SAM
https://sam.ensam.eu:443
The DSpace digital repository system captures, stores, indexes, preserves, and distributes digital research material.Sun, 21 Apr 2024 09:26:26 GMT2024-04-21T09:26:26ZCross-structures Deep Transfer Learning through Kantorovich potentials for Lamb Waves based Structural Health Monitoring
http://hdl.handle.net/10985/23690
Cross-structures Deep Transfer Learning through Kantorovich potentials for Lamb Waves based Structural Health Monitoring
POSTORINO, Hadrien; REBILLAT, Marc; MECHBAL, Nazih; MONTEIRO, Eric
In Lamb Waves based Structural Health Monitoring (LWSHM) of composite aeronautic structures, Deep Learning (DL) methods have proven to be promising to monitor damage using the signals collected by piezoelectric sensors (PZTs). However, those data driven algorithms are strongly problem dependent: any structural change dramatically impacts the accuracy of the predictions and the generalization of the learnt algorithms to other structures within the fleet is impossible. Transfer Learning (TL) promises to face that issue by capitalizing on the knowledge acquired on a given structure to transfer it on another
from the fleet. An original TL approach based on the Optimal Transport (OT) theory is proposed here to handle this issue. OT provides a rigorous mathematical framework for TL that can be practically implemented using Input Convex Neural Networks modelling Kantorovich potentials but that has never been used for LWSHM. Using OT, the knowledge acquired on a rich LW database is transferred to poorer LW databases collected on different structures with rising structural divergences. A Structural
Index (SI) is defined and used to compute the gap between those different structures and can be used to estimate a priori the necessity of the use of TL methods. The proposed OT based TL method for LWSHM manages to reduce by almost 50% the predictions errors between numerical structures with strong differences (bias in mechanical properties and erroneous PZT position) in comparison with standard approaches. That leads to a promising approach to combine rich numerical database with
poorer database in order to build robust algorithms for LWSHM of a fleet of aeronautical composite structures.
Sat, 01 Apr 2023 00:00:00 GMThttp://hdl.handle.net/10985/236902023-04-01T00:00:00ZPOSTORINO, HadrienREBILLAT, MarcMECHBAL, NazihMONTEIRO, EricIn Lamb Waves based Structural Health Monitoring (LWSHM) of composite aeronautic structures, Deep Learning (DL) methods have proven to be promising to monitor damage using the signals collected by piezoelectric sensors (PZTs). However, those data driven algorithms are strongly problem dependent: any structural change dramatically impacts the accuracy of the predictions and the generalization of the learnt algorithms to other structures within the fleet is impossible. Transfer Learning (TL) promises to face that issue by capitalizing on the knowledge acquired on a given structure to transfer it on another
from the fleet. An original TL approach based on the Optimal Transport (OT) theory is proposed here to handle this issue. OT provides a rigorous mathematical framework for TL that can be practically implemented using Input Convex Neural Networks modelling Kantorovich potentials but that has never been used for LWSHM. Using OT, the knowledge acquired on a rich LW database is transferred to poorer LW databases collected on different structures with rising structural divergences. A Structural
Index (SI) is defined and used to compute the gap between those different structures and can be used to estimate a priori the necessity of the use of TL methods. The proposed OT based TL method for LWSHM manages to reduce by almost 50% the predictions errors between numerical structures with strong differences (bias in mechanical properties and erroneous PZT position) in comparison with standard approaches. That leads to a promising approach to combine rich numerical database with
poorer database in order to build robust algorithms for LWSHM of a fleet of aeronautical composite structures.Transfer Learning to close the gap between experimental and numerical data
http://hdl.handle.net/10985/23347
Transfer Learning to close the gap between experimental and numerical data
POSTORINO, Hadrien; REBILLAT, Marc; MECHBAL, Nazih; MONTEIRO, Eric
The deployment of Deep Learning (DL) strategies is particularly advantageous in Structural Health Monitoring (SHM) based of lamb Wave (LW) propagation due to the high quantity of data collected by the network of piezoelectric transducers (PZT) during all the life cycle of the composite structure. However, such strategies rely on large training databases, difficult to collect experimentally. The use of numerical simulations faces that issue, but the modelsnever fit perfectly to the real structures, leading to error of diagnostic.
We propose here to use Transfer Learning (TL) approaches to reduce the predictions errors of a Convolutional Neural Network (CNN) trained with numerical data. The network predicts the size and the position of a damage on a composite plate equipped with PZT. It is trained on a large source database composed of different damage scenarios on a composite plate. A second smaller target database is generated with small variations on the mechanic properties and PZT positions to simulate manufacturing uncertainties. Those uncertainties lead to prediction errors of the CNN. A Domain Adaptation (DA) based on Optimal Transport (OT) is used to project the target data on the source domain and therefore reduces the predictions error of the CNN. These TL approach should allow us to close the gap between experimental and numerical data.
Sun, 01 May 2022 00:00:00 GMThttp://hdl.handle.net/10985/233472022-05-01T00:00:00ZPOSTORINO, HadrienREBILLAT, MarcMECHBAL, NazihMONTEIRO, EricThe deployment of Deep Learning (DL) strategies is particularly advantageous in Structural Health Monitoring (SHM) based of lamb Wave (LW) propagation due to the high quantity of data collected by the network of piezoelectric transducers (PZT) during all the life cycle of the composite structure. However, such strategies rely on large training databases, difficult to collect experimentally. The use of numerical simulations faces that issue, but the modelsnever fit perfectly to the real structures, leading to error of diagnostic.
We propose here to use Transfer Learning (TL) approaches to reduce the predictions errors of a Convolutional Neural Network (CNN) trained with numerical data. The network predicts the size and the position of a damage on a composite plate equipped with PZT. It is trained on a large source database composed of different damage scenarios on a composite plate. A second smaller target database is generated with small variations on the mechanic properties and PZT positions to simulate manufacturing uncertainties. Those uncertainties lead to prediction errors of the CNN. A Domain Adaptation (DA) based on Optimal Transport (OT) is used to project the target data on the source domain and therefore reduces the predictions error of the CNN. These TL approach should allow us to close the gap between experimental and numerical data.Experimental Damage Localization and Quantification with a Numerically Trained Convolutional Neural Network
http://hdl.handle.net/10985/23342
Experimental Damage Localization and Quantification with a Numerically Trained Convolutional Neural Network
POSTORINO, Hadrien; REBILLAT, Marc; MECHBAL, Nazih; MONTEIRO, Eric
Structural Health Monitoring (SHM) based on Lamb wave propagation is a promising technology to optimize maintenance costs, enlarge service life and improve safety of aircrafts. A large quantity of data is collected during all the life cycle of the structure under monitoring and must be analysed in real time. We propose here to use 1D-CNN to estimate the severity and the localisation of a damage with the signals measured on a composite structure monitored with piezoelectric transducers (PZT). Two architectures have been tested: one takes for input the difference of the time signals of two different states and the second takes for in-puts temporal damage indexes. Those simple networks with a few layers predict with high precision the position and the severity of a damage in a composite plate. The evaluations on different cases show the robustness to simulated manufacturing uncertainties and noise. An evaluation on experimental measurement shows promising results to localise a damage on a real plate with a CNN trained with numerical data.
Wed, 01 Jun 2022 00:00:00 GMThttp://hdl.handle.net/10985/233422022-06-01T00:00:00ZPOSTORINO, HadrienREBILLAT, MarcMECHBAL, NazihMONTEIRO, EricStructural Health Monitoring (SHM) based on Lamb wave propagation is a promising technology to optimize maintenance costs, enlarge service life and improve safety of aircrafts. A large quantity of data is collected during all the life cycle of the structure under monitoring and must be analysed in real time. We propose here to use 1D-CNN to estimate the severity and the localisation of a damage with the signals measured on a composite structure monitored with piezoelectric transducers (PZT). Two architectures have been tested: one takes for input the difference of the time signals of two different states and the second takes for in-puts temporal damage indexes. Those simple networks with a few layers predict with high precision the position and the severity of a damage in a composite plate. The evaluations on different cases show the robustness to simulated manufacturing uncertainties and noise. An evaluation on experimental measurement shows promising results to localise a damage on a real plate with a CNN trained with numerical data.Towards an industrial deployment of PZT based SHM processes: A dedicated metamodel for Lamb wave propagation
http://hdl.handle.net/10985/19570
Towards an industrial deployment of PZT based SHM processes: A dedicated metamodel for Lamb wave propagation
POSTORINO, Hadrien; REBILLAT, Marc; MONTEIRO, Eric; MECHBAL, Nazih
Numerical simulations of Structural Health Monitoring processes based on wave propagation can be very costly in terms of computation time, especially for complex aeronautic composite structures, and therefore strongly limits the deployment of industrial applications. Metamodels build a relatively simple relationship between inputs and outputs from a set of data and thus can overcome that difficulty. A metamodel based on radial basis functions interpolation is build in order to predict a Lamb Wave measurement on a damaged composite plate equipped by a network of 3 piezoelectric elements. The input parameters describe the position of the damage. This surrogate model is used to predict the measured signals for new damage configurations with a limited computational cost. Moreover, this metamodel is used in a reverse way to solve the inverse problem. A swarm particle optimisation algorithm seeks to find the position of a damage from a set of simulated signals.This approach allows us to identify correctly the damage localisation for an unknown configuration, providing therefore a new method for damage localization.
Wed, 01 Jan 2020 00:00:00 GMThttp://hdl.handle.net/10985/195702020-01-01T00:00:00ZPOSTORINO, HadrienREBILLAT, MarcMONTEIRO, EricMECHBAL, NazihNumerical simulations of Structural Health Monitoring processes based on wave propagation can be very costly in terms of computation time, especially for complex aeronautic composite structures, and therefore strongly limits the deployment of industrial applications. Metamodels build a relatively simple relationship between inputs and outputs from a set of data and thus can overcome that difficulty. A metamodel based on radial basis functions interpolation is build in order to predict a Lamb Wave measurement on a damaged composite plate equipped by a network of 3 piezoelectric elements. The input parameters describe the position of the damage. This surrogate model is used to predict the measured signals for new damage configurations with a limited computational cost. Moreover, this metamodel is used in a reverse way to solve the inverse problem. A swarm particle optimisation algorithm seeks to find the position of a damage from a set of simulated signals.This approach allows us to identify correctly the damage localisation for an unknown configuration, providing therefore a new method for damage localization.