Can I use an Artificial Intelligence system on a machine not instrumented and without data history?
Just a quick post here as it seems to be a confusion – when we talk about predictive maintenance – between the definition of “Artificial Intelligence”, and “Machine learning”.
If this is true that machine learning really has been draining the interests of all scientists lately (lots of data coming from everywhere on which they can develop new kick-ass algorithms), this is not the only form of Artificial Intelligence available for predictive maintenance. If this is what makes scientists vibrates, an Artifical Intelligence System is more than that. According to google dictionary it is a “Computer system able to perform tasks normally requiring human intelligence”. Good news is, there are solutions available to tell you automatically what’s going on with your machine.
At ACOEM we have brought explicit solutions such as the Accurex™: Using a Bayesian network, we’ve been able to successfully reproduce the approach of vibration of our experts. No history needed, no learning phase needed. Just describe your machine (e.g. motopump 1500RPM 100kW), and you’re good to go!
For sure it’s not designed for the most complex machines (you know, the ones already instrumented on which everybody – including ourselves – are doing some machine learning experimentations?), but more for that poor non instrumented machine (which btw has hundreds of sisters in your factory hitting your maintenance budget), those machines for which you cannot afford to instrument all sensors of the planet and for which you cannot wait months before knowing if something is really happening…
Artificial Intelligence is not a just luxury for instrumented machines that are still today a minority out there. We believe that, capitalizing on experts knowledge, it will make the job of every maintenance operator easier, empowering him with new abilities he never though he had and contributing to more reliable installations, and more efficient maintenance.