The first rule of efficiency is to accurately measure equipment utilization and quality rates
"You can’t fix what you don’t measure.” The oft-repeated quote, itself an abbreviated take on several business maxims credited to management guru Peter Drucker in the 1970s and ‘80s, is more applicable than ever in the era of Big Data, ubiquitous connectivity, and other smart technologies taking manufacturing by storm.
Such appraisals weren’t new then, or even especially profound. Shigeo Shingo, the renowned Japanese consultant who popularized the Toyota Production System, long ago opined: “The most dangerous kind of waste is the waste we do not recognize.”
These commonsense truisms have stood the test of time. But they often go unheeded. That’s a problem. For shop owners, many simply don’t know what they don’t know.
For example, machine utilization typically only averages about 28 percent of a manufacturer’s production capacity, according to Graham Immerman, vice president of marketing at MachineMetrics Inc., in Northampton, Mass. “Many manufacturers assume they’re running at 70-80 percent, when it’s closer to 20. They’re buying new machines, assuming that they’re capacity constrained, but it’s actually just a product of not having real-time insights into how often their machines are actually running or not,” said Immerman.
As depressing as that 28 percent capacity utilization figure appears, it doesn’t even account for scrap. The most useful and well-defined measure of manufacturing efficiency is overall equipment effectiveness (OEE), which considers quality a necessary component of productivity. Thus, an ideal OEE of 100 percent—for performance, availability, and quality—equates to machines running at their maximum design speed, without interruption, producing only good parts. So, if you’re using only 28 percent of the available machine time, any defective output would depress your OEE below 28 percent, even if you’re running your machines at 100 percent.
OEE was originally intended as a quick, back-of-a-napkin way to analyze a production operation, noted Jim Finnerty, ShopFloorConnect product manager for the Wintriss Controls Group, Acton, Mass. “Somebody would go out with a stopwatch and a clipboard and do a one-time test of machinery, (however) OEE was never meant to be something that was tracked constantly.”
But OEE data is exactly what’s needed if companies are going to get a handle on the complexities of the shop floor. At the same time, it’s not reasonable to expect machine operators and other production personnel to be constantly typing up cycle time studies, downtime reports, and QC data.
Companies such as Wintriss and MachineMetrics offer a better way.
The Wintriss approach places a small touchscreen control, called ShopFloorConnect Machine Interface (SMI), on a machine to track output. It fits any machine type and it’s uniformity reduces the need for training. “The box tracks uptime and parts count, if that’s applicable,” and communicates that information to a central PC running ShopFloorConnect software,” said Finnerty.
The SMI can also connect with a handheld scanner, so the operator can input the job they’re about to run without having to key anything in. And when it’s time to change jobs to match a user’s production schedule, he added, it’s as simple as pressing a button.
Capturing part count sometimes means you have to get creative. Finnerty recounted the unusual example of making the yellow tape that says, “POLICE LINE, DO NOT CROSS,” noting the customer wanted to measure the meters of tape it was producing. “We put an encoder on one of the rollers in the machine and determined how many pulses from the encoder equated to a meter. So, every 237 pulses out of this encoder became our count for a meter of production.”
Perhaps the most illuminating information captured by the SMI is determining why there is downtime. If the machine control can provide a code explaining a downtime condition, Finnerty said, Wintriss can create a connection that will automatically capture this information. In a typical installation, the operator records the information. It’s crucially important that the SMI makes this easy and fast, while ensuring that the data are useful. And the simple solution is to present the operator with a menu of choices that make sense for that machine, such as “broken tool,” or “unscheduled maintenance.”
Finnerty illustrated: “Let’s say you want to detect any stoppage that lasts longer than five minutes. If the SMI detects that the machine has stopped for longer than five minutes, we prevent further operation and the ‘downtime reason menu’ pops up on the screen. A message tells the operator to select the downtime reason in order to restart the machine. You might also set it up so that any downtime longer than 15 minutes alerts the supervisor.”
The captured data lets management see the cause of the downtime and whether there are serious problems. Conversely, if the operator has to type in a reason for the downtime, spelling errors could lead to inconsistent and unquantifiable results.
The MachineMetrics approach differs in two main respects. For starters, the company doesn’t necessarily install an HMI on each machine. Secondly, and more importantly according to Immerman, MachineMetrics created software that automatically references thousands of machine maps to identify each machine, then pull detailed data from its control. Whether it’s “turning, milling, stamping, fabrication, injection molding, cold heading, lasers, grinders, welders, robots, or cobots,” he said, “our universal translation engine takes all those different data streams and makes them all the same.” It doesn’t matter what system— MTConnect, OPC UA, Fanuc Focus, Ethernet IP, or Modbus TCP—the machines are using to communicate. “We’ve spent tens of millions of dollars and thousands of development hours making this possible,” Immerman noted. “It’s our superpower.”
MachineMetrics is able to capture various data, including spindle load and which programing is running on each machine. “We can automatically scrape that program header to know exactly what operation, part, material, or cycle time estimation is happening on that machine in real time, without any operator entry required,” Immerman said. “You can also transfer real-time production data and insights from these machines into an MES or ERP without requiring any manual data entry from your operators.”
Like Wintriss, MachineMetrics offers an operator interface for information that can’t be gleaned from the control. This typically is a tablet interface that works on a phone or computer, Immerman explained. The company also can provide tablets that snap onto the side of the machine.
What does all this data tell you? First, you’ll probably find out that you’re wasting a lot of valuable production time. What’s more, Finnerty said, “it’s not the big stuff that everybody remembers that’s causing the problems. It’s the little, ongoing, day-after-day problem that’s killing you.” He cited the example of a high-mix, low-volume production house that was very good at reducing setup time for all the job changes that occurred each day. “But once the operator was done setting up a job, the QC person would have to come over and sign off on it. Only when the QC person signed off on the job could the operator start the machine and run the parts.” Furthermore, each operator was responsible for running four machines, so they weren’t necessarily in the immediate area when QC signed off on a job.
After adding “Waiting for QC” and “Waiting for operator” as possible downtime reasons in ShopFloorConnect, the data indicated that operators were actually very efficient at their job, but the waiting was the bottleneck. In fact, slow handoffs cost the company 15 to 20 minutes per job. With 20 machines and an average of eight jobs per machine per day, this equated to losing 40 to 54 hours of machine time per day. To put it in OEE terms: This one factor was responsible for an 8.3 percent hit to the ideal machine availability.
ShopFloorConnect includes an alert function, so the QC person gets an immediate text when the operator has finished the setup and selects “Waiting for QC” as the downtime reason. The operator gets one when QC has signed off and selects “Waiting for operator.” And if the handoff still takes longer than 10 minutes, the supervisor gets a text. Incorporating this simple change has shortened the 15- to 20-minute handoff time down to five, according to Finnerty. “Multiply that by the number of changes they did in that shop every day, and they became quite a bit more efficient with no additional setup investment.”
Another way to reveal hidden capacity is to more accurately predict completion times, which is something both ShopFloorConnect and MachineMetrics purport to do. “It’s highly likely that manufacturers scheduling for different production runs are totally incorrect,” Immerman explained, “because they’re based on other systems and people’s estimations, and not real-time machine data. So, cycle time standard optimization is a classic use case that can reveal remarkable capacity in factories.” Immerman added that “the key to being efficient” is knowing “what you are really good at and what you’re not. As a result, many OEMs are starting to use MachineMetrics to decide whether or not to outsource some of their own production.”
In addition, Immerman said MachineMetrics is working on ways to predict when certain behaviors will happen. “If you pull data fast enough from a modern machine control, certain things start becoming visible that were never visible before. For example, micro fractures in tools may lead to chipped parts that might need to be reworked or scrapped.”
BC Machining in Brasstown, N.C., is a good example. BC runs automated Swiss turns 24/7. Before installing MachineMetrics, the company would often scrap as much as one-third of a shift’s parts from a given machine, owing to a broken endmill that went unnoticed. Then they’d lose several more hours as the operator sorted through parts picking out rejects. With MachineMetrics, a predictable pattern emerged in the data that indicated with 99 percent accuracy when a tool was likely to fail. Now the operator needed only to tend to the machine to replace the questionable tool before it caused problems. Shop Floor Manager Tim Merrill said the system works so well he tends to “ignore” the machines “unless there’s an issue,” adding that the scrap rate has also been “reduced considerably.”
Manufacturing Engineer Mike Driskell agreed, noting the “time savings at our Swiss turn machines has been monumental, to say the least.”
While predicting machine faults before they occur will undoubtedly improve both efficiency and part quality, manufacturers probably will never be able to completely forgo in-process or final quality control checks. And companies will never know how efficiently they’re producing parts unless they can confirm part quality. So, it’s critically important that the inspection process itself be efficient.
One way to greatly improve measurement efficiency is to remove human “fat fingers” from transferring the data, explained Tim Cucchi, product manager, precision hand tools at The L.S. Starrett Co., Athol, Mass. To that end, Starrett has introduced the DataSure 4.0 data collection system, which can fully integrate with precision measuring hand tools. Cucchi said they offer wireless hand tools with embedded radios, and also tools with an output port that connects to a backpack radio. Both types in turn connect to a PC or mobile device (Android or Apple). Either way, “you can capture every measurement, and with a push of a button send that to an Excel spreadsheet. Or if you have a specific SPC, MRP, or ERP system, we can tie into that.” MachineMetrics interfaces with such hand gauges, CMMs, and other measuring devices to roll quality data into their calculations.
Removing humans from data transfer is both faster and more accurate. Starrett recently commissioned a study in which the company measured 500 parts. “In the first trial, the operators hand wrote the results. We also recorded the time it took to make each measurement, turn around to write it down—everything they had to do,” said Bobby Williams, central regional sales manager. “In that setup, they made 62 errors over 500 parts, and each measurement required 37 motions over 29 seconds. When they went to full wireless data collection in the second trial, they reduced the motions to 17, the measurement time averaged 6.6 seconds, and the error entries were zero.”
Exclusive Ford Performance engine builder Roush Yates Engines (RYE) of Mooresville, N.C., presents another example of how data transfer can be simplified. RYE disassembles factory-built Ford V-8s, then machines and reassembles them to NASCAR specifications. The company was inspecting engine block cylinders using eight standard dial indicators in a fixture. “They were taking eight to ten seconds per measurement, and another eight to ten seconds to write it down,” Cucchi said, noting the inspectors had to be careful to attribute each value in the table to the correct indicator.
By switching to Starrett’s W2900-01 electronic digital indicators along with DataSure 4.0, Williams said, the measurements were transferred automatically and accurately into an organized database that saved from eight to ten seconds per measurement. With RYE conducting some 3,000 measurements per day, Williams calculated the company saved about seven and a half hours of measurement time every day.
Those are the kind of numbers that make an OEE-obsessed manager smile.
Connect With Us
