[논문]스마트 팩토리 모니터링을 위한 빅 데이터의 LSTM 기반 이상 탐지

김진술

doi:10.9728/dcs.2018.19.4.789

스마트 팩토리 모니터링을 위한 빅 데이터의 LSTM 기반 이상 탐지
LSTM-based Anomaly Detection on Big Data for Smart Factory Monitoring 원문보기

디지털콘텐츠학회 논문지 = Journal of Digital Contents Society, v.19 no.4, 2018년, pp.789 - 799

(전남대학교 전자컴퓨터공학과) , (전남대학교 전자컴퓨터공학과) , 김진술 (전남대학교 전자컴퓨터공학과)

초록
AI-Helper

이 논문에서는 이러한 산업 단지 시스템에서의 비정상적인 동작이 일어날 때, 시간 계열의 데이터를 분석하기 위하여 Big 데이터를 이용한 접근을 기반으로 하는 머신 러닝을 보여줍니다. Long Short-Term Memory (LSTM) 네트워크는 향상된 RNN버전으로서 입증되었으며 많은 작업에 유용한 도움이 되었습니다. 이 LSTM 기반 모델은 시간적 패턴뿐만 아니라 더 높은 레벨의 시간적 특징을 학습 한 다음, 미래의 데이터를 예측하기 위해 예측 단계에 사용됩니다. 예측 오차는 예측 인자에 의해 예측 된 결과와 실제 예상되는 값의 차이입니다. 오차 분포 추정 모델은 가우스 분포를 사용하여 관찰 스코어의 이상을 계산합니다. 이러한 방식으로, 우리는 하나의 비정상적 데이터의 개념에서 집단적인 비정상적 데이터 개념으로 바뀌어 갑니다. 이 작업은 실패를 최소화하고 제조품질을 향상시키는 Smart Factory의 모니터링 및 관리를 지원할 수 있습니다.

Abstract ▼ AI-Helper

This article presents machine learning based approach on Big data to analyzing time series data for anomaly detection in such industrial complex system. Long Short-Term Memory (LSTM) network have been demonstrated to be improved version of RNN and have become a useful aid for many tasks. This LSTM based model learn the higher level temporal features as well as temporal pattern, then such predictor is used to prediction stage to estimate future data. The prediction error is the difference between predicted output made by predictor and actual in-coming values. An error-distribution estimation model is built using a Gaussian distribution to calculate the anomaly in the score of the observation. In this manner, we move from the concept of a single anomaly to the idea of the collective anomaly. This work can assist the monitoring and management of Smart Factory in minimizing failure and improving manufacturing quality.

주제어

AI 본문요약
AI-Helper

* AI 자동 식별 결과로 적합하지 않은 문장이 있을 수 있으니, 이용에 유의하시기 바랍니다.

문제 정의

We prove that the management and maintenance of smart systems are tricky, so we need an efficiency procedure covering the real-time information transferring, analysis and decision-making framework with potential support from big data analytics. Such mention is the primary objective of this paper that aims at making an anomaly identification and timely alarming. Moreover, we would like to use modern machine learning technologies in order to generate the knowledge about anomaly happened in the system to announce a warning/alarm instantly.
As keep in mind, the machine learning (ML) should be one of the core parts of modern industrial information and control system [2]. This researching work will explore implicit information about degradation state of the sub-system in complex systems and identify anomaly event during operation, to enable real-time maintenance planning programs [5]. The real-time maintenance is very useful in cost savings and increasing availability of the facilities.

가설 설정

. We consider anomaly candidates belong to positive class and normal candidates belong to negative class. We consider anomaly candidates belong to positive class and normal candidates belong to negative class.

제안 방법

From existing big data infrastructure of industry 4.0, this study introduces a big data-based framework for anomaly detection in the smart factory. Both benefits of big data and machine learning are utilized in industrial Supervisory Control and Data Acquisition (SCADA) based management systems to understand time series data.
We use LSTM-RNN as sequence prediction model which feeds input data underlying time series form. In this study, the trained model is used to estimate the distribution of prediction error. The prediction error model verifies the likelihood of anomaly behavior.
The stacked LSTM network trained on only normal data to generate predictor over some time steps. The distributed probability of error signal was estimated and was used to assess the likelihood of anomaly score. We can solve anomaly by LSTM based classifier on two classes including normal label and anomalous one.

대상 데이터

The architecture is composed of three stacked LSTM layer with the number of LSTM cells {64, 256, 100} respectively as shown in Fig. 6, that followed by one fully connected layer. We also regularize between each layer with dropout operation 0.
Because of the collective anomalous score using an unsupervised manner, both unsupervised and supervised approaches can be used for scoring individual anomaly. The proposed model was evaluated on moving crane and fuel consumption dataset. On acoustic novelty detection, [17] presented a novel approach based on non-linear predictive denoising autoencoders (DA) with LSTM for isolating abnormal acoustic signals.

이론/모형

The error vector calculated on the normal validation-1 time series are used to estimate μ and ∑ of Normal distribution using Maximum Likelihood Estimation (MLE) algorithm.

참고문헌 (21)

B. Chen, J. Wan, L. Shu, P. Li, M. Mukherjee, and B. Yin, "Smart Factory of Industry 4.0: Key Technologies, Application Case, and Challenges," IEEE Access, Vol. 6, pp. 6505-6519, 2017.
T. Skripcak and P. Tanuska, "Utilisation of on-line machine learning for SCADA system alarms forecasting," Proceedings of 2013 Science and Information Conference, SAI 2013, pp. 477-484, 2013.
S. Mohanty, M. Jagadeesh, and H. Srivatsa, " 'Big Data' warehouse, 'BI'implementations and analytics," Apress, 2013.
I. Garitano, R. Uribeetxeberria and U. Zurutuza, "A review of SCADA Anomaly Detection System," Soft Computing Models in Industrial and Environmental Applications, 6th International Conference SOCO 2011, Vol. 87, 2011.
M. Shafiee and M. Finkelstein, "An optimal age-based group maintenance policy for multi-unit degrading systems," Reliability Engineering \& System Safety, Vol. 134, pp. 230-238, 2015.

상세보기
Kv, R. Satish, and N. P. Kavya, "Trend Analysis of E-Commerce Data using Hadoop Ecosystem," International Journal of Computer Applications, Vol. 147, No. 6, pp. 1-5, 2016.
J. Dean and S. Ghemawat, "MapReduce: simplified data processing on large clusters," Communication of ACM, Vol. 51, No. 1, pp. 107-113, 2008.
Flume: http://flume.apache.org/
Hive:https://hive.apache.org/
Zeppelin: https://zeppelin.apache.org/docs/0.6.2/
L. Zhou, S. Pan, J. Wang, and A. V. Vasilakos, "Machine learning on big data: Opportunities and challenges," Neurocomputing, Vol. 237, pp. 350-361, 2017.

상세보기
Jordan, I. Michael and M. M. Tom, "Machine learning: Trends, perspectives, and prospects," Science , Vol.349, No. 6245, pp. 255-260, 2015.

상세보기
V. Chandola, A. Banerjee, and V. Kumar, "Anomaly detection: A survey," ACM Computing survey (CSUR), Vol. 41, No. 3, pp. 1-58, 2009.
P. Malhotra, L. Vig, G. Shroff, and P. Agarwal, "Long short term memory networks for anomaly detection in time series," Proceedings, pp. 89, 2015.
S. Hochreiter and J. Urgen Schmidhuber, "Long Short-Term Memory," Neural Computation, Vol. 9, No. 8, pp. 1735-1780, 1997.

상세보기
T. Olsson and A. Holst, "A Probabilistic Approach to Aggregating Anomalies for Unsupervised Anomaly Detection with Industrial Applications," in FLAIRS Conference, pp. 434-439, 2015.
E. Marchi, F. Vesperini, F. Weninger, F. Eyben, S. Squartini, and B. Schuller, "Non-linear prediction with LSTM recurrent neural networks for acoustic novelty detection," in Neural Networks (IJCNN), 2015 International Joint Conference, pp. 1-7, 2015.
L. Bontemps, J. McDermott, N.-A. Le-Khac, and others, "Collective anomaly detection based on long short-term memory recurrent neural networks," in International Conference on Future Data and Security Engineering, pp. 141-152, 2016.
S. D. Anton, D. Fraunholz, C. Lipps, F. Pohl, M. Zimmermann, and H. D. Schotten, "Two decades of SCADA exploitation: A brief history," in Application, Information and Network Security (AINS), 2017 IEEE Conference, pp. 98-104, 2017.
Understanding LSTM Network [Internet]. Available: http://colah.github.io/posts/2015-08-Understanding-LSTMs
Xisom solution: http://www.xisom.com

저자의 다른 논문 :

LOADING...

내보내기 구분	파일저장 인쇄 메일전송
구성항목	기본정보 상세정보 관리번호, 논문명, 저널/프로시딩명, 저자 , 발행년, 권, 호, 시작페이지, 끝페이지, 발행기관 관리번호, 논문명, 대등논문명, 저자 , 저널/프로시딩명, 발행기관, 발행년, 발행언어, 권, 호, 시작페이지, 끝페이지, ISBN, ISSN, 주제분야, 키워드, 초록(한글), 초록(영문), 저자(소속기관)
저장형식	Text(ASCII format) Excel format RefWorks Direct Export RIS format (for Reference Manager, ProCite, EndNote), Scholar's Aids, Mendeley
메일정보	받는사람 (필수) @ 보내는사람 (선택) @ 제목 내용 KISTI 검색결과 이메일 서비스
안내	총 건의 자료가 검색되었습니다. 다운받으실 자료의 인덱스를 입력하세요. (1-10,000) 검색결과의 순서대로 최대 10,000건 까지 다운로드가 가능합니다. 데이타가 많을 경우 속도가 느려질 수 있습니다.(최대 2~3분 소요) 다운로드 파일은 UTF-8 형태로 저장됩니다. 파일의 내용이 제대로 보이지 않을실 때는 웹브라우저 상단의 보기 -> 인코딩 -> 자동선택 여부를 확인하십시오. ~ Text(ASCII format) Excel format

연합인증