Hopefully, your manufacturing operations are like shooting fish in a barrel. But without robust machine monitoring and the IIoT, you’re more likely going to be shooting in the dark.
It’s just a lowly chip conveyor. Its sole function is to carry metal shavings out of the CNC machine tool and into a waiting drum, hopefully for disposal at a nearby recycling center. It’s an essential piece of equipment, but hardly a critical component of a shop’s manufacturing strategy, right?
Now imagine investing tens or even hundreds of thousands of dollars into lights-out manufacturing, only to have the conveyor motor burn out long after everyone’s gone home for the night. The chips keep piling up and it isn’t long before they’re jamming up the works.
Best case: There will only be about 30 minutes of unexpected downtime in the morning while the operator breaks out the shovel and pliers, then gets digging. But if the pile is high enough, operators could have to contend with some broken tools.
Worst case? There could be a fire.
Similar calamities might arise from minor, often everyday occurrences—a dull end mill, a chipped insert or a damaged punch—any one of which can lead to hours of downtime and potential machine damage. Furthermore, such events can occur in less automated, fully tended machine shops and fabricating houses, particularly those where manufacturing processes are pushed to the limit.
Rob Caron can help machine tool owners and operators avoid these types of bad days. The owner and president of Caron Engineering Inc. in Wells, Maine, he and his team spend their days developing the technology to address both of these needs—machine monitoring and process optimization—by collecting real-time machine data, then either presenting it to the operator or taking actions based on that information.
“One of the cutting tool parameters we track is power over time,” Caron said. “We calculate a single number that changes as the tool dulls, providing insight into its life and health that helps people understand when it’s time to change the tool, for example, or adjust the feed and speed values.”
Analyzing such data can also help identify issues with the machine tool itself, he added. Operators can compare power over time and similar metrics across multiple part runs or even multiple machines cutting the same part.
“If one machine’s trend differs from the others, it’s a reason to investigate further,” Caron said “We also have the ability to analyze bearing health through vibration analysis, which is accomplished by installing sensors on the spindle to analyze vibration and temperature in the machine tool.”
Caron noted that most shops—at least those with a maintenance department—periodically check machine spindles with diagnostic tools. But this is often insufficient or sometimes overlooked, leading to unplanned downtime and, depending on the machine tool, a very costly repair.
“As automation and unattended operation become more prevalent, there will be even less opportunity for manual maintenance,” he said. “Our system can test spindle bearings daily and provide continuous, unattended bearing analysis. This information can be sent to the maintenance department automatically. If you have critical components, you can then plan for repairs or replacements instead of discovering a seized spindle unexpectedly.”
Sound maintenance routines are critical to the success of any manufacturing company. So are sound processes. This is especially true for lights-out shops, where predictable tool life, chip control and cutting fluid health can spell the difference between success and failure. As Caron pointed out, installing sensors for these and other machining variables—chip conveyors and coolant systems included—is a simple, relatively low-cost insurance policy; but there’s also the possibility of intelligent, automated process adjustments based on feedback from these advanced monitoring systems.
“We offer several products in our lineup, with the most well-known being TMAC, our tool monitoring adaptive control system,” Caron said. “But we also have products like ToolConnect, which uses RFID tags containing presetter information to load tools, and AutoComp, which gathers dimensional data during the machining process and automatically adjusts tool offsets or calls up alternate tools based on that information.”
One company well acquainted with these products is Austin, Texas-based Wolfram Manufacturing Inc., which President Nathan Byman describes as a “working machine shop and technology consulting firm.” After many years in the oil and gas industry, he opened the doors of what has become an established machine shop by purchasing an Okuma Multus B400 multitasking lathe in 2011.
Along the way, Byman learned a valuable lesson: Machine tools are powerful, precise and amazing, but they are also just dumb. “Given any opportunity, they will injure themselves, break themselves and make bad parts without telling you,” he lamented. “They’re inherently risky.”
This viewpoint, together with a technical mindset and the desire to “do things differently,” led the Texas shop to Caron Engineering and TMAC. “We started with the idea that we were going to teach machines to take care of themselves with data. Caron’s products were absolutely key to that, along with things like multifunction machine tools, which minimize part handling. We also focused on promoting uniformity through high-pressure coolant systems, in-process probing, and toolpath simulation. These were the core technologies that we built the business on.”
Using data to analyze processes for more effective decision making has also been a huge piece of the Wolfram team’s approach. To this end, the company collects, analyzes and archives massive amounts of information from across its shop floor. Some would claim it’s overkill.
“Every cutting tool we run knows how many parts and program executions it has done,” Byman explained. “We can look at an end mill we used two months ago and see that it ran 400 of this part, 25 of the next and 45 of the one after that. However, we’ve built the system so that no additional effort is needed to collect this information.
“Piles of data that you’re not doing anything with are wasted by definition,” he continued, “which is why we have a dashboard that we can run our shop by, with machines calling for help when needed. This allows our staff to focus on designing processes and thinking about new ways to do things.”
Doing so creates jobs that are better and more respectful of human beings, Byman noted, adding that they also pay better. “It’s a virtuous circle. When something does go wrong, all that data helps us immediately identify the cause without having to rely on traditional methods of asking people questions. We can examine the data, such as tool changes, feature analysis and horsepower curves to find the root cause of the problem.”
The Wolfram team has become so good at data management that it developed its own system, OnTakt, which the company recently commercialized. Wolfram also gained Rob Caron’s attention, and became a distributor and integrator of his products about six years ago.
When asked whether this is helping his competition, Byman seemed nonplussed. “We knew early on that what we do here would not be as easy as it might look, but for people who get technology and appreciate it—even those who might be quoting on the same parts—I’m happy to help.
“I hate to say it, but I sometimes feel that the generations before mine were too focused on spreading U.S. manufacturing to other parts of the world,” Byman confided. “Our goal now should be to bring it back home and help people get good at it again.”
Apparently, there’s plenty of interest in doing just that. Jeff York has been the marketing manager at Scytec Consulting Inc. for about two years, and he estimates that the Englewood, Colo.-based software development and consulting firm’s competition has doubled during that time.
“The benefits of data collection and analysis don’t just come from generating a bunch of graphs and dashboards,” York said. “The benefits come from utilizing those charts and dashboards and empowering operators by fostering communication between all facets of the company, whether it’s the entry-level worker packing boxes on the factory floor or the C-level manager in the corner office working to improve the bottom line.”
Big data is all about continuous improvement, he asserted. The ability to monitor productivity metrics such as machine cycles, feedrate overrides, part counts and defect levels in real time helps users find hidden success paths, and often brings immediate results. In fact, it’s not uncommon for shops to identify problems within minutes of connecting machines to the network, and enjoy utilization increases of 10% to 30% soon thereafter.
York also noted that making these connections and then gathering the data is easier than most people expect; what’s harder is the culture change needed to implement the resulting improvement opportunities. “A lot of shops are still operating with pen and paper, so getting them to adopt a digital, highly automated data collection strategy requires significant discussion and education of everyone involved,” he said.
It’s because of this potential trepidation that Scytec recommends starting small. Invest in a basic system, e.g., the bronze level of the company’s DataXchange for basic downtime tracking, then add to it as needed. This approach eases the pain of a large-scale implementation, while providing tangible benefits quickly and with minimal disruption.
“Whether you’re using Modbus TCP, OPC UA, or MTConnect, you can put the results on a mobile device, a large monitor out on the shop floor, or pull them into a third-party system using an API [application programming interface] or through direct integration with CGTech’s VERICUT, for example,” York said. “We also have the ability to monitor robots and even complete manufacturing cells. Whatever the approach, there’s an immediate benefit to just implementing this stuff and seeing what’s going on out there.”
Graham Immerman, chief commercial officer at MachineMetrics Inc., Northampton, Mass., shares similar views, but suggested that Job 1 in any such data collection project might not be what many expect. “Factories aiming to move towards advanced analytics and derive insights from their operations often face challenges in obtaining reliable real-time data,” he said. “There are several reasons for this, one of which is networking. Despite the advent of Wi-Fi and 5G and the excitement these technologies provide, networking a factory using Ethernet remains the best and most effective way to gain reliable access to those machines, without which improvement opportunities will remain limited.”
That said, Immerman and his competitors rarely care what network protocol or communication technology is used, provided it’s reliable and has sufficient bandwidth for the task at hand: bringing data from the shop floor into the company’s production monitoring system. Those who invest in such capabilities, he suggests, stand to see a very quick return on investment (ROI). “I had a conversation with a customer recently who has experienced multi-million-dollar-utilization improvements by leveraging our platform across four of its facilities.”
Immerman continued, “We’ve found that real-time visibility into operations typically creates ROI for our customers in a matter of weeks. For example, some larger organizations find that capacity utilization analysis alone can lead to tens of millions of dollars of bottom-line value from CapEx savings. It’s stunning how much opportunity and low-hanging fruit there is when the data is in the right context that helps describe what is actually going on out there. It’s like being blind, but finally being able to see for the first time.”
But what does “basic analytics” mean in this context? Perhaps more importantly, when does it become advanced, and what should shops expect from both? Answering these questions frequently comes down to the make and vintage of the machine tools being monitored. Advanced analytics often require more advanced data science, Immerman explained, pulling data from machine controls at higher sampling rates or outfitting legacy machines with sensors capable of revealing signatures invisible to more basic monitoring solutions.
For a typical mom-and-pop shop with a mix of machine tools—some new, some old, some commodity, some premium—it’s important to set expectations and have clear objectives early on as to what the captured data can be used to achieve. But even with older machinery, there’s likely to be a firehose of data coming in—far more information than humans will have the time or inclination to analyze manually.
“That’s why it’s crucial to also discuss A.I. and machine learning and the ability to automatically spot trends in this context,” said Immerman. “Technologies like MachineMetrics greatly simplify the data analysis process, enabling shops to benefit even if they have a mix of older and newer machines. Either way, the future of analytics is about helping customers understand what’s going to happen next and provide recommendations for what they can do about it. There’s tremendous value in both respects, but the bottom line is that missing out on the low-hanging fruit can preclude companies from experiencing the initial success and value that inspires them to continue investing in these types of technologies.”
Jordan Kathe, continuous improvement manager of Chicago-based Flexible Steel Lacing Co. (Flexco), can speak to this value. For nearly a century, the company has manufactured solutions designed to optimize the efficiency, productivity and safety of belt-conveyor systems for a wide variety of global industries such as mining, food processing and parcel handling. One of Kathe’s responsibilities is to assist with the collection and analysis of production data from the shop floor, a role he’s had for the past five of his seven years with the company.
“We realized early on that there wasn’t an effective way to understand our machine capacity beyond napkin math (scratching notes down as your go) and rough estimations,” he said. “Given the diverse range of machined, stamped and cold-headed products we manufacture here, it was difficult to schedule and observe these processes as they happened. This led us to consider how we could better utilize the information within the machine control.”
Kathe and his colleagues were well aware that machine tools “know” whether they’re on or off, and the number of cycles they’ve completed. However, they faced challenges in extracting this data in-house due to their limited expertise in this area, coupled with the difficulty in finding a scalable, comprehensive solution.
“We sought help from external vendors, but most specialized in predictive analytics and couldn’t help us with data collection,” Kathe said. “That’s when we turned to Graham and MachineMetrics, who promised a one-size-fits-all solution compatible with our various equipment types.”
That was in 2019. Since then, Immerman’s promises have borne fruit. Flexco installed the system on 20 of its high-volume machines, a process Kathe described as “very straightforward.” The team has focused on getting comfortable with the technology, gathering baseline data, recording downtime codes and studying long-term trends. The company also reevaluated its lean-manufacturing strategy and placed MachineMetrics at the center of production management activities, with regular discussions about downtime codes serving as the lead-in for conversations about potential improvement opportunities.
A recent Kaizen event highlighted the value of this approach. “We have automated assembly equipment that places rivets into collated plastic strips,” Kathe said. “But this process experienced frequent jams, causing nuisance stops where the machine had to be shut down and then turned back on after resolving the issue. The recurring nature of this problem prompted us to gather a team to investigate, and soon discovered that the machine’s air settings were improperly calibrated. After making some adjustments to the process, we saw a 42% increase in weekly throughput for that equipment. This improvement all started with the ability to visualize the stack up of nuisance stops, emphasizing the importance of the data we collect.”
While the MachineMetrics application has additional capabilities beyond the collection of machine status data and downtime reason codes—and Kathe plans to explore these features soon—the power of these simple data points has proven itself to be invaluable.
“The ability to see a timeline of when the machine was on and off throughout the day is critical to understanding and optimizing our processes,” Kathe said.
“And surprisingly, it has also reduced friction between management’s wondering about the causes of machine downtime and the operator having to explain what happened three days ago. Everything is right there in the system.”
Connect With Us
