Non-destructive pile testing and data analysis at Bothkennar

An artificial neural network-based system and its use in processing data fr om traditional non-destructive sonic echo tests on installed concrete found ation piles is described. The piles tested formed part of the EPSRC's soft clay test site at Bothkennar. The neural network pile profile predictions a re compared to the design/as-build profiles. Finite element analysis-based predictions of the piles' shapes are also provided. The neural network gave sensible results for field test data upon which it had not been trained. T he derived pile lengths for piles tested at Bothkennar were correct to +/-5 %. The derived pile, radii were correct to +/-20%. This error was predomina ntly due to errors in modelling the pile head where th surrounding soil's e ffect on the pile head's dynamic stiffness influenced the system response.