Filtern
Dokumenttyp
- Zeitschriftenartikel (8)
- Vortrag (8)
- Beitrag zu einem Tagungsband (7)
- Posterpräsentation (1)
Sprache
- Englisch (23)
- Mehrsprachig (1)
Schlagworte
- Machine learning (20)
- Structural health monitoring (9)
- Fiber optic sensors (8)
- Hydrogen (6)
- BOFDA (5)
- Artificial neural networks (4)
- Composites (4)
- Fiber optics sensors (4)
- Brillouin distributed sensing (3)
- Pressure vessels (3)
Organisationseinheit der BAM
Eingeladener Vortrag
- nein (8)
We propose to use artificial neural networks (ANNs) for raw measurement data interpolation and signal shift computation and to demonstrate advantages for wavelength-scanning coherent optical time domain reflectometry (WS-COTDR) and dynamic strain distribution measurement along optical fibers. The ANNs are trained with synthetic data to predict signal shifts from wavelength scans. Domain adaptation to measurement data is achieved, and standard correlation algorithms are outperformed. First and foremost, the ANN reduces the data analysis time by more than two orders of magnitude, making it possible for the first time to predict strain in real-time applications using the WS-COTDR approach. Further, strain noise and linearity of the sensor response are improved, resulting in more accurate measurements. ANNs also perform better for low signal-to-noise measurement data, for a reduced length of correlation input (i.e., extended distance range), and for coarser sampling settings (i.e., extended strain scanning range). The general applicability is demonstrated for distributed measurement of ground movement along a dark fiber in a telecom cable. The presented ANN-based techniques can be employed to improve the performance of a wide range of correlation or interpolation problems in fiber sensing data analysis and beyond.
Distributed humidity fiber-optic sensor based on BOFDA using a simple machine learning approach
(2022)
We report, to our knowledge for the first time, on distributed relative humidity sensing in silica polyimide-coated optical fibers using Brillouin optical frequency domain analysis (BOFDA). Linear regression, which is a simple and well-interpretable algorithm in machine learning and statistics, is utilized. The algorithm is trained using as features the Brillouin frequency shifts and linewidths of the fiber’s multipeak Brillouin spectrum. To assess and improve the effectiveness of the regression algorithm, we make use of machine learning concepts to estimate the model’s uncertainties and select the features that contribute most to the model’s performance. In addition to relative humidity, the model is also able to simultaneously provide distributed temperature information addressing the well-known cross-sensitivity effects.
In the last few years, the use of machine learning has emerged in the field of distributed fiber optic sensors as a promising approach to enhance their performance and provide new capabilities. In this study, we use machine learning for simultaneous measurements of temperature and humidity in polyimide (PI)-coated optical fibers based on Brillouin Brillouin optical frequency domain analysis (BOFDA). Different non-linear machine learning algorithms are employed, namely polynomial regression, decision trees and artificial neural networks (ANNs), and their discrimination performance is benchmarked against that of the conventional linear regression. The performance is evaluated using leave-one-out cross-validation to ensure that the models are reliable and able to generalize well on new data. We show that nonlinear machine learning algorithms outperform the conventional linear regression and thus could pave the way towards simultaneous cost-effective temperature and humidity distributed sensing, which has the potential to find attractive new applications in the field of civil and geotechnical engineering, from structural health monitoring of dikes and bridges to subsea cables and long pipelines corrosion detection.
Brillouin Optical frequency domain analysis (BOFDA) is a well-known technique in distributed Fiber sensing for temperature and strain monitoring. In this study, we applied a signal post-processing method based on convolutional neural networks (CNNs) and we show that in comparison to conventional methods based on Lorentzian curve fitting (LCF), our CNN model is more robust against noise and can shorten the measurement time by more than 9 times. This will open the way for applications, where faster monitoring is needed.
We propose a method based on artificial neural networks to extract strain information from wavelength-scanning coherent optical time domain reflectometry (C-OTDR) data. Our neural network algorithm performs more than two orders of magnitude faster than the conventional approach.
This is due to the highly parallel evaluation of the neural networks on a GPU accelerated computer and the fact that conventional correlation and interpolation analysis needs many Iteration steps. This opens the way for real-time C-OTDR strain sensing because the neural Network strain predictions require less time than the measurements themselves. Real-time data Analysis enables long-term sensing e.g. in structural health monitoring, because the large amount of raw data does not have to be stored but can immediately be reduced to the strain data of interest.
Time-Efficient Convolutional Neural Network-Assisted Brillouin Optical Frequency Domain Analysis
(2021)
To our knowledge, this is the first report on a machine-learning-assisted Brillouin optical frequency domain analysis (BOFDA) for time-efficient temperature measurements. We propose a convolutional neural network (CNN)-based signal post-processing method that, compared to the conventional Lorentzian curve fitting approach, facilitates temperature extraction. Due to its robustness against noise, it can enhance the performance of the system. The CNN-assisted BOFDA is expected to shorten the measurement time by more than nine times and open the way for applications, where faster monitoring is essential.
Short presentation of the PhD project in machine learning based Brillouin distributed sensing. Machine learning can be used to enhance the performance of BOFDA and reduce considerably the measurement time. Apart from this, ML can also be used to extract more information from the Brillouin gain spectrum in order to render the temperature and strain discrimination possible
Machine learning predictions of surface migration barriers in nucleation and non-equilibrium growth
(2021)
Machine learning is playing an increasing role in the discovery of new materials and may also facilitate the search for optimum growth conditions for crystals and thin films. Here, we perform kinetic Monte-Carlo simulations of sub-monolayer growth. We consider a generic homoepitaxial growth scenario that covers a wide range of conditions with different diffusion barriers (0.4–0.55 eV) and lateral binding energies (0.1–0.4 eV). These simulations are used as a training data set for a convolutional neural network that can predict diffusion barriers and binding energies. Specifically, a single Monte-Carlo image of the morphology is sufficient to determine the energy barriers with an accuracy of approximately 10 meV and the neural network is tolerant to images with noise and lower than atomic-scale resolution. We believe this new machine learning method will be useful for fundamental studies of growth kinetics and growth optimization through better knowledge of microscopic parameters.
In the last few years, the use of machine learning has emerged in the field of distributed fiber optic sensors as a promising approach to enhance their performance and provide new capabilities. In this study, we use machine learning for simultaneous measurements of temperature and humidity in polyimide (PI)-coated optical fibers based on Brillouin Brillouin optical frequency domain analysis (BOFDA). Different non-linear machine learning algorithms are employed, namely polynomial regression, decision trees and artificial neural networks (ANNs), and their discrimination performance is benchmarked against that of the conventional linear regression. The performance is evaluated using leave-one-out cross-validation to ensure that the models are reliable and able to generalize well on new data. We show that nonlinear machine learning algorithms outperform the conventional linear regression and thus could pave the way towards simultaneous cost-effective temperature and humidity distributed sensing, which has the potential to find attractive new applications in the field of civil and geotechnical engineering, from structural health monitoring of dikes and bridges to subsea cables and long pipelines corrosion detection.
Brillouin Optical Frequency Domain Analysis (BOFDA) is a powerful and well-established method for static distributed sensing of temperature and strain. Recently, we demonstrated a BOFDA system based on convolutional neural network which shortens the measurement time considerably. In this paper, we apply leave-one-out cross validation to evaluate the generalization performance and provide an unbiased and reliable machine learning model for a time-efficient BOFDA system.