Since the Industrial Revolution began, companies have worked to improve their production processes. They have focused primarily on increasing production speed, reducing errors and controlling costs. In the 1950s, with the work of W. Edwards Deming, continuous improvement (CI) became accepted as a critical part of the production management process.
In subsequent years, consumer safety and traceability concerns have required the incorporation of inspection and coding equipment into packaging lines. In addition, production line equipment has become driven by digital control systems. The CI process has advanced from manually gathering and analyzing data and making system adjustments to using digital technology to automate the collection of that data and artificial intelligence (AI) to make enhancements to the application.
That evolution is ongoing: digital technology and AI are becoming an active part of the improvement process, playing a vital role in limiting human error, supporting predictive maintenance and improving output quality.
Limiting human error
The majority of packaging lines today still do not operate under the direction of a master control system. Discrete systems may be linked, but all pieces of equipment on the line are not always interconnected. As a result, when changes to a system need to be made, an operator’s intervention is often required.
As an example, take the sell-by dates printed on packaged food products. Expiration dates for foods vary and sell-by dates are more conservative than spoilage dates. Sell-by dates relate more to moving products off retail shelves than directly to food safety. In this simple operation, a printer’s software adds (for instance) eight days to today’s date and prints that as the date by which the product must be sold. When a holiday intervenes and retailers are closed, the date needs to be extended to allow the retailer adequate time to sell the product. Typically, an operator makes that change manually, creating an opportunity for human error.
On a line controlled from a single central source, the master control system can be programmed with both the normal sell-by date and scheduled holidays. The control system then can automatically make the date change required by a holiday. This will eliminate the operator intervention and the chance for error, as well as allowing the retailer time to sell the product.
Supporting predictive maintenance
Predictive maintenance to reduce or eliminate line shutdowns is a constant focus of line supervisors, who are always trying to improve the accuracy of their forecasts. A company schedules re-calibration of an in-line scale and the replacement of a printer’s ink supply at a specific frequency based on experiential and historical data. This input data is valuable because each production environment has unique operating conditions. Ambient temperature, air quality and other factors play a part in determining maintenance needs.
Master control systems continuously capture digital data from these individual systems and store it in a central location, usually via the cloud. Within that data is valuable knowledge about each system. Is the scale on a downward trend in accuracy, even though it is not time for its scheduled maintenance? Is the printer using higher-than-expected levels of solvent due to high temperatures or low humidity? Has the printer been left in "run" mode when it is not printing?
Using direct input from each system will then change the company’s maintenance program from one based on analogue information drawn from experience to one based on direct digital input. The master control system, using that data, can now alert the line operator to make the required changes immediately. In this way, AI overrides schedules based on experiential data and makes decisions based on current situational data.
Improving print quality verification
With the status of each production system being remotely monitored, what remains to be improved is output quality. In a printer, this boils down to: Is the information being printed complete and is it readable? If not, the packaged product represents a liability for the company and can cause rework and waste.
Machine vision systems are extremely reliable at inspecting detectable features of packages (cocked caps, missing labels, fill levels, etc.). However, only those systems — whether external systems or systems integrated into printers — specifically programmed to confirm or reject printed alphanumeric codes have proved reliable at confirming printer output.
The digital data captured by a camera and evaluated by the system software activates a pass/fail decision. If a package is identified as a rejection, the operator is alerted. On a line controlled by AI, several consecutive failures would initiate more aggressive action, to the extent of bypassing alerts and shutting down the line.
Advancing continuous improvement
The ultimate objective of CI is to maximize overall equipment effectiveness (OEE) by improving the availability, performance and output quality of production line equipment. The ability to collect, analyze and use system data has transferred gradually from discrete systems alerting operators to more integrated control systems and decision-making AI.
The companies that invest in using this wealth of data to improve system performance and OEE will gain a significant advantage over their competitors. Those that do not will be missing a great opportunity.
Kathryn Fox is a global marketing manager for Videojet Technologies Inc., a world leader in the product identification market, providing in-line printing, coding and marking products, and application-specific fluids. She works with key Videojet leaders to market the digital advancements that affect the Industrial Internet of Things.
