Many manufacturers are planning some sort of AI or IIoT implementation on the factory floor. Many already have -- but according to Microsoft, almost 30% of IoT PoCs fail because of unclear ROI. What's going wrong in these cases?
Well, a lot, potentially. But one mistake is common to many of these misimplementations: They mistook data for knowledge.
Data is all around us. Every condition of every thing at every moment is a data point. We capture it with our eyes and ears and other senses. In manufacturing and Industrial IoT settings, data is captured with an astounding number of devices and sensors: We capture it with temperature probes, and with accelerometers that measure machinery vibration. We capture it with machine vision systems that image articles on the assembly line.
That’s all data. But until data is properly parsed, there’s no knowledge attached to it. Until we ask questions of our data, it has no answers at all; and until we ask the right questions, it very often has wrong answers -- or potentially even worse, it randomly has some right answers, but you don’t know which ones or how they came to be.
Which means, of course, that transforming data into knowledge requires far more than just collecting large amounts of data: You also have to ask the right questions in the right way.
If that just seems like common sense, that’s because it is. And it’s why people who are looking to bring Digital Transformation or Industry 4.0 implementations into the manufacturing environment should be looking at discrete initiatives with targeted goals: Because in focusing down it’s much easier to see where you want to end up, and thus much easier to understand what data to collect and what data to analyze. And it makes asking the right questions in the right way much easier, too.
Focus Helps Us Collect the Right Data
If I’m looking out of my window, there’s an endless number of things to see, and every one of those things has a huge number of data points of which I could make use. Trying to understand and use them all simultaneously would be impossible. But suppose that when I look out my window, I have a purpose: I want to know whether a squirrel is going to raid my bird feeder.
Having that focus then causes me to filter my data down so that I can continuously analyze the situation: Is there a squirrel in the yard? Where is the squirrel? Where is the feeder? Is the squirrel moving towards the feeder? Those are the right questions to ask to answer my question, and they allow me to constantly collect and assess the data that I need to know as I monitor whether a squirrel is going to raid the feeder; they help me to turn data into knowledge.
It’s a process that comes naturally to us, and we’re using that process over and over, all day, every day, as we go about our lives. We’re constantly presented with an infinite number of data points, and we’re constantly deciding what data to collect and how to use it. We’re asking the right questions in the right way.
It’s a simplistic analogy, to be sure, but one that too often gets lost when AI and IIoT projects are being considered as part of Digital Transformation or Industry 4.0 initiatives for the factory floor: Collecting vast amounts of data is put first, even before figuring out the problems that need to be solved and the questions that need to be asked in order to solve those problems.
That’s a basic disconnect between how we naturally problem-solve – i.e., identifying the problem, and then gaining the knowledge to solve it – and how many manufacturers go about their Smart Factory initiatives. And that basic disconnect is why so many IIoT initiatives fail to generate meaningful ROI: They generate terabytes of data without a clear, upfront picture of what the data is for.
Start with Identifying a Problem that AI / IIoT Can Solve
So as you go about planning your Digital Transformation, start with the critical steps of identifying a problem and the right questions to ask to solve that problem. Only then move on to planning how you’ll get the data you need to solve it. Your path to ROI will be more clear and happen more quickly, because you’ll be able to quickly turn data into the knowledge needed to succeed.
Because it’s not the size of your Big Data that matters – it’s how you use it.
If you're using machine vision for defect detection and it's identifying too many false defects, then it's not really doing the job you hired it to do -- and you probably already know it's a problem that needs to be solved. Our Spyglass Visual Inspection solution sits behind your machine vision system, eliminates false defects, removes the need for costly human re-inspection, and allows production lines to run faster. Check it out -- it's AI that turns your machine vision inspection data into immediately actionable knowledge.