Model Scale Tests of Laterally Loaded Piles in Sand
- Author(s): Favaretti, Camilla
- Advisor(s): Lemnitzer, Anne
- et al.
Deep foundations are one of the most common foundation systems used in engineering practice. Prediction of pile response requires a good understanding of the mechanism that governs interaction between the pile and the soil. In order to obtain efficient and economical pile designs, geotechnical engineers utilize experimental studies combined with traditional analytical models. Most formulations used today are based on elastic solutions or Winkler formulations (e.g. p-y curves) that were derived and calibrated with limited amounts of existing test data. Controversies exist with respect to the influence of various parameters, such as head fixity, pile installation techniques, soil profiling, and axial- lateral load. Moreover, the interpretation of the data from pile instrumentation is not straightforward, given the numerical errors that incur in the double differentiation of the bending moment profile. Specifically, this research pursues the following objectives:
1. Selected p-y curves limitations, including the influence of different concrete, reinforcement, and tip restraint, are addressed through combined experimental and numerical studies on model-scale test specimens.
2. Advanced construction materials, such as high-strength polymer concrete and innovative polymeric reinforcement materials, are also examined to assess their suitability for commercial introduction into routine foundation design practice.
3. An innovative strain gauge based instrumentation will help 1D instrumentation (i.e. longitudinal strain measurements).
4. The results obtained from the model-scale lateral loaded tests on concrete piles are evaluated to provide insights into the soil-pile interaction behavior. An optimization technique implemented in a genetic algorithm framework is proposed to facilitate data interpretation and to derive p-y curves even in presence of disturbed data readings and pile nonlinearity. The proposed genetic algorithm targets directly p-y curves and evaluates them through the minimization of a fitness function, represented by the explained variance (EV) between raw p-y values and fitted p-y function. Moreover, this approach allows the generation of p-y curves from an ensemble of different statistical methods.
The ultimate intention of this study is to incrementally eliminate uncertainties associated with pile analysis and provide new understanding for design in practice.