Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Low-Power Vibration-Based Predictive Maintenance for Industry 4.0 using Neural Networks: A Survey

Alexandru Vasilache, Sven Nitzsche, Daniel Floegel, Tobias Schuermann, Stefan von Dosky, Thomas Bierweiler, Marvin Mußler, Florian Kälber, Soeren Hohmann, Juergen Becker

TL;DR

This survey analyzes neural-network-based approaches for low-power, vibration-based predictive maintenance in Industry 4.0, contrasting Spiking Neural Networks (SNNs) and Artificial Neural Networks (ANNs) in on-device edge settings. It shows that most SNN work uses shallow, feed-forward architectures with various spike-encoding schemes and limited hardware demonstrations, while ANNs explore deeper, more diverse models and more extensive hardware accelerators. A key finding is the absence of standardized benchmark datasets for PM, with researchers relying on a mix of custom and a few public datasets like CWRU. The paper argues for developing practical edge hardware implementations and a common benchmark to accelerate deployment of low-power PM solutions.

Abstract

The advancements in smart sensors for Industry 4.0 offer ample opportunities for low-powered predictive maintenance and condition monitoring. However, traditional approaches in this field rely on processing in the cloud, which incurs high costs in energy and storage. This paper investigates the potential of neural networks for low-power on-device computation of vibration sensor data for predictive maintenance. We review the literature on Spiking Neural Networks (SNNs) and Artificial Neuronal Networks (ANNs) for vibration-based predictive maintenance by analyzing datasets, data preprocessing, network architectures, and hardware implementations. Our findings suggest that no satisfactory standard benchmark dataset exists for evaluating neural networks in predictive maintenance tasks. Furthermore frequency domain transformations are commonly employed for preprocessing. SNNs mainly use shallow feed forward architectures, whereas ANNs explore a wider range of models and deeper networks. Finally, we highlight the need for future research on hardware implementations of neural networks for low-power predictive maintenance applications and the development of a standardized benchmark dataset.

Low-Power Vibration-Based Predictive Maintenance for Industry 4.0 using Neural Networks: A Survey

TL;DR

This survey analyzes neural-network-based approaches for low-power, vibration-based predictive maintenance in Industry 4.0, contrasting Spiking Neural Networks (SNNs) and Artificial Neural Networks (ANNs) in on-device edge settings. It shows that most SNN work uses shallow, feed-forward architectures with various spike-encoding schemes and limited hardware demonstrations, while ANNs explore deeper, more diverse models and more extensive hardware accelerators. A key finding is the absence of standardized benchmark datasets for PM, with researchers relying on a mix of custom and a few public datasets like CWRU. The paper argues for developing practical edge hardware implementations and a common benchmark to accelerate deployment of low-power PM solutions.

Abstract

The advancements in smart sensors for Industry 4.0 offer ample opportunities for low-powered predictive maintenance and condition monitoring. However, traditional approaches in this field rely on processing in the cloud, which incurs high costs in energy and storage. This paper investigates the potential of neural networks for low-power on-device computation of vibration sensor data for predictive maintenance. We review the literature on Spiking Neural Networks (SNNs) and Artificial Neuronal Networks (ANNs) for vibration-based predictive maintenance by analyzing datasets, data preprocessing, network architectures, and hardware implementations. Our findings suggest that no satisfactory standard benchmark dataset exists for evaluating neural networks in predictive maintenance tasks. Furthermore frequency domain transformations are commonly employed for preprocessing. SNNs mainly use shallow feed forward architectures, whereas ANNs explore a wider range of models and deeper networks. Finally, we highlight the need for future research on hardware implementations of neural networks for low-power predictive maintenance applications and the development of a standardized benchmark dataset.
Paper Structure (37 sections, 1 figure, 2 tables)

This paper contains 37 sections, 1 figure, 2 tables.

Table of Contents

  1. Introduction
  2. Focused Applications
  3. Sensors
  4. Spiking Neural Networks
  5. Fundamentals
  6. Preprocessing / Event Conversion
  7. Approaches
  8. Ultra-low Power Machinery Fault Detection Using Deep Neural Networksnitzsche_ultra-low_2021
  9. Online Detection of Vibration Anomalies Using Balanced Spiking Neural Networks dennler_online_2021
  10. Damage Detection in Structural Health Monitoring with Spiking Neural Networks zanatta_damage_2021
  11. Spiking Neural Network-Based Near-Sensor Computing for Damage Detection in Structural Health Monitoring barchi_spiking_2021
  12. A spiking neural network-based approach to bearing fault diagnosis zuo_spiking_2021
  13. Research on Fault Diagnosis Based on Spiking Neural Networks in Deep Space Environment li_research_2022
  14. Machine Hearing for Industrial Acoustic Monitoring using Cochleagram and Spiking Neural Network zhang_machine_2022
  15. Efficient Time Series Classification using Spiking Reservoir dey_efficient_2022
  16. ...and 22 more sections

Figures (1)

  • Figure 1: Taxonomy of spike encoding techniques. Inspired by auge2021survey.