Analytics for Industrial Operations

Recently, I attended Cisco’s Data and Analytics Conference, where Jim Green, their CTO for Analytics, spoke about trends in Data Science. With a focus on the workflow of data science professionals, Cisco has spent a lot of time understanding the process. Mr. Green’s findings show that Data Scientists typically spend about 60% of their time just finding data. Before a theory can be tested, the necessary data to inform the test has to be located and extracted. Even more surprising, most of the rest of the workflow involved prepping that data – cleansing, re-factoring and normalizing it – to enable analysis.

Consider that in the context of Big Data ("datasets so large or complex that traditional data processing applications are inadequate"): industrial automation devices are the birth place of huge volumes of data useful for analysis, but even for those manufacturers who employ actual data scientists, the job of uncovering insights from a sea of tag values over time is a daunting one. For the average plant operations personnel, most Analytics likely lands outside the reach of their in-house capabilities.

In response, new service and software offerings have sprung up, attempting to provide Analytic solutions to industrial operations customers, promising that their offerings can work against unstructured data, integrate easily with their control network, or somehow create value just by storing the data in the Cloud. Solutions do exist, and some of the tools out there are genuinely useful, even for people without a data science background. But the projects to implement these solutions are expensive, slow, and often so narrow in scope that they aren't directly re-usable for the next problem that requires Analytics. Manufacturers know that there must be value in their data, and are willing to spend time and money to find it – but they're chasing solutions that haven't been optimized for their problems.

Is there a way to bring some sanity to an industrial information marketplace full of noise, but relatively devoid of solutions? I'd like to suggest that we need to begin to look at industrial automation devices as participants in the true Internet of Things – not an alternate, or less-capable “Industrial IoT.” Most the expectations that consumers have begun to apply to IoT devices in their home should be applicable to industry as well: things like ease-of-use, instant access to data, and modern interaction models are due for an appearance in the manufacturing world. If you'll indulge me in a few posts, I’d like to explore some of these expectations, and the trends related to them, and look at how these things can begin to apply to automation devices – and to Rockwell Automation devices in particular.

In different words, there are different teams at Rockwell Automation working on these problems – with real results. At Automation Fair next week, we'll show how we're beginning to unify those efforts into a cohesive solution that provides instant Analytic value for our customers, and positions us to take leadership in this next iteration of the Internet, where the Connected Enterprise is informed by meaningful data that is easier to find and use for our customers. Hope to see you there -- if you're not sure what the future looks like, you can always #AskShelby