Related Experiment Video
Updated: Aug 17, 2025
11:18
Closed-loop Neuro-robotic Experiments to Test Computational Properties of Neuronal Networks
Published on: March 2, 2015
10.4K
Spiking Neural Networks for Structural Health Monitoring
George Vathakkattil Joseph1, Vikram Pakrashi1
1UCD Centre for Mechanics, Dynamical Systems and Risk Laboratory, School of Mechanical and Materials Engineering, University College Dublin, 4 Dublin, Ireland.
Sensors (Basel, Switzerland)
|December 11, 2022
Summary
This study introduces spiking neural networks (SNNs) for extracting neural cepstral coefficients from structural vibration signals. This neuromorphic approach enables efficient anomaly detection in structural health monitoring (SHM).
Area of Science:
- Neuromorphic Computing
- Structural Health Monitoring
- Signal Processing
Background:
- Traditional structural health monitoring (SHM) faces challenges in power efficiency for sensor nodes.
- Extracting features like cepstral coefficients from vibration signals is crucial for SHM.
- Neuromorphic computing offers potential for low-power, efficient processing.
Purpose of the Study:
- To implement a spiking neural network (SNN) for cepstral coefficient extraction in SHM.
- To demonstrate the effectiveness of SNNs for feature extraction from operational vibration signals.
- To evaluate the application of SNN-derived features for anomaly detection.
Main Methods:
- Developed and implemented an SNN for cepstral coefficient extraction.
- Deployed the algorithm on specialized neuromorphic hardware (Intel® Loihi).
- Benchmarked performance using numerical and experimental data from a degraded single-degree-of-freedom system.
Main Results:
- Successfully extracted neural cepstral coefficients using SNNs.
- Demonstrated the efficacy of these coefficients for anomaly detection.
- Validated the approach on a single-degree-of-freedom system with stiffness changes.
Conclusions:
- Spiking neural networks are effective for cepstral coefficient extraction in SHM.
- Neuromorphic hardware enables power-efficient implementation of SHM algorithms.
- This work pioneers a non-Von Neumann computing approach for SHM applications.
