job talk q&a Flashcards
Why did you use the Winkler approach?
Its easy to deploy, since most engineers already have a Winkler model coded for pile design
Its easy to understand, engineers are familiar with the concepts of Winkler soil reactions
It is efficient
Can you explain how the soil reactions are extracted?
The soil reactions are obtained by integrating the tractions of the stress components of the surrounding soil elements at each depth
they represent the net force or moment applied by the soil on a cross section of the caisson
So what are the implications of not having the coupling terms in the Winkler matrix?
My results show that not having these coupling terms reduce the generalisation capabilities of your model.
Let me explain. these soil reactions are calibrated using 3dfe simulations based on a fully rigid caisson.
if I include the coupling terms the soil reactions are general enough that if it works well for predicting the global stiffness of a caisson with rigid or flexible skirts.
however if i exclude the coupling terms, it does not work well for predicting the global stiffness of caisson with flexible skirt,
Why do you need the shear strain ref, when your nle model uses dev strain ref?
You can easily convert shear strain ref to deviatoric strain.
e.g. for simple shear test, deviatoric strain is shear strain/ sqrt(3)
can you explain how the sos optimisation works?
first I presume that the yield function is of polynomial form
Then I use semi definite programming to find the coefficients for the polynomial that best fits the data while ensuring that the polynomial is a convex function
It is the latter part that is difficult. It is np hard to apply convexity constraints to a polynomial of degree 4 and above
The SOS technique offers a computationally tractable manner to achieve this
how realistic is von mises and drucker prager?
as a first estimate, they are good enough to model the salient features of undrained clay and drained sand, which are pressure independency and dependency of the yield criterion
However, i do plan to explore more advanced soil models in the future, when I develop cyclic design methods for caissons.
Can I explain how your mp model provide the uncertainty measure
The ml model I’m using here is a probabilistic model called gaussian process regression model.
This model assumes that the data points are random variables following a multivariate gaussian distribution and the covariance is determined by how close the data points are in the input space
The uncertainty measure is then the confidence interval of the predictions based on this gaussian distribution
Do you think offshore wind would fly in Australia
I’m glad you ask that. Personally I’m optimistic but I must say I can predict the behaviour of foundations but I can’t predict what the Australian populace attitude towards offshore wind would be in the next 10 years
However I do believe that our roles as researchers is to come up with innovative solutions to drive down the cost of offshore renewables and make the economics so compelling that u cannot ignore it
So that’s what I’m focusing on. Driving the cost down and make offshore wind competitive with other sources of energy
Can u describe the sequential swipe technique
How is it diff from side swipe
First you apply a large displacement to reach capacity
Now the current load state is on the yield surface
Then u apply small incremental changes in displacement directions so that the load state slowly travel along the yield surface
Side swipe is a special case of the sequential swipe where u apply a big change in displacement direction
My results show that u need to keep the incremental change small in order for the load state to stay on the yield surface
That is why the side swipe is unreliable as the load state tends to leave the yield surface
Why apply data driven not to entire system
First, it did not get the buyin from the industry partner as they have a well-tuned optimisation process for the wind farm design using the Winkler framework and they are happy with the beam approximation of the structural response. They are just unhappy with the soil reaction approximation of the soil response and they wanted a more surgical solution for this.
Why go offshore wind when onshore enough
Offshore wind provides several benefits over onshore wind such as
pros
stronger and steadier winds
Greater economies of scale
less opposition from community
cons
more expensive infrastructure
How are the existing caisson-based wind farms designed now?
They are designed using 3dfe models but due to computational limitations, they cannot thoroughly explore the foundation parameter space.
so they have to use engineering judgement to preselect some potentially critical design cases and investigate them using the 3dfe model.
this usually results in a feasible solution but it may be suboptimal.
why is it important that the yield function be convex?
It is important because if the yield function is not convex, you cannot use the implicit elastoplastic stress-update algorithm reliably and you may face convergence problems.
And that is unfortunate as the implicit algorithms tend to be much more efficient than the explicit algorithms, especially if the yield function is convex
