Vials move into vision

Quality, accountability and product reconciliation are production mandates at the Indianapolis factory of Eli Lilly in the US, which is why this leading pharmaceutical manufacturer has automated its mechanical vial counting process.

The company enlisted Clarke Engineering Services, an Indianapolis-based systems integrator that specialises in vision system integration, commissioning, installation, and validation for the pharmaceutical and diagnostic industries.

Together, Eli Lilly and CES co-developed a tray inspection system using a colour visual inspection process powered by In-Sight machine vision sensors from Cognex. This quality inspection system, dubbed TIS-3000, is reportedly the first fully integrated tray inspection system for the pharmaceutical industry that combines machine vision with automatic tray handling, to enable 100% accountability of products.

While Eli Lilly’s mechanical counter provided only vial count, TIS-3000 uses machine vision to perform a visual inspection that detects fallen vials, identifies missing flip-seals, and verifies cap colour on each vial. As a result, the tray inspection system improves product quality, in addition to increasing vial-counting efficiency for complete product reconciliation.

By counting vials in trays at the front end of a packaging line, TIS-3000 helps the company eliminate count-related deviations and resultant rework as specific job lots are run. This saves money, time, and resources required for costly reworks. ‘The TIS-3000 establishes the initial count on our packaging line,’ explains Rob Stapleton, operations team leader at Eli Lilly. ‘If the numbers do not reconcile at the end of the line, we have to open up the entire job and find out why a discrepancy exists. This can be very costly and time consuming considering packaging order sizes range up to 360 000 vials.’

The Indianapolis plant also noted that the existing mechanical counters were prone to jams and that operators could easily mishandle vials around the counters resulting in a count discrepancy. This made identification of root causes for mechanical counter errors difficult to identify. A major benefit to the new tray inspection system is that root causes to count deviations are easier to identify, correct, and explain.

In selecting a vision system, two primary criteria were used, according to CES president Chris Clarke. ‘First, the system chosen would need powerful vision tools to meet Eli Lilly’s demand for accuracy and repeatability to comply with ever more stringent regulatory requirements,’ he explains. ‘Second, it had to be very simple to use, so operators with no vision experience could easily modify the inspection on-the-fly during product changeovers.’

Several criteria were needed for the system to be successful in meeting Eli Lilly’s requirements. The vision system would have to establish a tray count, look for vials that have fallen over, check that flip seals are in place, and inspect for vials with the wrong colour caps, called strangers.

After evaluating a number of systems, CES selected the Cognex In-Sight 1000C colour machine vision sensor. This combines a DSP-based vision-processing unit with a 640x480, 8-bit progressive scan digital vision camera in a housing with built-in communications, and a library of vision software tools.

‘We tested several systems to prove which could do the job, then selected the vision system based on the results,’ says Clarke. ‘Cognex tools were more robust for this application than the other vision tools we evaluated, and the powerful pre-processing filters helped us achieve the desired inspection results. The Cognex vision tools provided the most robust solution.’

The image pre-processing filter tools that other systems offered were not extensive enough to accomplish the inspection goals for the vial trays used at Eli Lilly. The total count of each tray was not repeatable and therefore not reliable.

Due to Cognex’s own spreadsheet interface the user can also use one filter to look through another filter, resulting in a further pre-processed image for blob counting. The software is user-friendly to navigate through and make changes to the program, or add a new product.

To set up the application, CES used the In-Sight vision spreadsheet interface. The process involved selecting vision tools and parameters from drop-down menus. The vision spreadsheet then automatically generated tool results into worksheet cells, which were then linked together to set up the inspection.

CES used In-Sight’s blob analysis tools to count the number of vials and determine cap shape, and colour histogram tools to verify cap colour in order to identify strangers. Additionally, the vision sensor was trained to distinguish specific colours using hue, saturation, and intensity.

The TIS-3000 features Allen-Bradley controls and also incorporates an industrial PC from Advantech. The TIS-3000 can run Lilly’s largest tray configuration at over three trays per minute, which equates to over one thousand vials per minute. The system speed capabilities are well within the packaging line speed requirements of 320 vials per minute.

During the inspection process, the operator first selects the program and inputs the data required for a given product run and then loads a tray onto the load station. The TIS 3000 uses three counting programs, each designed to handle a different type of count. The most common is a fixed tray count where the operator selects tray style, the specific count for that tray type, and cap colour. ‘Trays that don’t meet these test criteria, would be rejected,’ explains Clarke.

Operators can also choose a manual-input program scripted for counting new tray styles that aren’t yet programmed into the system. For example, if the manufacturer comes out with a new style tray designed to hold 100 vials, the operator programs the system to handle this count by selecting manual count, and entering 100. Once again, if the count isn’t met, the tray is rejected.

The third type of count handles random counts, such as when the operator must establish counts for various tray styles each containing a different number of vials. In this instance, the system inspects as usual, but accepts all counts, while recording individual tray counts as well as the total count of all of the trays in the production run. After selecting the count type, and entering the necessary data for a given product run, the operator places a tray onto the load station.

An indexing conveyor belt with cleats carries the tray full of capped vials into the inspection area. Inside the hood, linear high-frequency fluorescent lamps, complemented by a custom designed dome, provide cloudy day illumination to enhance the features of interest. A single In-Sight vision sensor, suspended in the hood above the tray, captures the image of the tray and vials.

If the tray contains the correct number of vials, it passes the inspection area, and is indexed to the unload station. If the tray fails, a servo-controlled reject arm transfers it to the reject conveyor, unless cap of the wrong colour is identified. ‘If a stranger is detected, the system locks up and requires supervisor intervention in order for the tray to proceed to the reject station,’ says Clarke. ‘This is an important feature in order to error-proof the process to prevent the wrong medication from being packaged into a lot.’

The first systems that have been installed are being used for accountability purposes on packaging lines to reconcile incoming lot counts with final order counts.