Introduction & see attachment
VERHAERT is an SME with a clear focus on innovation and development of new
products for international customers in various sectors. We offer a broad range
of product development services, from business consultancy to series
production.
Predictive maintenance is growing in popularity as anomaly detection becomes
more cost-effective to implement. Anomaly detection can be done in the cloud,
on an edge server, or on an edge device (IoT device). Nowadays, companies are
becoming more aware of operational costs associated with a centralized cloud
solution and are looking for an economically viable alternative.
One potential solution is to run a neural network directly on the sensor. This
may require a more expensive sensor, but it can enable the hardware after the
sensor, such as edge devices, edge servers, and the cloud to be more cost
effective. Chip manufacturers like STM have released a new type of IMU sensor
with this capability, called Intelligent Sensor Processing Units (ISPUs). They can
run an AI inference on a MEMS IMU and offer solutions to train and port the
neural network to the hardware. Migrating this intelligence to the sensor is a
trend in which we as Verhaert see great potential.!
