Seminar 1, 11 June 2020, 4pm (GMT+1, local time Amsterdam)

Talk 1: Towards the efficient design and manufacturing of lightweight structures with favorable vibro-acoustic properties

Talk 2: Design, analysis and experimental validation of a gradient-based periodic composite structure for broadband, wide angle of incidence absorption enhancement

Seminar 2, 18 June 2020, 4pm (GMT+1, local time Amsterdam)

Talk 1: Machine Learning for Engineering Applications

Talk 2: Gaussian Process in non-Gaussian likelihood regression and the data association problems

Seminar 3, 25 June 2020, 4pm (GMT+1, local time Amsterdam)

Talk 1: Nonlinear Identification: Towards Large Scale Systems

Talk 2: : Robust Nonlinear State Space Identification using Deep Encoder Neural Networks

Seminar 4, 2 July 2020, 4pm (GMT+1, local time Amsterdam)

Talk 1: Shifting the assessment paradigm: from inanimate to self-aware infrastructures

Talk 2: Coupling Bayesian filtering with Reduced Order Modeling for Virtualization

The recordings of these seminars can be found here: https://chatzi.ibk.ethz.ch/smm-news/2020/07/unconventional-seminar-series.html

Seminar 5, 9 July 2020, 4pm (GMT+1, local time Amsterdam)

Talk 1: Grey-box models: Putting some physics back into machine learning models for structural health monitoring

Talk 2: Constraining Gaussian Processes for Grey-Box Damage Localisation

Building physical insight into machine learning algorithms offers the ability to enhance inference capabilities of traditional black-box methods. This talk will investigate how a Gaussian Process framework for localising damage can be constrained, allowing knowledge of a structure’s boundary conditions to be incorporated into the model. A number of cases where the integration of physical knowledge returns an improved performance will then be explored, demonstrating the benefit of taking a grey-box approach.

19 November, 4pm (GMT+1, local time Amsterdam)

Talk 1: Using Digital twin for improve safety: challenges and opportunities

slides available here: https://www.dropbox.com/s/z5xxv0leuv2hwit/EP_DigitalTwin_challenges_opportunity.pdf?dl=0

Talk 2: A structured approach for quantitative assessment of critical infrastructure resilience

26 November, 4pm (GMT+1, local time Amsterdam)

Talk 1: Computational Vibration applied to Aircraft Engines

High Level of Vibration in Aircraft Engines can cause the failure of the whole engine which can lead to disaster. It is necessary to develop accurate methods that permits to predict the risks of vibration during the design process. Experimental test of vibration for a full scale engine is very expensive and it will be interesting to be able to validate and certify most of the design choices using computational methods. In this talk we will introduce the approaches we have been developing for more than 10 years to model vibration in complex structures. The approach relies on accurate and efficient numerical methods for vibration analysis of large scale model. Challenges, success and new trends of computational vibration will be presented.

Talk 2: Non-Intrusive reduced order modelling of geometrically nonlinear finite elements structures: Direct computation of invariant manifold via normal form

The increase of flexible and lightweight structures in modern mechanical design, motivates the development of accurate, yet efficient, numerical methods which can model the geometric nonlinear behaviour of such structures as they experience large amplitude vibration. Being full order models of geometrically nonlinear structures computationally expensive, reduced-order modelling has recently attracted a large amount of attention in the engineering community. Normal form approach applied to nonlinear vibratory systems defines a nonlinear mapping from modal coordinates to normal coordinates which span the invariant manifolds of the system, thus allowing one to derive efficient reduced-order models. In the context of finite element structures, the drawback of this approach is the need to compute the complete basis of all possible linear modes of the system. In this talk, we will propose a direct method to compute the coefficients of the normal form and the reduced-order models without the need of prior computation of the modal basis. The method computes the nonlinear modes of a structure with an exact formulation up to the third order directly from the system in physical coordinates. Moreover, its non-intrusive implementation is possible in any finite element software. The results provided by the direct normal form method will be compared to those obtained with another non-intrusive and direct method, the quadratic manifold method, and advantages and limitations will be discussed.

Relevant papers:

https://link.springer.com/article/10.1007/s00466-020-01902-5

https://link.springer.com/article/10.1007/s11071-020-05813-1

https://arxiv.org/pdf/2009.12145

3 December, 4:30pm (GMT+1, local time Amsterdam)

Talk 1: Sound insulation: from test based design to numerical optimization

The acoustic development of building systems is usually based on general design rules and many experimental prototype tests. This is costly and time consuming, and leads to suboptimal designs. This talk provides an overview of some recent research efforts that aim at an efficient yet sufficiently accurate numerical prediction framework for the acoustic design of building systems, which takes all relevant uncertainties into account and which opens the way for design optimization.

Talk 2: Structural health monitoring based on operational modal analysis from long gauge dynamic strain measurements

Vibration-Based Monitoring (VBM) is a non-destructive method for structural damage identification and condition assessment. Natural frequencies and mode shapes are probably the most commonly used modal characteristics in VBM applications. However, they can exhibit a low sensitivity to certain types of damage while the sensitivity of natural frequencies to environmental influences may be sufficiently high to completely mask the effect of even severe damage. Modal strains are a promising alternative for damage identification and their use for VBM applications is explored in this research. Fiber-optic Bragg gratings (FBG) are used to measure the dynamic strains of the tested structures from which the modal strains can be obtained. The damage detection and localization capabilities of the modal strains and of related quantities that are directly obtained from them, such as neutral axis positions, are investigated through operational modal analyses on prestressed concrete beams that are subjected to progressive damage tests. It is demonstrated that the cracks that are induced in the beams by the progressive damage tests do have a clear and local influence on the strain mode shapes and neutral axis positions, even at low damage levels. Furthermore, experimental investigations are conducted in laboratory and on a railway bridge to identify also the influence of temperature on modal strains.