Making the Most of Vision
Machine vision systems can now discover error trends and adjust machinery and processes accordingly. Users, however, are slow to take advantage of the new benefits.
by Erik Swain, Senior Editor
Machine vision technology has entered a new era in the healthcare industry. For the past 10 years or so, pharmaceutical and medical device manufacturers have added vision systems to their packaging lines. For the most part, the additions have been stand-alone systems that are equipped to verify that label characters, date and lot codes, and other information are correct on every package. When they catch something incorrect, they reject the sample.
New systems being developed, however, go far beyond machine vision in its simplest form. They can check for more than just correct characters, and they can be networked to make other machines on the line more efficient. They can even play a crucial role in activities from statistical process control to corporate planning. Finally, their costs are coming down.
DVT's SmartImage sensors perform on-line QA inspection.
"We are trying to redefine the idea of a vision system being an 'island' on the floor," says Michael Valverde, director of communications for DVT Corp. (Norcross, GA). "It was a pass-fail device that did one thing regardless of the sophistication of the application. Now it can be a key communication tool in a network environment."
However, only a few healthcare product companies have started to embrace such benefits. Suppliers and consultants predict that the combination of expanded capabilities and reasonable prices should bring such advanced systems into wider use in the industry. But no one knows for sure.
SALES UP BUT PRICES DOWN
In 1998, according to a study by the Automated Imaging Association (AIA; Ann Arbor, MI), the numbers revealed a unique situation. The unit sales of application-specific machine vision systems to the pharmaceutical industry rose to 34 from 25, but dollar sales declined from $17.1 million to $11.9 million. A similar trend occurred for general-purpose machine vision systems sold to the industry, says Nello Zuech, president of consulting firm Vision Systems International (Yardley, PA), which conducted the North American market research for the study.
The decline in dollar sales is not a reflection of flagging demand, however. Indeed, this year's buyer intention report from the Packaging Machinery Manufacturers Institute (Arlington, VA) ranked inspection and detection systems the second most likely kind of packaging machine to be purchased in the pharmaceutical industry this year.
Users can program GSMA-Parish Automation's Q-Spec OCV vision system to verify three character strings.
Rather, Zuech says, the numbers reflect the changing technology. "More systems have become PC-based and are dependent on software rather than hardware," he says. "That has reduced the amount of truly proprietary systems, which tend to be more expensive, and it has reduced the actual engineering costs." In addition, he says, the hardware, especially cameras and lighting, has become a commodity rather than a specialty and as such has dropped dramatically in price.
Mac Hashemian, president of supplier Xyntek Inc. (Yardley, PA), foresees vision systems as a crucial component of a comprehensive system not just to inspect the packaging lines, but to manage a firm's entire enterprise. "Algorithms are being written to allow the machines to make smarter decisions, based on inputs from the sensors," he says. "You can tie the vision system into the central control system running the line, then tie it into the management information or management executive system."
The newest vision systems, he says, are capable of plotting error trends and of feeding such trend information into the process control system, which could then react by making the line faster or slower, for example. Through microprocessors, they can then feed that data—from the most common errors to the line efficiency at any given shift to the operators working when errors occurred—to a common corporate database. There, the data can be analyzed by anyone from the quality assurance department to the head of the company. This "closed-loop control," as it is called in the trade, is what industry has been seeking for years.
DVT has a new system that it hopes will accomplish such control through the Ethernet, Valverde says. "In the past, statistical process control was possible, but not practical, during everyday inspection," he says. "Now it is because of the rapid decision-making capabilities and the ability to collect information in real time across the whole network."
In that kind of a system, the infrastructure is in place to enhance performance significantly because the data can be used for everything from monitoring tool wear to reducing scrap to cutting back downtime, he says. And the information can be accessed over the Internet, meaning a solution discovered at one of a company's plants can be implemented at others without any personnel travel.
The recent past has seen other innovations that represent significant advances in a machine vision system's capabilities.
For example, PPT Vision Inc. (Eden Prairie, MN) has developed an all-digital machine vision system and network. It features a digital camera that acquires images at 70 frames per second instead of the standard 30 and is set up through a software network that improves timing and prevents accidental altering. Special image-processing hardware is designed to meet the fast speeds demanded by in-line machine vision.
GSMA-Parish Automation's vision system uses a Fanuc robot to present products for date and lot code verification and bar code scanning.
"Applications are enhanced because the all-digital system doesn't suffer from the video noise that plagues analog systems," says Chuck Bourn, director of marketing communications. "It is fast, it offers greater accuracy, and it allows you to easily add up to 16 cameras where you need them, even at great distances."
In another digital advancement, Global Vision Inc. (Montreal) recently introduced a digital proofreading system for electronic package artwork. "A lot of pharmaceutical companies want to eliminate paper," says Renee Symonds, marketing and sales manager at Global Vision. "They also want to be able to approve artwork a lot faster. The digital system can connect to a network, allowing users to communicate with their printers. And it is compatible with document management software."
There are also systems sophisticated enough to discern critical errors from noncritical ones. This kind of vision is often applied to a label before it hits the packaging line, whether on a printing line or off-line in a finishing department, says Ignatius Manning, general manager, VRP Web Technology Inc. (Mississauga, ON, Canada), which specializes in those kinds of applications. "You can multizone a label to allow the QA department to design an inspection task for each part of the job," says Manning. Applying a different algorithm to each part of the label, for example, would catch a missing decimal point in the dosage instructions, which is critical, but pass a missing dot of an i in the list of inactive ingredients, which is not critical. That, Manning says, improves the speed and efficiency of the line.
TO BUY OR NOT TO BUY
The question remains, however, whether the healthcare industry will embrace these new technologies or whether it will focus on the immediate bottom line and invest only in the most economical and basic systems. The response could well vary from one company to another.
"My experience is that most individuals in the pharmaceutical arena have not viewed vision as a process control kind of application," says Jeff Snyder, packaging applications engineering manager at RVSI Acuity CiMatrix (Canton, MA). "Some companies tend to use the information for some process control, with a human being in the loop. Pharmaceutical packaging operations are still rather people-intensive and are not highly automated as they are in other industries."
Gary Parish, president of Parish Automation Group, a division of GSMA-Parish (Palm Bay, FL), agrees. "The feedback loop is certainly there, but most companies are not using it," he says. "If a label on the bottle is crooked, there is the capability to send word back to the labeler to adjust its positioning. But most don't do it yet."
As much as suppliers push the new technology, healthcare product companies won't commit to it until they can be sure they will get a high return on their investment. That involves vendor information as well as internal information, so suppliers can only argue the point to a certain extent. "Like any new technology, the cost of starting up is high," says Gene Fuchs, vice president of Genesis Machinery Products (Exton, PA). "But when you amortize it down the road and consider how long the machine lasts, and the labor savings, and the increased accuracy, you should come out miles ahead."
One issue sure to drive demand will be validation requirements. The easier the new systems are to validate, the more likely they will have a place in the healthcare packaging market.
Validation expert Joseph Busfield, principal of Pharmaceutical Technical Services (Warrington, PA), says that the new vision systems can be validated much like computers. "If you follow GAMPs (good automated manufacturing practices) and guidelines on computer validation in general, that's a good approach," he says. "That's what a vision system is—a computer with quirky input. Most of the computer validation stuff will be the same. The twist is verifying the input. You have to test that it does respond to an alien label. And that if the system reads wrong, the package fails on the reject side. A misread that passes is no good."
The technology is used widely enough that "having machine vision on a new line is a foregone conclusion today," as RVSI's Snyder puts it. Now the question becomes whether healthcare companies are ready to take the next step and adopt systems that offer far more capabilities than they used to.
And the next step is becoming more affordable. "With the migration to the Windows format and microprocessing, you are able to do more for the same price, and you are getting more capabilities free," Zuech says.