Publications
AeroMDO Lab's purpose is to develop leading-edge aerospace & defense technology and to move these out of the lab into the real world.
DISCOVERThe establishment of AeroMDO Lab actually dates back to the founding of the Multidisciplinary Optimization Group by Prof. Nikbay in 2006. AeroMDO Lab officially opened in the Faculty of Aeronautics and Astronautics in 2019. Our current research interests mainly focus on a number of disciplines and techniques: Aeroelasticity, Aerodynamics, Sonic Boom Minimization, Reduced Order Modeling, MDO, UQ, and Deep Learning.
AeroMDO Team aims to train highly qualified researchers ambitious for the core values such as being a pioneer and society-oriented. We work hard to convert scientific knowledge into the leading aerospace & defense technology with academic, industrial, national, and international collaborations.
By focusing on needs-based advanced knowledge, we develop and integrate cutting-edge tools in the field of computational aerodynamics, structural mechanics and acoustics, with the aid of multidisciplinary optimization techniques for robust design solutions.
We thrive as we spread our research outcomes and accomplishments originating from our university lab to the world, collaborate with industry & public sectors and international institutions, and publish in prestigious journals and proceedings.
ITU1773 Technology Transfer Office
Sep 2024 - Sep 2025 (On-going)
ITU1773 Technology Transfer Office
Sep 2024 - Sep 2025 (On-going)
ITU1773 Technology Transfer Office
Dec 2023 - Dec 2025 (On-going)
TUBITAK 1001 Scientific and Technological Research Projects Funding Program
Apr 2023 - Apr 2026 (On-going)
ITU - BAP-Scientific Research Program
Dec 2022 - Dec 2024 (On-going)
GE Global Services Gmbh UK Branch funding program
Feb 2022 - Aug 2022
TUBITAK 1001 Scientific and Technological Research Projects Funding Program
2019 - 2022
GE Global Services Gmbh UK Branch funding program
2021 - 2022
ITU - BAP-Scientific Research Program
2019 - 2021
GE Global Services Gmbh UK Branch funding program
Feb 2022 - Aug 2022
TUBITAK 1001 Scientific and Technological Research Projects Funding Program
2019 - 2022
GE Global Services Gmbh UK Branch funding program
2021 - 2022
ITU - BAP-Scientific Research Program
2019 - 2021
Reliability and robustness requirements of aerospace engineering systems are of major importance and need to be assessed accurately and efficiently during the early phases of the design process to ameliorate the certification tasks mandatory for the final design. Numerical assessment of the reliability and robustness of a system is possible with uncertainty quantification (UQ) methods that input the effect of uncertain variables to propagate them to the overall system performance.
Multifidelity approaches to design and analysis for complex systems include both low- and high-fidelity data so as to maximize the accuracy of model estimations while minimizing the cost computationally.
The application areas of these methods can be listed as wing-design optimization, robotic learning, even recently being extended to human-in-the-loop simulation.
Deep learning is an important subfield of machine learning and artificial intelligence (AI) that imitates the way humans gain certain types of knowledge and is inspired by the function of the brain called artificial neural networks.
Advanced applications of multi-fidelity surrogate modelling techniques provide significant improvements in optimization and uncertainty quantification studies in many engineering fields. Multi-fidelity surrogate modelling can efficiently save the design process from the computational time burden caused by the need for numerous computationally expensive simulations. However, no consensus exists about which multi-fidelity surrogate modelling technique usually exhibits superiority over the other methods given for certain conditions. Therefore, the present paper focuses on assessing the performances of the Gaussian Process-based multi-fidelity methods across selected benchmark problems, especially chosen to capture diverse mathematical characteristics, by experimenting with their learning processes concerning different performance criteria. In this study, a comparison of Linear-Autoregressive Gaussian Process and NonlinearAutoregressive Gaussian Process methods is presented by using benchmark problems that mimic the behaviour of real engineering problems such as localized behaviours, multi-modality, noise, discontinuous response, and different discrepancy types. Our results indicate that the considered methodologies were able to capture the behaviour of the actual function sufficiently within the limited amount of budget for 1-D cases. As the problem dimension increases, the required number of training data increases exponentially to construct an acceptable surrogate model. Especially in higher dimensions, i.e. more than 5-D, local error metrics reveal that more training data is needed to attain an efficient surrogate for Gaussian Process based strategies.
RESEARCHGATE BACKThe design of unmanned combat aerial vehicles (UCAVs) used for military applications is dictated by their low- observability characteristics rather than the aerodynamic performance. Owing to this reason, the UCAVs experience flow separation during takeoff and landing and exhibit stability issues. Moreover, the wing design significantly influences the aerodynamic performance of the UCAV model. The configurations of UCAV models are still under exploration, and the conceptual design of UCAV is still undergoing various developments, so the dimensions and shape of the UCAV design have not been defined yet. Hence, the assessment of the conceptual configurations and their design are worthy issues that need to be investigated. In the present work, the initial weight, aerodynamic sizing, planform selection and then the conceptual design of a non-constant leading edge UCAV configuration was performed. Later, the obtained conceptual design was optimized using multi-fidelity surrogate models with a vortex-lattice method to achieve a better lift and drag ratio. Lastly, the optimized design was validated using the computational fluid dynamic (CFD) model to verify the accuracy of the surrogate model. It was found that the optimized design exhibited superior lift and drag characteristics compared to the reference design.
AIAA BACKThis paper proposes two novel multi-fidelity neural network architectures tailored for high-dimensional inputs such as computational flow fields. The proposed methods are compared with the multi-fidelity deep neural networks from the literature using a 2-dimensional flow-varying supercritical airfoil problem. The main objective of this study is to generate a multi-fidelity prediction of aerodynamic coefficients using pressure coefficient fields around the airfoil. To generate the dataset, a coarse grid is solved using SU2 Euler solver for low-fidelity data whereas a relatively finer grid is utilized for high-fidelity data to obtain viscous solutions using the Spallart-Allmaras turbulence model. The performance metrics to compare the methods are determined as the test accuracy, physical training time, and the size of the high-fidelity samples. Results demonstrate that the proposed multi-fidelity neural network architectures outperform the multi-fidelity deep neural networks in predictive modeling using high dimensional inputs by improving the multi-fidelity prediction accuracy up to 78.7%.
AIAA BACKThis paper presents preliminary results of an ongoing research in prediction of time-dependent flow fields by focusing on data-driven surrogate modeling using artificial neural networks for unsteady aerodynamic problems. The aim of this research is to model unsteady flow fields with learning in low-dimensional space and reconstruct with recurrent autoencoders. Within the scope of this paper, we separately share our findings in viscous unsteady flow field reconstruction of a 2D cylinder in a channel with a deep autoencoder and unsteady aerodynamic-acoustic time-series prediction of the supersonic NASA C25D aircraft with shallow long short-term memory networks. Satisfactory results are achieved in both unsteady applications, yet further improvements and validations are needed to be achieved to establish the desired surrogate unsteady aerodynamic modeling for supersonic aircraft maneuvers.
RESEARCHGATE BACKThis paper presents risk-based design optimization of a supersonic axisymmetric outwardturning engine inlet with geometric uncertainties by exploiting a convolutional neural network. Due to the exorbitant computational burden, a convolutional neural network-based surrogate model is implemented to be used in the prediction of engine performance parameters which are total pressure recovery and mass flow ratio. To generate a dataset, 256 unique configurations for the inlet are parametrically designed in Engineering Sketch Pad while the SU2 Suite is used to obtain a solution of supersonic flow domains. For sought-after optimum reliable design, gradient-free particle swarm optimization is incorporated with a first-order reliability method. Inlet buzz is considered as the critical phenomenon in computations of the reliability of the engine while the maximum probability of failure is limited with 1−E7 in the optimum inlet configuration.
RESEARCHGATE BACKWe present several approaches on propulsion system airframe integration design concepts for supersonic aircraft to obtain an appreciable decrease in sonic boom loudness level. Propulsion system integration affects the ground pressure signature of a supersonic aircraft through three main factors which are; distortion of the free-stream flow by engine inlet/outlet flow conditions due to high temperature and pressure values, the location of the engine(s), and the structural deformations caused by the propulsion system loading. In this paper, we focus solely on the effects of engine inlet/outlet flows and the location of the engine on the aircraft. CFD simulations are performed for benchmark models provided by AIAA Second Sonic Boom Prediction Workshop for these two factors. Firstly, four different engine locations are considered on the Jaxa-Wing-Body (JWB) low-boom model, and the effect of engine location on the sonic boom is investigated. The inlet and outlet flow conditions are selected the same as the ones for the C25D concept aircraft. Next, two different configurations of 2-D external compression supersonic inlets and a minimum length nozzle are designed using the compressible flow relations and the method of characteristics respectively. To predict the effective inlet and nozzle performances before the airframe integration process in a 3-D model, empirical formulas are used to calculate the spillage and boat-tail drags for the inlet and nozzle components. The two-ramp configuration resulted in a lower spillage drag than the three-ramp configuration. A correlation between empirical and numerical solutions is observed through the design and integration phases of the propulsion system. Finally, the propulsion system components are integrated into a 3-D JWB model with a designed pylon to assess the sonic boom loudness. A difference of 10 dB in-vehicle loudness level is encountered for the aircraft models with and without propulsion system integration. The effect of different inlet configurations on the sonic boom of the overall aircraft is seen to be less significant. However, a two-ramp configuration provided better total pressure recovery and mass flow ratio. The effect of the pylon on the nozzle performance and the near field pressure distribution is observed to be more significant and further investigation is recommended as a future study.
RESEARCHGATE BACKIn this paper, surrogate based approaches for multifidelity uncertainty quantification are implemented in a sonic boom prediction framework for improving the supersonic aircraft design process under uncertainties. The sonic boom prediction framework requires output from multidisciplinary analyses such as obtaining the flow field pressure distribution solution from a flow solver to generate the near-field pressure signature of the aircraft and then propagating this near-field pressure signature throughout the atmosphere to the ground by using aeroacoustic methods. The open-source SU2 suite is employed as a high fidelity flow analysis tool to obtain the aerodynamic solution while in-house post-processing scripts are developed to generate the necessary near-field pressure signature. For low-fidelity flow analysis, A502 PANAIR, a higher-order panel code to solve flows around slender bodies in low angles of attack for subsonic and supersonic regimes, is used. For nonlinear aeroacoustic propagation, NASA Langley Research Center code sBOOM is incorporated with the near-field pressure signature for enabling both high-fidelity and low-fidelity sonic boom calculations. Efficient uncertainty quantification tools are developed in-house by implementing multifidelity polynomial chaos expansion and multifidelity Monte Carlo methods. Several atmospheric parameters are considered to comprise randomness and these uncertainties are propagated into the sonic boom loudness prediction of a low boom aircraft called the JAXA wing-body. Finally, an assessment of multifidelity uncertainty quantification methods is presented in terms of their performances and numerical accuracies.
RESEARCHGATE BACKSonic boom signature prediction is a primary criterion for low-boom commercial supersonic aircraft design and requires a multidisciplinary analysis framework. Due to high computational time required in multidisciplinary optimization processes, we study sonic boom prediction within a multi-fidelity approach. In this study, sonic boom loudness is computed by employing multi-fidelity aerodynamic and acoustics solvers. First, PANAIR panel code is used to determine the aerodynamic characteristics and the nearfield pressure signature with low-fidelity. Meanwhile, at high-fidelity, SU2 multi-physics solver is employed to solve for the near-field flow region around the aircraft. As a highfidelity acoustic solver, NASA’s sBOOM code is coupled with the SU2 code via a Python script to automate the aerodynamic and acoustic solution processes sequentially. Various wing-body combinations and several test cases were analyzed with both low- and highfidelity solvers in our former studies. Here, the near-field pressure signature and sonic boom ground signature of a complete low-boom aircraft configuration are studied with more focus to get a better understanding of the physical and geometrical limitations of the combined low-fidelity solution methods. This study aims to address the low-fidelity sonic boom prediction methods with respect to high-fidelity solutions in support of multi-fidelity design exploration and multidisciplinary optimization studies for supersonic aircraft design.
RESEARCHGATE BACKThis paper presents an uncertainty quantification study for the aeroelastic analysis of the High Reynolds Number Aerostructural Dynamics (HIRENASD) wing. The computational aeroelastic analysis employs an open-source multi-physics suite SU2 with a fully-coupled fluid-structure interaction capability. The surrogate model which is used for the uncertainty quantification study is constructed by integrating an Active Learning procedure into the Gaussian Process Regression Method for improving efficiency and accuracy. The current Active Learning assisted uncertainty quantification approach is assessed with respect to the conventional uncertainty analyses which are based on surrogates generated with Polynomial Chaos Expansion, Kriging, and Polynomial Chaos-based Kriging metamodel methods. The root mean square error and maximum absolute error verification metrics demonstrated that the Active Learning assisted Gaussian Process method provided more successful results than other methods in capturing both global and local features during this aeroelastic uncertainty quantification study.
RESEARCHGATE BACKIn this paper, a sonic boom minimization framework which gathers efficient computational strategies via multi-fidelity and multi-objective optimization methods has been developed to leverage low boom aircraft design concepts. In this multidisciplinary framework, sonic boom prediction requires transfer of the near field pressure data obtained from the aerodynamic solver to the aeroacoustic solver to propagate this pressure signature through the atmosphere to the far-field. For multi-fidelity analyses in the aerodynamic domain, SU2 code is employed as the high-fidelity flow solver through Euler equations, while A502 PANAIR, a higher-order panel code, is employed as the low-fidelity flow solver. For low-fidelity and highfidelity sonic boom calculations, the linear and non-linear solvers of NASA’s sBOOM code are coupled with the near-field pressure data processed from the flow solutions of SU2 and PANAIR codes. For sonic boom minimization, wing planform shape parameters of a supersonic aircraft model are represented by the Class-Shape Transformation method while engineering sketch pad ESP is used for geometric model generation and parametric design update during optimization. As multi-objective optimization algorithm, an in-house code implementing Davidon-Fletcher-Powell Penalty Function method with a multi-start strategy is used for global optimum search. A multi-fidelity surrogate model is constructed with the co-kriging method using linear auto-regressive information fusion scheme and employed in optimization. In-house scripts are developed to couple and drive these analysis tools in this multidisciplinary framework and finally a parametric wing shape design optimization study is conducted for a supersonic wing-body configuration to demonstrate the sonic boom minimization performance.
RESEARCHGATE BACKBu çalışmada Amerikan Havacılık ve Uzay Enstitüsü AIAA tarafından düzenlenen Sonik Patlama Çalıştayı'nda kullanılan 69 Delta Kanat uçak geometrisi üzerinde şekil eniyilemesi yapılmıştır. Ön şok basınç artışını minimize etmek için bu uçak geometrisinin gövdesinin şekli değiştirilerek sonik patlama eniyilemesi yapılmıştır. İlk olarak başlangıç geometrisi için ağ yapısı oluşturulmuş ve akış analizleri gerçekleştirilmiştir. Bu analizler deney verileri ile doğrulanıp analiz kabiliyeti test edilmiştir. Daha sonra bu analiz sonucundan elde edilen yakın irtifa basınç izi kullanılarak sonik patlama hesaplaması yapılmış ve doğrulanmıştır. Doğrulama işleminin ardından uçağın ön gövdesi Bezier eğrileri kullanılarak değişkenleştirilmiş ve temsili model temelli eniyileme sürecine başlanmıştır. Öncelikle aerodinamik ve sonik patlama analizlerinde kullanılacak temsili modellerin kurulması için, belirlenen geometrik değişkenler için sınırlandırılmış bir tasarım uzayında Latin hiperküp örnekleme yöntemi ile belli sayıda örnekleme yapılmıştır. Örnekleme noktalarında yüksek doğruluklu analiz yöntemleri ile aerodinamik sürükleme ve sonik patlama değerleri hesaplanmıştır. Ardından temsili modeli eğitmek amaçlı üretilen örnekleme noktalarındaki analiz sonuçları kullanılarak Kriging yöntemi aracılığıyla temsili model oluşturulmuş ve doğrulanmıştır. Üretilen temsili modellerin genetik algoritma ile entegrasyonu ile sonik patlama minimizasyonu için eniyileme çalışması gerçekleştirilmiştir. Eniyileme çalışmasında sürükleme katsayısı kısıtlama kriteri olarak tanımlanmıştır.
RESEARCHGATE BACKCurrent Members
Prof. Dr. Melike Nikbay
Principal Investigator
Sihmehmet Yildiz
Research Assistant
Dilan Kilic
Research Assistant
Emre Guler
Research Assistant
Dr. Pranesh Chandrasekaran
Post-Doctoral Researcher
Burak Berkan Bedir
Graduate Researcher
Murat Kurnaz
Graduate Researcher
Part-Time Members
Enes Cakmak
Graduate Researcher
Berkay Pirlepeli
Graduate Researcher
Our continuous goal is to establish and develop Public-University-Industry collaborations by employing both national and international funding resources.
You can stop by our office for a cup of coffee and just use the contact form for any questions and inquiries
Copyright © 2022. All Rights Reserved. Designed by AeroMDO