What might cause the U-Curve of Fault Rate vs Intermittency?

Options
Dov Stern
Dov Stern
edited July 2023 in Share Your Ideas

Hi all! My name is Dov Stern and I'm one of the product managers at Augury!

Using our leading Machine Health database, we can discover new relationships with fault rates in machines.
With your collaboration, we can turn those relationships into insights to help you bring fault rates down.

We found that intermittency has a U-Curve relationship with fault rates.
As machines become less continuous in operation, fault rates first peak on machines that are off 2-10 hours per day, before reaching a low point for machines off 10+ hours a day.

Help us discover why fault rates peak for machines at the intersection of continuous and intermittent operation:

Do you have certain machines that are designed for continuous operation while others are designed for intermittent operation?

How do you protect against startup stresses on the machine? How do you determine for which machines is this a concern?

How do your maintenance practices differ across machines that have more or less continuous operation?

Thanks in advance for your feedback!

Tagged:

Comments

  • Tal Gurevich
    Options

    Hi Everyone,
    My name is Tal, and together with Dov, we lead the AI product domain at Augury Machine Health, along with several teams of algorithm developers, physicists and machine learning data engineers. The data we share here is based off of our vast network of machines and hundreds of millions of vibration recordings.

    We invite you to comment share and discuss!

  • Jeff_meyers
    Options

    @Tal Gurevich Is it common to observe vibration changes following the greasing of a bearing? Alternatively, can the act of greasing a bearing be deduced from analyzing recorded vibration patterns?

  • Tal Gurevich
    Options

    @Jeff_meyers correct, it is common to see reduction in overall vibration features along with the lowering of confidence on detectors of anomalies or faults like bearing wear. We have also trained our models to discern improvements in machine condition and severity. To deduce that is definitely from lubrication is a product of training it on more sharp and tagged data for this exact maintenance event.

    tell me more about how and where this would meet your specific needs? is it that you would feel more confident knowing there's an improvement post-lubrication? that you need to know exactly what to act upon when alerted for lubrication?

Categories