Machine Vision for Packaging Blends Science, Engineering and Art

Feb. 8, 2006
Vision systems are essential for maintaining product and packaging integrity and quality, but their integration requires a healthy mix of art and science.

The machine vision systems used to automate pharmaceutical packaging processes stand right at the intersection of science, engineering and art. Successful applications don’t depend exclusively on any of the three, but rather a bit of each.

Automated vision systems, once considered too expensive and impractical for drug packaging, are now mainstays for many drug manufacturing facilities, allowing users to monitor line performance, eliminate operator error, and ensure that the supply chain is free of counterfeits. “Machine vision systems are contributing to pharmaceutical product quality to the point where production lines do not run when the vision systems are down,” says Pedro Santiago, a packaging engineer at Eli Lilly del Caribe, Carolina. “That is how important they have become.”

Today, vision sensors, smart cameras and embedded vision processors sell for well under $10,000 yet provide many capabilities. “Most modern vision systems perform optical character recognition (OCR), optical character verification (OCV), and can read bar codes and 2-D Matrix, maximizing the short- and long-term capital investment,” says Iván Avilés, a packaging engineer from Pfizer, Caguas.

Manufacturers can justify installing machine vision at many locations along a packaging line, with each performing a single inspection to verify quality following each specific value-adding function. This article will outline the critical uses of machine vision systems for pharmaceutical packaging and discuss best practices from some of the leading packaging professionals in the Puerto Rican pharmaceutical industry.

Machine vision (MV) is not computer vision (CV), nor artificial intelligence (AI), digital image processing (DIP), or pattern recognition (PR). However, for most pharmaceutical processes, these technologies must be integrated to meet time, cost and quality demands, consistently. The U.S. Food and Drug Administration's 21 CFR Parts 210 and 211 require extensive monitoring and documentation of pharmaceutical production lines. Machine vision systems monitor product and packaging integrity, and ensure that packaging fulfills its fundamental requirements: to contain, protect, dispense and communicate (see below).

Primary Packaging: Solid Dosage and Liquids Primary Packaging: Solid Dosage Blister and Pouch Secondary and End-of-Line Packaging
Bottle, vial integrity PVC/aluminum film integrity Components integrity
Net content Product Cartoning, tray packing, case packing, palletizing operations
Closure integrity Foil printing  
Seal integrity Die-cutting/CR perforation  
Labeling integrity    

Verifying integrity

Using machine vision systems to ensure the integrity of product and packaging components is one of the most important applications today. A vision system can inspect a bottle or vial for the correct size, color, shape, dimensions and other critical properties such as neck finish or the presence or absence of particulate contamination. When, for example, any passing bottle diameter is out of specification, the vision system will detect it, reject the non-conforming unit, shut down the operation if consecutive faults are detected, display an "error" message and sound an alarm.

Tablet or capsule integrity can be guaranteed by a machine vision system that determines whether the product is in the correct position, and also detects whether the product is broken, partial or contaminated with a foreign particle.

Blister packaging’s complexities

Inspecting tablets in blister packs has become more complicated as branding efforts have prompted pharmaceutical companies to package products in colors close to that of the tablet itself.

The keys to successful color, shape and size inspection include:

  • homogeneous lighting;
  • high-resolution optical elements (lenses, filters, cameras);
  • high-speed image processors with corresponding software.

“All these elements are equally important, and none should be underestimated,” says Luis Reyes, packaging engineer of Merck Sharp & Dohme, Arecibo. “Each needs to be carefully addressed in Design Specification documents issued by the systems’ provider,” adds José Rivera, another packaging engineer, from GlaxoSmithKline, Cidra.

These documents vary. User Requirement Specifications (URS) tend to be fairly general: “The Labeler shall have a Vision System integrated to verify Lot Number, Expiration Date, Part Number and label presence.”

Detailed Design Specifications (DDS) are more specific: “The Labeler incorporates a high spatial resolution camera of 1024 x 1024 pixels using Optical Character Verification (OCV) methods to inspect and verify the Lot Number, Expiration Date as printed on the label by the steered beam laser printer with Futura 14 font, the pre-printed label Part Number, and label presence.”

Vision systems are also used to verify the correct assembly of packaging components, which requires full knowledge of the characteristics to be inspected. Monochromatic inspection can be useful when components’ colors are standardized. However, color inspection is more commonly used today because requirements vary in different international markets. For example, a pump assembly for the Japanese market requires a blue color clip, while markets in the rest of the world require a white color clip.

Verifying quality

While on-line verification can guarantee the physical integrity of products and minimize process failures, machine vision systems are also important for verifying quality attributes. Verification can be accomplished in various ways, from measuring net content to checking seal integrity and labeling.

The machine vision system helps verify that the package meets containment requirements. Short counts are one of the major challenges in solid dosage packaging today. They can occur in a number of different situations:

  • For low- or high-volume products;
  • With slat filling or electronic counters;
  • With low- or high-speed packaging lines.

Today, there is increased interest in studying the mechanical root cause of short counts, in order to minimize and eventually eliminate this problem. Even when a short-count problem is under control, other containing issues such as the absence or presence of foreign particulates need to be addressed.

At this point, there is no perfect solution to short counts, and inspection systems offer the best way to guarantee counts.

Vision systems can also be used to detect skewed, high or missing closures on bottle packaging. This application is becoming increasingly important on high-speed packaging lines to detect defects at different inspection windows within a limited processing time.

In blister packaging lines, seal integrity can be confirmed or challenged by vision systems based on pre-established sealing patterns, allowing vision systems to verify the package’s protection and dispensing functions.

Finally, vision systems assure that the package meets its communication requirements. The most common use of vision systems in this regard is inspecting variable information and component version numbers on container labeling. In this application, optical character recognition (OCR) and verification (OCV) are typically used to track product in process. There are some new variations on the theme, including bar code inspection and high-resolution label inspection, but inspection for presence, correctness and legibility of lot numbers, expiration dates and manufacturing dates is still the most critical labeling application.

Printing system quality is critical

However, the success of variable data character verification depends upon the quality and consistency of the printing system. Most OCR/OCV systems’ libraries are supported by printed characters from the system to be inspected, whether it uses hot stamp, thermal transfer, inkjet or laser (mask or steered-beam) printing technologies, or uses different surface substrates.

Machine vision systems monitor product positioning, defects and the presence or absence of foreign product. Photo courtesy of Micron Pharmaworks.

“We prefer the laser steered-beam printing systems, because of their consistency and adequacy to our high speed lines,” says Reyes of Merck Sharp & Dohme. “Hot stamp printing is suitable to most of our labels at a maximum rate of 200 units per minutes,” notes Santiago of Lilly.

However, even such well-tested applications as OCR and OCV are changing to meet new customer needs, and algorithms have been developed to determine, automatically, whether a problem is due to poor printing or an incorrectly printed character. Distinguishing between these two issues is very important. If a label has been poorly printed, the bottle should be rejected off the line without operator involvement. However, if an incorrectly printed character is detected, the entire line must stop immediately.

To minimize downtime when a character or label falls below acceptable thresholds, operators can search against the model library. If no viable match is found, the product become a “no-read,” not a “no-match,” and can be rejected without the concern of cross-contamination. Optimizing the algorithms to search through an entire model set quickly has challenged vendors but the latest technology is much easier for the user to apply.

Integration and validation

A user-friendly interface and complete integration with the machine’s control system are important for any machine vision system. Therefore, it is imperative that URS provided to machine OEMs specify expectations regarding vision system integration.

The human-machine interface (HMI) must provide a simple and logical navigation screen that clearly shows the system’s configuration, set up and operation. Recipes and inspection formats that are created must be object-oriented to provide a friendly environment for training tools and reporting. “A well-designed application software front end is a must to facilitate user’s training,” believes GSK’s Rivera.

From an operational and maintenance perspective, keeping things simple pays off. “We believe in simplicity by providing all the main system elements from a single source, rather than from different suppliers,” states Peter Buczynsky, president of Micron Pharmaworks (Tampa, Fla.) “It helps in operation, maintenance and troubleshooting.”

The qualification of vision systems is governed by cGMP 211.68 (b) and guided by the most recent version of GAMP established by the International Society of Pharmaceutical Engineering (ISPE). Most commissioning and qualification practices for modern systems are standardized by ISPE’s C&Q guidelines. The most relevant aspects of documentation are:

  • URS;
  • System-Level Impact Assessment;
  • Component-Level Impact Assessment;
  • Validation Plan;
  • Commissioning and Qualification Tests;
  • Business Continuity Plans;
  • Part 11 Gap Analysis;
  • Validation Reports.

“As long as the URS and DDS are properly documented, the vision systems’ validation cycle is quite simple,” says José Santiago, president of the packaging engineering and validation firm VALTEC (Aibonito). “There are no major complexities when requirements are made crystal clear.”

Training for the future

Machine vision systems are critical to product security, quality compliance and process automation. Education and training on machine vision technology is evolving, with cooperation among the industry, suppliers and academia. Working with industry and suppliers is essential, since that is where most of the machine vision expertise resides, says Cuauhtemoc Godoy, associate dean of the School of Engineering at Polytechnic University (San Juan). The school is expanding its post-graduate and certification programs to accommodate innovations in machine vision, with the goal being to increase the level of education and expertise, he says.

In Puerto Rico, engineers, OEMs and system providers test new vision systems and applications constantly, and share their knowledge with each other. This is the only way to ensure that the right balance of science, engineering and art is achieved for machine vision applications.

About the Author

César Hernández, P.E., is packaging engineer at Pfizer in Caguas, Puerto Rico. He has 20 years of experience in the confectionary, consumer healthcare and pharmaceutical packaging industries. He has been project manager for several packaging automation projects in the pharmaceutical industry, and has spoken and written extensively on packaging line integration. He has a B.S. in Industrial Engineering from the University of Puerto Rico, Mayaguez.

Machine Vision 101

Twenty years ago, automated vision systems were considered futuristic and, to a great extent, unaffordable for most pharmaceutical packaging applications. In the last decade, they gained acceptance, but many companies still considered them costly and unnecessary. Within the past few years, however, costs have fallen, technologies have improved, and the need for vision systems is widely recognized as manufacturers increase their focus on quality, performance monitoring and anticounterfeiting. Thus, it is rare to walk into a pharmaceutical facility without seeing such systems.

Machine vision is an electronic alternative to human or manual inspection that helps companies reduce defective products or packages. The systems are 100% accurate in most cases. What make machine vision cameras “smart sensors” is a built-in computer processor and a CCD (charge-coupled device) that captures digital images that can be inspected down to their smallest unit: a pixel. Using Windows-based Framework software, product features can be evaluated based on presence/absence, orientation, size, position and other specifications. This information can then be reported to direct robots, HMI/SCADA devices, printers or other machines along the production line.


About the Author

César Hernández | PE