Remote Vision System Monitoring Unleashes Predictive Maintenance Capabilities
| By: Winn Hardin, Contributing Editor
The Industrial Internet of Things (IIoT) offers an alluring premise for automakers and other high-value manufacturers: the ability to detect machine failures before they occur. It’s a worthy goal, considering that one minute of automotive factory downtime can cost a manufacturer $20,000 on high-profit vehicles.
Achieving this objective not only requires collecting the right performance data from these machines but also properly analyzing it to spot — and act upon — inefficiencies or signs of degradation. While data collection is a cornerstone of machine vision, industrial end-users don’t always know what to do with this information, or they let it languish on servers.
Integrators who provide remote support on vision systems are starting to evaluate how remotely monitoring system performance to predict failure and help make sense of existing data could give their manufacturing customers an edge on the road to IIoT implementation. As with any new service, integrators are treading that road cautiously.
Parameters to Measure
Robust remote support services can build a potential pathway to remote monitoring because integrators already have access to the vision system from offsite. “Our first line of defense is remote support,” says Starke Farley, senior sales engineer for Integro Technologies (Salisbury, North Carolina).
For each of the systems it integrates, Integro Technologies installs TeamViewer software on the PC to enable remote control and access. “As long as the customer has Internet access, we can look right into the system and troubleshoot that way,” Farley says. “Customers are happy because they get a much faster response, and most of the time we can figure out what’s going on remotely rather than sending someone to the site.”
Integrators track a variety of variables to measure the health of the vision system, including camera gain, light output, and environmental conditions. “We normally build the light output monitor into the field of view of the inspection window and monitor the intensity from a histogram tool,” Farley says.
As for camera gain and environmental conditions, “we only monitor temperature in systems installed where there is no air conditioning, and that is typically done via a thermostat installed inside the electrical enclosure,” Farley continues. “We have had to put individual heat sinks directly onto 3D heads before since you can’t enclose those and not affect the image quality.”
To turn data collection into corrective action, system integrators must be unobtrusive and collaborative with the remote monitoring services they provide. That means working within the customer’s existing IT infrastructure to modify or develop databases to store the vision system’s information, says Nick Tebeau, manager of the Vision Solutions product group at LEONI Engineering Products & Services, Inc. (Lake Orion, Michigan).
When creating a new database structure, “we have the ability to control and best organize the data that is provided by the vision system,” Tebeau says. “We then work with our customers to decide at what point we push the information that is pertinent to them through their existing Supervisory or MES software.”
LEONI’s software can measure any quality point that is coming out of vision to both improve processes and identify potential component or system failures. Tebeau uses the example of inspecting a soft drink can, where vision is looking at surface finish and label presence and placement, as well as reading barcodes and batch codes.
“The data that is collected can tell customers that if they push their equipment to a higher run rate because they’re trying increase production volumes, scratches on the cans go up by 20%, or even that the majority of their quality issues are happening at 2 a.m. on Sunday and therefore they should provide some additional training for their staff,” says Tebeau.
Although the end-user is responsible for monitoring the data internally, LEONI can review parameters remotely to make changes. In the case of the soft drink can, the manufacturer sees dents on various parts of the can and may ask LEONI to remotely access the database to split that data off with the goal of determining how frequently the dents happen.
The data captured during inspection can also help predict maintenance needs not only of the vision system itself but of the surrounding equipment as well. For example, the vision system may detect that the graphic sleeves on the 20-ounce bottles are being stretched.
“We know that stretching of the graphic sleeve is usually caused by the main drive belt slipping,” Tebeau says. “We can measure the amount of stretch in the graphic sleeve, then inform the customer that they are still within specs but will want to schedule maintenance to look at it over the weekend before the stretching gets any worse.”
Preventing Failure Systematically
Machine vision’s predictive proficiencies rely on a comprehensive approach to data collection and management. ATS Automation’s (Cambridge, Ontario) ATS SmartVision system is integrated with its data collection and analytics solution called ATS Toolkit.
“Toolkit is not only capable of collecting and managing vision data results and images, it is fully tied into all of the control systems included in our automation and collects data on everything from sensor states and status to robot feedback and servo positions,” says Steve Wardell, director of imaging at ATS Automation. “All of these data points are timestamped and correlated to the machine running state and operator feedback to provide a rich suite of machine performance metrics.”
Monitoring the status of lighting components demonstrates just how valuable predictive maintenance capabilities can be, as compromised lighting sources often indicate the start of failure in a vision system. “We measure, record, and monitor lighting levels and contrast differences on every snap so that we can track whether there are either abrupt or gradual degradation of the lighting intensities in the images,” Wardell. “By catching the degradation before it leads to image analysis failures, we can proactively warn the users/operators that some maintenance is required before production is jeopardized.”
Wardell notes that the root causes of the failures can be unexpected, such as an operator leaving a piece of paper on the light itself. ATS’s system includes the results of these investigations, as well as recording and analyzing maintenance efforts to further provide current potential failure causes in the future.
Like integrators, OEMs are seeing the benefits of predicting maintenance on their machines. While industrial robots from FANUC America (Rochester Hills, Michigan) are known for their high reliability, the company sometimes has to dispatch workers and replacement parts immediately to the customer site when a component unexpectedly fails during production.
Wanting to take a proactive rather than reactive approach to maintenance issues, FANUC developed a program called ZDT, which stands for Zero Down Time. The robot sends data about machine and controller performance to the cloud, where analytics developed by FANUC identify potential equipment problems before unexpected downtime occurs. The ZDT platform sends information about the potential failure to the service department, which sends the parts to the customer for installation during a planned outage.
ZDT collects vision images from the cameras performing error checking or part location, as well as some meta data from the process on the robot that was used to capture the vision image. However, FANUC doesn’t currently analyze vision data to predict failures but hopes to do so in the future.
“We are just getting into what is possible with vision data, so it is still too early to say what we will be able to detect,” says Joe Gazzarato, director of engineering at FANUC America.
During ZDT’s 18-month pilot, FANUC deployed the platform on 7,000 robots in 38 automotive factories across six countries. It has detected and prevented 72 component failures. Estimated savings resulting from the program run in the hundreds of millions of dollars.
Barriers to Entry
Just as manufacturers are navigating the complexities of the IIoT, so, too, are system integrators. For starters, accessing a manufacturing customer’s information remotely isn’t necessarily guaranteed.
Through its software, i4 Solutions (St. Paul, Minnesota) currently monitors values such as camera gain, camera exposure time, and CPU temperature. All alarms, such as a warning that parts are barely passing, are logged. This information is displayed to authorized users — most often only to the customer’s personnel.
“Of course, we don’t have remote access to this data without an Internet connection to the system,” says Brian Durand, president of i4 Solutions. “In this world of evolving security threats, it can be hard to get permission to connect production equipment to the Internet.”
The decision to remotely monitor vision system performance parameters for predictive maintenance purposes depends on the environment and application. “Some types of vision system installations are more susceptible to environmental conditions — for example, dust, heat, moisture — and some get hosed down on a regular basis with no issues,” says Integro’s Farley.
Furthermore, “there has to be a justifiable need for remote monitoring of those conditions in order to justify the cost,” Farley adds.
LEONI’s Tebeau acknowledges that remote monitoring can be a difficult sell for some customers “because it is solving a problem that hasn't happened and it isn’t as obvious on an ROI. Companies are usually reactive with spending. They’ll say, ‘This is a major issue, so therefore here is some money to make sure that it doesn't happen again.’ When you talk about predictive maintenance, you are preventing the issues from happening in the first place.”