Two Minute Tips  

Artificial Intelligence vs Vibration-based Diagnostics

Pawel Lecinski | Reliability Lead, CMRP

Artificial Intelligence vs Vibration-based Diagnostics

Condition-based maintenance (CBM), predictive maintenance (PdM), and most of all prescriptive maintenance (RxM) can significantly improve planning and scheduling,  increase the lifespan of the equipment and its overall reliability, and reduce spare parts in warehouse and these affect the entire business.

Vibration-based diagnostics is one of the most useful CBM (PdM) techniques that allows you to detect defects in the industrial equipment related to physical wear and tear of its components and other related factors. Application of prediction methods, based on in-depth analysis of vibration signals, allows for increasing the intervals between repairs, as well as for preventing issues associated with unacceptable condition of equipment mechanical components.

However, what is important to know is that vibration monitoring techniques that are widely used at present generate the domain of informative features of equipment defects in a form that is extremely problematic for automatic systems. The main issue is that the scope of this description is not sufficient at all for steady work of machine learning algorithms. Traditional methods of vibration signal processing were proposed more than 50 years ago and are still actively used. Moreover, vibration-based condition monitoring has traditionally been an expert-oriented field of knowledge, and a human as expert is not physically able to work with a large amount of data.

Additionally, at the same time, the condition monitoring systems are closed software and hardware environments.

It is not an easy job to create a good machine learning algorithms-based condition monitoring system. Of course in the market we can find many solutions that are ‘Industry 4.0’ with AI and machine learning, but what does it mean in reality?

Vibration analysis as CBM can provide a lot of information about problems with the equipment like misalignment, unbalance, bearing problems, etc. We can detect also electrical (stator eccentricity, shorted laminations and/or loose iron, motor broken bars, etc.) or hydraulic (cavitation, turbulence) issues.

So we should avoid the following in our automatic systems:

  • condition monitoring systems that collect the data and support generalized notifications such as ‘something is wrong’ – detailed and precise defect detection and reporting (rotor imbalance, shaft runout, wear of the outer ring of the bearing/gear, etc.) solution is required,
  • non-automatic systems that rely upon manual diagnostics by a team of experts: (‘notify human experts’),
  • systems that give information about probability of failure and time when it happens, however saying nothing of what will happen, which part (element) will fail,
  • non-scalable automatic systems which require a long-term and expensive R&D for each specific type of equipment.

To take a real step forward, we need truly automatic diagnostics, where no humans are involved. The new system must have detailed and precise defects detection and reporting (rotor imbalance, shaft runout, wear of the outer ring of the bearing/gear, etc.) and has to support most types of equipment. Only then can we say that AI is more useful than the PdM with human experts we have now.

Notify of

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Inline Feedbacks
View all comments

About the Author

Pawel Lecinski Reliability Lead, CMRP

Pawel is EU Predictive Maintenance SME and Reliability Lead North-Central Europe II in a food industry company. He has 10 years of experience in maintenance and reliability engineering. Pawel holds a master’s degree in Automatic Control & Robotics and postgraduate in Accounting & Financial Management and Project Management. He is certified ultrasound and vibration analyst with practical experience in IR, motion amplification, MCE and tribology. He has also high interest in Industry 4.0, new technologies and big data management. For his work and contribution to the maintenance, reliability and physical asset management profession he received many corporate and industrial awards Rising Leader CMRP of the Year 2017 including.