Prediction of in situ state parameter of sandy deposits from CPT measurements using optimized GMDH-type neural networks

• Published in: Acta geotechnica Vol. 17; no. 10; pp. 4515 - 4535
• Main Authors: Duan, Wei, Congress, Surya Sarat Chandra, Cai, Guojun, Zhao, Zening, Liu, Songyu, Dong, Xiaoqiang, Chen, Ruifeng, Qiao, Huanhuan
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.10.2022; Springer Nature B.V
• Subjects: Complex Fluids and Microfluidics; Cone penetration tests; Earthquakes; Engineering; Foundations; Geoengineering; Geotechnical Engineering & Applied Earth Sciences; Group method of data handling; Hydraulics; Laboratory tests; Liquefaction; Mathematical models; Methods; Neural networks; Parameters; Performance evaluation; Research Paper; Sand; Seismic activity; Seismic stability; Sensitivity analysis; Soft and Granular Matter; Soil Science & Conservation; Soil stability; Solid Mechanics; Stability analysis; Structural stability; Cone penetration test; State parameter; Liquefaction; In situ state; GMDH neural network
• ISSN: 1861-1125; 1861-1133
• DOI: 10.1007/s11440-022-01540-6

Increase in seismic events around the world has necessitated the evaluation of soil liquefaction potential since it adversely influences the stability of adjacent structures and poses a threat to lives and property. The first step in evaluating the liquefaction potential involves the estimation of the in situ state of sand and the state parameter as an index to characterize the behavior of sand. Due to the difficulty in obtaining high-quality undisturbed sandy samples for laboratory testing and evaluation, simplified equations based on the cone penetration test (CPT) are used for reasonable estimation of field behavior. In the present study, an optimized group method of data handling (GMDH)-type neural network was proposed to estimate the state parameter from CPT data obtained from the historical liquefaction database. A comparison was made between the measured and the predicted values of the state parameter to evaluate the performance of the proposed GMDH neural network method. A sensitivity analysis of the proposed model was also carried out to study the effect of input variables on the output variable of the proposed model. Additionally, the evaluation of in situ state and liquefaction potential of sand based on the state parameter was also presented and compared with the existing methods. Overall, the use of the GMDH model in evaluating the in situ state of sand and subsequent liquefaction potential assessment has shown promising results.