Abstract
Infrastructures such as roads, railway tracks, buildings,
bridges and tunnels etc. suffer from degradation over
a period of time. Phenomena such as ageing, hazards and
natural disasters significantly affect the structure’s sustainability.
Structures require timely monitoring and maintenance to assure
long term structural integrity, health and safety. Structural
Health Monitoring (SHM) helps in early damage detection and
improvement in structures reliability by enabling retrofitting
of wireless sensors to capture data to help detecting potential
failures. In this work, an innovative design concept for an
autonomous system based on Self-repairing Spiking Artificial
Neural Networks (SANNs) with Self-powered Sensor Nodes has
been described. SANNs offer an energy efficient computing
approach with the capability to mimic the self-repair principles
of the human brain for highly-reliable information processing.
It is adept to tolerating failures by adapting the network
topology and re-learning the input-output mapping, whereby
it can bypass a significant number of non-operational nodes
whilst still maintaining an adequate SHM sensory infrastructure.
The design also provides an opportunity to reduce the power
consumption of the wireless sensor by implementing the model
as a Big-little architecture. This enables extremely low energy
always-on ‘little’ sensor nodes (SN) to operate in combination
with ‘big’ on-demand, neuromorphic anomaly detectors (NAD).
Energy harvesting through vibrational or solar ambient energies
can help supply sufficient power for the sensor node to maintain
an ‘always-on’ state, while periodic analysis of sensor output data
exceeding a threshold-level could help in anomalies detection for
self-repairing algorithm. A pulsed mode with on-time and eventbased
duty cycle management helps in minimizing the average
power consumption of sensor nodes. The NAD powered typically
by solar energy harvesting, is activated on threshold attainment
and analyses the incoming data for further anomalies.
bridges and tunnels etc. suffer from degradation over
a period of time. Phenomena such as ageing, hazards and
natural disasters significantly affect the structure’s sustainability.
Structures require timely monitoring and maintenance to assure
long term structural integrity, health and safety. Structural
Health Monitoring (SHM) helps in early damage detection and
improvement in structures reliability by enabling retrofitting
of wireless sensors to capture data to help detecting potential
failures. In this work, an innovative design concept for an
autonomous system based on Self-repairing Spiking Artificial
Neural Networks (SANNs) with Self-powered Sensor Nodes has
been described. SANNs offer an energy efficient computing
approach with the capability to mimic the self-repair principles
of the human brain for highly-reliable information processing.
It is adept to tolerating failures by adapting the network
topology and re-learning the input-output mapping, whereby
it can bypass a significant number of non-operational nodes
whilst still maintaining an adequate SHM sensory infrastructure.
The design also provides an opportunity to reduce the power
consumption of the wireless sensor by implementing the model
as a Big-little architecture. This enables extremely low energy
always-on ‘little’ sensor nodes (SN) to operate in combination
with ‘big’ on-demand, neuromorphic anomaly detectors (NAD).
Energy harvesting through vibrational or solar ambient energies
can help supply sufficient power for the sensor node to maintain
an ‘always-on’ state, while periodic analysis of sensor output data
exceeding a threshold-level could help in anomalies detection for
self-repairing algorithm. A pulsed mode with on-time and eventbased
duty cycle management helps in minimizing the average
power consumption of sensor nodes. The NAD powered typically
by solar energy harvesting, is activated on threshold attainment
and analyses the incoming data for further anomalies.
| Original language | English |
|---|---|
| Title of host publication | 2022 IEEE Zooming Innovation in Consumer Technologies Conference (ZINC) |
| Publisher | IEEE |
| Pages | 1-4 |
| Number of pages | 4 |
| ISBN (Electronic) | 978-1-6654-8374-2 |
| ISBN (Print) | 978-1-6654-8375-9 |
| DOIs | |
| Publication status | Published (in print/issue) - 8 Aug 2022 |
| Event | IEEE Zooming Innovation in Consumer Technologies Conference (ZINC): ZINC - Serbia Duration: 22 May 2022 → 26 May 2022 |
Conference
| Conference | IEEE Zooming Innovation in Consumer Technologies Conference (ZINC) |
|---|---|
| Period | 22/05/22 → 26/05/22 |
Keywords
- Spiking Neural Networks
- Energy harvesting
- structural health monitoring
- autonomous nodes
