The proposed use of generalized regression neural network to predict the entire static load test
MetadataShow full item record
In the UAE, continuous flight auger piles (CFA) are the most commonly used type of foundations to construct high rise buildings, bridges, and other heavy structures due to the high groundwater table and the weak soil/rock layers near the ground surface. To minimize the risk of failure, of these CFA piles, mandatory expensive field tests need to be performed and the most important one is the Static Pile Load Test (SPLT). To minimize the number of tests required for a particular project in the field, this paper proposes using General Regression Neural Network (GRNN) to predict the pile performance ahead of any test. The data collected from thousands of loading points in over one hundred projects from Dubai, Abu Dhabi, and Al Ain cities are used to develop a GRNN capable of predicting SPLT curves with reasonable accuracy. The friction angle, unconfined compressive strength, depth, soil type, groundwater table, pile’s diameter, and pile’s length are the parameters that are input to predict the load-displacement curves of the SPLT. This approach can complement conventional SPLT and provide engineers with sufficient insight on the pile performance ahead of the actual test.