17-10-2024 - Lagerhuysch

PostDoc presentation

A validated DEM modelling framework on plate and pile penetrations in a double-layer scour protection

Hao Shi

Monopiles are the dominant foundation type for offshore wind turbines, accounting for approximately 80% of the installed capacity. Installing offshore monopile foundations requires monopiles to penetrate several pre-installed scour protection rock layers before securing them into the seabed. With the increasing power of the wind turbines, both the size of the monopile and the scour protection rocks increase. The accurate prediction of the penetration resistance is crucial to ensure successful monopile installations. To facilitate or even replace the costly and labour-intensive experimental small-scale penetration tests, a numerical model has been developed using the Discrete Element Method (DEM) that captures the discrete nature of interactions between rocks and piles and predicts the resistance during the penetration process. The developed DEM model includes armour and filter rocks represented by multispheres and sand particles by spheres. A multistage calibration, verification and validation DEM modelling framework is proposed and examined with small-scale penetration tests conducted using plates and piles in a bilayer scour protection configuration. The sand material model is calibrated and verified using penetrometer tests and the rock material models are calibrated and verified using a plate penetration test. The verified model predicts the penetration resistance well in small-scale pile penetration tests and proves the validity of the proposed framework. The proposed framework facilitates the modelling area where traditional continuum-based numerical methods give less accurate predictions and allow us to achieve further insights that are hardly or nearly impossible to obtain by experiments.

17-10-2024 - Lagerhuysch

PostDoc presentation

Efficient Constraint Multi-Objective Optimization with Applications in Ship Design

Roy de Winter

In this talk I will introduce the work that I conducted during my PhD at C-Job Naval Architects and the Natural Computing group of Leiden University. The main contribution of my thesis are efficient multi-objective optimization algorithms that can also handle constraints. I worked on and discovered several different strategies that speed up the optimization process despite the expensive nature of the objective and constraint functions. The algorithm that I will introduce in this talk is the IOC-SAMO-COBRA algorithm that has been used to optimize the damage stability criteria and the cargo hold volume of a single hold cargo ship by adjusting the bulkhead positions. The IOC-SAMO-COBRA algorithm incorporates all the techniques and recommendations discovered during my PhD, which include: 1) Use an as small as possible initial design of experiments for faster convergence, 2) Adaptively chose the best surrogate fit, 3) The use of uncertainty quantification methodologies is redundant when doing multi-objective optimization, 4) I t is possible to propose multiple solutions per iteration by using the hypervolume performance indicator, and finally 5) Combining inexpensive constraint and objective functions with surrogate models for expensive functions leads to better Pareto frontier approximations.