Pattern recognition in 2d graphical data

I collect data from machinery such as vibration data. This would be time series data - where the x axis would be TIME and the y axis would be something such as 'vibration' or 'voltage' or 'energy' etc. I want to be able to tell the AI tool what certain patterns mean - for example, when a machine starts up, it will have a distinct vibration signature, or a amperage usage signature.

Once the AI is 'trained' on different signatures, I want to then feed it realtime data and then the AI would be analyzing this realtime data (at a configurable time period) to look for these signatures.

I should be able to tell the tool to find me signatures with an 80% similarity to the 'baseline' that was fed in earlier.

This tool can then be used to analyze data to look for potential problems. For example, if a machine starts up and the vibration signature is 90% similar, but the amperage signature is 60% similar then we may be looking at a potential issue.