Defectnet: Multi-Class Fault Detection on Highly-Imbalanced Datasets

Anantrasirichai, Nantheera; Bull, David

doi:10.1109/icip.2019.8803305

Cited by 15 publications

(8 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The most common architecture of a CNN has the convolution layer connected to a pooling layer, which combines the outputs of neuron clusters at one layer into a single neuron. Some architectures omit pooling layers to obtain dense features [22,23]. Subsequently, activation functions such as tanh (the hyperbolic tangent) or ReLU (Rectified Linear Unit) are applied to introduce non-linearity into the networks [24].…”

Section: Convolutional Neural Networkmentioning

confidence: 99%

A deep learning approach to detecting volcano deformation from satellite imagery using synthetic datasets

Anantrasirichai

Biggs

Albino

et al. 2019

Remote Sensing of Environment

Self Cite

129

View full text Add to dashboard Cite

Satellites enable widespread, regional or global surveillance of volcanoes and can provide the first indication of volcanic unrest or eruption. Here we consider Interferometric Synthetic Aperture Radar (InSAR), which can be employed to detect surface deformation with a strong statistical link to eruption. Recent developments in technology as well as improved computational power have resulted in unprecedented quantities of monitoring data, which can no longer be inspected manually. The ability of machine learning to automatically identify signals of interest in these large InSAR datasets has already been demonstrated, but data-driven techniques, such as convolutional neutral networks (CNN) require balanced training datasets of positive and negative signals to effectively differentiate between real deformation and noise. As only a small proportion of volcanoes are deforming and atmospheric noise is ubiquitous, the use of machine learning for detecting volcanic unrest is more challenging than many other applications. In this paper, we address this problem using synthetic interferograms to train the AlexNet CNN. The synthetic interferograms are composed of 3 parts: 1) deformation patterns based on a Monte Carlo selection of parameters for analytic forward models, 2) stratified atmospheric effects derived from weather models and 3) turbulent atmospheric effects based on statistical simulations of correlated noise. The AlexNet architecture trained with synthetic data outperforms that trained using real interferograms alone, based on classification accuracy and positive predictive value (PPV). However, the models used to generate the synthetic signals are a simplification of the natural processes, so we retrain the CNN with a combined dataset consisting of synthetic models and selected real examples, achieving a final PPV of 82%. Although applying atmospheric corrections to the entire dataset is computationally expensive, it is relatively simple to apply them to the small subset of positive results. This further improves the detection performance without a significant increase in computational burden (PPV of 100%). Thus, we demonstrate that training with synthetic examples can improve the ability of CNNs to detect volcano deformation in satellite images, and propose an efficient workflow for the development of automated systems.

show abstract

Section: Convolutional Neural Networkmentioning

confidence: 99%

A deep learning approach to detecting volcano deformation from satellite imagery using synthetic datasets

Anantrasirichai

Biggs

Albino

et al. 2019

Remote Sensing of Environment

Self Cite

129

View full text Add to dashboard Cite

show abstract

“…Other approaches which do not use analytics of the SCADA data, employ visual inspection of the turbine through a drone and apply Convolutional Neural Networks (CNNs) to process the images and detect common external damages such as erosion or missing teeth in the vortex generator [10]. The image datasets, like the SCADA data, have in common that they are highly imbalanced, which requires specific architectures for the CNNs [11]. Moving back to turbine sensor data, multiclass classification has been attempted on simulations of turbine data through Support Vector Machines (SVMs), succeeding in the isolation of different faults [12].…”

Section: State Of the Artmentioning

confidence: 99%

Machine Learning for Wind Turbine Fault Prediction through the Combination of Datasets from Same Type Turbines

Bösch

2022

Floating Offshore Energy Devices

View full text Add to dashboard Cite

Abstract. Early fault detection in wind turbines is key to reduce both costs and uncertainty in the generation of energy and operation of these structures. The isolation of many wind farms, especially those offshore, makes scheduled maintenance very costly and on many occasions inefficient. In addition, the downtime of these structures is typically long and a predictive solution is much needed to 1) help prepare for the maintenance procedure beforehand, for instance to avoid delays when waiting for the required resources and components for maintenance to be available and, 2) avoid the possibility of more destructive system failures. Predicting failures in such complex systems requires modeling of multiple components in isolation and as a whole. Physics-based and data-based models are used for this purpose, which have been proven useful in this regard. Specifically, Machine Learning algorithms are proven to be a valuable resource in a wide range of problems in this industry, however a solution capable of accurately predicting the range of faults of a particular type of wind turbine is still a challenge. In this paper, we will introduce the capabilities of machine learning for wind turbine fault prediction, as well as a technique to predict different types of faults. We will compare the performance of two well established machine learning algorithms (namely K-Nearest Neighbour and Random Forest classifiers) on real wind turbine data which have produced great levels of prediction accuracy. We also propose data augmentation methods to help enhance the training of ML models when wind turbine data is scarce by merging data from turbines of the same type.

show abstract

“…Data augmentation is also important to reduce the class imbalance. In this context, some works propose new loss functions (or adaptations to existing ones) to avoid training problems [9][10][11][12][13]. These techniques can also be employed in union with ChessMix.…”

Section: Related Workmentioning

confidence: 99%

ChessMix: Spatial Context Data Augmentation for Remote Sensing Semantic Segmentation

Pereira¹,

Santos²

2021

Preprint

View full text Add to dashboard Cite

Labeling semantic segmentation datasets is a costly and laborious process if compared with tasks like image classification and object detection. This is especially true for remote sensing applications that not only work with extremely high spatial resolution data but also commonly require the knowledge of experts of the area to perform the manual labeling. Data augmentation techniques help to improve deep learning models under the circumstance of few and imbalanced labeled samples. In this work, we propose a novel data augmentation method focused on exploring the spatial context of remote sensing semantic segmentation. This method, ChessMix, creates new synthetic images from the existing training set by mixing transformed minipatches across the dataset in a chessboard-like grid. ChessMix prioritizes patches with more examples of the rarest classes to alleviate the imbalance problems. The results in three diverse well-known remote sensing datasets show that this is a promising approach that helps to improve the networks' performance, working especially well in datasets with few available data. The results also show that ChessMix is capable of improving the segmentation of objects with few labeled pixels when compared to the most common data augmentation methods widely used.

show abstract

Defectnet: Multi-Class Fault Detection on Highly-Imbalanced Datasets

Cited by 15 publications

References 18 publications

A deep learning approach to detecting volcano deformation from satellite imagery using synthetic datasets

A deep learning approach to detecting volcano deformation from satellite imagery using synthetic datasets

Machine Learning for Wind Turbine Fault Prediction through the Combination of Datasets from Same Type Turbines

ChessMix: Spatial Context Data Augmentation for Remote Sensing Semantic Segmentation

Contact Info

Product

Resources

About