Improve Pile Design by Genetic Algorithm-遗传算法在提高桩基承载力预测中的应用


报告题目:Improve Pile Design by Genetic Algorithm-遗传算法在提高桩基承载力预测中的应用

报  告  人:加拿大Ryerson大学  刘金元  教授




This presentation evaluates the accuracy of existing design methods for the axial ultimate capacity of driven piles in Ontario and improves existing design methods by applying a Genetic Algorithm (GA). Piles have been used for more than 2,000 years. However, there are still many uncertainties in predicting the axial ultimate capacity in design. Glacial deposits covering Ontario, Canada make this design process even more challenging due to their unsorted and heterogeneous characteristics. A total of 43 H or pipe piles (30 of them subjected to both tension and compression load tests) were selected from a database of pile load tests collected by Ministry of Transportation of Ontario (MTO). On average, the exiting design methods overestimate the pile capacity by a factor of 1.16 to 3.00. GA is applied to develop two new correlation formulas between the skin resistance and standard penetration test blow count. The proposed correlations are slightly conservative with the capacity but provide errors within ± 30 % of the measured skin resistance. In summary, the new design formulas by the GA substantially improve the design of piles.


Dr. Jinyuan Liu, P.E., P.Eng., is a Professor of Geotechnical Engineering at Ryerson University in Toronto, Canada. He is trained with two doctorate degrees in Geotechnical Engineering. Before joining Ryerson, he had worked for five years in the US and 2 years in China. He has published more than 100 technical articles and reports. His research interests are in Applied Geotechnical Research. Currently, his research team is working on two main projects: Improving pile design in Ontario and Investigating deep soil mixing to stabilize sensitive Champlain Sea clay. More about his research can be found at