A machine learning framework for in situ sensing of pile length from seismic cone penetrometer data

Several hundred high mast light towers (HMLTs) throughout the state of Minnesota have foundation systems, typically concrete-filed steel pipe piles or steel H-piles, with no construction documentation, e.g., pile lengths. Reviews of designs within current standards suggest that some of these foundations may have insufficient uplift capacity in the event of peak wind loads. Without knowledge of the in situ pile length, an expensive retrofit or replacement program would need to be conducted. Thus, developing a screening tool to determine in situ pile length – as compared to a bulk retrofit of all towers with unknown foundations – would provide significant cost savings. The goal of the project is to establish a non-destructive evaluation (NDE) technique, including the testing setup and data analysis algorithms, for determining in situ pile lengths by way of seismic waves. This work specifically focuses on (i) the use of ground vibration waveforms captured by a seismic cone penetrometer; (ii) three-dimensional visco-elastodynamic finite element method (FEM) used as a tool to relate the sensory data to in situ pile length; and (iii) the use of machine learning (ML) algorithms, trained with the outputs of FEM simulations, to solve the germane inverse problem. In principle, the scarcity of existing HMLT configurations with known pile depth creates an absence of field data needed to adequately train an ML algorithm. However, we demonstrate that the use of FEM simulations as proxy training data for the ML algorithms leads to a robust data interpretation scheme.