Anomaly Diagnosis Using Autoencoder in Edge Computing Systems

Mert Kışlakçı; Mahmut Durgun

doi:10.47897/bilmes.1132562

TR EN

Anomaly Diagnosis Using Autoencoder in Edge Computing Systems

Abstract

IoT sistemleri geleneksel buluta bağlı bir mimaride çalışır. IoT cihazlarında oluşturulan veriler buluta aktarılır, orada depolanır ve daha sonra anlamlı bilgiler çıkarmaya çalışarak işlenir. Ancak tercih edilen bu yapıda sürekli buluta bağımlı olmanın dezavantajları oldukça yüksektir. Her bir bilgi parçasının ham olarak buluta aktarılması ağ trafiğini artırırken, verileri yalnızca bulut katmanında işlemek için yüksek donanım gücü gerektirir. UBISOKKAT (Edge Computing Systems Kullanarak Otomatik Kodlayıcı Kullanarak Anomali Teşhisi) sistemi yukarıda belirtilen sorunlara çözüm olarak ortaya çıkmıştır. UBISOKKAT sistemi, IoT sistemleri ve bulut sistemleri arasında bir ara katman görevi görür. IoT noktalarında üretilen her veri önce orta katmandaki UBISOKKAT sistemine gönderilir ve burada bulut katmanına iletilir. Makine öğrenimi modeli daha sonra bulut katmanına yerleştirilir ve ara katman yazılımından aldığı verileri kullanarak kendini eğitmeye başlar. Eğitim süreci tamamlanan modelin çıktıları UBISOKKAT sistemine gönderilir ve otomatik kodlayıcı bulutta değil ara katman yazılımı yazılımında çalıştırılır. Bunun en büyük avantajı, gerçek zamanlı sistemlerde verilerin buluta gönderilmemesi, yerel noktalarda analiz edilmesi, ağ trafiğinin azaltılması ve gecikmenin azaltılmasıdır. Aynı zamanda her veri bulutta analiz edilmediği için yerel noktalarda analiz edilerek bulut ihtiyacı azaltılmakta, yüksek maliyetler düşürülmekte ve sistemin canlılığı arttırılmaktadır. Bu çalışmada son katmanda otomatik kodlayıcı modeli çalıştırılmış ve tek fazlı elektrik motorundan elde edilen verilere dayanarak UBISOKKAT sisteminin uç noktalardaki anomalileri teşhis edebildiği görülmüştür.

Keywords

Makine Öğrenimi, Nesnelerin İnterneti, Bulut Bilişim, Uç Bilişim, Anomali Teşhisi

Thanks

Proje sürecinde bana her konuda rehberlik eden saygıdeğer Dr. Öğr. Üyesi Yeliz DURGUN ve Dr. Öğr. Üyesi Mahmut DURGUN hocalarım ile verdiği destekler sebebiyle annem Semra KIŞLAKÇI ile babam Hikmet KIŞLAKÇI’ ya çok teşekkürü borç bilirim.

Anomaly Diagnosis Using Autoencoder in Edge Computing Systems

Abstract

IoT systems operate in a traditional cloud-connected architecture. The data generated on IoT devices is transferred to the cloud, stored there, and then processed, trying to extract meaningful information. However, in this preferred structure, the disadvantages of being constantly dependent on the cloud are quite high. While transferring each piece of information raw to the cloud increases network traffic, it requires high hardware power to process data only at the cloud layer. UBISOKKAT (Anomaly Diagnosis Using Automatic Encoder in Edge Computing Systems) system has emerged as a solution to the problems mentioned above. The UBISOKKAT system acts as an intermediate layer between IoT systems and cloud systems. Every data generated at IoT points is first sent to the UBISOKKAT system in the middle layer, where it is transmitted to the cloud layer. The machine learning model is then placed in the cloud tier and begins to train itself using the data it receives from the middleware. The outputs of the model whose training process is completed are sent to the UBISOKKAT system and the automatic encoder is run in the middleware software, not in the cloud. The biggest advantage of this is that in real-time systems, data is not sent to the cloud but analyzed at local points, reducing network traffic and reducing latency. At the same time, since not every data is analyzed in the cloud, it is analyzed at local points, reducing the need for the cloud, reducing high costs and increasing the viability of the system. In this study, an automatic encoder model was operated in the last layer and it was seen that the UBISOKKAT system could diagnose anomalies at the extreme points based on the data obtained from the single-phase electric motor.

Keywords

Edge Computing, Machine Learning, Anomaly Diagnosis, Cloud Systems, IoT

References

[1] Q. Xu, J. Zhang, ve B. Togookhuu, “Support mobile fog computing test in piFogBedII”, Sensors (Switzerland), 2020, doi: 10.3390/s20071900.
[2] Aydemir, F., “IoT Based Indoor Disinfection Coordinating System Against the New Coronavirus”, International Scientific and Vocational Studies Journal, 4(2),81 - 85. doi: 10.47897/bilmes.751995.
[3] O. Kayode, D. Gupta, ve A. S. Tosun, “Towards a Distributed Estimator in Smart Home Environment”, IEEE World Forum Internet Things, WF-IoT 2020 - Symp. Proc., Haz. 2020, doi: 10.1109/WF-IOT48130.2020.9221083.
[4] Y. Liu vd., “Deep Anomaly Detection for Time-Series Data in Industrial IoT: A Communication-Efficient On-Device Federated Learning Approach”, IEEE Internet Things J., c. 8, sayı 8, ss. 6348–6358, Nis. 2021, doi: 10.1109/JIOT.2020.3011726.
[5] D. Utomo ve P. A. Hsiung, “Anomaly Detection at the IoT Edge using Deep Learning”, 2019 IEEE Int. Conf. Consum. Electron. - Taiwan, ICCE-TW 2019, May. 2019, doi: 10.1109/ICCE-TW46550.2019.8991929.
[6] O. Kayode ve A. S. Tosun, “LIRUL: A Lightweight LSTM based model for Remaining Useful Life Estimation at the Edge”, Proc. - Int. Comput. Softw. Appl. Conf., c. 2, ss. 177–182, Tem. 2019, doi: 10.1109/COMPSAC.2019.10203.
[7] S. Nandi, H. A. Toliyat, ve X. Li, “Condition monitoring and fault diagnosis of electrical motors - A review”, IEEE Trans. Energy Convers., c. 20, sayı 4, ss. 719–729, Ara. 2005, doi: 10.1109/TEC.2005.847955.
[8] A. Baghbanpourasl, D. Kirchberger, ve C. Eitzinger, “Failure prediction through a model-driven machine learning method”, 2021 IEEE Int. Work. Metrol. Ind. 4.0 IoT, MetroInd 4.0 IoT 2021 - Proc., ss. 527–531, Haz. 2021, doi: 10.1109/METROIND4.0IOT51437.2021.9488550.
[9] M. Adkisson, J. C. Kimmell, M. Gupta, ve M. Abdelsalam, “Autoencoder-based Anomaly Detection in Smart Farming Ecosystem”, Proc. - 2021 IEEE Int. Conf. Big Data, Big Data 2021, ss. 3390–3399, 2021, doi: 10.1109/BIGDATA52589.2021.9671613.
[10] Y. H. Park ve M. J. Kim, “Design of Cost-Effective Auto-Encoder for Electric Motor Anomaly Detection in Resource Constrained Edge Device”, Proc. 3rd IEEE Eurasia Conf. IOT, Commun. Eng. 2021, ECICE 2021, ss. 241–246, 2021, doi: 10.1109/ECICE52819.2021.9645739.

[11] L. Njilla, L. Pearlstein, X. W. Wu, A. Lutz, ve S. Ezekiel, “Internet of Things Anomaly Detection using Machine Learning”, Proc. - Appl. Imag. Pattern Recognit. Work., c. 2019-October, Eki. 2019, doi: 10.1109/AIPR47015.2019.9174569.
[12] “Intelligent Systems with Internet of Things (IoT) Part 1 & 2 & 3 | by Mert Kışlakçı | Medium”. https://medium.com/@mertkslkc/intelligent-systems-with-internet-of-things-iot-part-1-2-3-7353dec3cbf8 (erişim Haz. 12, 2022).
[13] “What is AC Motor? Definition & Types - Circuit Globe”. https://circuitglobe.com/ac-motor.html (erişim Haz. 12, 2022).
[14] “What is a Raspberry Pi and How Does it Work? | Pi Day”. https://www.piday.org/whats-a-raspberry-pi-and-how-does-it-work/ (erişim Haz. 11, 2022).
[15] “Alışverişe Devam Edin”. https://market.samm.com/raspberry-pi-3-b-plus (erişim Haz. 11, 2022).
[16] “8-Channel 12-Bit ADC for Raspberry Pi (STM32F030) - Seeed Studio”. https://www.seeedstudio.com/8-Channel-12-Bit-ADC-for-Raspberry-Pi-STM32F030.html (erişim Haz. 12, 2022).
[17] “Analog Piezoelektrik Seramik Titreşim Sensörü Satın Al | Robotelektronik.com”. https://www.robotelektronik.com/urun/analog-piezoelektrik-seramik-titresim-sensoru (erişim Haz. 15, 2022).
[18] “What is cloud computing? Everything you need to know about the cloud explained | ZDNet”. https://www.zdnet.com/article/what-is-cloud-computing-everything-you-need-to-know-about-the-cloud/ (erişim Haz. 12, 2022).
[19] “What is edge architecture?” https://www.redhat.com/en/topics/edge-computing/what-is-edge-architecture?sc_cid=7013a000002pu3zAAA&gclid=CjwKCAjwnZaVBhA6EiwAVVyv9E3Yr7ZeN8b6zMO9bT_lRpEGCm8UpVe6hmw5oWRnMNvApOyoDOAqUxoCN0YQAvD_BwE&gclsrc=aw.ds (erişim Haz. 12, 2022).
[20] “Edge AI Anomali Algılama Bölüm 2 - Özellik Çıkarma ve Model Eğitimi”. https://www.digikey.com/en/maker/projects/edge-ai-anomaly-detection-part-2-feature-extraction-and-model-training/70927a6e439b49bea7305953a3c9bfff (erişim Haz. 11, 2022).
[21] “Google Colab Nedir ve Nasıl Kullanılır? – Global AI Hub”. https://globalaihub.com/blog/google-colab-nedir-ve-nasil-kullanilir/ (erişim Haz. 12, 2022). [22] “About Keras”. https://keras.io/about/ (erişim Haz. 12, 2022).
[23] “TensorFlow”. https://www.tensorflow.org/ (erişim Haz. 12, 2022).
[24] V. Kotu ve B. Deshpande, “Anomaly Detection”, Data Sci., ss. 447–465, Oca. 2019, doi: 10.1016/B978-0-12-814761-0.00013-7.

Details

Primary Language

Turkish

Subjects

Artificial Intelligence

Journal Section

Research Article

Authors

Mert Kışlakçı ^*
0000-0001-8506-7498
Türkiye

Mahmut Durgun
0000-0002-5010-687X
Türkiye

Publication Date

June 29, 2022

Submission Date

June 18, 2022

Acceptance Date

June 28, 2022

Published in Issue

Year 2022 Volume: 6 Number: 1

DOI

https://doi.org/10.47897/bilmes.1132562

IZ

https://izlik.org/JA79RY95ZL

APA

Kışlakçı, M., & Durgun, M. (2022). Anomaly Diagnosis Using Autoencoder in Edge Computing Systems. International Scientific and Vocational Studies Journal, 6(1), 41-50. https://doi.org/10.47897/bilmes.1132562

AMA

1.Kışlakçı M, Durgun M. Anomaly Diagnosis Using Autoencoder in Edge Computing Systems. ISVOS. 2022;6(1):41-50. doi:10.47897/bilmes.1132562

Chicago

Kışlakçı, Mert, and Mahmut Durgun. 2022. “Anomaly Diagnosis Using Autoencoder in Edge Computing Systems”. International Scientific and Vocational Studies Journal 6 (1): 41-50. https://doi.org/10.47897/bilmes.1132562.

EndNote

Kışlakçı M, Durgun M (June 1, 2022) Anomaly Diagnosis Using Autoencoder in Edge Computing Systems. International Scientific and Vocational Studies Journal 6 1 41–50.

IEEE

[1]M. Kışlakçı and M. Durgun, “Anomaly Diagnosis Using Autoencoder in Edge Computing Systems”, ISVOS, vol. 6, no. 1, pp. 41–50, June 2022, doi: 10.47897/bilmes.1132562.

ISNAD

Kışlakçı, Mert - Durgun, Mahmut. “Anomaly Diagnosis Using Autoencoder in Edge Computing Systems”. International Scientific and Vocational Studies Journal 6/1 (June 1, 2022): 41-50. https://doi.org/10.47897/bilmes.1132562.

JAMA

1.Kışlakçı M, Durgun M. Anomaly Diagnosis Using Autoencoder in Edge Computing Systems. ISVOS. 2022;6:41–50.

MLA

Kışlakçı, Mert, and Mahmut Durgun. “Anomaly Diagnosis Using Autoencoder in Edge Computing Systems”. International Scientific and Vocational Studies Journal, vol. 6, no. 1, June 2022, pp. 41-50, doi:10.47897/bilmes.1132562.

Vancouver

1.Mert Kışlakçı, Mahmut Durgun. Anomaly Diagnosis Using Autoencoder in Edge Computing Systems. ISVOS. 2022 Jun. 1;6(1):41-50. doi:10.47897/bilmes.1132562

Cited By

Machine Learning-Assisted Wearable Thermo-Haptic Device for Creating Tactile Sensation

Bitlis Eren Üniversitesi Fen Bilimleri Dergisi

https://doi.org/10.17798/bitlisfen.1434202

INTERNATIONAL SCIENTIFIC AND VOCATIONAL STUDIES JOURNAL will publish the content under Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) license, which gives permission to copy and redistribute the material in any medium or format other than commercial purposes, as well as remix, transform, and build upon the material by providing appropriate credit to the original work.