Semester of Graduation
Master of Science in Civil Engineering (MSCE)
Department of Civil and Environmental Engineering
Over the years, numerous design methods were developed to evaluate the undrained shear strength, Su, ultimate pile capacity and pile set-up parameter, A. In recent decades, the emphasis was given to the in-situ cone and piezocone penetration tests (CPT, PCPT) to estimate these parameters since CPT/PCPT has been proven to be fast, reliable and cost-effective soil investigation method. However, because of the paucity of a vivid comprehension of the physical problem, some of the developed methods incorporate correlation assumptions which might compromise the consistent accuracy. In this study, the Artificial Neural Network (ANN) was exerted using CPT data and soil properties to generate a better and unswerving interpretation of Su, ultimate pile capacity and ‘A’ parameter. In this regard, a data set was prepared consisting of CPT/PCPT data as well as relevant soil properties from 70 sites in Louisiana for the evaluation of Su. For ultimate pile capacity, a database of 80 pile load tests was prepared. Lastly, data was collected from 12 instrumented pile load tests for the interpretation of the ‘A’ parameter. Corresponding CPTs along with the soil borings were also collected. Presenting these data to ANN, models were trained through trial and error using different feed-forward network techniques, e.g. Back Propagation method. Different models of ANN were explored with cone sleeve friction, fs, and tip resistance, qt, as well as plasticity index, PI, effective overburden pressure, σ’vo, etc. as input data and were compared to the conventional methods. It was found that the ANN model with qt, fs, and σ’vo as inputs performed satisfactorily and was found to be better than the conventional empirical method of evaluation of Su. On the other hand, ANN models with pile embedment length, pile width, qt, and fs as inputs, outperformed the best-performed direct pile-CPT methods in the interpretation of ultimate pile capacity. Similarly, the ‘A’ parameter predicted by the ANN models (PI, OCR, and Su as inputs) was also in good agreement with the actual one. These findings, hence, fortifies the applicability of ANN for estimating the undrained shear strength, ultimate pile capacity and ‘A’ parameter from CPT data and soil properties.
Mojumder, Md Ariful Hassan, "Evaluation of Undrained Shear Strength of Soil, Ultimate Pile Capacity and Pile Set-Up Parameter from Cone Penetration Test (CPT) Using Artificial Neural Network (ANN)" (2020). LSU Master's Theses. 5145.