Vibration measurements, temperature maps, surface images… Sensor data is a gold mine of information about the health of machines. But you need to know how to use it.
One way to do this is to have historical data on the machine. This identifies two things.
First, abnormal diets that have not been seen before. These are situations that may or may not be normal and deserve to be alerted.
Going further, it is possible to recognize and generalize from sensor data obtained during previous failures. The A.I. will learn to recognize the signs that may lead to a malfunction.
On some systems, it is possible to model the physical behavior of the machine in order to have a digital twin. The interest of this digital twin is to be able to calculate the potential impacts of new situations.
Finally, it is possible to perform real-time acquisition from A.I. technologies such as computer vision, on data that cannot easily be monitored by traditional sensors. For example, vision can be used to identify corrosion on a surface, or to control the pollution of the lubricating liquid.