonlinearly with the decrease in dynamic deviator stress and moisture content and increases with the increase in confining pressure and compaction degree, and M r is uniformly and inversely related to I p. Acoording to Spearman correlation coefficient values between various input features and the resilient modulus, M r has the strongest correlation with the moisture content, dry density, plasticity index and fine particle content. Compared with the standard universal model, the model proposed in this paper has an average prediction error of 5.74%, and the maximum prediction error is not more than 15%, which means the prediction effect is improved. Therefore, for a specific highway subgrade construction project, it is valuable to carry out targeted dynamic resilient modulus in advance to establish a prediction model based on physical properties indicators.