Mixing Augmentation Research Articles

Research on intelligent recognition method for self‐blast state of glass insulator based on mixed data augmentation

AbstractAutomatically and accurately detecting the self‐blast state of glass insulators is of great significance to operation and maintenance of transmission lines. To solve the shortcomings of the existing open‐loop cognitive models to detect the self‐blast state of glass insulators, this study explores a mixed data augmentation‐based intelligent recognition method to detect the self‐blast state of the glass insulator, by imitating the human cognitive mode. Firstly, generative adversarial network is utilised to obtain the high‐quality generative self‐blast samples of the glass insulator, and the non‐generative data augmentation techniques is used to obtain rich sample features. Secondly, considering the characteristics of aerial images such as large scale variations, variable shooting angles and complex backgrounds, feature maps with strong semantics and adaptive multi‐scale fusion are extracted using the feature pyramid network with adaptive hierarchy and the multi‐deformable convolutional network. Then, the extracted feature maps are transmitted to a two‐dimensional stochastic configuration network that can adaptively generate hidden nodes and basis functions so as to develop the self‐blast state classification criteria with universal approximation capability. Thirdly, based on the generalised error and entropy theory, the semantic error entropy evaluation indices of recognition results are defined to evaluate in real time, the credibility of the uncertain recognition results for the self‐blast state of the glass insulator. Then, based on transfer learning and the established self‐optimising feedback mechanism for feature pyramid network, the self‐optimising adjustment and reconstruction of the feature map space with strong semantics and multi‐scale fusion and its classification criteria are realised. Finally, the stacking method is applied to integrate the recognition results of the feature pyramid network with adaptive hierarchy and multi‐channel deformable convolutional networks to improve the robustness of the recognition model. Results of experimental comparison with other machine learning and deep learning methods verify the feasibility and effectiveness of the proposed method.

Few-Shot Continuous Authentication for Mobile-Based Biometrics

The rapid growth of smartphone financial services raises the need for secure mobile authentication. Continuous authentication is a user-friendly way to strengthen the security of smartphones by implicitly monitoring a user’s identity through sessions. Mobile continuous authentication can be viewed as an anomaly detection problem in which models discriminate between one genuine user and the rest of the impostors (anomalies). In practice, complete impostor profiles are hardly available due to the time and monetary cost, while leveraging genuine data alone yields poorly generalized models due to the lack of knowledge about impostors. To address this challenge, we recast continuous mobile authentication as a few-shot anomaly detection problem, aiming to enhance the inference robustness of unseen impostors by using partial knowledge of available impostor profiles. Specifically, we propose a novel deep learning-based model, namely a local feature pooling-based temporal convolution network (LFP-TCN), which directly models raw sequential mobile data, aggregating global and local feature information. In addition, we introduce a random pattern mixing augmentation to generate class-unconstrained impostor data for the training. The augmented pool enables characterizing various impostor patterns from limited impostor data. Finally, we demonstrate practical continuous authentication using score-level fusion, which prevents long-term dependency or increased model complexity due to extended re-authentication time. Experiments on two public benchmark datasets show the effectiveness of our method and its state-of-the-art performance.

Towards blind detection of steganography in low‐bit‐rate speech streams

To prevent the abuse of low-rate speech-based steganography from threatening cyberspace security, the corresponding steganalysis approaches have been developed and received significant attention from research community. However, most existing steganalysis methods assume that steganography methods are known in advance, which in practice is impractical. That is why, in this paper, we present three blind detection schemes suitable for steganography in low-bit-rate speech streams. The first is based on mixed sample data augmentation. It randomly selects a certain proportion of steganographic samples from the sample set of each steganographic method to form a training set together with the original carrier samples for training to enhance the robustness of the model. The second relies on decision fusion where first step is to train a dedicated classification model for each steganography method and then use a majority voting mechanism in the detection stage to fuse the outputs of each model to give the final detection result. Compared to the other two steganalysis schemes, the third one design the detection model based on self-paced ensemble according to the distribution characteristics of speech samples. Its main idea is to fully train multiple base classifiers through multiple iterations as well as under-sampling processes, and organically fuse them to form a powerful ensemble classifier. In each iteration, differing from the traditional ensemble classifier solution, we put more attention to the steganographic samples at the decision boundary for the under-sampling process of the steganography set composed of multiple steganography methods, rather than randomly selecting steganographic samples. The steganographic samples at the decision boundary are searched using the classification hardness given by the ensemble classifier trained in the last iteration, which is more informative and more conducive to improve the performance of base classifiers. The experimental results show that the proposed three schemes can achieve efficient blind detection for low-bit-rate speech-based steganography, and the steganalysis scheme based on the self-paced ensemble has the best performance. Specifically, when the embedding rate is at 30%, the accuracy of the steganalysis scheme based on self-paced ensemble is more than 85%, while the accuracy of the other two steganalysis method is less than 80%. Additionally, the steganalysis scheme based on the self-paced ensemble learning even outperforms dedicated detectors for specific steganographic methods in terms of recall for steganographic sample detection.

Numerical Study of the Induced Shock on the Mixing Augmentation of Hydrogen Counter-Flow Jet in the Supersonic Flow

In the scramjet engine, the injection and mixing of fuel are the basis for organizing efficient combustion. The shock wave/jet shear layer interaction is one of the methods used to enhance fuel mixing in supersonic flow. In this research, the effect of the induced shock wave on the mixing augmentation of fuel counter-flow jet is studied in the supersonic crossflow. The results show that the symmetrical setting of the shock wave generators (SWGs) has the highest mixing efficiency but brings a greater total pressure loss. In the specific range, the mixing efficiency increases with the increase in the angle of the SWG, while the total pressure recovery coefficient is the opposite. The induced oblique shock wave acting on the front of the fuel jet has an obvious influence on the flow field, resulting in the highest mixing efficiency. In the end, the analytic hierarchy process (AHP) is used to evaluate the extent of mixing augmentation by mixing parameters, which provides a decision-making idea for the scheme decision of mixing augmentation.

LMix: regularization strategy for convolutional neural networks

Deep convolutional neural networks perform well in the field of computer vision, but exhibit undesirable behaviors such as memorization and sensitivity to adversarial examples. Therefore, proper regularization strategies are needed to alleviate these problems. Currently, regularization strategies with mixed sample data augmentation perform very well, and these algorithms allow the network to generalize better, improve the baseline performance of the model. However, interpolation-based mixed sample data augmentation distorts the data distribution, while masking-based mixed sample data augmentation results in excessive information loss for overly regular shapes of masks. Although mixed sample data augmentation is proven to be an effective method to improve the baseline performance, generalization ability and robustness of deep convolutional models, there is still room for improvement in terms of maintaining the of image local consistency and image data distribution. In this paper, we propose a new mixed sample data augmentation-LMix, which uses random masking to increase the number of masks in the image to maintain the data distribution, and high-frequency filtering to sharpen the image to highlight recognition regions. We applied the method to train CIFAR-10, CIFAR-100, SVHN, and Tiny-ImageNet datasets under the PreAct-ResNet18 model to evaluate the method, and obtained the latest results of 96.32, 79.85, 97.01, and 64.16%, respectively, which are 1.70, 4.73, and 8.06% higher than the optimal baseline accuracy. The LMix algorithm improves the generalization ability of the state-of-the-art neural network architecture and enhances the robustness to adversarial samples.

3LPR: A three-stage label propagation and reassignment framework for class-imbalanced semi-supervised learning

Semi-supervised learning (SSL) has been studied widely in standard benchmark datasets; however, real-world data often exhibit class-imbalanced distributions, which pose significant challenges for deep semi-supervised models. To address this issue, we design a three-stage learning framework, 3LPR, by combining unsupervised feature extraction, graph-based Label Propagation, and mixed data augmentation (MDA)-based label Reassignment. Specifically, we first explore the performance of supervised and unsupervised learning for feature extraction of class-imbalanced data and then establish our first stage of feature extraction through unsupervised learning. Then, we adopt graph network-based offline label propagation and sieving to effectively expand the labeled set to overcome the excessive label bias in the classifier during the training process. Finally, we propose a label reassignment (LRA) algorithm for class-imbalanced semi-supervised learning (CISSL) to train the expanded dataset, where the MDA strategy is adopted but with the label reassigned. The experimental results demonstrate that the proposed 3LPR framework for CISSL outperforms other state-of-the-art methods on various datasets.

Mixing enhancement of supersonic flow induced by splitter plate cavity

Fast and efficient mixing of fuel and oxidizer in a RBCC combustor is crucial for the efficiency improvement of engine propulsion system. Whereas, the existence of compressibility effects of high-speed flow notably suppresses the mixing process. In present work, the strategy of splitter plate cavity is proposed to promote mixing. The influences of splitter plate cavity with different length to depth ratio (r) on mixing process are researched. Fine vortex structures are well captured and the newly found T-shaped structures and their evolution dynamics induced by the cavity instability are thoroughly analyzed. The results indicate that tremendous growth of the mixing layer in flow transition region is realized. For the dimensionless streamwise velocity distribution, the recognized appearance of multiple peaks at far flow field for splitter plate cavity with r=4 and 8 leads to great velocity gradient in local position, which can promote mixing process of upper and lower streams. The distributions of streamwise turbulence intensity, turbulent kinetic energy and transversely integrated turbulent kinetic energy suggest that vigorous dynamic behaviors of T-shaped structures can obviously increase local turbulence, especially for splitter plate cavity with r=8. The structure topology analysis results reveal that larger size of the evolutionary T-shaped structure and more intense rotation process of these structures can undoubtedly engulf more main stream into the mixing region to achieve mixing augmentation.

An overview of mixing augmentation methods and augmentation strategies

Deep Convolutional Neural Networks have made an incredible progress in many Computer Vision tasks. This progress, however, often relies on the availability of large amounts of the training data, required to prevent over-fitting, which in many domains entails significant cost of manual data labeling. An alternative approach is application of data augmentation (DA) techniques that aim at model regularization by creating additional observations from the available ones. This survey focuses on two DA research streams: image mixing and automated selection of augmentation strategies. First, the presented methods are briefly described, and then qualitatively compared with respect to their key characteristics. Various quantitative comparisons are also included based on the results reported in recent DA literature. This review mainly covers the methods published in the materials of top-tier conferences and in leading journals in the years 2017–2021.

From fundamental to practical aspects of bubble-induced agitation

<h2>Abstract</h2> While the buoyancy-driven rise of bubbles in a liquid seems to be a simple problem, the bubble dynamics, interfacial phenomena, and the interaction between the rising bubbles and surrounding liquid flow turned out to be very complex, which have fascinated the multiphase flow society for a long time. For example, the rise dynamics of a bubble is governed by the wake structure behind the bubble, which is also known to be the origin of bubble-induced agitation (turbulence) on the liquid flow. The understanding of this interaction is critical because some wish to utilize bubbles for various engineering purposes while others to detect and remove them. Here, the fundamental nature of bubble-induced flow or disturbance that we have learned will be first discussed briefly, which will be followed by the discussion of some of our recent efforts to utilize the rising bubbles as a strategy of flow control such as the fluid mixing enhancement and heat transfer augmentation.

Effects of circular and non-circular nozzle exit geometries on subsonic and supersonic jet propagations

The mixing enhancement and core length reduction of a jet without significant loss of thrust are essential for reducing infrared radiation, mitigating aeroacoustic noise, improving combustion characteristics, and thrust vectoring. The jet mixing can be improved by manipulating the flow behavior. In subsonic and sonic jets, the flow manipulation may be achieved by utilizing nozzles with non-circular geometries that shed vortices of varying size due to their non-uniform azimuth curvatures. Non-uniform vortices generate differential spreading along the nozzle’s perimeter, causing axis switching and improving entrainment characteristics. Therefore, the present study examines the effects of two non-circular nozzle exit shapes (elliptic and square) on the mixing augmenting efficacy at subsonic and sonic flow conditions. The circular nozzle is tested for comparison. Both quantitative and qualitative analyses evaluate the efficacy of nozzles with non-circular exit geometries. Among the configurations investigated, the elliptic nozzle is superior in shortening the potential core length and enhancing the jet spread. A maximum reduction of 18.75% in core length with rapid jet decay was accomplished with the elliptic nozzle. The measurement of pressure profiles at different streamwise locations reveals that the spread rate is greater for elliptic and square jets than their circular counterpart. The elliptic jet exhibits the highest spread along the minor-axis direction compared to the major-axis direction. The differential jet spread rate in the elliptical jet causes an early axis-switching––direct evidence of mixing augmentation. Shadowgraph images show the asymmetric pattern of shock cell structures and differential spreading in elliptic and square jets.

Hydrogen mixing augmentation mechanism induced by the vortex generator and oblique shock wave in a scramjet engine

The mixing process between the fuel and the incoming air is extremely important for the engineering implementation of the scramjet engine. In the current study, the vortex generator coupled with the oblique shock wave is utilized to promote the hydrogen mixing process in a supersonic crossflow. The configurations of the vortex generator are put into investigation, namely typical ramp, split ramp and ramp vane. Some parameters are provided to evaluate the flow field properties quantitatively. The obtained results predicted by the three-dimensional Reynolds-average Navier-Stokes (RANS) equations show that the method of shock wave/jet shear layer interaction coupled with the vortex generator can effectively improve the mixing efficiency. Different vortex generator structures all have great effect, especially for Case SR (split ramp), with the mixing efficiency raised by 36.27%. The streamwise vorticity plays an important role in the mixing process.

Characteristics of enhanced mixing induced by plate jet actuation in supersonic flow

The fast and efficient mixing of fuel and oxidizers under supersonic conditions is of great importance for improving the performance of scramjet engines. The mixing process in the inner flow of a scramjet combustor is heavily inhibited by compressibility effects. In this paper, the novel strategy of plate jet actuation is proposed, and its effects on mixing augmentation are analyzed by employing numerical programs developed in-house. The fine vortex structures induced by the plate jet actuation are well captured, and the dynamic behaviors of newly observed T-shaped structures are analyzed in detail. It is found that in plate jet actuation flow, Kelvin–Helmholtz (K–H) vortices induced by K–H instability coexist with T-shaped structures induced by jet actuation instability. The interaction of adjacent T-shaped structures leads to the distortion and breakup of large-scale structures, which can obviously improve the interfaces of upper and lower streams. The distribution of the turbulence intensity along the streamwise direction suggests that with the introduction of plate jet actuation, more intense fluctuations occur in the flow. The growth process of mixing layer thickness indicates that with plate jet actuation, a sharp increase in mixing thickness can be achieved in the near flow field. The results of structural topology analysis show that upper plate jet actuation can produce structures with larger sizes, and the distortion and penetration process of these structures can entrain more upper and lower streams into the mixing region. It is suggested that the present proposed strategy is a good candidate for mixing enhancement with the application of scramjet combustors.

Class Activation Map-Based Data Augmentation for Satellite Smoke Scene Detection

With global warming, increasingly severe forest fires have caused serious damage to the Earth&#x2019;s ecological environment. Remote sensing (RS) technology has become an important means of forest fire monitoring owing to its unique advantages of multiple cycles and wide coverage. In smoke scene detection, the current simple mixed sample data augmentation (MSDA) methods lead to the loss of key objects in images, and training the samples whose images do not match the labels is the main reason for the decline in classifier accuracy. We propose a Class Activation Map-based Mixing method (CAMMix), a new MSDA method. In contrast to previous data augmentation (DA) methods, CAMMix can select the region and degree of mixing throughout the significance map. With the aid of an auxiliary classifier, CAMMix generates a mixed mask with class significance, such that the newly generated mixed sample distribution is close to the original data distribution. We also propose an intervention for the coverage method to further prevent the loss of smoke objects. On a smoke dataset (USTC&#x005F;SmokeRS), CAMMix achieves the best accuracy of 94.95% and 83% for 64% and 8% training set respectively. It results that CAMMix and variants provides high classification accuracy and preserves key object information, outperforming the Mixup, CutMix, ResizeMix and FMix.

Mixing augmentation induced by the combination of the oblique shock wave and secondary recirculation jet in a supersonic crossflow

The mixing process of fuel-air in the supersonic crossflow is a pivotal technology for the scramjet engine. In this paper, numerical simulation of the transverse sonic hydrogen jet into a supersonic Mach 3 crossflow with the mixing augmentation strategy induced by the combination of the oblique shock wave and secondary recirculation jet has been carried out. Detailed flow field structures, hydrogen mass fraction distributions, vortex structures, heat flux and some parameters have been explored in order to investigate its mixing enhancement mechanism. Results of the three-dimensional Reynolds-average Navier-Stokes (RANS) equations coupled with the two-equation shear stress transport (SST) κ-ω turbulence model show that the combined strategy of the oblique shock wave and secondary recirculation jet device can effectively improve the mixing speed and mixing efficiency with little total pressure loss. Also, the secondary recirculation jet device can reduce the peak of the heat flux effectively. In this study, the case with the single bleed hole owns the best effect with improving the mixing efficiency by 82.75% locally and reducing the maximum heat flux by 15.24% respectively。

Towards efficient network compression via Few-Shot Slimming

While previous network compression methods achieve great success, most of them rely on the abundant training data which is, unfortunately, often unavailable in practice due to some reasons, e.g., privacy issues, storage constraints, and transmission limitations. A promising way to solve this problem is to perform compression with a few unlabeled data. Proceeding along this way, we propose a novel few-shot network compression framework named Few-Shot Slimming (FSS). FSS follows the student/teacher paradigm, and contains two steps: (1) construct the student by inheriting principal feature maps from the teacher; (2) refine the student feature representation by knowledge distillation with an enhanced mixing data augmentation method called GridMix. Specifically, in the first step, we employ normalized cross correlation to perform the principal feature analysis, and then theoretically construct a new indicator to select the most informative feature maps from the teacher for the student. The indicator is based on the variances of feature maps which can efficiently quantitate the information richness of the input feature maps in a feature-agnostic manner. In the second step, we perform the knowledge distillation for the initialized student in first step with a novel grid-based mixing data augmentation technique which greatly extends the limited sample dataset. In this way, the student is able to refine its feature representation and achieves a better result. Extensive experiments on multiple benchmarks demonstrate the state-of-the-art performance of FSS. For example, by using 0.2% label-free data of full training set, FSS yields a 60% FLOPs reduction for DenseNet-40 on CIFAR-10 with only a loss of 0.8% in top-1 accuracy, achieving a result on par with that obtained by the conventional full-data methods.

Investigation on the impact of the induced shock wave on the hydrogen mixing augmentation in a supersonic crossflow: A numerical study

The mixing process between the fuel and the incoming air flow plays an important role for the engineering implementation of a scramjet engine. In the current study, the shock wave/jet shear layer interaction approach is utilized to promote the hydrogen mixing enhancement in a supersonic flow. The action location between the induced shock wave and the air inflow, as well as the intensity of the shock wave, is investigated to capture the mixing enhancement mechanism. Some parameters are provided to evaluate the flow field properties quantitatively, namely the mixing efficiency and the total pressure recovery coefficient. The obtained results predicted by the three-dimensional Reynolds-average Navier-Stoke (RANS) equations coupled with the two-equation shear stress transport (SST) κ-ω turbulence model show that the shock wave/jet shear layer interaction approach can effectively improve the mixing efficiency. The mixing speed and the mixing efficiency both increase with the increase of the intensity of the shock wave. The mixing enhancement mechanism is that the enhancement of the streamwise and spanwise vorticity upstream of the jet promotes the enlargement of the upstream separation zone and recirculation zone. When the shock wave position moves downstream, the spanwise vorticity increases, and the downstream recirculation zone increases significantly to promote mixing.

Design exploration on the mixing augmentation induced by the oblique shock wave and a novel step in a supersonic flow

The mixing process in the supersonic flow is very important for the engineering implementation of the scramjet engine, and many novel mixing augmentation schemes have been proposed and investigated in the past few years. In this paper, the combination between the oblique shock wave and a novel step has been proposed to promote the mixing process between the supersonic airflow and the fuel, and the oblique shock wave has been induced by a ramp located on the upper wall of the channel. The impacts of the jet location and the jet pressure ratio have been both evaluated in order to explore their influences on the mixing enhancement mechanism. The obtained results show that the oblique shock wave is beneficial to prolong the residence time of the hydrogen in the supersonic flow, and the hydrogen distribution moves upstream with the oblique shock wave. At the same time, when the jet pressure ratio is large enough, the recirculation zone formed downstream of the step would increase, and this is beneficial for the mixing process between the fuel and the air as well. This may imply that the proposed combination scheme is a good candidate for the fuel mixing augmentation strategy, and its physical mechanism should be explored much deeper in the future, especially for its three-dimensional effect.

Parametric study on mixing augmentation mechanism induced by cantilevered ramp injectors in a shock-induced combustion ramjet engine

The mixing process is very important for the shock-induced combustion ramjet engine, and the fuel and air should be mixed completely before entering into the combustion chamber. In the current study, the studies on standard and swept cantilevered ramp injectors are provided to reveal the mixing enhancement mechanism and the formation of the complex shock wave system. At the same time, some parameters are also taken into consideration to evaluate the flow field properties, including the mixing efficiency, the intensity of streamwise vorticity and total pressure recovery efficiency. The obtained results predicted by the three-dimensional Reynolds-average Navier-Stokes (RANS) equations coupled with the two equation shear stress transport (SST) k-ω turbulence model show that the mixing enhancement mechanism of the cantilevered ramp injector is the shear action in the near flow field and the large-scale vortex structure caused by the cantilevered configuration in the far flow field. The cantilevered ramp injector is beneficial for the mixing enhancement, and it can reduce the total pressure loss as well.

Phenomenology of Turbulence Production in Merging Process of Supersonic Streamwise Vortices

Merging of corotating vortices is of primary interest in the field of supersonic mixing due to the ability of two or more structures to coalesce and form a larger one which enhances the stirring motion and straining in the flow, thus promoting turbulence production and mixing augmentation. During experimental efforts aimed at quantifying mean and fluctuating flowfields produced by the merging of two supersonic streamwise corotating vortices through stereoscopic particle image velocimetry, it was noticed that both the strain rates and turbulent kinetic energy would peak within the elongated merger’s vorticity patch. The article discusses how this is in fact a counterintuitive phenomenon and shows that a possible explanation can be given by studying the flow within a frame of reference which rotates with the merger itself. It will be shown how the merger is surrounded by a different flowfield than that measured in the laboratory frame of reference, characterized by the presence of recirculation zones referred to as ghost vortices. These vortices introduce large rates of strain within the merger, thus providing a possible justification for the high levels of turbulence attained within the rotating structure. The experimental findings are strongly corroborated by reduced-order, inviscid simulations, showing the formation of similar structures and clarifying the presence of peaks of strain rates within the merger.

Secondary fuel jet strategies on mixing enhancement performance of rocket-based combined cycle engine

The mixed augmentation performance between secondary fuel jet and uneven supersonic inflow is a significant evaluation index of the rocket-based combined cycle engine. In this study, the effects of the secondary fuel jet on the flow field structure and mixing enhancement performance are discussed by changing both the secondary fuel jet angle and spacing of multiple rows of jet holes. The three-dimensional coupled implicit Reynolds averaged Navier-Stokes equations, SST k-ω turbulence model, and eddy dissipation concept (EDC) reaction model were adopted to evaluate the flow field structure with a sonic hydrogen jet. It is indicated that the 135° fuel jet has the largest penetration depth value, followed by the 90° fuel jet, whereas the 45° fuel jet has the smallest penetration depth value. However, the 45° fuel jet possesses the best mixing performance, it easily induces a pair of downstream counter-rotating vortex, and the oblique torque is further generated to aggravate the tearing ability to the jet. The three jet holes induce a stronger shock wave and form a larger barrier under the 90° jet. Additionally, when the spacing between the multiple rows of jet holes reduces, penetration depth is easily created. Further, when the spacing of the three rows of jet holes is larger, the fluctuation of the mixing coefficient increases.