TY - GEN

T1 - A Rayleigh-Ritz Autoencoder

AU - Terbuch, Anika

AU - O'Leary, Paul

AU - Ninevski, Dimitar

AU - Hagendorfer, Elias Jan

AU - Schlager, Elke

AU - Windisch, Andreas

AU - Schweimer, Christoph

PY - 2023

Y1 - 2023

N2 - This paper presents a new architecture for unsupervised hybrid machinelearning, called a Rayleigh Ritz Autoencoder (RRAE). It is suitable forapplications in instrumentation and measurement where the system beingobserved by multiple sensors is well modelled as a boundary value problem.The embedding of the admissible functions in the decoder implements a trulyphysics-informed machine learning architecture. The RRAE provides an exactfulfilment of Neumann, Cauchy, Dirichlet or periodic constraints. Only theencoder needs to be trained; consequently, the RRAE is numerically moreefficient during training than traditional autoencoders.The new Rayleigh-Ritz Autoencoder has been applied to an instrumentationand measurement problem in structural monitoring. It involves the fusionof data from multiple sensors and the solution of a boundary value problem.A 1-norm minimization has been chosen to minimize the effects ofnon-Gaussian perturbations and to demonstrate the non-linear abilities ofthe RRAE. The results from the tunnel monitoring application over months ofwork are presented in detail."

AB - This paper presents a new architecture for unsupervised hybrid machinelearning, called a Rayleigh Ritz Autoencoder (RRAE). It is suitable forapplications in instrumentation and measurement where the system beingobserved by multiple sensors is well modelled as a boundary value problem.The embedding of the admissible functions in the decoder implements a trulyphysics-informed machine learning architecture. The RRAE provides an exactfulfilment of Neumann, Cauchy, Dirichlet or periodic constraints. Only theencoder needs to be trained; consequently, the RRAE is numerically moreefficient during training than traditional autoencoders.The new Rayleigh-Ritz Autoencoder has been applied to an instrumentationand measurement problem in structural monitoring. It involves the fusionof data from multiple sensors and the solution of a boundary value problem.A 1-norm minimization has been chosen to minimize the effects ofnon-Gaussian perturbations and to demonstrate the non-linear abilities ofthe RRAE. The results from the tunnel monitoring application over months ofwork are presented in detail."

KW - Physics informed machine learning

KW - Rayleigh-Ritz

KW - Structural monitoring

M3 - Conference contribution

SP - 1

EP - 6

BT - 2023 IEEE International Instrumentation and Measurement Technology Conference (I2MTC) (I2MTC 2023)

PB - Institute of Electrical and Electronics Engineers

CY - Kuala Lumpur, Malaysia

T2 - 2023 IEEE International Instrumentation and Measurement Technology<br/>Conference (I2MTC) (I2MTC 2023)

Y2 - 22 May 2023 through 25 May 2023

ER -