Defining Industrial Data

Industrial data comes from industrial equipment. In order to improve equipment performance and to keep industrial equipment running, we need to instrument the equipment with sensors that give us data about things like vibration, temperature, energy draw and flow rates. This data comes in the form of a continuous wave form, so it must be sampled at time intervals so that we can store the reading in a database. The frequency at which we sample a signal determines how fast we can respond to anomalies. For example, if you are sampling a temperature reading at the plant once a day, and ambient temperature has negative impact on production, then you need to wait until tomorrow to get the next reading to see if you need to adjust temperature. The sampling rate then is really a function of how fast we need to detect anomalies so we can react to them in positive ways.

Large pieces of equipment, like turbines or even production lines that make diapers have thousands, or sometimes, tens of thousands of sensor readings. To control high speed equipment, you need to sample the sensors at high fidelity. For example, if you sampled every millisecond, you’d generate 1,000 samples per second, per sensor. If you had 1,000 sensors and wanted to sample at this frequency for  eight hours a day, five days a week for one year, you would generate almost 7.5 billion samples per year, per equipment type. Of course, most industrial firms have hundreds of large pieces of equipment. As you can see, based on the sampling rate, you can generate a lot of data.

Once the data is sampled, industrial firms use the data in one of two ways. First, they look at it in real time to see if there are any anomalies they need to catch right away. For example, if you know that a piece of equipment is running too hot and could break down, you may want to slow down production to allow the equipment time to cool down, or you may want to investigate the cooling systems to see why the cooling system is allowing the equipment to overheat. The second way industrial firms use this data is for historical analysis to see if they can determine adverse trends or relationships that can help them look for additional failure modes that might be more complex, or esoteric.

Industrial data is then consumed by real-time applications like SCADA systems or it may be stored in industrial databases, called Historians, to facilitate long term analysis.

How are you using Industrial Data in your business?

defining industrial data

Brian Courtney

A recent transplant to the MidWest, Brian thinks Big Data “rocks.” He’s recently taken the Analytics piece of GE’s business under his wing, so if you have thoughts on any of these – MidWest, Big Data or Predictive Analytics – even rocks – follow Brian on Twitter @brianscourtney.

More Posts