A fabless semiconductor company, POLYN provides application specific Neuromorphic Analogue Signal Processing (NASP) technology and Neuromorphic Front End (NFE) chips for always-on sensor-level solutions. NFE enables ultra-low power consumption, low latency, and high resiliency for AI products deployed at the thin edge.
VibroSense extracts unique patterns from sensor’s raw signal and passes the valuable data only for classification at the next compute point. It has been designed to be integrated into existing solutions to improve the ROI and OPEX of the deployment and is particularly suited to Industry 4.0 applications.
“Predictive Maintenance solutions usually require cloud services and are resource-hungry,” said Eugene Zetserov, Vice President of Marketing for POLYN. “The sensor node consumes a lot of electricity on vibration data transmission. Collecting data from many sensors requires considerable resources such as the sensor node infrastructure, radio bandwidth, data processing and storage in the cloud. VibroSense reduces the need for these resources through a thousandfold reduction of sensor data to be sent for analysis in the central cloud or at the edge.”
POLYN provides a framework for trained neural networks conversion into an analogue neuromorphic chip for inference. It supports a hybrid architecture where unique patterns of the specific signal extracted in the analogue portion, leaving classification for the digital element. According to POLYN, VibroSense provides much greater flexibility along with power consumption savings and specific machine adaptation to allow various deployments within the same chip.
“VibroSense is the only analogue neuromorphic solution currently on the market that extracts vibration signal patterns at the sensor level. It not only saves IIoT network bandwidth and reduces total cost of ownership, it enables faster adaptation of predictive maintenance solutions, better performance and sustainability,” explained Zetserov.
