With the spread of modern additive manufacturing techniques, topology optimisation represents an advanced methodology for structure optimisation. Topology optimization algorithms address the problem of structure optimisation, by targeting the optimal distribution of material and void regions within a predefined design space.
As in other fields of optimisation, also in topology optimisation, gradient based optimisation techniques have the well-known limitations for engineering applications (need of a smooth model, convergence to local solutions), while stochastic methods, even if able to handle black-box models they can tackle problems in limited size. The neuro-evolution approach is the one proposed in this research and is aiming at bridging the gap between those two families of techniques. Topology optimisation is performed by optimising the parameters of a neural network that models the material thickness and distribution. Finite element analysis is performed at each step of the optimisation to evaluate the structural performance of the current solution.
This research project is about the development of neuro-evolution topology optimisation techniques and their application to the design of hydraulic devices, or their component
components - to achieve the best structural design approach to optimise the shape of existing hydraulic devices.
Project start date: January 2018