Can you get more value out of your legacy SCADA?
Is your SCADA (Supervisory Control & Data Acquisition) system too old to incorporate Machine Learning?
Put your hand up if your SCADA system was last upgraded before Big Blue beat humans at Jeopardy! (10 years ago). SCADA systems are often upgraded on a time scale of decades while Machine Learning tool sets are being upgraded monthly. Please reply in the comments as to reasons in your experience that it takes so long to update a SCADA system, it can be the subject of a future blog.
Ok, so Machine Learning is shiny and new, is it possible to use new Machine Learning (ML) with an old SCADA system? The short answer is yes, the long answer will take a little longer to explain.
The first requirement is that you need a mechanism/protocol to get the data out of the SCADA system. This might not be available in all SCADA systems, but it is fairly common to have installed something like this to get the data out to the corporate WAN. Managers and executives like reports generated from SCADA data, which is why this kind of link often exists. If you have a corporate historian the data might already be coming out of the SCADA system and into the corporate historian. There is a chance that it is exactly the data that ML can use. Of course you might have to increase the number of data points or increase the frequency of the data, but that's got to be (at least slightly) easier than replacing the SCADA system. If you don't have a corporate historian link already in place, maybe your SCADA system has a link for this purpose that you can use to extract data directly to the ML server.
Ok, at this point we have a mechanism to get near real-time data out of your SCADA system, and in most cases it can be done with little to no changes to the SCADA system. That's an important first step, because you don't want to put another straw onto the camel's back!
The second question is: are you willing to let your data go out to the cloud? There are some very secure methods of transmitting and storing your data in the cloud, but there are quite a few companies that really don't want their data on the cloud. For the most part this is an IT issue now, not a SCADA issue. This is the reason that Rogue7 built the Crusher: it's a server specially designed for ML in an industrial control room. The Crusher means that your data never leaves your computer room.
The third question is: what is the ML going to do for you? I'll leave that for future blogs because there are a lot of options.
The final question is: who wants to see the results? Operators? Managers? Execs? Which departments? SCADA, Maintenance, Scheduling, head office? This interface is pretty easy to implement using standard web browsers like Chrome that are available on all computers, tablets and phones. But you can also generate Excel files or PDF reports. If SCADA operators are on your list then you can put the web interface on a computer monitor right next to the SCADA screens (probably the one that he is using to read his email).
I admit that I left out a few details in all of the steps, but I think this answers the question that you can incorporate new ML into an old SCADA system. On a future blog I will talk about other sources of data for ML, because ML is very good at processing large volumes of data and condensing it down to a few useful facts.
Please leave comments below.
- Douglas F.