The secret will soon reach the masses of process engineers working in manufacturing plants across the country -process improvement is changing.
- There won’t be a better algorithm for a PID loop.
- There won’t be a new measurement tool to clearly identify product tolerances.
- And there won’t be an outside team that will come in and install new control equipment.
An extremely holistic approach to control is approaching and washing up on the shores of plants across the world – all thanks to the Industrial Internet of Things and Machine Learning. The idea is simple – measure absolutely everything, and compare that big data with results.
Let’s take an example production line that makes PVC pipe. There are dozens of these across the country churning out countless miles of pipe used for new homes, drain and sewer lines across the country.
Typical Control systems monitor and control items directly involved in the manufacturing process, including chemistry mixing ratios, extruder temperature zones and speeds, inner and outer diameter measuring tools, pipe cutters, pullers, and the like.
Tuning all of these values and continuously monitoring values and set points becomes as much an art as a science on the production line. Variations in these measurements and control outputs are all reacting to unmeasured and unknown environmental conditions.
Unknowns? That doesn’t sound right, but it is true..
Where do these unknowns come from?
- The amount of time the line has been running
- The shift that is running the manufacturing line
- The outside and inside temperatures, the rate of those changes
- The purity of the PVC pellets purchased
- The temperatures of…everything
- The age of the equipment and/or consumables.
The list can go on and on. A process engineer from 1998 may tell you that there would be no way to manage and sift through that level of data.
Enter Macro Process Improvement.
Macro Process Improvement – Gathering Big Data and Letting Loose Machine Learning to Identify New Opportunities
With the advent of low-cost Industrial Internet of Things devices and the connection of those with a myriad of wireless technology choices, gathering data is now incredibly simple.
Thanks to cloud computing, storing data is almost trivial.
Lastly, thanks to Machine Learning Tools from a variety of vendors, trends can be identified and determined through sifting through gigabytes of seemingly unrelated data.
This combination of massive data analysis sitting well above the manufacturing line process control, I bravely call Macro Process Improvement.
An Example of Macro Process Improvement Changing the World
Last month’s article about Hershey at Tech Republic dot com came across my screen.
The article discusses process improvements done at Hershey to improve the accuracy of their Twizzler lines, and enable them to tighten tolerances of their product output. This tighter tolerance allowed them to more accurately fulfill weight they deliver in individual bags. This improvement saved them up to 10% of product per bag yielding a $500k/year in material cost savings – while staying within the product promises written on consumer labels.
So how exactly did they do it?
By looking at where they put 14,000 pounds of liquid licorice.
Each liquid licorice holding tank in the Twizzler manufacturing lines holds 14,000 pounds of material and it takes a while for every ounce that goes in to get through the tank into production.
By taking temperature readings across 22 different locations on every holding tank every minute of the day, and analyzing the data sitting on Azure servers through ready-made algorithms, they were able to correlate specific holding tank conditions to optimal output results. Sure, this took 60 Million data points, and everyone is happier for it.
Continuous Process Control
Aside from labeling this type of analysis as macro process improvement, the steps can be taken to go further. Continuous Process Control at this level means connecting the data coming from the Internet of Things network of sensors and applying it to controlling machines on the manufacturing line. No longer a PID loop managing a couple pieces of data, the Industrial Internet of Things allows engineering teams to employ hundreds of data points to manage control. If you would indulge me, I call this meta process control. What do you think?
The Things in the Industrial Internet of Things
Hallsten Innovations develops the devices that will make up the Internet of Things at large. We design and manage fleets of devices just like the temperature monitors used by Hershey. Our devices are specifically designed for process improvement as well as asset tracking and preventative maintenance.
Hallsten Innovations has industrial clients with device deployments that measure, track and improve performance, often well outside the scope of standard control system packages – due to device accessibility, quantity, geography or power requirements.