AI Types

There are three basic types of AI that Cognition can create for you.

A State AI is one that determines the instantaneous state of an asset or assets.

It can't capture things happening over time, but at any one moment it can tell you if the asset is in a particular state.

There are two sub-types of state AI - the Support Vector Machine which works like a check engine light for an asset, letting you know if there's a general problem, or a Neural Network which can learn to recognise specific asset states.

Event AIs detect events happening over time.

These are generally Neural Networks.

Each event-detecting AI works over a fixed window of time, at a fixed resolution. For example, the default event detecting AI works over 10 seconds, at a resolution of one-sample-per-second. However, it's possible to create AIs that work over a much larger time-frame with a coarser resolution, when required.

Model AIs produce values to calculate expected results.

These are generally Neural Networks.

These are used in two key scenarios…

• To produce a 'model' value that you can compare against, to indicate when an asset has output that is not expected, or
• To produce a value that is only indirectly measured and has no known formula, based on manual readings.