GNGTS 2023 - Atti del 41° Convegno Nazionale

Session 1.1 GNGTS 2023 indication on the reduction of surface waves and development of trapped waves in the sample. Coalescence of structure is marked by rapid increases in scattering values. d) Velocity and incremental strain. The red line represents the average. Normalisation is required for more robust training We are currently assessing the importance of individual parameters by omitting one at a time from the training routine. In this approach, the more important the omitted parameter, the larger the misfit will be when comparing the network output and the target timeseries. Over- and under-estimations are likely related to mechanical similarities between separate phases of the experiment. Crack Closure and Failure demonstrate similar magnitudes of strain increase. As we also vary the confining pressure, which affects the deformation and failure mode, we further estimate how parameter importance may change with time and experimental conditions. Approximately 100 models are trained in this way. Depending on model setup, the output can significantly vary but some trends are consistent. Critical points are consistent with key deformation phases reported in the literature. Our model does not understand the magnitude of coalescence phases (i.e. misfit) but it recognises that there should be stress drops and strain increases. The relative importance of individual training parameters is shown in Fig. 3 for target parameters of stress and strain as a percentage of Ultimate Compressive Strength (UCS). Figure 3: Training parameter importance vs. % of Ultimate Compressive Strength (UCS) for target parameters of stress and strain