Manufacturing data. As The Simpsons’ Lyle Lanley once quipped to the assembled residents of Springfield, it’s “a little like the mule with a spinning wheel. No one knows how he got it and dang if he knows how to use it.” Except this mule has, like, a LOT of spinning wheels. Industrial Historians are now ubiquitous at most manufacturers, data storage is cheap and many folks grab any and all tags they can find and send them to the Historian in case they ‘need them.’
Now, I have no problem with this approach; in fact, I encourage it. Everyone has a Historian (or can easily gain access to a cloud one if you want to forgo the capital expense and a bit of IT headache) and, as stated earlier, data storage is cheap. And data is more valuable the more history and transitions you associate with it. You never know what will be valuable in future problem solving / variance identification applications. Heck, that was Gmail’s strategy long ago in giving folks free email accounts that could store and send large (in those days) amounts of data. Gmail then had the opportunity to access all of that email data that, while not immediately valuable , would provide greater value later on.
The problem is most manufacturers never get to the ‘greater value later on’ stage. They are recording tons of data every minute, but no one actually bothers to make sense of it. There might be a couple of trends or folks might refer back to it when there is a problem and they want to reconstruct the ‘scene of the crime,’ but often that data goes without analysis to better understand the source(s) of different variabilities or inefficiencies. We should use that data to find those unearthed opportunities for improvement.
At Avanceon, we have developed and are enhancing a number of machine learning and statistical tools to make better sense of the data our customers collect, and provide valuable and actionable insight into what it’s trying to tell us. How are you using your data? Are you the mule with a spinning wheel or do you really know how to use it?
Image Source: pch.vector