Building a Smarter Planet. A Smarter Planet Blog.

By Christopher Conradi, Business Analytics, IBM

If you own a car in North America, you’re told to change the oil every 3,700 miles or six months. This applies whether you are living in Florida, driving peacefully to work, or living in Minnesota with frequent subzero temperatures in the winter. In Scandanavia, where I live, we change the oil in our vehicles less frequently because of concern about the environment.

But no matter what schedule you use, the point is that old-fashioned service manuals are not smart. Cars are used in different ways, and should therefore be serviced in different ways. And the same goes for any type of machinery.

Just because machines start out the same way doesn’t mean we should service them the same way. To determine how often we get vehicles serviced, we need to consider the environment in which the machines operate, and how they are being used. The trick, off course, is to figure out just that: where and how are they being used?

It all starts with collecting data. Sensors are becoming increasingly sophisticated. Using heat cameras, we can detect wear inside a ball bearing. Microphones can help us detect the slightest change in frequency of a motor and with accelerometers, which are small sensors that measure acceleration, we can record motion of robotic arms that will give away inconsistencies.

These sensors work much like the nervous system in our body. Each sensor on it’s own is somewhat useful, but when you start combining the sensory data from multiple sources along with statistics and previous recordings, you really start to leverage the potential. Feeling the ground tremble, hearing a train horn, and seeing that you are standing on train tracks, are of little value on their own, but combining the information might prove life saving. With such input, you know that taking one step to the side is smarter than running along the tracks.

This is what we call predictive maintenance. Measuring, in real-time, how machines are doing and combining it with statistics and knowledge to fix things before they break, not after. This gives customers the chance to plan for down time, and do repairs before faulty parts affects others. In many cases, they can limit repairs to a few dollars instead of thousands.

This can also be applied to the products already sold. A car manufacturer could put sensors in its  cars, which would report on how the car is performing. This would give us a large dataset to find faults and errors, which would help evolve future products or make the servicing the cars smarter. In other words, letting the customer know that a part is about to break before it actually does.

On a smarter planet, we will stop treating cars — or machines — as a homogenous group. Since each one is used differently, it should be serviced by looking at the health of each part, and not when the booklet tells you it’s due for servicing.

Please continue the conversation with Chris on Friday, September 12 on IBM’s People for a Smarter Planet Facebook to learn more on predictive maintenance.  

Technorati Tags: Analytics, IBM, predictive analytics, predictive maintenance