hig.sePublications
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • harvard-cite-them-right
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • sv-SE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • de-DE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Low-Latency Collaborative Predictive Maintenance: Over-the-Air Federated Learning in Noisy Industrial Environments
University of Gävle, Faculty of Engineering and Sustainable Development, Department of Electrical Engineering, Mathematics and Science, Electronics.ORCID iD: 0000-0002-4284-6691
University of Gävle, Faculty of Engineering and Sustainable Development, Department of Electrical Engineering, Mathematics and Science, Electronics.ORCID iD: 0000-0001-5429-7223
2023 (English)In: Sensors, E-ISSN 1424-8220, Vol. 23, no 18, article id 7840Article in journal (Refereed) Published
Abstract [en]

The emergence of Industry 4.0 has revolutionized the industrial sector, enabling the development of compact, precise, and interconnected assets. This transformation has not only generated vast amounts of data but also facilitated the migration of learning and optimization processes to edge devices. Consequently, modern industries can effectively leverage this paradigm through distributed learning to define product quality and implement predictive maintenance (PM) strategies. While computing speeds continue to advance rapidly, the latency in communication has emerged as a bottleneck for fast edge learning, particularly in time-sensitive applications such as PM. To address this issue, we explore Federated Learning (FL), a privacy-preserving framework. FL entails updating a global AI model on a parameter server (PS) through aggregation of locally trained models from edge devices. We propose an innovative approach: analog aggregation over-the-air of updates transmitted concurrently over wireless channels. This leverages the waveform-superposition property in multi-access channels, significantly reducing communication latency compared to conventional methods. However, it is vulnerable to performance degradation due to channel properties like noise and fading. In this study, we introduce a method to mitigate the impact of channel noise in FL over-the-air communication and computation (FLOACC). We integrate a novel tracking-based stochastic approximation scheme into a standard federated stochastic variance reduced gradient (FSVRG). This effectively averages out channel noise's influence, ensuring robust FLOACC performance without increasing transmission power gain. Numerical results confirm our approach's superior communication efficiency and scalability in various FL scenarios, especially when dealing with noisy channels. Simulation experiments also highlight significant enhancements in prediction accuracy and loss function reduction for analog aggregation in over-the-air FL scenarios.

Place, publisher, year, edition, pages
MDPI , 2023. Vol. 23, no 18, article id 7840
Keywords [en]
analog aggregation; channel noise; low latency; over-the-air federated learning; predictive maintenance
National Category
Electrical Engineering, Electronic Engineering, Information Engineering
Identifiers
URN: urn:nbn:se:hig:diva-43081DOI: 10.3390/s23187840ISI: 001074274300001PubMedID: 37765899Scopus ID: 2-s2.0-85172735809OAI: oai:DiVA.org:hig-43081DiVA, id: diva2:1801068
Available from: 2023-09-29 Created: 2023-09-29 Last updated: 2024-01-10Bibliographically approved
In thesis
1. Collaborative Predictive Maintenance for Smart Manufacturing: From Wireless Control to Federated Learning
Open this publication in new window or tab >>Collaborative Predictive Maintenance for Smart Manufacturing: From Wireless Control to Federated Learning
2024 (English)Doctoral thesis, comprehensive summary (Other academic)
Abstract [en]

Industry 4.0 represents a significant shift in the industrial landscape, aimed at improving efficiency, productivity, and competitiveness. This shift involves the digitalization of industries, impacting manufacturing and maintenance processes. A pivotal element of this transformation is the development of Cyber-Physical Systems (CPS) that seamlessly connect the physical factory floor with the digital realm. These systems monitor real-time data from the physical world and prepare feedback from the digital space, necessitating the harmonious integration of computation and communication, especially through wireless technology. Simultaneously, Machine Learning (ML) methods are advancing across various domains. The proliferation of wireless sensors and the Internet of Things, particularly within the CPS framework, generates substantial data. To address challenges such as latency, device resource limitations, and privacy concerns associated with centralized cloud processing, there is a shift towards edge computing, enabling distributed learning algorithms.

This dissertation tackles these challenges with four innovative methods that combine wireless technology, control systems, and distributed ML in the context of Industry 4.0. These methods aim to harness the potential of this digital transformation, making Predictive Maintenance (PdM) in industries smarter and more efficient. The first method, parallel event-triggering, is designed for multi-agent systems in industrial environments. It utilizes distributed event-based state estimation to enhance control performance and reduce network resource consumption. The second and third methods are developed for collaborative PdM using wireless communication in a federated approach. The second method focuses on real-time anomaly detection while preserving asset privacy at the edge level, and the third method optimizes remaining useful life prediction from sequential data within a federated learning framework. Both federated approaches enhance efficiency, simplify communication, and improve local model convergence. The fourth method introduces an innovative approach to collaborative PdM, utilizing over-the-air computation at the edge level. This approach offers low latency and improved spectral efficiency. The optimization challenges at the edge level are addressed by using a modified gradient descent approach, which effectively handles noisy communication channels and improves the convergence of ML algorithms. All four methods proposed in this thesis underwent a comprehensive evaluation,and the experimental findings demonstrate their effectiveness in achieving their intended objectives.

Abstract [sv]

Industri 4.0 representerar en betydande förändring i det industriella landskapet i syfte att förbättra effektivitet, produktivitet och konkurrenskraft. Denna förändring innebär en digital transformation av industrier, vilket påverkar tillverknings- och underhållsprocesser. En central del av denna transformation är utvecklingen av Cyber-fysiska system (CFS) som sömnlöst kopplar sammandet fysiska fabriksgolvet med den digitala världen. Dessa system övervakar realtidsdata från den fysiska världen och analyser från den digitala sfären, vilket kräver harmonisk integration av beräkning och kommunikation, särskilt genom trådlös teknik. Samtidigt utvecklas maskininlärningsmetoder inom olika områden. Spridningen av trådlösa sensorer och Internet of Things, särskilt inom CFS-ramverket, genererar betydande data. För att hantera utmaningar som latens, begränsade resurser hos lokala enheter och integritetsbekymmer kopplade till centraliserade molntjänster sker en övergång till beräkningar på nätverkets kant, även kallat edge computing, vilket möjliggör distribuerade inlärningsalgoritmer.

Denna avhandling tacklar dessa utmaningar med fyra innovativa metoder som kombinerar trådlös teknik, styrsystem och distribuerad maskininlärning inom ramen för Industri 4.0. Dessa metoder syftar till att utnyttja potentialen i denna digitala transformation och göra prediktivt underhåll i industrier smartare och effektivare. Den första metoden, parallell händelseutlösning, är utformad för fleragentssystem i industriella miljöer. Den använder distribuerad händelsebaserad tillståndsestimering för att förbättra reglerprestanda och minska förbrukning av nätverksresurser. De andra och tredje metoderna är utvecklade för att prediktivt underhåll och trådlös kommunikation skall samarbeta med ett federativt tillvägagångssätt. Den andra metoden fokuserar på avvikelsedetektering i realtids samtidigt som dataintegriteten bevaras på edge-nivå, och den tredje metoden optimerar förutsägelse av kvarvarande användbar livslängd från sekventiella data inom en federativ inlärningsram. Båda federativa tillvägagångssätten förbättrar effektiviteten, förenklar kommunikationen och förbättrar lokalt konvergensen av modeller. Den fjärde metoden introducerar ett innovativt tillvägagångssätt för samarbetsbaserat prediktivt underhåll, där over-the-air beräkningar används på edge-nivån. Denna metod erbjuder låg latens och effektivare utnyttjande av frekvensband. De utmaningar som finns på edge-nivån hanteras genom att använda en modifierad gradient descent metod, som effektivt hanterar brusiga kommunikationskanaler och förbättrar konvergensen av ML-algoritmer. Alla fyra metoder som föreslås i denna avhandling genomgick en omfattande utvärdering, och de experimentella resultaten visar deras effektivitet för att uppnå sina avsedda mål.

Place, publisher, year, edition, pages
Gävle: Gävle University Press, 2024. p. 94
Series
Doctoral thesis ; 42
National Category
Electrical Engineering, Electronic Engineering, Information Engineering
Identifiers
urn:nbn:se:hig:diva-43564 (URN)978-91-89593-23-7 (ISBN)978-91-89593-24-4 (ISBN)
Public defence
2024-04-04, 12:108, Kungsbäcksvägen 47, Gävle, 13:00 (English)
Opponent
Supervisors
Available from: 2024-03-14 Created: 2024-01-10 Last updated: 2024-03-14Bibliographically approved

Open Access in DiVA

fulltext(1297 kB)56 downloads
File information
File name FULLTEXT01.pdfFile size 1297 kBChecksum SHA-512
2535a4e905ad9f0837a64d590a0fa640528ee71950562a9736f5fc183422924c97485d6ca5d1cd398f613e5e3e0034f81e02ca22b4cfa56d09fdab8f447967da
Type fulltextMimetype application/pdf

Other links

Publisher's full textPubMedScopus

Authority records

Bemani, AliBjörsell, Niclas

Search in DiVA

By author/editor
Bemani, AliBjörsell, Niclas
By organisation
Electronics
In the same journal
Sensors
Electrical Engineering, Electronic Engineering, Information Engineering

Search outside of DiVA

GoogleGoogle Scholar
Total: 56 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

doi
pubmed
urn-nbn

Altmetric score

doi
pubmed
urn-nbn
Total: 154 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • harvard-cite-them-right
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • sv-SE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • de-DE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf