interesting topic ?
@stanislaus.schymanski here is a new result. I don't know if it is relevant and worthy to dig into:
When we looked at the computed g_s and compared to the theoretical one obtained by inverting the PM model, we found that sometimes the ratio between the computed and inverted g_s was different. We made 2 hypothesis (i) the ideal stress factor is not linear and (ii) there are other variables to take in line.
To answer this question we set up a new stress factor function, with 3 degrees of freedom. The mathematic expression is inspired from Combe et al., 2015 and is an exponential with a shape factor (see the stress_factor_shape notebook in the branch having the same name). This additional parameter, let's call it \alpha
decides if the stress factor is linear (\alpha = 0
), concave ($\alpha < 0$) or convexe ($\alpha > 0$). In the notebook there is a sensitivity analysis plot (see attached) that show as a heat map the sensitivity of the models (varying g_s, constant g_s and PT) to \theta_3
and \alpha
. We found out that the most relevant stress factor shape for the varying surface resistance is linear (the yellow star indicates the best performing model), concave for the constant g_s model and convex for the PT model. It might mean something in term of interpreting the meaning of the stress factor coefficient ... What do you think ?