top of page

Why ultrasound will benefit most from modern AI technologies (part 1)

For years, maintenance has been able to use different technologies for predictive maintenance and condition monitoring. Essentially, the "Big 5" are the following technologies, all with their strengths and weaknesses.

  1. Oil analysis

  2. Vibration analysis (simple)

  3. Vibration measurement (complex)

  4. Ultrasound (passive)

  5. Infrared/Thermography

Putting effort and added value/depth of insight into a chart, one may end up looking at something like this:



The high effort forVibration analysis is mostly driven by investments in permanent sensors as well as their constant analysis effort to derive the desired insights. However, companies who have used vibration technology for many years will outperform other technologies such as Ultrasound and Thermography as analysis of them is still a very manual process. For Ultrasound this is humans listening to sound files and for Thermography, this is humans looking at and analysing infrared images. Oil analysis can creates an unique view and "back up" opinion if stakes are high.


Each of these technologies have earned their spot in the Condition Monitoring Portfolio and ideally a maintenance department masters all of them. However, due to the complexity and restricted ressources, this is a rather unrealistic undertaking. Therefore the question every company needs to face:


If I need to choose one, which Condition Monitoring Technology do I use?

Three weighty reasons speak for Ultrasound:

  1. It can detect failures as early as no other technology and therefore NASA also uses it

  2. Ultrasound is a technology with a very good cost-benefit ratio

  3. Ultrasound technology will get the greatest "boost" by Artificial Intelligence due to its unique characteristics

The main reason for the very good benefit/effort ratio are the simple and non-invasive measurement possibilities. The second reason is the possibility to avoid permanent sensors and use handheld devices. Systems such as motors can be measured non-invasively during normal operation and the application of Ultrasound also extends to other use cases (e.g. detection of electrical faults and leaks), which are not covered in this article.


So why does Ultrasound benefit most from modern AI Technologies?

Artificial Intelligence is only as good as the data on which it can train itself. The data Ultrasound provided is like a perfect fit, so here are the three main reasons:

  1. Ultrasound stays where it is. Two motors in immediate proximity to each other have no negative influence on the measurement. Ultrasound does not "migrate" from the housing of one motor into the air to then penetrate into the housing of the next motor again. So the measurements are not location-specific.

  2. Factories produce little Ultrasonic noise, so the data is very "pure" and free of background noise. Air-born sources of Ultrasound (e.g. leakages) do not mix up with the strucrtural born Ultrasound noises caused by mechanical parts.

  3. Ultrasound has a high information density and a few seconds of measuring time at a bearing location are sufficient for data analysis.

Due to the formentioned characteristics, it is possible for the AI to learn from different data sources in different factories and plants. One can combine similar engines from different factories and different ages to a "virtual" engine. On this basis one then has a data-driven representation of an engine over the lifetime. How exactly these "virtual" engines can estimate Remaining Useful Lifetime (RUL) has still to be shown in practice.


To sum it up, a few years from now, our Condition Monitoring chart may very well look like this, where Ultrasound combined with AI will be the most reliable Condition Monitoring Technology to provide a reasonable accurate estimation of the Remaining Useful Lifetime (RUL) of assets.




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

Email: markus@senzoro.com

0 comments

Комментарии

Оценка: 0 из 5 звезд.
Еще нет оценок

Добавить рейтинг
bottom of page