Full title: ML enhanced analysis and adjoint-based optimization tools for 
cavitating flows in hydraulic turbines


The Doctoral Candidate (DC10) will be hired for 34 months as part of the 
Industry empowerment to Multiphase fluid dynamics simulations using Artificial 
intelligence and Statistical methods on modern hardware architectures at Scale 
(SCALE) project being funded through the Horizon Europe Marie Skłodowska-Curie 
Actions (MSCA) Doctoral Networks. DC10 will be enrolled in the PhD program of 
the School of Mechanical Engineering of the National Technical University of 
Athens (NTUA), Greece. The main objective of DC10 is to develop physics-informed 
and numerics-informed DNNs (Deep Neural Networks) and apply them to shape 
optimization of hydraulic turbomachines. The objective by itself defines the 
scientific fields which DC10 will be working on: (a) Machine Learning (ML), (b) 
Computational Fluid Dynamics (CFD) for multiphase flows and (c) gradient-based 
optimization based on the adjoint method. Applicants should have adequate skills 
in all three of them; with regard to CFD, in particular, the candidate should 
have experience in programming, not only using CFD tools. As written above, the 
application domain will be that of hydraulic turbomachinery. A 4-month 
secondment (within the 34-month project) at ANDRITZ HYDRO (AHS-AT), Linz, 
Austria is scheduled; the purpose of this secondment is to collect data related 
to previous designs and apply the developed tools to industrial cases.

DC10 will use and enhance the in-house GPU-enabled CFD solver PUMA of NTUA for 
the simulation of steady-state, cavitating flows. PUMA is based on the 
Transport-Equation-Mixture model (TEM) where liquid and vapor phases are 
considered as incompressible fluids and the mass transfer rates across them are 
determined using the Kunz model. PUMA also features a continuous adjoint method 
developed for shape optimization purposes. Both the prediction and adjoint-based 
optimization tools have been optimized for GPUs architectures. On the ML side, 
DC10 work will be based on available tools that use a specific two-branch DNNs 
to predict scalar fields; these DNNs have been used as surrogates of modeling 
tools in the framework of multi-disciplinary optimization, for example to 
replace turbulence and/or transition models to reduce the cost of RANS solvers. 
Enhancing, modifying and collating the mentioned codes, but also developing new 
methods to produce an efficient, accurate, reliable and low-cost tool for the 
optimization of hydraulic turbines will be the central task of DC10.


Requirements
Research Field Engineering: Mechanical engineering
Education Level: Master Degree or equivalent

Skills/Qualifications
Candidates should have a strong background in CFD, especially for hydraulic 
turbomachines. In order of importance, they should possess good programming 
skills in C++ (mandatory) and CUDA-C (i.e. the programming environment of PUMA) 
and experience in using high-performance computing centers. Experience in 
using/developing optimization methods is welcome. This is a position in the 
field of CFD for incompressible/cavitating flows, in rotating machines, not 
based on commercial CFD s/w. Though experience in using commercial CFD s/w is 
welcome and surely helps, this position requires good programming skills and is 
addressed to programmers, rather than just users, of CFD s/w. Background 
knowledge regarding Machine Learning and Deep Neural Networks is required. 
Programming skills in Python and deep learning frameworks (preferably TensorFlow 
and TensorFlow C++ API) are mandatory (the first) and welcome (the second).

Languages ENGLISH
Level Excellent

Benefits
The selected candidate will receive a salary in accordance with the MSCA 
regulations for DCs. The minimum gross salary includes a living allowance 
(€3374.4 per month), a mobility allowance (€600 per month) and a family 
allowance (€660 per month), if the researcher has family (‘Family’ means persons 
linked to the researcher by (i) marriage or (ii) a relationship with equivalent 
status to a marriage recognised by the legislation of the country where this 
relationship was formalized or (iii) dependent children who are actually being 
maintained by the researcher). The guaranteed (EC) funding is for 34 months.

Eligibility criteria
Applicants can be of any nationality and must hold a Master of Science degree 
(or equivalent) in engineering. They need to fully respect the following 
eligibility criteria: (a) Must be doctoral candidates, i.e. not already in 
possession of a doctoral degree at the date of the recruitment. (b) Must 
undertake transnational mobility. Researchers must not have resided or carried 
out their main activity (work, studies, etc.) in Greece for more than 12 months 
in the 36 months immediately before their date of recruitment. Compulsory 
national service, short stays such as holidays, and time spent as part of a 
procedure for obtaining refugee status under the Geneva Convention are not taken 
into account.

Selection process
The candidates should send a CV, cover letter (in which the applicant’s 
experience in CFD development should become clear), BSc and MSc degrees 
(certified copies plus translation in English), and two letters of 
recommendation are necessary. Copies of publications could be sent later on, 
upon request. Personal interviews might be asked.

All applications should be mailed to kgianna@mail.ntua.gr (email subject: 
“SCALE-DC10-Application”).

The outcome of the evaluation process will be announced by mid of October 2024. 
The 34-month contract is expected to start around 01/11/2024.

Additional comments
DC10 will be working (excluding the secondment) at the Zografou Campus of NTUA 
in Athens, Greece. The PCOpt/NTUA group consists of about 15 people, including 4 
experienced researchers, among which the major developers of the PUMA and its 
adjoint code. Apart from the PhD thesis supervisor (Prof. K. Giannakoglou), 
researchers of the PCOpt/NTUA Unit with previous experience in similar tasks 
(CFD code running on GPUs, adjoint methods, shape parameterization, cavitation 
models, turbomachinery) will support DC10 in her/his project/PhD. The PCOpt/NTUA 
Unit possesses a powerful multiprocessor platform, including both CPU and GPU 
clusters, which is expected to be upgraded during the project life; this will 
support research to be performed by DC10.

Work Location
Company/Institute: National Technical University of Athens
Country: Greece
City: Athens