Killing WC Softly with AI

Great article here! Another way to kill the art faster, if its not dead already. How do you teach nuances, feel, the thinking behind how to make the art come alive and apply it. Below is the video showing the launching of the book for this project :-

In the field of engineering I’ve had companies talk about how they want to develop AI software to do machine maintenance. They think that with big data they can anticipate machine problems. One local company even used AI software to spot problems by examining data patterns.

This is what happens when today’s engineers and software programmers jump into the fray without the benefit of actual hands-on maintenance work. It is like nobody wants to look at a problem any more, much less try to solve it. They just want to throw everything to big data.

In the world of machine problems there are also nuances and small details to look into to solve a problem. Data can tell you its one problem or the other depending on the type of data collected, how you collect it and how you intepret the data.

I remember one story of a ship engineer calling for service because the machine was vibrating a lot. When the engineer I sent boarded the ship he asked the ship engineer to show him what he did. The ship engineer put his measurement probe on the machine cover and showed that the vibration was high.

And that exactly was the source of the problem, except there was no problem at all. Its more of a case of what is commonly known as GIGO – garbage in, garbage out. The probe was placed on a part that would vibrate more than usual so it is not unusual for the vibration to be high. But placed at the proper place to take data there was no issue.

A second example is that of a plant engineer arguing that a machine situated below a huge tank was causing vibration to be high. A measurement showed that the tank vibration was fine but the enginner insisted that this was the cause but data showed otherwise. Finally, my engineer found the cause – it was the other way round in that the huge tank caused the machine to vibrate.

However, since the tank is not a machine that moves how did it cause the vibration? When we think of movement we frequently think of visible movement. A tank that does not seem to move is always moving, except the motion is not visible to the naked eye. You can only see the movement if you have a high speed camera.

In this cause the tank is old and huge. If a tank is old its structure may no longer support it the way it is supposed to work. It won’t collapse but the tank may press against the machine below it. Plus, more investigative work uncovered that the tank was supposed to be reinforced as per the original design. But as-built the tank was never reinforced.

So therein are the possible causes which rectification work can confirm if this is the cause. It was interesting that the plant engineer was not aware that the tank was supposed to be reinforced much less know why the reinforcement was never carried out.

This case is an example of how big data would not be able to solve the problem because the cause is due to something that was not there in the first place. So even if a baseline measurement was carried out on commissioning the data would be flawed because of the omission of the reinforcement.

The idea of AI is very appealing. I know many customers want to use AI because its what we call “new toilet”, basically they want it cause its new and they think it will be useful but unless they understand their own problems they will not know if a new approach is truly superior.

Last example which I use to explain to customers about solving problems. One plant claimed that their new pump has a problem. The vendor that sold the pump to them took their word for it and called for someone to solve the problem. Yes, the vibration was high but the pump didn’t seem problematic so what could be the cause?

The data couldn’t explain the problem away. Then my engineer, the human AI, asked the right question about the pump’s load. And there was the answer. The problem was not so much caused by the pump but by how the customer was using the pump. In layman terms we can think of a person of 250 lb lying on top of a bed designed to take the weight of a 150 lb person, causing the bed to collapse and finding fault with the person instead of the person who laid on the bed. In technical terms the pump was designed to run above 90% load but the customer was running at 85% load, causing the pump to have vibration problem (if you want the full technical explanation you’ll have to attend two levels of an engineering course to learn about this). Once the explanation was given the customer could easily verify the cause. Again, data couldn’t solve this problem, merely point out that the cause is not due to the typical causes.

So in using AI to teach WC I would say interesting, nice, great but there is no way to teach the nuances, small details and feel of a movement. I am not against technology, just that not everything can be solved by technology.

Oh, the technology we use for solving machine problems have been around for more than 20 years. AI can’t improve on them, just collect a lot more data which someone or a software programmed not necessarily by a machine engineer (we typically call them mechanical engineer) who has to teach the AI how to intepret the data. This in turn depends on the input they get from the people they consulted when designing the software and the limitations of the software.

It would be nice if this was another revolution like “De revolutionibus orbium coelestium”.

But I somehow feel as it this was more like what it was before Nicolaus Copernicus introduced his revolutionary model that came to be known as Copernican heliocentrism.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s