Giving feedback on performance has been a management technique ever since there’s been management, and signs at the workplace are a favorite tool to furnish that feedback. But today’s version of industrial feedback goes way beyond the “X Days Without an Accident” sign.
As automation takes hold in food and beverage, it brings the potential to make available ever more detailed metrics in near-real time. Letting workers know how they’re doing, while they’re doing it, can be a powerful tool to motivate and correct. Setting up such a feedback system requires determining which metrics will be presented and how they will be compiled.
Kellogg uses digital billboards to provide real-time feedback at all of its manufacturing facilities, says Reuben Calow, VP for global Kellogg Work System and digital supply chain. “The system takes data directly from our equipment, systems and other sources, and displays it into actionable formats,” Calow says. “Technicians and managers use this information to determine our line performance and where abnormalities may exist. These insights help maximize the performance and efficiency of each plant.”
Presenting the right metrics is obviously critical, but determining what they are is sometimes not so obvious, says Jerry Bouma of Toma & Bouma Management Consultants.
“Every business has a few highly critical variables/metrics that are key to the success of that business – it may be throughput per hour, some measures of quality, food safety measures, etc.,” Bouma says.
“It is important for management to determine what the key ‘lead’ variables are and not focus on those that are the result of doing something else well. Often the focus is on cost per unit or margin per unit, but these are the results of other actions.”
As an example, Bouma mentions Little Potato Co., which processes whole specialty potatoes and potato products. One of the biggest concerns was keeping quality ratings consistent through washing and sorting. The percentage of potatoes that meet quality standards during initial sorting has to be as close as possible to the percentage that meet those standards after the potatoes are washed and sorted again; if there’s wide variation, it’s a sign that something is wrong, either with initial inspection or downstream processing. Another critical metric is meeting packaged weight limits, which is especially challenging for whole potatoes.
“Since you always have to pack the minimum weight standard (a two-pound bag must have at least two pounds), one always packs a slight overweight since potatoes do not come off an industrial press – the two-pound bag may be 2.05, 2.1 or 2.15 pounds,” Bouma says. “Keeping this as close to two pounds is critical – that’s what is being paid for.”
Bryan Sapot, CEO of SensrTrx, a supplier of industrial analytical software, agrees that proper filling is often a vital metric. Collecting data from checkweighers and comparing it with scrap rates is a good way to determine exactly what is going wrong.
Other common KPIs in food processing, Sapot says, include “units per employee, units per hour, units per hour per employee, throughput, a performance calculation measuring actual units vs. target. Downtime and availability – measuring how much downtime a line had during a shift, this includes number of stops, duration and they reasons why. From a quality perspective, most companies want to track yield and rework.”
Whatever metrics are chosen, Bouma says, “Clear objectives are critical. Staff need to know what they are aiming for, where they currently stand and what progress has been made. That being said, the tone of how these metrics are presented is important – they should not be presented in big black or red letters which appear punitive or threatening.”
Throughput is probably the most important metric, but there’s a danger in concentrating too much on that, says Alec Levenson, a senior research scientist at the University of Southern California’s Marshall School of Business. It can lead the workforce to give short shrift to other aspects, like quality.
“You have to be careful what you wish for,” Levenson says. “When you focus people on specific productivity numbers, they will focus on getting those numbers. But the problem is you can’t, in real time, measure everything that you need to meet your operational objectives.”
Path to the billboard
No matter what metrics a company decides to use, it has to figure out how to get them into the billboard. That depends heavily not just on the metrics chosen but on the state of the plant’s automation and digital control.
For a real-time billboard to work, raw data has to be extracted from multiple, and often heterogeneous, sources, and somehow piped into the app that will display it. There are different ways to do this, depending on what kinds of hardware and software are available.
Generally speaking, the more centralized a plant’s software, the easier it is to set up a real-time display of production stats. Many enterprise resource planning or manufacturing execution systems have dashboard modules that can display selected data in a variety of formats, including on a shop-floor billboard. These systems almost always are already being fed with the required data; setting up a floor display is a simple matter of installing a screen and routing the information from the right module of the central software.
As for getting at the data in the first place, software centralization helps there too, says Doug Cornwell, controls engineering manager at design-build firm Barnum Mechanical Inc.
“The best scenario would be if an end user had a centralized supervisory control and data acquisition (SCADA) system with a data historian,” Cornwell says. In such a case, data would pass from the programmable logic controllers (PLCs) to the historian, from which it can be extracted and shaped into whatever display is desired.
However, in many food & beverage plants, automation is piecemeal at best. Real-time displays in such cases must draw their data from a cobbled-together system.
When there is no SCADA or other centralized app, the desired data for the billboard display has to be drawn directly from a variety of devices: most often PLCs, but also other types of controllers, devices and sensors. The simplest way to do this is to dedicate a PLC to that purpose for each line.
This PLC will extract raw data that is often binary, such as “running/not running,” from devices like photoelectric sensors, says Steve Malyszko, president and CEO of Malisko Engineering.
“Each is a data point with its own unique tag name or data register assigned to it in each controller,” Malyszko says. “Many of the more robust applications ‘feeding the billboard’ have algorithms built in where the software can be configured to calculate uptimes, downtimes, number of occurrences, length of occurrences and OEE [overall equipment effectiveness] based on nothing more than periodically reading the tags or registers (status of each raw data point) in each controller.”
Once data is extracted, it must be collected for routing into whatever app will process it. One common way to do this is to route it through a gateway server. These can be set up to receive and send data wirelessly, because robustness isn’t a big issue; if the feedback billboard goes down, it won’t affect operations short-term.
Different devices will sometimes use different communication protocols, which can be collected by a server using Open Platform Communications (OPC), says Brent Robertson, a customer leader for Aptean, a vendor of industrial software.
“In the OPC layer, it allows us to communicate with all these different drivers,” Robertson said in a Food Processing webinar. “It allows us to communicate with those drivers using tags, and then we flow those tags into our factory database.”
A feedback digital sign can get its information from multiple points on the line.
A controller like the DXM wireless controller from Banner Engineering can collect the data using wireless radio technology and then can process the data with a simple program and push it to the cloud or to another server, says Mark Schmid, Banner's senior global business development manager for consumer packaged goods. Data can go to an in-house centralized app, or can be pushed across the internet into a cloud-based app, such as SensrTrx’s.
“The data flows from the machines to a gateway installed in the plant,” Sapot says. “The gateway pushes data to our cloud. Then we built an app for the Amazon Fire Stick, which can be plugged into any large screen TV. Our cloud application pushes the data to the TV in near-real time.”
Bouma says real-time feedback works best if employees know why it’s being furnished.
“Keep it fresh. And always remind people why this is being done,” he says. “These measures are not just an existential or bureaucratic exercise. A percentage point change quickly turns into a $1 million loss or gain. Everyone needs to know that every small improvement is critical. All of the time. Measures are not abstract. They matter!”