top of page

Storing raw data in an AI-friendly way? A real competitive advantage!

You know how to effectively and efficiently maintain the machines in production. For this reason, you no longer replace bearings, gears, etc. based on time, but on a forecast or the actual condition. Maybe you work with an external service provider who regularly sends you a report on the condition of the spare parts or you have built up internal know-how within your own team. Are you sure that you have already exhausted everything that is possible?

You can be sure if your service provider has already offered you an AI model to make the service offered cheaper. You can also be sure if you are given a remaining lifetime with each report or a data set that is suitable for AI training. After all, these are precisely the requirements that give you a real advantage and, of course, save you money.

You surely have reasons for preferring to hire an external service provider instead of building up the know-how regarding predictive maintenance or condition monitoring in your own company. You are fine with someone coming by regularly, measuring your systems and sending you a report as a result after a few days. Hand on heart: unfortunately, such a report is usually not self-explanatory and what it says is more of a "description of the data" and less of a "you have to do this now", is it not? Your maintenance staff rightly ask themselves whether they have to replace the bearing if they hear the statement "The outer ring on the bearing is damaged" or whether they can wait (for how long?). Also with the statement "According to the standard, the threshold value has been exceeded.” your maintenance staff ask themselves what exactly to do, what is the best decision.

Is a bearing damaged or the motor not screwed on properly? Whatever - the external service providers usually do not answer these kinds of questions because the depth of knowledge of the analysis is not sufficient for this. Honestly, though, that is exactly what you and your maintenance staff want: Know when to replace a component.

And the management adds: Exchange yes, but please as close as possible to the actual time of failure, because then we also save costs. But you also have to be able to live with that: Your external service providers use devices that are usually not capable of training artificial intelligence. They train on these devices for weeks, listening to the ultrasound with headphones.

From the collected data should result clear recommendations for action

Special evaluations follow - without recommendations for action. In this way, valuable measurement results are collected and accumulated over years - but it is not possible to work with them because they are not stored in a suitable form, i.e. suitable for AI. But that is the future. Go ahead and ask your external service provider whether it is possible to train an AI with the data you have collected. The answer will most likely be 'no'. Currently, none of the established providers stores the valuable system data collected over many years in an AI-compliant manner. A valuable opportunity for digitalisation is therefore not only not being used, but is being missed.

It is therefore worthwhile to think carefully about which device to start the predictive maintenance journey with. But it is never too late to rely on Senzoro and its Ultrasound Condition Monitoring System...

Author: Dipl. Ing. Mag. Markus Loinig, Senzoro GmbH




Rated 0 out of 5 stars.
No ratings yet

Add a rating
bottom of page