Uncertainty aware anomaly detection to predict errant beam pulses in the SNS accelerator

Blokland, W.; Ramuhalli, Pradeep; Peters, Charles C.; Yucesan, Yigit A.; Жуков, А. В.; Schram, M.; Rajput, Kishansingh; Jeske, Torri

doi:10.48550/arxiv.2110.12006

Cited by 3 publications

(3 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For example, previous study on using discrete cosine transform for anomaly detection in HVCM waveforms was done by (Pappas, Lu, Schram, & Vrabie, 2021), while (Radaideh, Pappas, Walden, et al, 2022) developed advanced recurrent neural network autoencoder models for time series anomaly detection in the HVCMs powering the RFQ section. Further efforts on applications of machine learning for fault detection in particle accelerators include application of vari-ety of binary classifiers (Rescic, Seviour, & Blokland, 2020), Siamese neural networks (Blokland et al, 2021), adaptive neural networks for time-varying beam control (Scheinker, 2021), and similar others (Edelen et al, 2016). Overall, neural networks have demonstrated a promising potential in the field of fault identification and diagnosis as described in this comprehensive survey (Mohd Amiruddin, Zabiri, Taqvi, & Tufa, 2020).…”

Section: Introductionmentioning

confidence: 92%

Application of Convolutional and Feedforward Neural Networks for Fault Detection in Particle Accelerator Power Systems

et al. 2022

View full text Add to dashboard Cite

High voltage converter modulators (HVCM) provide power to the accelerating cavities of the Spallation Neutron Source (SNS) facility. HVCMs experience catastrophic failures, which increase the downtime of the SNS and reduce beam time. The faults may occur due to different reasons including failures of the resonant capacitor, core saturation due to the magnetic flux, insulated-gate bipolar transistor (IGBT) failures, and others. We recently have setup a HVCM test stand to develop and test machine learning models for anomaly detection and fault prognostics. In this work, we propose binary classifiers and autoencoder architectures based on convolutional (CNN) and feedforward neural networks (FNN) to facilitate distinguishing normal from faulty waveforms coming from the HVCM during operation. The results indicate that the CNN binary classifier is the best model among the four showing very stable performance in the training and testing sets with impressive precision and recall metrics, reaching up to 99% with a very small uncertainty. The FNN classifier shows the least performance with a large uncertainty in its metrics. The performances of the two autoencoders based on CNN and FNN were in between, showing very good performance nonetheless.

show abstract

Section: Introductionmentioning

confidence: 92%

Application of Convolutional and Feedforward Neural Networks for Fault Detection in Particle Accelerator Power Systems

et al. 2022

View full text Add to dashboard Cite

show abstract

“…A novel uncertainty-aware Siamese model to predict upcoming faults (Blokland et al, 2021) that combines the use of uncertainty quantification and a deep Siamese architecture was developed to predict the similarity between beam pulses and provide an uncertainty that includes out of domain errors. The results show that this model outperforms previous results by 4 times in operational regions of interest.…”

Section: B Anomaly Detection and Classificationmentioning

confidence: 99%

Colloquium: Machine learning in nuclear physics

et al. 2022

View full text Add to dashboard Cite

TABLE I. Table of ML methods discussed in this Colloquium with an indication of the main type of learning (S, supervised; U, unsupervised; semi-S, semisupervised). Acronym Method Description Learning type AE, VAE Autoencoders, Variational autoencoders ANN capable of learning efficient representations of the input data without any supervision U ANN Artificial neural network Models for learning defined by connected units (or nodes) and hidden layers with well-defined inputs and outputs S BED Bayesian experimental design Bayesian inference for experimental design S BM Boltzmann machine Generative ANN that can learn a probability distribution from sets of changing inputs U BMA, BMM Bayesian model averaging, Bayesian model mixing Bayesian inference applied to model selection or the combined estimation, or performed over a mixture model S BNN Bayesian neural network ANN where the parameters of the network are represented by probabilities learned by Bayesian inference S BO Bayesian optimization Optimization of functions without an a priori knowledge of functional forms. S and semi-S CNN Convolutional neural network ANN where convolution is used to reduce dimensionalities S EMB Ensemble methods and boosting Methods based on collections of decision trees as simple learners S GAN Generative adversarial network System of two ANNs where a generative network generates outputs while a discriminative network evaluates them U GP Gaussian process Collection of random variables that have a joint Gaussian distribution used in Bayesian inference Semi-S KNN k-nearest neighbors Nonparametric method where inputs consist of the k closest training examples in a dataset S KR Kernel regression Extension of linear regression methods to include nonlinear function kernels S LR Logistic regression Convex optimization method based on maximum likelihood estimate for classification problems S LSTM Long short-term memory RNN capable of learning long-term dependencies S PCA Principal component analysis Dimensionality reduction technique based on retaining the largest eigenvalues of the covariance matrix U REG Linear regression Linear algebra methods used for modeling continuous functions in terms of their explanatory variables S RL Reinforcement learning Learning achieved by trial and error of desired and undesired events Neither S nor U RNN Recurrent neural network ANN where connections between nodes allow for temporal dynamic behavior S SVM Support vector machine Convex optimization techniques with efficient ways to distinguish features in datasets S

show abstract

“…The interest in machine learning for control applications in particle accelerators can be seen in these studies (Nguyen, Lee, Sass, & Shoaee, 1991;Edelen et al, 2016). Uncertainty-aware anomaly detection framework of the errant beam pulses was developed by (Blokland et al, 2021) using Siamese neural networks with ResNet blocks. For a beam-based study with real measured data, the authors of (Rescic, Seviour, & Blokland, 2020) employed different machine learning binary classifiers (e.g.…”

Section: Introductionmentioning

confidence: 99%

Early Fault Detection in Particle Accelerator Power Electronics Using Ensemble Learning

Radaideh¹,

Pappas²,

Wezensky³

et al. 2023

IJPHM

View full text Add to dashboard Cite

Early fault detection and fault prognosis are crucial to ensure efficient and safe operations of complex engineering systems such as the Spallation Neutron Source (SNS) and its power electronics (high voltage converter modulators). Following an advanced experimental facility setup that mimics SNS operating conditions, the authors successfully conducted 21 early fault detection experiments, where fault precursors are introduced in the system to a degree enough to cause degradation in the waveform signals, but not enough to reach a real fault. Nine different machine learning techniques based on ensemble trees, convolutional neural networks, support vector machines, and hierarchical voting ensembles are proposed to detect the fault precursors. Although all 9 models have shown a perfect and identical performance during the training and testing phase, the performance of most models has decreased in the next test phase once they got exposed to realworld data from the 21 experiments. The hierarchical voting ensemble, which features multiple layers of diverse models, maintains a distinguished performance in early detection of the fault precursors with 95% success rate (20/21 tests), followed by adaboost and extremely randomized trees with 52% and 48% success rates, respectively. The support vector machine models were the worst with only 24% success rate (5/21 tests). The study concluded that a successful implementation of machine learning in the SNS or particle accelerator power systems would require a major upgrade in the controller and the data acquisition system to facilitate streaming and handling big data for the machine learning models. In addition, this study shows that the best performing models were diverse and based on the ensemble concept to reduce the bias and hyperparameter sensitivity of individual models.

show abstract

Uncertainty aware anomaly detection to predict errant beam pulses in the SNS accelerator

Cited by 3 publications

References 12 publications

Application of Convolutional and Feedforward Neural Networks for Fault Detection in Particle Accelerator Power Systems

Application of Convolutional and Feedforward Neural Networks for Fault Detection in Particle Accelerator Power Systems

Colloquium: Machine learning in nuclear physics

Early Fault Detection in Particle Accelerator Power Electronics Using Ensemble Learning

Contact Info

Product

Resources

About