Part 2: Reality check - AI/ML applied Machine Health

Thank you for your likes, responses and shares by the kick-off with part 1 what fueled me to share some more insights and to remove some of the hype around applied AI/ML for Machine Health condition tools with electrical rotating equipment.

The basics are the same - Back to the future - Complexity eliminated & improved ROI

Going back, half a decade ago, trough my private venture, we did solve asset reliability for production lines with ML tools and reusing data out of existing workflow historian databases and time series based structured data and state-full data from machines. For that time, in the early days of IoT maturity, it was pretty cool "future stuff" where today these ML tool sets have become a commodity where army's of data scientists are labeling unstructured data sets to train the models what was a pretty complex service. Start the time laps at 0:20 on below trailer and it provides the insights how we "assembled" the required components to deliver the solution.

Big PROBLEMS, BIG business

Inside part 1 of the Reality check - AI/ML applied Machine Health vibration analysis various issues did pop-up in the case to apply ML out of massive amounts for vibration data and concluded that it has very limited opportunity to scale those kind of applications. The last couple years various Motor OEM's like #ABB with Ability Smart Sensor, #Siemens Simotics IQ and #WEG Motor scan entered the Asset Management market positioning their (I)IoT transition with new offerings including compact and easy to fix sensors mounted to the outside of the motors and promising a wealth of insights. Anyone that has some basic understanding of vibration data en information that is collected from a surface is able to conclude that it is great marketing stuff for the exhibition and it likely does not excel. The reasoning is due to searching for the attractive ROI models for the millions of machines like pumps, blowers / compressors and other electrical motor driven rotating equipment in our brown field world. Yes, the size and scale of inefficient and ineffective maintained installed assets is beyond our imagination, not even to go into the subject of wasted energy and water, many are searching to answer the question if repairing or replacing these painful objects makes sense.

What is neXt

Motor Power Monitoring has proven to be a highly valuable maintenance tool promoted by companies like #Fluke. Although current signature analysis (MCSA) is a relatively young, rarely utilized technology, it is rapidly gaining acceptance in industry today. Mechanical faults related to belts, couplers, alignment and more are easily found through the use of a demodulated current spectrum. MCSA is simply the process by which motor current readings are recorded and analyzed in the frequency domain. It has been around since 1985 and proven itself well over the years in locating rotor faults and air gap problems in motors.

The motor current signature is recorded in a time domain format. The current is represented in a graph form with the amplitude shown on the “Y” axis and the time on the “X” axis. The result is a typical current sinewave shown in Figure 1. (left)

In order to analyze the data, a Fast Fourier Transform (FFT) is performed. An FFT is a mathematical operation designed to extract the frequency information from the time domain and transform it into the frequency domain. An example of an FFT spectrum is shown in Figure 2. (right).

While the FFT spectrum is a great source for identification of rotor bar problems in motors, it proved difficult to analyze most other frequencies. In order to address this problem, the demodulated current spectrum was developed. With all that demodulated MCSA can do, there are still many questions as to how it will fit into and benefit a PdM program. Frequently asked questions might include: Why do I care about finding mechanical faults with a demodulated current signal if I can find them with technologies like vibration analysis or infrared thermography? How reliable is the data that is generated and can it take the place of vibration? How often should MCSA be completed on equipment?

Rep.AI.r or replace?

To answer the question if an asset needs to be repaired or replaced it would also be helpful if the information could be provided what the condition of the asset is and information is provided within the alarm context what the kind of failure is expected. In case we would apply ML/AI to MCSA in a 24/7 EDGE condition monitoring environment it can be done.

A fundamentally different approach to exploiting the opportunities around artificial intelligence is to redesign the entire business around AI. This means that the core of the business is fundamentally re-evaluated from the perspective of a fully automated, AI-driven approach. The intent is to identify which business processes are truly needed and which are superfluous. Subsequently, the remaining processes are then redesigned to become fully automated without human involvement. The more central the business process is, the more energy should be put into automating it and driving it through ML or DL techniques. An example of how MCSA is analysed and AI/ML toolset applied is provided in the below animation.

Credit to Jan Bosch whom recently nicely addressed an illustrative example of the above is the rapid growth of companies like Uber and Lyft. Rather than putting a human in a taxi call center and having people call that phone number, the core business process is fully automated and AI driven. It involves humans only as drivers who can accept or reject a ride and passengers who order through their mobile device.

In the next chapter we will dive deeper into that with the relationship of Asset Health Maintenance processes and how you can to take full benefits from it with real world solutions.

Thank you for your valuable time to read and commend,