The talk [1] to be given will address challenges associated to the generation of a Hybrid FEM/test twin model of an automotive electric car engine in partnership with Stellantis:
Describing test outputs
The electric engine case study being detailed combines strong dominance of harmonic responses and un-measured inputs. The harmonic balance vector signal model chosen gives a space/time/frequency approximation of the response.
Choosing model parameters
Geometry, contacts in bolted joints and laminated stacks, non-linear viscoelastic bushings have here a notable impact.
Building a reduced parametric model
This provides a 2 to 3 orders of magnitude speedup that is necessary for any practical application.
Building a Hybrid FEM/test twin model
Test and FEM are combined using an expansion-based state/parameter estimation process.
Etienne Balmes made a presentation entitled “Challenges in structure born vibration models, Hybrid FEM/test twin building, an electric engine case history” at the SIA/AFM meeting https://www.sia.fr/evenements/324-mecanique-electrique
SDT dockId used to perform Experimental Modal Analysis
Experimental modal analysis seeks to extract shape and
resonance properties from test data. While identification algorithms have been
well documented for a long time, at least under the assumption linear behavior,
significant differences continue to exist in the implementation details. In
particular, non-linear optimization of poles and work by sub-bands is an often
overlooked necessity to avoid bias. Once modal data available, the next step is
to obtain test/analysis correlation. This requires topology correlation which
brings its share of errors, which despite being usually small should be
addressed. Global correlation criteria, such as the MAC, are introduced next
but for efficient use should be complemented by a set of procedures to localize
measurement, topology and correlation errors. While the simple evaluation of
correlation is often an industrial objective, the best exploitation of test
results is achieved using hybrid approaches combining test with a model, which
does not need to be exact to usefully complement the measurements. In
particular, expansion methods estimating the full Finite Element responses from
data measured at sensors are particularly useful. Energy based criteria on the
model side (Minimum Dynamic Residual Expansion or the various variants of Error
in Constitutive Relation) have been long known to provide excellent solutions
but deployment has been scarce due to an important numerical cost. Simple model
reduction strategies are shown to give excellent results for industrial models
and open the way for model error localization and updating.
The tutorial is illustrated using standard procedures
implemented in the Structural Dynamics Toolbox which provides experimental
modal analysis, finite element modeling, model reduction and correlation tools
in the MATLAB environment. A main brake squeal application serves a red line
and is complemented by illustrations from other industries.
Etienne Balmès, CEO of SDTools was invited as a keynote speaker at the ISMA (International Symposium on Modal Analysis) conference in Leuven, Belgium on September 17, 2018.
ISMA 2018 is the 28th international conference in a series of annual courses and biennial conferences on structural dynamics, modal testing and noise and vibration engineering, organised by the Department of Mechanical Engineering of the KU Leuven., and is a reference in the field of structural dynamics and vibro-acoustics.
Shapes & DOF: on the use of modal concepts in the context of parametric non-linear studies
Abstract:
Physical responses tend to lie within restricted subspaces even for parametric problems. For a given subspace, the choice of a basis defines Degree Of Freedom (DOF) and this choice may give interesting meaning to the associated amplitudes. Classical modal analysis builds subspaces combining modeshapes and static responses. Parametric loads for non-linear, damped, variable, … structures are discussed to extend the theory and illustrated for test and simulation cases. Challenges in shape extraction and basis generation techniques are then detailed. Introducing the ability to manipulate models with variable junction properties, component material and geometry, load and operating conditions, … opens new questions on the quantification and tracking of changes and objectives throughout design exploration. The definition of a reference linear system and the use of global and/or local modal DOF are shown to provide an interesting perspective.
