Vibration sensors on the machinery are linked to a gateway incorporating edge analytics. The sensor data is initially processed in the gateway and communicated to the cloud where analytical algorithms reside.
The advanced machine-learning algorithms allow potential failures to be predicted, based on capturing and analysing abnormal patterns of vibrations and rotational speeds, which are mapped to failure mechanisms. This approach provides continuous, around-the-clock monitoring without needing the full cost of networking, storage and a datacentre infrastructure.
By implementing statistical machine-learning algorithms in the gateways, the amount of data that needs to be transmitted to the cloud is greatly reduced. Continuous monitoring of all the rotating machinery in a plant can result in Terabytes of data every day, but by using edge computing, the amount of data that needs to be stored, analysed and transmitted is said to be reduced by more than 99.9% to a few megabytes per day.
Data can be sent from the gateway to control systems using hard-wired connections, or to the cloud via Ethernet, Wi-Fi, or cellular networks.
The Quick Predict system can also be configured to send alerts and show only data that indicates an issue, thus saving time and effort.
“We originally designed the vibration analytical algorithm for our own fabrication plants to improve maintenance and uptime,” explains Chet Hullum, general manager of Intel’s Industrial Solutions Division. The technology “is now commercially available to the growing number of enterprises that need to enhance their IoT deployments with a solid solution that helps increase productivity and save costs by preventing full failure.”
Harman and Intel say that their new technology will help organisations to boost uptime, cut spare parts costs, and optimise the use of their workforces by reducing emergency repairs.
Maintaining rotating equipment is, they point out, expensive and resource-intensive. Even with spare parts and machines in place, a pump failure can cause costly production delays, necessitating emergency work and hurried scheduling of maintenance crews. And the spot vibration readings collected manually by technicians under weekly or monthly preventative maintenance programmes, “simply do not provide the data needed to identify all problems early enough to allow for planned repair”.