Transfer learning for radioactive particle tracking

Lindner, Guilherme; Shi, Sai; Vučetić, Slobodan; Mišković, Sanja

doi:10.1016/j.ces.2021.117190

Cited by 10 publications

(6 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This possibly can be achieved through common Monte Carlo based codes available in the literature for photon transport simulations such as MCNP [ 27 , 28 ] and GEANT4 [ 29 ]. To the best of the authors’ knowledge, the only studies applying such codes for the RPT calibration process are so far reported by [ 21 , 30 , 31 ]. The authors of [ 21 , 32 ] applied the so-called Monte Carlo N-Particle code version 5 (MCNP5) to simulate the RPT calibration process and proposed a new algorithm for the reconstruction process.…”

Section: Introductionmentioning

confidence: 99%

“…The GIPPE-RPT software also is integrated by a computational fluid dynamics (CFD) solver to estimate, and transfer to the GEANT4, the phase holdup distribution of the investigated multiphase flow system. The authors of [ 31 ] also applied the GIPPE-RPT software with the help of the transfer learning (TL) approach to examine the re-using of historical calibration data when the operation conditions of the system have been changed. They concluded that the TL approach can be used to exploit historical RPT calibration data collected under various conditions when training an RPT model under a new condition.…”

Section: Introductionmentioning

confidence: 99%

“…Example of that can be found in high-pressure or high-temperature multiphase flow systems or at the presence of dense internals inside the system. Additionally, it was concluded that the inclusion of historical data by means of the TL approach, when sufficient RPT calibration data for a new condition is available, showed less reconstruction accuracy, and hence is not recommended [ 31 ]. However, despite major progress made using these Monte Carlo-based models [ 21 , 30 , 31 ], the experimental validation for such models is a must for a more reliable RPT calibration process.…”

Section: Introductionmentioning

confidence: 99%

“…Additionally, it was concluded that the inclusion of historical data by means of the TL approach, when sufficient RPT calibration data for a new condition is available, showed less reconstruction accuracy, and hence is not recommended [ 31 ]. However, despite major progress made using these Monte Carlo-based models [ 21 , 30 , 31 ], the experimental validation for such models is a must for a more reliable RPT calibration process. The main process in the RPT technique is the actual tracking, which is an absolute experimental work.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

GEANT4 Simulation for Radioactive Particle Tracking (RPT) Technique

Alghamdi

Al-Juwaya

Alomari

et al. 2022

Sensors

View full text Add to dashboard Cite

In the past two decades, the radioactive particle tracking (RPT) measurement technique has been proven to visualize flow fields of most multiphase flow systems of industrial interest. The accuracy of RPT, and hence the data obtained, depend largely on the calibration process, which stands here as a basis for two subsequent processes: tracking and reconstruction. However, limitations in the RPT calibration process can be found in different experimental constrains and in assumptions made in the classical Monte Carlo approach used to simulate number of counts received by the detectors. Therefore, in this work, we applied a GEANT4-based Monte Carlo code to simulate the RPT calibration process for an investigated multiphase flow system (i.e., gas–liquid bubble column). The GEANT4 code was performed to simulate the number of counts received by 28 scintillation detectors for 931 known tracer positions while capturing all the types of photon interaction and overcoming solids’ angle limitations in classical approaches. The results of the simulation were validated against experimental data obtained using an automated RPT calibration device. The results showed a good agreement between the simulated and experimental counts, where the maximum absolute average relative deviation detected was about 5%. The GEANT4 model typically predicted the number of counts received by all the detectors; however, it over-estimated the counts when the number of primary events applied in the model was not the optimal.

show abstract

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

GEANT4 Simulation for Radioactive Particle Tracking (RPT) Technique

Alghamdi

Al-Juwaya

Alomari

et al. 2022

Sensors

View full text Add to dashboard Cite

show abstract

“…This method is particularly advantageous when the data provided for inference is not sufficient or is difficult to label (Jaini et al, 2017). Transfer learning has been integrated into various signal processing applications, including trajectory tracking and radioactive particle tracking (Pereida et al, 2018;Lindner et al, 2022). Whereas many machine learning methods are applicable to learning a set of parameters of parametric models, Bayesian nonparametric methods allow for probability models from infinite dimensional families.…”

Section: Introductionmentioning

confidence: 99%

Bayesian Nonparametric Learning and Knowledge Transfer for Object Tracking Under Unknown Time-Varying Conditions

Alotaibi

Papandreou-Suppappola

2022

Front. Signal Process.

View full text Add to dashboard Cite

We consider the problem of a primary source tracking a moving object under time-varying and unknown noise conditions. We propose two methods that integrate sequential Bayesian filtering with transfer learning to improve tracking performance. Within the transfer learning framework, multiple sources are assumed to perform the same tracking task as the primary source but under different noise conditions. The first method uses Gaussian mixtures to model the measurement distribution, assuming that the measurement noise intensity at the learning sources is fixed and known a priori and the learning and primary sources are simultaneously tracking the same source. The second tracking method uses Dirichlet process mixtures to model noise parameters, assuming that the learning source measurement noise intensity is unknown. As we demonstrate, the use of Bayesian nonparametric learning does not require all sources to track the same object. The learned information can be stored and transferred to the primary source when needed. Using simulations for both high- and low-signal-to-noise ratio conditions, we demonstrate the improved primary tracking performance as the number of learning sources increases.

show abstract