Professor Daniel J. InmanDistinguished Professor,University of Michigan, Ann Arbor, USADaniel J. Inman received his Ph.D. from Michigan State University in Mechanical Engineering in 1980 and is the Harm Buning Collegiate Professor of former Chair of the Department of Aerospace Engineering at the University of Michigan. Formerly he was the Director of the Center for Intelligent Material Systems and Structures and the G.R. Goodson Professor in the Department of Mechanical Engineering at Virginia Tech and is the Brunel Chair in Intelligent Materials and Structures at the University of Bristol, UK (visiting). A former Department Chair of the Department of Mechanical and Aerospace Engineering, State University of New York at Buffalo, he has held adjunct positions in the Division of Applied Math at Brown University and in math at the University of Southern California. Since 1980, he has published eight books (on vibration, energy harvesting, control, statics, and dynamics), eight software manuals, 20 book chapters, over 395 journal papers and 688 proceedings papers, given 72 keynote or plenary lectures, graduated 67 Ph.D. students and supervised more than 75 MS degrees. He works in the area of applying smart structures to solve aerospace engineering problems including energy harvesting, structural health monitoring, vibration suppression and morphing aircraft. He is a Fellow of American Institute of Aeronautics and Astronautics, American Society of Mechanical Engineers, International Institute of Acoustics and Vibrations, Society of Experimental Mechanics and the American Academy of Mechanics, and an honorary professor at Nanjing University of Aeronautics and Astronautics.
He is currently Technical Editor of the Journal of Intelligent Material Systems and Structures (since 1999), Technical Editor of the Shock and Vibration Digest (1998-2001), and Technical Editor of the journal Shock and Vibration (1999-2004). He has served as Technical Editor of ASME Journal of Vibration and Acoustics (1990-1999), and as Associate Editor of the following: ASME Journal of Vibration and Acoustics (1986-89), ASME Journal of Applied Mechanics (1988-94), Mechanics of Machines and Structures (1986-98), International Journal of Analytical and Experimental Modal Analysis (1986-1990) and Journal of Intelligent Material Systems and Structures (1992-1999) and Smart Materials and Structures (1991-2001).
He won the ASME Adaptive Structures Award (2000), the ASME/AIAA SDM Best Paper Award (2001), SPIE Smart Structures and Materials Life Time Achievement Award (2003), the 2007 ASME/Boeing Best Paper by the ASME Aerospace Division's Structures and Materials Committee, the ASME Den Hartog Award for lifetime achievement in teaching and research in vibration (2007), the Lifetime Achievement Award in Structural Health Monitoring (2009), the AIAA Structures, Structural Dynamics and Materials Award (2014) and the 2015 Rayleigh Lecture award (ASME Noise Control and Acoustics Division).
His research interests lie in using smart material systems and structures to solve problems in vibration and in shape changing structures. Over the last decade he has become interested in bio inspired flight using morphing concepts.
Inverse Theory in Vibration Applications
It often happens in science, engineering and mathematics that different discipline specific researchers ignore problems and techniques in those disciplines not directly related to their own. The purpose of this talk is to recognize the important similarities between various vibration topics and that these similarities can be exploited to provide new results. This talk examines the commonality between five vibration research areas:
1) Mechanical Metamaterials2) Structural Health Monitoring3) Damage Detection4) Vibration Suppression5) Model Updating
In order to limit the scope of this presentation it is limited to structures describing linear structural systems that can be well defined in second order form resulting from the direct application of Newton’s law. These five areas of research can all be cast as inverse problems and hence can be approach by a common theory. One particular advantage of looking at this set of problems as an inverse problem is that it highlights the importance of existence and uniqueness in trying to obtain results. The lack of uniqueness in vibration problems is one of the reasons there are so many methods of solution of a given problem. Each of these problems will be defined and recast as an inverse eigenvalue problem and solutions suggested.
Professor Livija Cveticanin
University of Novi Sad, Novi Sad, Serbia
Obuda University, Budapest, HungaryReceived her Ph.D. from the University of Novi Sad, Novi Sad, Serbia and Doctor of Sciences from the Hungarian Academy of Sciences, Budapest, Hungary. Full-time professor of Mechanics and Theory of Machines and Mechanisms at the University of Novi Sad. Head of the Department of Technical Mechanics at the University of Novi Sad. Head of the Doctoral School of Safety and Security Sciences at the Obuda University in Budapest, Hungary. Major research in: Dynamics of Systems with Variable Mass, Nonlinear Vibrations, Rotor Dynamics. She published 6 books (on analytical solutions in nonlinear vibration, dynamics of bodies and mechanisms with variable mass, dynamics of non-ideal mechanical systems), 15 book chapters, over 400 journal and proceeding papers and given more than 20 keynote or plenary lectures. She is the member of Editorial Boards of a few journals and was the Editor-in-Chief for special issues of journals. She is the Head of the Academy of Sciences and Arts of Vojvodina and the Fellow of Serbian Academy of Nonlinear Sciences, Serbian Academy of Engineering, Hungarian Academy of Engineering and Serbian Scientific Society. She won the Scopus Award for the most cited paper in Serbia, October Prize of Novi Sad and Lifetime Achievement Award of the University of Novi Sad.
On Mechanical Metastructures Applied in Vibration Suppression
The problem of vibration suppression is one of the most long-lasting problem in dynamics. Recently, various types of mechanical metastructures are developed for reducing or elimination of vibration. Mechanical metastructures are usually designed as the macro version of metamaterials, which are artificially produced materials which have to satisfy certain physical requirements. Mechanical metastructures are periodical with various unit cells connected into a complex structure which is three, two or one dimensional. Two types of unit cells are considered. In the so called ‘mass-in-mass’ unit a mass is added to absorb the vibration. However, vibration absorption is possible without adding the mass into the unit. Then, the certain deformation of the unit cell is required for vibration suppression. Usually, metastructures are divided into three groups depending on their properties:
1) metastructureswith negative effective mass
2) metastructures with negative effective stiffness and
3) metastructures with negative Poisson’s ratio.
Elastic metamaterials exhibit negative effective mass, metastructures for isolation have negative effective stiffness and the auxetic structures have negative Poisson’s coefficient. All the mentioned structures are nonlinear. Various techniques for solving the nonlinear vibration of structures are applied. Amplitude-frequency characteristics of systems are determined. Although the existing metastructures are shown to be very suitable for vibration absorption or suppression (depending on the structure), tests have also shown a major drawback. Future investigation are necessary for their elimination.
Professor Mehdi Vahdati
Principal Research Fellow
Imperial College, London, UK
Professor Mehdi Vahdati has over 30 years of experience in developing models for turbomachinery aerodynamics and aeroelasticity. He is the pioneering author of CFD aeroelasticity Code AU3D, which is used by Rolls-Royce pls. He recognised that large scale CFD can replace experimental testing (1990) for real engineering.
- First CFD code to perform analyses of whole fan blade assemblies (1993)
- First CFD code to compute surge for an entire 8 stage compressor with 400 mil points (2018)
- First CFD code to model the entire 3 shaft engine (2019)
Beyond the successful 30 years of experience, the new research path will be "Smart Solution to Fluid Flow", which recognises that the generation of CFD model will be data-driven and ML-based. The model will include manufacturing tolerenaces which is fundamental for accurate prediction of performace and stability in turbomachinery. The goal of the ML-CFD approach will be to develop a model that is:
- More accurate than conventional models
- More computationally efficient than conventional ones
- Able to extract features/gain insights on physical phonemena
Application of Machine Learning in Turbomachinery
Reynolds-averaged Navier–Stokes (RANS) is the most common approach for solving real engineering problems in turbomachinery. RANS techniques rely completely on modelling assumptions by representing turbulent characteristics as additional stresses, which results in considerably lower computational cost than Direct Naiver-Stoke simulations (DNS). RANS models are constructed using a formal averaging procedure applied to the exact governing equations of motion and require closures to represent the turbulent stresses and scalar fluxes emerging from the averaging process. With the explosive growth of available data and computing resources, recent advances in machine learning and data analytics have yielded transformative results across diverse scientific disciplines,including image recognition, cognitive science, and genomics . However, more often than not, in the course of analysing complex physical, biological or engineering systems, the cost of data acquisition is prohibitive, which results in decision making under partial information. In this small data regime, most state-of-the-art machine learning techniques (e.g. deep/convolutional/recurrent neural networks) lack robustness and the convergence is not guaranteed. There is a vast amount of prior knowledge in aerodynamics that can be utilised in modern machine learning practice. These include conservations laws that govern the time-dependent equations of fluid flow, empirically validated rules, experimental data, existing conventional CFD data and the understanding of the principles of fluid flow. The above knowledge can act as an agent that constrains the space of admissible solutions to a manageable size. Therefore, learning process (such as the neural
networks) can be constrained to respect these principles which originate from the physical laws. This powerful construction allows the introduction of a new technology in CFD leading to the development of new data-efficient and physics-informed CFD solvers. Using the physical insight gained from past, a set of ML models are developed which can be used to assess the reliability and stability of an aero-engines during the development stages. The models can help detect potential problems during early stages of the development of an engine, and hence reduce
the need for re-designs. To achieve this goal, supervised learning such as classification and neural network can be implemented. A large amount of user input data (produced by CFD) are fed into the algorithms in two-fold: training samples used to improve the “fitting” of the model to known input data, and testing samples used for evaluating the prediction accuracy of the trained model in unknown potential applications. The achievement of this goal will lead to
1) A method which is more accurate than RANS solvers (as it is trained on high-fidelity data such
as DNS and LES)
2) Computationally faster than RANS
3) It can be used to extract features/gain insights on physical phenomena
In this presentation the application of ML to the following problems will be repented
1) Turbulence models
2) Manufacturing tolerances
3) Flutter modelling of compressors
Professor Jérôme Antoni
Laboratoire Vibrations Acoustique
University of Lyon, France
Jerome Antoni (M.S. in Mechanical Engineering, 1995, Ph.D. in Signal Processing, 2000) joined the University of Technology of Compiègne in 2001, after completing his PhD at the University of Grenoble (France). He currently holds a full professor position at the University of Lyon, France, Laboratoire Vibrations Acoustique. The main direction of his research activity is concerned with the development of signal processing methods in mechanical applications, with a special interest to the resolution of inverse problems in acoustics and vibrations. This includes applications in machine and structural health monitoring (MHM & SHM), identification and imaging of acoustic and vibration sources. He has published more than one hundred journal papers in these domains. Jerome Antoni has been with the editorial boards of the International Journal of Condition Monitoring, the International Journal of Rotating Machinery, Diagnostika, Mechanical Systems and Signal Processing and Applied Sciences. He is the Head of the research group Laboratoire Vibrations Acoustique at University of Lyon.
Aeroacoustic Signal Processing
This talk addresses the processing and analysis of aeroacoustic signals. By aeroacoustic signals, it is meant the measurement of sound produced by the turbulent motion of a fluid or the interaction of a fluid with obstacles. Such measurements are encountered in numerous applications. Of particular concern here is sound produced by rotating machines, as encountered in aeronautics. The processing of aeroacoustic signal poses several challenges that are inherent to the nature of the signals. First, aeroacoustic signals typically comprise a superposition of several components, with different statistical properties. Second, they often suffer from a very low signal-to-noise ratio. The talk addresses these two topics. A rather general decomposition is introduced that separates aeroacoustic signals into a tonal part, a cyclostationary broadband part attached to rotating components, and a residual stationary broadband part that typically embodies background noise. Examples are shown where this decomposition allows the analysis of different physical phenomena that would have been difficult to isolate otherwise. Besides, another approach is introduced that directly decomposes aeroacoustic signals into one part relating to their acoustical content and another part relating to their aerodynamic content. This is found especially useful in applications where the former is of primary interest, while being completely masked by the latter. Examples of application are shown in aeronautics, including in-flight and wind-tunnel tests.
