The Transformative Journey of Smart Connected Operations
Industrial Iot | 10 Minute Read
A Chat with Laura Proctor, PTC University’s Education Development Manager
In 1848, California became a hot spot of activity when gold was discovered near the Sierra Nevada Mountains. Fast forward to present day, and a modern-day gold rush has companies across the technology industry all angling for a piece of the pie. And this time, it’s not just in California – it’s on a global scale.
What is this hot commodity? It’s data. And there’s a lot of it. According to statista.com, an online market research and business intelligence portal, by 2020, the installed base of Internet of Things (IoT) devices is forecasted to grow to almost 31 billion worldwide. So, with all that available data, comes the question: how can companies extract value to improve a product’s lifecycle while also generating high-level predictions of behavior? The answer resides in the concept of Smart Connected Operations (SCO).
Laura Proctor, PTC University’s IoT Curriculum Group Lead, is here to talk to us about the potential impact. Laura has an impressive technical background with undergraduate degrees in Mathematics and Mechanical and Aerospace Engineering from the University of Missouri and graduate degrees in Mechanical Engineering and Computation and Design for Optimization from the Massachusetts Institute of Technology. She now leads an innovative content development team here at PTC University focusing on IoT and SCO solutions.
KS: Laura, when we talk about the massive amounts of data IoT offers to businesses, specifically manufacturers, what is the potential benefit for those early adopters of an SCO approach?
LP: If we want to get to the bottom line, the benefits are really focused on increasing a metric called Operational Equipment Effectiveness (OEE). The reason this one metric is so powerful, is that it’s a product of three productivity metrics: availability, performance, and quality.
Availability is the amount of time the line is operating – 100% means that no equipment is down during planned production time. Performance measures that the equipment is running at the rate expected – 100% indicates that the equipment is running as fast as possible. Quality indicates the absence of defects – 100% means that only good parts are being made.
So, even if just one of the metrics that make up OEE is improved, value will increase. Now, this said, improving OEE isn’t the only reason that one would adopt SCO and there is a lot of nuance in this one metric, but it is one of the top use cases. And specifically, when you talk about industry competition, those who prioritize establishing an increased OEE metric ahead of their competitors may very well establish a competitive edge in their markets.
KS: Once the enormous opportunity is realized, how do companies build an effective SCO framework? What are the main components to consider?
LP: First things first, the business problem must be outlined and the goals clearly defined. It is often the C-level suite that outlines that strategic direction, but everyone working towards the goal must commit to the journey and understand the importance of digital transformation or else it will not be achieved.
While visiting manufacturers, I often ask why they are doing whatever task at hand. I ask what problem they are trying to solve. In the past, I’ve heard people tell me that they need to get “all the data” because there’s an initiative at their company to collect and historize as much data as they can. When I ask what they will do with the data, they don’t know. Yes, big data can be useful to understand business problems. But, it’s important to have an objective. You must first understand which problem you want to solve.
Once that business problem has been identified and measurements for success have been defined, that’s when execution begins. The good news is that companies don’t need to build their SCO framework from scratch. In PTC University’s forthcoming course on SCO, we have a validated framework on building a proof-of-concept that includes instructions, guidelines, and templates for each step along the way.
But before starting execution, it is important to complete a “technology template” to better understand how the user experience maps to the available technology and to identify what resources will be needed. After that, you can really get into the nuts and bolts of solving the problem. This will help you determine which applications you will use to execute on your SCO pilot, or proof-of-concept, then get started.
KS: For the business problem at hand, what kind of analytics are necessary?
LP: Many companies aren’t even connected yet, so a good starting point for them is remote asset monitoring with descriptive analytics. When we talk about SCO, it’s so easy to get caught up in the latest buzz around creating artificial intelligence that uses machine learning algorithms on big data to promote adaptive analytics. But really at this stage, what matters most is improved operational performance, no matter how that’s achieved.
Right now, there’s a large focus on asset performance management. I’ve spoken to some customers who have seen an improvement in throughput just by showing a monitor with live statistics on their lines right there on the plant floor. Problems are being recognized sooner, fixed sooner, and resulting in fewer overall defects and higher overall throughput. Simply connecting the factory in a secure, intelligent way is already realizing benefits.
Let’s consider the four capabilities of smart, connected products which are: monitor, control, optimize, and automate (from How Smart, Connected Products are Transforming Competition by James E. Heppelmann and Michael E. Porter). The first one alone, monitor, is a great first step. Simply getting to the point where you can monitor the factory is an achievement. Then, the next step could be to layer in more control and optimization with the goal being to reach a point where it’s automated utilizing artificial intelligence. So, when you’re first getting started, it’s critical for those developing their first proof-of-concept projects, or pilots, to think about how they can execute it in an iterative fashion.
KS: I see. So for those companies who are just getting started on their transformative journey, where are the predicted difficulties? What are the trending growing pains that organizations have already experienced?
LP: The interesting thing about the manufacturing world is the amount of potential there is for growth. Do you realize that only about a third of companies polled by LNS Research in 2018 even use descriptive and diagnostic analytics? In plain English, this is the field of analytics which is largely focused on applying statistical formulas to the data to better understand what happened (descriptive) and correlate why (diagnostic).
Interestingly, about the same number of manufacturers are trying to apply predictive maintenance. But, when I speak to people in the field and hear that not much leverage has been made, it begs the question, are you trying to run before you learn to walk? Back this up with paper after paper about organizations having problems implementing their solution in a timely manner, or not seeing a return on the hundreds of thousands of dollars invested, it’s clear to see that failure is an option, and without proper planning, is very much a likelihood.
So, my advice is to dream big, but start small. Once you’ve gotten the connectivity and monitor pieces in place, then, layer onto that the exciting potential of data analytics. And, for those companies just beginning, laying out the proper architecture to do this is imperative. This is where PTC’s solutions are undeniable with built in prescriptive maintenance functionality.
KS: Once SCO has been successfully adopted, what will be the end results for technicians? How will augmented reality (AR) contribute or influence the interpretation of performance data? What will it mean for efficiency in servicing?
LP: There are so many implications for AR in the factory space as this space is really starting to gain traction. AR is finding a strong footing when used to assist humans in their tasks and decision making. The big things that AR will do, in my opinion, are: (1) reduce training time, (2) reduce error rates, and (3) reduce completion time of tasks.
I could imagine a scenario in which a high priority machine goes down. First, with IoT, the maintenance technician would receive a text or alert about this almost instantaneously. If it were a big field or factory, the technician receives directions either via an AR overlay or by a GPS map system. They go to the parts inventory and pick up what they need. Then, they make their way to the equipment that needs servicing. Or, perhaps a robot picks up the parts they need and delivers it directly to the asset that needs servicing. I mean, it’s the future, right? Why carry things when we have robots to do the heavy lifting.
The technician, through their wearable, sees the exact spot on the equipment that they need to focus on. Perhaps they have a robotic tool that configures itself automatically to disassemble the equipment to get to the point of issue. While working, they see clear instructions of each step until they get to the broken part that needs to be replaced. They fix it and receive all the instructions they need to reassemble the equipment. All with very little reliance on memory or expertise. Imagine, further, if they skip a step and hear a buzz or see a red screen alerting them to the issue. There is so much potential.
KS: One clear advantage is that managers will now be able to make analytically driven decisions. How do you see that frameshift improving operational performances and fine-tuning business models?
LP: Well, with the idea of big data paired with machine learning algorithms, there are many implications. I once heard someone posit that machine learning algorithms aren’t going to replace the blue-collar workers, but they will more likely replace the c-level executives. Personally, I don’t know that I would go that far. But, when we talk about predictive analytics in a factory sense, we are generally trying to predict a specific outcome. In other words, we’re using input data to predict a final value or event.
And to do this well, we need to leverage machine learning algorithms that result in associated outcomes. And that’s where the training data comes in. To ensure an accurate model, we typically need lots and lots of data. Hence, the gold rush. Now there are dozens and dozens of supervised machine learning algorithms out there – from naïve Bayes to Neural Networks all of which are comparatively nuanced.
But the takeaway here is that the implications for asset performance management are huge. You would then be able tostart performing some randomized feature modifications such as, nudging the algorithm input feature data and monitoring the results for optimal value. If done well, you’d wind up with important prescriptive advice on next steps. At that point, the business model will present with clear statistics about the risk management.
KS: It sounds like the more data leveraged, the better the outcome. But when we talk about mining historical data to make better predictions, how do companies ensure that all necessary security concerns have been addressed?
LP: It is very important to address security – especially with so many industrial protocols that don’t have security built in and with the number of cyber-attacks on the rise. People are cautious about any exposure of industrial assets online, and they should be!
The first thing to do is get IT on board and assemble a cross-functional security team making sure that everyone knows the current architecture, processes, and practices. Perform a security audit to make sure that you know what you have and what needs to be protected. Each company achieves security in a different way, but no matter what you do, implement the highest level of security, change your passwords often, and patch and update your software.
KS: Preventative maintenance is seen as a big upside to using an SCO approach, but many companies who have invested in IoT have not perfected the ideal product development loop. What examples of success have you seen so far?
LP: Preventative maintenance is a tremendous benefit to SCO. At LiveWorx this year, I saw a company that built logic based off of hundreds and hundreds of work orders so that when something seemingly minor happened, like a gasket was starting to wear, sensor data would pick up on it and automate a work order for their maintenance technician to go replace it. All done without predictive analytics, just from utilizing meticulous historical work orders and coding it up logically within ThingWorx. That’s the key. It was done online, using logic, and got the information to the right person faster than if the factory were not connected.
KS: It feels like we’re at the beginning of the largest transformative industrial journey the world has ever seen. If you could gaze into a crystal ball, what do sophisticated SCO environments look like to you in 10 years? How will machine learning factor in?
LP: If we’re getting to a point where we can display work process through AR wearables, I don’t think it’s going to take much more to actualize maintenance robots that can complete tasks with very little, if any, human intervention. Computer vision and self-directed autonomous robots will become more commonplace. And, while I think that robots and other AI-driven interventions will be doing more and more of the heavy lifting and performing the unsafe jobs, I don’t think that we’re quite going to see the anthropomorphic style robots that I fantasized about when I was a child.
Although I’m very excited to see how the landscape is changing, I’m even more excited to see the good that can come of this, the lives that can be saved, and the amazing technology that will spring forward from all the technological advancements.
KS: Thanks so much for your time, Laura.