Resource-efficient Edge AI solution for predictive maintenance

被引:2
作者
Artiushenko, Viktor [1 ]
Lang, Sebastian [2 ]
Lerez, Christoph [3 ]
Reggelin, Tobias [1 ]
Hackert-Oschaetzchen, Matthias [3 ]
机构
[1] Otto von Guericke Univ, Modeling & Simulat Working Grp, Unv Pl 2, D-39106 Magdeburg, Germany
[2] Fraunhofer Inst Factory Operat & Automat, Sandtorstr 22, D-39106 Magdeburg, Germany
[3] Otto von Guericke Univ, Chair Mfg Technol Focus Machining, Fac Mech Engn, Univ Pl 2, D-39106 Magdeburg, Germany
来源
5TH INTERNATIONAL CONFERENCE ON INDUSTRY 4.0 AND SMART MANUFACTURING, ISM 2023 | 2024年 / 232卷
关键词
Edge AI; predictive maintenance; tool condition monitoring; milling;
D O I
10.1016/j.procs.2024.01.034
中图分类号
TP301 [理论、方法];
学科分类号
081202 ;
摘要
Data-driven predictive maintenance (PM) is an approach that leverages advanced analytics, artificial intelligence (AI), and sensor data to predict when equipment failure might occur, and to perform maintenance just in time to prevent it, reducing downtime and maintenance costs. In manufacturing, one of the biggest potential applications for intelligent PM systems is tool condition monitoring (TCM). TCM aims to monitor tool wear in real-time, ensuring the quality of the manufactured products and the safety of the surrounding people and equipment. In recent decades, many studies have been carried out on tool condition monitoring for different machining operations such as milling, drilling or turning, and have demonstrated the effectiveness of various AI algorithms. Recent advances in hardware have made it possible for edge devices to run complex AI algorithms locally. This technology is called Edge AI. The Edge AI approach has several key benefits such as reduced latency, scalability, data privacy and security that can accelerate the integration of the PM solution for TCM at the production level. This paper presents the design of an Edge AI system for tool condition monitoring, consisting of state-of-the-art, low-cost components and using open-source software. Based on the proposed design, a prototype was built and tested during milling process. Four machine learning (ML) models and one deep learning (DL) model were run on a low performance edge device. Their predictions were validated. Challenges faced in the implementation are concluded along with directions and suggestions for future research. (c) 2023 The Authors. Published by Elsevier B.V.
引用
收藏
页码:348 / 357
页数:10
相关论文
共 21 条
[1]   Tool breakage detection using support vector machine learning in a milling process [J].
Cho, S ;
Asfour, S ;
Onar, A ;
Kaundinya, N .
INTERNATIONAL JOURNAL OF MACHINE TOOLS & MANUFACTURE, 2005, 45 (03) :241-249
[2]  
Fernandez A, 2018, Learning From Imbalanced Data Sets, DOI DOI 10.1007/978-3-319-98074-4
[3]   In-process Tool Wear Prediction System Based on Machine Learning Techniques and Force Analysis [J].
Gouarir, A. ;
Martinez-Arellano, G. ;
Terrazas, G. ;
Benardos, P. ;
Ratchev, S. .
8TH CIRP CONFERENCE ON HIGH PERFORMANCE CUTTING (HPC 2018), 2018, 77 :501-504
[4]   Application of backpropagation neural network for spindle vibration-based tool wear monitoring in micro-milling [J].
Hsieh, Wan-Hao ;
Lu, Ming-Chyuan ;
Chiou, Shean-Juinn .
INTERNATIONAL JOURNAL OF ADVANCED MANUFACTURING TECHNOLOGY, 2012, 61 (1-4) :53-61
[5]   1D convolutional neural networks and applications: A survey [J].
Kiranyaz, Serkan ;
Avci, Onur ;
Abdeljaber, Osama ;
Ince, Turker ;
Gabbouj, Moncef ;
Inman, Daniel J. .
MECHANICAL SYSTEMS AND SIGNAL PROCESSING, 2021, 151
[6]  
Kochenderfer MJ, 2019, ALGORITHMS FOR OPTIMIZATION
[7]  
Lee YL, 2018, INT SYMPOS VLSI DES
[8]  
Mallisetty Sai Balaji, 2023, 2023 International Conference on Intelligent Data Communication Technologies and Internet of Things (IDCIoT), P798, DOI 10.1109/IDCIoT56793.2023.10053520
[9]   A comprehensive tool-wear/tool-life performance model in the evaluation of NDM (near dry machining) for sustainable manufacturing [J].
Marksberry, P. W. ;
Jawahir, I. S. .
INTERNATIONAL JOURNAL OF MACHINE TOOLS & MANUFACTURE, 2008, 48 (7-8) :878-886
[10]   Design and Development of an Edge-Computing Platform Towards 5G Technology Adoption for Improving Equipment Predictive Maintenance [J].
Mourtzis, Dimitris ;
Angelopoulos, John ;
Panopoulos, Nikos .
3RD INTERNATIONAL CONFERENCE ON INDUSTRY 4.0 AND SMART MANUFACTURING, 2022, 200 :611-619