Squeezed Convolutional Variational AutoEncoder for unsupervised anomaly detection in edge device industrial Internet of Things

Kim, Dohyung; Yang, Haneul; Chung, Minki; Cho, Sungzoon; Kim, Huijung; Kim, Minhee; Kim, Kyungwon; Kim, Eun-Seok

doi:10.1109/infoct.2018.8356842

Cited by 38 publications

(15 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this model, we utilize a relatively long sliding window in order to capture temporal dependencies of periodic time series data. Besides, each coming data point instead of batch data in a time window adopted in [28] will be evaluated to be a normal or an anomaly, in order to respond to anomalies as soon as possible. The overall structure is shown in Figure 1.…”

Section: Methodsmentioning

confidence: 99%

Unsupervised Anomaly Detection of Industrial Robots Using Sliding-Window Convolutional Variational Autoencoder

et al. 2020

View full text Add to dashboard Cite

With growing dependence of industrial robots, a failure of an industrial robot may interrupt current operation or even overall manufacturing workflows in the entire production line, which can cause significant economic losses. Hence, it is very essential to maintain industrial robots to ensure high-level performance. It is widely desired to have a real-time technique to constantly monitor robots by collecting time series data from robots, which can automatically detect incipient failures before robots totally shut down. Model-based methods are typically used in anomaly detection for robots, yet explicit domain knowledge and accurate mathematical models are required. Data-driven techniques can overcome these limitations. However, a major difficulty for them is the lack of sufficient fault data of industrial robots. Besides, the used technique for anomaly detection of robots should be required to not only capture the temporal dependency in collected time series data, but also the inter-correlations between different metrics. In this paper, we introduce an unsupervised anomaly detection for industrial robots, sliding-window convolutional variational autoencoder (SWCVAE), which can realize real-time anomaly detection spatially and temporally by coping with multivariate time series data. This method has been verified by a KUKA KR6R 900SIXX industrial robot, and the results prove that the proposed model can successfully detect anomaly in the robot. Thus, this work presents a promising tool for condition-based maintenance of industrial robots.INDEX TERMS Anomaly detection, industrial robots, sliding window, variational autoencoder, convolutional neural network.

show abstract

Section: Methodsmentioning

confidence: 99%

Unsupervised Anomaly Detection of Industrial Robots Using Sliding-Window Convolutional Variational Autoencoder

et al. 2020

View full text Add to dashboard Cite

show abstract

“…Anomaly detection methods have been applied to a variety of industrial processes from system health monitoring in largescale power generation [8], intelligent maintenance scheduling in smaller production plants [9], fault detection in residential Heating Ventilation and Air Conditioning (HVAC) systems [10] and quality control techniques in manufacturing [11]. Large and high-value installations can justify the expense of human analysts or specifically tailored solutions, however as the scale and value of the installation falls the need for more generalised and automated approaches becomes clear.…”

Section: A Industrial Iot and Industry 40mentioning

confidence: 99%

“…Convolutional Variational Autoencoders (CNN-VAE) are utilised in an IoT inspired environment in [9], here the authors demonstrate a method of reducing the size, complexity and training cost of the autoencoder without damaging its ability to identity anomalous instances. This makes the Squeezed CNN-VAE (SCVAE) more suitable for deployment in edge devices within an IoT network.…”

Section: B Anomaly Detection On Multivariate Time-series Datamentioning

confidence: 99%

“…Similarly due to the imbalance between the normal data and anomalous data classical multiclass machine learning approaches would be insufficient to capture the nature of the data stream. Therefore an approach which trains the candidate algorithm only on normal data with anomalies being reported when they fall outside of some region around this normal data would be the most viable [9]. • Multivariate data -As shown by a number of methods discussed in Section V-A the addition of contextual information can improve the suitability of a given detector this may be in the form of temporal information, environmental information or additional sensor streams.…”

Section: Research Challenges and Future Directionsmentioning

confidence: 99%

See 1 more Smart Citation

Anomaly Detection for IoT Time-Series Data: A Survey

Cook

Mısırlı

Fan

2020

IEEE Internet Things J.

491

174

View full text Add to dashboard Cite

Anomaly detection is a problem with applications for a wide variety of domains, it involves the identification of novel or unexpected observations or sequences within the data being captured. The majority of current anomaly detection methods are highly specific to the individual use-case, requiring expert knowledge of the method as well as the situation to which it is being applied. The IoT as a rapidly expanding field offers many opportunities for this type of data analysis to be implemented however, due to the nature of the IoT this may be difficult. This review provides a background on the challenges which may be encountered when applying anomaly detection techniques to IoT data, with examples of applications for IoT anomaly detection taken from the literature. We discuss a range of approaches which have been developed across a variety of domains, not limited to Internet of Things due to the relative novelty of this application. Finally we summarise the current challenges being faced in the anomaly detection domain with a view to identifying potential research opportunities for the future.

show abstract

“…The first level extracts features through the convolutional neural network (CNN) model, and the second level takes the extracted features as the input of the process state transition algorithm to construct the normal state process transfer model of an ICS to realize the detection of unknown attacks or 0-day attacks. Kim et al [17] applied the time series data anomaly detection to the edge calculation of the industrial Internet of things (IIoT) with a stacked compression variational autoencoder (SCVAE), which reduced the size of the model and the time spent in detection while ensuring a similar accuracy level. Because there are many controllers and sensors in industrial control systems, the data corresponding to different processes or scenarios are different.…”

Section: Introductionmentioning

confidence: 99%

Multivariate Abnormal Detection for Industrial Control Systems Using 1D CNN and GRU

et al. 2020

View full text Add to dashboard Cite

Currently, most anomaly detection approaches in industrial control systems (ICSs) use network event logs to build models, and current unsupervised machine learning methods rarely use spatiotemporal correlations and other dependencies between multiple variables (sensors/actuators) in a system to detect anomalies. Most of the existing anomaly detection technologies simply compare the current states with the predicted normal range. Due to the highly dynamic characteristic of industrial control systems, it is insufficient to simply compare the current states with the predicted normal range. As a result, these approaches have low detection rates for unknown or new types of attacks. In view of these shortcomings, this paper presents a network model for predicting sensor/controller parameters in industrial control systems. To predict the parameter values of the sensors and controllers more accurately, the 1D convolutional neural network (1D_CNN) and gated recurrent unit (GRU) are combined to fully learn the spatiotemporal correlation and other dependencies between the parameter values of the sensors and controllers at each moment. An abnormal state detection method based on the calculation of the statistical deviation is proposed to realize the anomaly detection of industrial control systems. The model is validated on the Secure Water Treatment (SWaT) dataset. The precision, recall and F1 scores are used to evaluate the effectiveness of this method in anomaly detection on the SWaT dataset. The experimental results show that the average precision and recall of this method are 0.99 and 0.85, respectively, and that the average F1 score is 0.91. The experimental results show that the proposed method can be successfully applied to anomaly detection systems in industrial control systems with lower false positive rates. INDEX TERMS Auto-encoder, 1D convolutional neural network, gated recurrent unit, industrial control system, anomaly detection, SWaT dataset.

show abstract

Squeezed Convolutional Variational AutoEncoder for unsupervised anomaly detection in edge device industrial Internet of Things

Cited by 38 publications

References 13 publications

Unsupervised Anomaly Detection of Industrial Robots Using Sliding-Window Convolutional Variational Autoencoder

Unsupervised Anomaly Detection of Industrial Robots Using Sliding-Window Convolutional Variational Autoencoder

Anomaly Detection for IoT Time-Series Data: A Survey

Multivariate Abnormal Detection for Industrial Control Systems Using 1D CNN and GRU

Contact Info

Product

Resources

About