Early Stage Researcher Plans
The AdMoRe research project consists in eight individual research projects based on the applications of interest of the mentioned advanced techniques (Advanced Model Reduction for real-time, inverse and industrial problems).
All admitted PhD students will have to satisfy the following minimum requirements:
- Degree equivalent to a Masters in Engineering or in Applied Mathematics, Physics or a similar science based subject, equivalent to at least 300 ECTS (60 ECTS =1 year of studies).
- Non-native English speaking candidates will be required a minimum IELTS score of 7.0 (in both the written and oral)
ESR1. Fast Hi-Fi CDF Vehicle Design
Accurate and fast evaluation of drag and aerodynamic noise around vehicles is still today an open computational problem. Accurate drag evaluation is important and today’s commercial CFD codes predict values of drag that match reasonably well experimental results. Nonetheless, being able to predict (and optimise) drag when the design variables are the vehicle geometry is also crucial and decisive because it has a major impact in the overall design time. This is known to be extremely expensive in production cycles due to the large amount of configurations tested and the high cost of each simulation involved in the shape optimization process
Reliable and fast modelling of aero-acoustics is in today’s frontier of knowledge and industry is in need of relevant contributions in this area. Aerodynamic noise requires capturing complex flow features with high-fidelity computational tools. This extra difficulty makes the complete aero-acoustics design cycle a major challenge for today’s engineers and scientists.
The aim of this project is to develop a methodology able to capture all the flow features that are required in aero-acoustic simulations. A technique capable of introducing the geometric design variables as extra parameters in the numerical simulation will be developed. The method will be applied to the numerical solution of the compressible Navier-Stokes equations.
ESR2. Validation and reliability for thermo-mechanical simulations
This project pertains to the verification and validation (V&V) techniques applied to the simulation of problems of high industrial relevance (additive manufacturing) to be solved with state-of-the-art numerical methodologies (reduced order models).
V&V is an essential approach to modern computational mechanics for industrial problems, aiming at certifying that 1) the model corresponds to the physical reality (Validation) and 2) the numerical solution is accurate enough to meet the requirement of the end user (Verification).
Fast, reliable and robust simulations are a major goal for industrial competitiveness. Once feasible computations for the desired problem are available, verification and validation (V&V) techniques assess reliability of both model & numerical approximation. However, both input data and measurements used for validation are most times subject to uncertainties that have to be also accounted for in the model. The reduced order solutions in multiparametric setups allow resolving a large number of stochastic dimensions describing uncertainty. The aim of this project is to develop a methodology able to joint the expertise of the academic partners on V&V and parameterized reduced order models. The longstanding and solid experience of the industrial partner guarantees relevant contributions where uncertainty is at the heart of the simulation. The parameterized generalized solutions (computational vademecums) obtained with the corresponding accuracy for both models and methods allow efficiently applying uncertainty quantification techniques.
ESR3. Fast Design Prototyping of MRI Scanners
The ability to simulate engineering outputs in an industrially relevant time scale is of critical importance. Often, industry requires an understanding of how specified output measures change under the variation of problem parameters. However, performing parameter sweeps is often computational prohibitive, particularly in an industrial design environment and when considering three-dimensional problems. To address this issue there is considerable academic and industrial interest in the provision of fast and computationally inexpensive reduced order models, which can be certified with credibility bounds to ensure reliability in the predictions.
In this PhD project the chosen application of interest is a clinical MRI system. Siemens Magnet Technology are interested in understanding the coupled electro-magneto-mechanical response of the conducting components making up an MRI scanner and the effects that changes in its design have on its performance and vibrations. To do this, the coupled set of elasticity equations, for describing the mechanical response, and Maxwell equations, for electromagnetic response, must be solved.
The PhD will first develop a computational framework based on advanced finite element technology and efficient algorithms for the solution of the resulting non-linear equation system. Given the physics involved and the complexity of the problem at hand, this tool will then be used for generating solution “snapshots” for certain designs. The PhD will then develop a low-fidelity reduced order modelling technique capable of rapidly producing simulation outputs for new design parameters of a realistic three-dimensional configuration. The resulting algorithm will then be transferred to Siemens Magnet Technology for use in an industrial context. The PhD will become a member of an active research group working on the development and application of computational techniques in coupled problems.
IRSP4. Nonlinear reduced computational mechanics
Model order reduction for multi-parametric models has proven to be very effective; in particular, by considering those parameters as model extra-coordinates in the Proper Generalized Decomposition. Then, the parametric solution can be post-processed online in real time and in deployed computing platforms (tablets and smartphones). Two major challenges persist: (i) the solution of non-linear parametric models within the PGD framework and (ii) the consideration of online simulation involving too large trajectories. The aim of this project is, precisely, to study these challenges and to discern the viability to applied novel approaches in industry, in particular for simulating the thermo-mechanical models involved in online simulation of welding.
IRSP5. PGD for interactive aerodynamic optimization
Accurate and fast evaluation, in a daily industrial production environment, of drag around vehicles is still today an open computational problem. Nonetheless, being able to predict (and optimise) drag when the design variables are the vehicle geometry is also crucial and decisive because it has a major impact in the overall design time. This is known to be extremely expensive in production cycles due to the large amount of configurations tested and the high cost of each simulation involved in the shape optimization process
This project is aimed to explore the viability of the proper generalized decomposition (PGD) in a daily industrial production environment. A technique capable of introducing the geometric design variables as extra parameters in the numerical simulation will be developed. The method will be applied to the numerical solution of the incompressible Navier-Stokes equations using both an in-house code for demonstrators and OpenFoam industrial applications.
ESR6. Fast Simulation-Assisted Shape Correction after Machining
Large and thick aeronautical structural parts frequently exhibit significant distortions after machining because of the residual stresses due to previous steps (heat treatment). A Post-machining shape correction phase, based on successive mechanical operations (for example, 3-point bending) has to be applied. This process is long and costly and cannot be standardized because distortions vary from one part to the other (non reproducibility). The reshaping process relies entirely on the know-how and experience of the operator.
There is a need to introduce numerical simulation in order to assist the operator and propose the optimal shape correction sequence.
The aim of this project is to demonstrate the feasibility of such an approach.
Two challenging issues have to be tackled:
- The building of a Reduced Order Model able to reproduce the results of the thermo-mechanical simulation (heat treatment, machining, reshaping).
- The identification of the uncertain parameters (material properties, boundary conditions), using the ROM, through the comparison of the response of the real part to the few 1st reshaping operations with the predicted behaviour.
A virtual demonstration of this approach will be performed where the “real” part will be a simulation, which uncertain parameters will be supposed not to be known.
ESR7. A high fidelity benchmarking tool in the design and optimisation of MRI scanners
Traditional approaches to engineering design employ low order finite element discretisations and traditional approaches to finite element analysis. However, such methodologies require dense discretisations and lead to expensive computations in order to ensure sufficient accuracy for today’s modern designs. The latest developments in finite element analysis offer the possibilities for high levels of accuracy within a computational competitive time scale, but, to date, have not been applied in an industrial context.
In this PhD project the chosen application of interest is a clinical MRI system. Siemens Magnet Technology are interested in understanding the coupled electro-magneto-mechanical response of its conducting components making up an MRI scanner and the effects that changes in its design have on its performance and vibrations. The magnetic coupling between the various components produces secondary magnetic fields which are also coupled back. An accurate prediction of these secondary magnetic fields is essential to predict the impact on imaging performance, especially in three dimensions. To do this, the coupled set of elasticity equations, for describing the mechanical response, and Maxwell equations, for electromagnetic response, must be solved.
This PhD will first develop a computational framework based on advanced hp-finite element technology and efficient algorithms for the solution of resulting non-linear equation system in a transient context. This will help Siemens Magnet Technology to improve its MRI coil designs and comparisons against industrial designs will be undertaken to ensure that the high-fidelity model of a realistic three-dimensional configuration is fully benchmarked. The PhD will become a member of an active research group working on the development and application of computational techniques in coupled problems.
ESR8. Real-time monitoring and control of additive manufacturing
Production of large complex parts is a challenging issue for today's industry. Additive manufacturing (3D printing) appears to be an appealing process where material is placed and progressively added on the substrate already formed. The desired properties and geometry are produced laying additional layers in a predefined trajectory. The thermal process is significantly affected by many material and process parameters (thermal power, velocity...). Moreover, heat conduction inside the part depends on contact resistances and change of phases, and induces residual stresses. Real-time monitoring techniques for such process are extremely valuable in industrial practice. An offline/online approach is considered. In the offline stage the thermal parametric model is solved, and it dialogs, in the online stage, from a smart enrichment within a generalized formulation of finite elements looking for a non-intrusive implementation.