AULA - Commissiekamer 2 Live Meeting
Jul 24, 2024 09:00 - 11:00(Europe/Amsterdam)
20240724T0900 20240724T1100 Europe/Amsterdam MS20.2: Physics-enhanced machine learning and data-driven nonlinear dynamics AULA - Commissiekamer 2 Enoc2024 n.fontein@tudelft.nl
41 attendees saved this session
A transition likelihood maximisation approach to identify nonlinear stochastic dynamicsView Abstract
MS-20 - Physics-enhanced machine learning and data-driven nonlinear dynamics 09:00 AM - 09:20 AM (Europe/Amsterdam) 2024/07/24 07:00:00 UTC - 2024/07/24 07:20:00 UTC
This talk aims to introduce a statistical approach to identify the stochastic dynamic system behind a measured time series data. The method is formulated as a maximum likelihood problem, where we estimate the parameters of the dynamical system generating the data by solving the corresponding optimisation problem. This flexible formulation allows the identification of stochastic dynamical systems even through noisy measurements. We demonstrate the method on synthetic datasets generated by stochastic nonlinear oscillators and investigate statistical properties of the estimators through numerical experiments.
Presenters
HS
Henrik Sykora
Assistant Professor, Budapest University Of Technology And Economics
Recovering Non-stationary Latent Forces by Nonlinear Bayesian FilteringView Abstract
MS-20 - Physics-enhanced machine learning and data-driven nonlinear dynamics 09:20 AM - 09:40 AM (Europe/Amsterdam) 2024/07/24 07:20:00 UTC - 2024/07/24 07:40:00 UTC
The Gaussian process latent force model (GPLFM) has been shown to be an effective and practical method for a number of tasks within dynamics, for example, joint estimation of inputs and states or, more recently, in the recovery of nonlinear restoring forces for system identification. One possible limitation of the GPLFM is that the estimated force is modelled a priori as a Gaussian process, hence is a stationary process. This work extends the model to account for non-stationary force estimation, whether an external or internal forcing, by means of a deep GPLFM formulation.
Presenters
MC
Max Champneys
Post-doctoral Research Associate, University Of Sheffield
Co-Authors
TR
Timothy Rogers
Senior Lecturer In Dynamics, The University Of Sheffield
JL
Joe Longbottom
PhD Student, University Of Sheffield
Quasipotential of nongradient systems via a combined data-driven techniqueView Abstract
MS-20 - Physics-enhanced machine learning and data-driven nonlinear dynamics 09:40 AM - 10:00 AM (Europe/Amsterdam) 2024/07/24 07:40:00 UTC - 2024/07/24 08:00:00 UTC
The quasipotential serves as a natural extension of the potential function in nongradient systems, and it enables predicting the maximum likelihood transition paths, transition rates and expected exit times of metastable dynamical states. In this work, the combination of a neural network and a sparse regression identification algorithm is employed to perform the symbolic discovery of quasipotential functions. The data-driven approach is benchmarked in an archetypal system with know exact quasipotential and applied to the dynamics of a nanomechanical resonator.
Presenters
PB
Pierpaolo Belardinelli
Polytechnic University Of Marche, Ancona, Italy
Co-Authors
BL
Bo Lin
National University Of Singapore
Reduced-order modeling and system identification of nonlinear dynamics through a varational approachView Abstract
MS-20 - Physics-enhanced machine learning and data-driven nonlinear dynamics 10:00 AM - 10:20 AM (Europe/Amsterdam) 2024/07/24 08:00:00 UTC - 2024/07/24 08:20:00 UTC
We present a data-driven, non-intrusive framework with embedded uncertainty quantification to build interpretable reduced-order models (ROMs) using variational autoencoders and variational identification of nonlinear dynamics.
Presenters
AF
Attilio Frangi
Professor, Politecnico Di Milano
Co-Authors Paolo Conti
PhD Student, Politecnico Di Milano
AM
Andrea Manzoni
Politecnico Di Milano
Attention is all you need - an interpretable artificial neural network architectureView Abstract
MS-20 - Physics-enhanced machine learning and data-driven nonlinear dynamics 10:20 AM - 10:40 AM (Europe/Amsterdam) 2024/07/24 08:20:00 UTC - 2024/07/24 08:40:00 UTC
This paper presents an innovative data-driven model order reduction approach for complex dynamical system through a piecewise artificial neural network architecture, which incorporates a state space-based attention mechanism, designed to enhance both predictive performance and interpretability. The proposed framework combines multiple sub-networks, each characterized by distinct architecture (hyper-parameters), to capture diverse patterns and features within the data. The attention mechanism at the culmination of the architecture further enriches the model’s interpretability by dynamically indicating which sub-network is most representative for a given input, offering an insightful decomposition of the system’s phase space. Our framework can be seen as a generalization of the concepts presented by Rewienski et al. [1], evolving the idea to encompass a broader range of models and complex scenarios. The novelty of our approach lies in its ability to learn directly from data both the model order reduction and the phase space partition, rather than relying only on physics knowledge. To demonstrate the efficacy and interpretive capabilities of our framework, we apply it to a network system of 20 Kuramoto oscillators. The results showcase not only good performance in terms of efficiency but also an enhanced ability to glean meaningful insights from the model’s predictions. This balance of high performance and interpretability positions our piecewise approach as a promising tool for applications where understanding the “why” behind predictions is as crucial as the predictions themselves.
Presenters Nico Novelli
PhD Student, Polytechnic University Of Marche, Ancona, Italy
Co-Authors
PB
Pierpaolo Belardinelli
Polytechnic University Of Marche, Ancona, Italy
SL
Stefano Lenci
Professor, Polytechnic University Of Marche, Ancona, Italy
FH
Frank Hellmann
Potsdam Institute For Climate Impact Research, Potsdam, Germany
Inverse Mapping Parameter Updating with Feature Selection based on Measured DataView Abstract
MS-20 - Physics-enhanced machine learning and data-driven nonlinear dynamics 10:40 AM - 11:00 AM (Europe/Amsterdam) 2024/07/24 08:40:00 UTC - 2024/07/24 09:00:00 UTC
To minimize the mismatch between a physical system and its digital twin (constituted by a model), model updating can be applied. In this work, an efficient, online model updating method is employed to update interpretable parameter values of nonlinear dynamics models. This method uses inverse mapping models that map a selected set of measured features to a set of parameter values. To do so, the inverse mapping model is constituted by a neural network that is trained based on simulated (training) data using supervisory learning. Here, the set of selected features is obtained by employing (joint) mutual information-based feature selection. To illustrate the method, it is applied to a high-tech industrial use case for which measurements are collected from a wire bonder. Here, it is shown that the updated models have higher predictive capacity than a reference model of which parameters are manually determined.
Presenters
BK
Bas Kessels
PhD Student, Eindhoven University Of Technology
Co-Authors
RF
Rob Fey
Associate Professor, Eindhoven University Of Technology
NV
Nathan Van De Wouw
Professor, Eindhoven University Of Technology
Professor
,
Politecnico Di Milano
PhD student
,
Eindhoven University Of Technology
PhD student
,
Polytechnic University Of Marche, Ancona, Italy
Polytechnic University Of Marche, Ancona, Italy
Post-doctoral Research Associate
,
University Of Sheffield
+ 1 more speakers. View All
Cambridge University
Professor
,
Politecnico Di Milano
Professor
,
Tokyo Metropolitan University
Dr. Peter Meijers
Post-doctoral researcher
,
Delft University Of Technology
Inverse Mapping Parameter Updating with Feature...
1701680751BMKessels_InverseMappingParameterUpdatingWithFeatureSelectionBasedOnMeasuredData.pdf View Download Presentation Submitted by Bas Kessels 6
Reduced-order modeling and system identificatio...
1704887908Abstract_ENOC_2024.pdf View Download Presentation Submitted by Paolo Conti 6
Attention is all you need - an interpretable ar...
1705251099ENOC_2024_Extended_Abstract__attention_is_all_you_need___an_interpretable_artificial_neural_network_framework.pdf View Download Presentation Submitted by Nico Novelli 4
Quasipotential of nongradient systems via a com...
1711390896Belardinelli_ENOC_2024.pdf View Download Presentation Submitted by Pierpaolo Belardinelli 4
Recovering Non-stationary Latent Forces by Nonl...
1705357682ENOC_Abstract1.pdf View Download Presentation Submitted by Timothy Rogers 2
A transition likelihood maximisation approach t...
1706557880ENOC_2024_ProbabilisticIdentificaiton.pdf View Download Presentation Submitted by Henrik Sykora 2
Upcoming Sessions
203 visits