Designing and Optimizing a Factory for IoT and Manufacturing

• Read original article here

Designing and Optimizing a Factory for IoT and Manufacturing
Modern factories need to be connected to compete.
SmartFactoryKL/Wikimedia Commons
Manufacturing in the 21st century is arguably easier than ever, however, a plethora of new technologies also mean that modern manufacturers need to be jacks of all trades in order to beat the competition.
The modern factory is increasingly being defined by how well it handles the massive volume of data it produces. The speed and complexity of data that manufacturing facilities produce are astronomical compared to years of past. The digitization of manufacturing has enabled many to gain competitive edges over older tools and methods. 
See Also
With everything moving towards digitization in the modern factory , this means that everything is also producing data. With this data comes analytical capability, enabling manufactures to make smarter, wiser decisions about their processes. This data collection is accomplished through the internet of things (IoT), a term we're likely all familiar with at this point. 
With all our manufacturing tools doubling as data producers now, managing a factory is no longer about just knowing the best tool for manufacturing, it's also about knowing how to leverage data to determine the best process. 
As this data flow increases, through IoT and other smart manufacturing tech, the question becomes, how do we as plant managers, operators, or just small manufacturers, manage our big data and extra value that impacts our bottom line?
Dealing with big data in a factory
Using big data in manufacturing can be overwhelming without the right understanding of how to manage it. The data can come in many forms, from hard, structured data like manufacturing files, all the way to completely unstructured data like error logs and machine logs. 
We can define the data produced in the modern factory in three ways: structured, unstructured, and real-time semi-structured. 
Structured data is data that fits into tables and is already formatted in such a way that insights can be drawn with ease. It's easy to manage and keep. Structured data, for example, can be our manufacturing data stored in databases.
Unstructured data is the type of big data we get from non-standard sources. These are things like shift logs for operators, or images of the plant or machines. All this data exists, but it needs to be decoded and organized before value can be extracted from it. 
Semi-structured data is data which doesn't conform to standard data models but does have headers, tags, and markers that differentiate the different parts of it into semi-interpretable documents. Examples of semi-structured data include sensors on machines, RFID information, motion controller data, and similar. 
Organizing, managing, and extracting useful information from the various types of data available from multiple sources, and in multiple states was once an impossible task. But today's IoT and data management platforms make it not only possible but relatively simple and scalable. 
An example of the cyber-physical systems pyramid in manufacturing. Explaining the idea of the levels of smart manufacturing of the future, Source: Behrad3D/Wikimedia Commons
Stepping back for a moment, I realize that just having all this data does not inherently mean that we understand its value, yet it certainly can have a lot of value. Machine data – feed rates, tool usage, RPMs, etc. – are all directly or indirectly correlated to yield and quality. If an operator on a CNC machine logs an error in machining every 1 in 10 parts, big data collection allows us to not only catch this with ease but by utilizing various problem-solving tools, like root cause analysis, we can also fix the core issue. 
Most Popular
RELATED: HOW WIFI6 IS ABOUT TO REVOLUTIONIZE THE INTERNET OF THINGS
If recognizing the value in big data is the first step, actually dealing with it is next. Managing all of this data in different formats, and enabling users (plant managers) to visualize how it can be used is the next step. At its core, this is a massive data structure and data science problem. 
I won't spend too much time delving into the solutions and tools used for analyzing and interpreting data in this post, as that warrants its own, much broader, research and discussion. I will note that there are a plethora of IoT tools our there, with Intel arguably being one of the smart factory leaders .
We've identified the problem, management of big data. We've recognized why it's important: insights and continuous improvement. And we've briefly mentioned that there are existing data tools to help manage all of this data. Next, we need to spend some time understanding the full complexities and insights we can garner from the use of big data in smart factories. 
Analytics are key to understanding the benefits of smart factories
I could spend all day talking about the high-level benefits of smart factories , but I'd argue that all of that would fly through the ears of just about every non-executive reading this. I think it's far more practical (and useful) to examine actual case studies and use cases around the implementation of big data analytics to improve our factories. 
I'll first reference a use case from Intel manufacturing. As with anything, though, keep in mind that it is in Intel's interest to sell you on IoT, especially considering their stake in the industry. Despite this, I feel the use-case stands. 
Intel wanted to determine a way to decrease the amount of false-negative indications that were being produced by a machine used to determine whether parts were good or bad. It was the job of this piece of equipment called the Automated Test Equipment (ATE) to perform tests on devices to evaluate their capabilities on a pass/fail basis.
A chart that depicts how developed each smart manufacturing trend is in the industry in 5 different phases, Source: ProfHvs/Wikimedia Commons
The problem was that the ATE would often wrongly categorize good units as faulty, impacting the factory's overall yield. Throwing out good units is obviously a problem you don't want to have, so Intel wanted to determine if there was a way to detect if the testing machine had a defect or fault that was causing it to label good units as faulty. 
After collecting data on the machines, they ran the data through an AI analytics tool that eventually predicted 90% of the potential failures in the testing machine, before they happened. This, in turn, reduced the yield losses from good parts being rejected by 25 percent, thus saving costs. 
RELATED: 11 FACTORY PROCESSES USED TO MAKE SOME OF YOUR FAVORITE PRODUCTS
You'll find more examples just like this from factories across the world that have implemented IoT and big data management tools. In many cases, the optimizations and improvements that IoT and big data management allow, can save factories enough money to cover the software expenses, and then some. 
IoT in the modern factory enables production visibility, operator improvements, reduced quality management costs, improved quality, and faster improvement cycles. All this is down to analytics and big data processing using AI. 
All this can sound scary to manufacturers who are unfamiliar with it. However, the tools available today make the process far simpler than ever before. 
The last thing I want to address is the immense bandwidth that the modern smart factory requires in order to operate effectively. In many cases, data collection needs to occur in real-time, something that can only be accomplished with fast networks. The roll-out of 5G will make this possible for a larger network of manufacturers.  
How 5G will improve smart factories
5G is essentially a new data infrastructure for wireless networks, one that can operate at exponentially faster speeds compared to 4G infrastructure. Data-intense technologies,  like artificial intelligence and the Internet of Things, will likely be brought into the public space through 5G . 
The Internet of Things is an area that seems particularly suited for 5G technology. Currently, IoT is being heavily used in the manufacturing sector to collect data in factories, as we've discussed in this article, as well as being used in the transportation sector to collect data on fleets. Smart home devices are also connected via the IoT, but lower-powered home wi-fi networks struggle when you have too many devices connected to them.