A Quality Strategy with Vision

Patient safety is of paramount importance in the pharmaceutical industry and if this is in any way compromised, pharmaceutical companies face the spectre of product recall. Considerable costs can be associated with recall notification, product retrieval and liability, in addition to the potential damage to the company’s brand and reputation. Industrial vision can play a key role in modifying the production process to help prevent the possibility of product recall.

Product recall issues
At the time of writing, the MHRA (Medicines & Healthcare products Regulatory Agency) had issued 119 alerts and recalls for drugs and medical devices during 2015 and recalled 26 drugs, some of which were class 2 recalls (to be implemented within 48hrs). Some of these are company-led recalls, others are as a result of problems identified in the field. Products can be recalled for a variety of reasons, such as:

  • Batch recalled due to particulate matter found in samples.
  • Batch distributed in the UK recalled because it is labelled for the Portuguese market.
  • Incorrect carton barcode identifies the product as 250mg tablets instead of 50mg
  • Carton and ampoule label incorrectly state product can be administered by intra-articular means
  • 25mg has been incorrectly printed in one position of foil packs however is correct in other positions
  • The coloured band on a carton package is incorrect although text is correct
  • Packs from a specific batch may have an over count or under count of one tablet
  • 50mg capsules are at the risk of fungal contamination.
  • A smell associated with some batches is believed to be a result of the foil packaging
  • The paper side of the primary packaging could be punctured compromising sterility

The ability of industrial vision systems to monitor, regulate, check, analyse, sort and classify with high precision, repeatability and safety, means that they can help to address all of these issues. Vision systems largely fall into two application areas: inspection during the manufacturing and packaging processes and inspection of labels, codes etc on the final products.

Vision and the integration process
Typical projects include the integration of cameras into existing production lines, combining vision systems with ancillary equipment such as conveyors for product rejection, pick and place, robotics, or the provision of stand-alone inspections separate to the process if integration is simply not possible. This requires expertise in fields as diverse as mechanical design, mechanical handling and transport systems, software, electronics, robotics, control systems, factory networks and CAD. In addition, it is important to understand specific requirements and standards for the industry such as the security and auditing requirements for 21 CFR Part 11 validation. Given the potential complexity of these projects, many companies turn to specialist vision systems integrators. The process begins by understanding the customer’s unique requirements in order to develop proposals to meet the specific manufacturing needs in terms of performance, reliability and adaptability. Typical factors to be considered could include the linear speed of the system; the number of parts per minute for inspection, the product spacing and orientation and whether they arrive singly or in an array. These latter factors are important for the reject process. There are a number of discrete stages for an integration project which would include proof of process, system build, installation, commissioning, training and post-installation support.  The following examples illustrate projects that have been undertaken by Olmec-UK in different areas of the production and packaging process.

Product and package integrity inspection
Woundcare “island” dressings feature a cut dressing pad attached to a laminate backing material. In a packaging machine, the dressing is fed into a top and bottom paper packaging web. The edges are heat-sealed to form a sterilisable pouch. Defects must be identified at the full line speed of 330 parts/min so that non-conforming product can be rejected. An 8-camera, two stage vision solution was implemented. Four 2K resolution line scan cameras inspect the incoming raw material for defects such as contamination. A 1.4MP resolution area scan camera measures the dressing pad material before cutting. The pad is placed onto the web and two further webs are laid on top before this laminate is cut into a dressing. A code on the paper packaging material is checked by another area scan camera. The dressing is fed into the packaging, illuminated from below and imaged with a line scan camera to check total dressing size in addition to defects such as: splices in the laminate, pad placement relative to the dressing cut, pad size, pad skew and position of the folds. All 7 cameras are networked together and produce minute-by-minute time-stamped data, including reject images, run statistics and useful data such as the product position in the pack. Defects detected at any stage of the process result in rejection of that particular dressing pack. The vision system outputs a result to the parent machine to allow a 1 or 2 mm advancement or retardation of the packaging end of the process to minimise the risk of the dressing being trapped in seal. However, this is further checked using a standalone infra-red backlit line scan imaging system. The paper packaging transmits more infra-red light than the dressings inside, revealing the position of the dressing with respect to the seal. Trapped in seal products are automatically rejected.

Tablet inspection
If cross contamination occurs on a pharmaceutical tablet packing line it is possible for an incorrect, or rogue, tablet to get into the final packaging. An industrial vision solution allows accurate inspection of the tablets before packaging. A redesign of the chutes carrying the tablets from a vibratory bowl feeder permits inspection of either falling tablets or tablets stacked up in the chute before they enter the pouches. A 1 mm slit on each side of the chute allows the front and the rear of the tablets to be simultaneously imaged by 4k resolution tri-linear colour line scan cameras as they pass through. Each slit is illuminated by two high intensity spotlights which are strobed by a high speed controller that also handles the triggering of the cameras. The system has to ensure that the lights for each camera are only on when that camera is acquiring its images. A PLC handles integration of the system into the packaging machine’s control system. A specially developed algorithm reads a slice of data from the front and rear of the tablet to confirm the correct colour combination, since different tablet batches may be homogenously white or blue in colour, or even bi-colour. Imaging in this way is important for bi-colour tablets since they can enter the chute in random orientations. The system automatically stops the packaging line as soon as a rogue tablet is identified, ensuring that no rogues can enter the final pouches. The camera system was mounted on a specially designed ‘C’-shaped frame which can be cantilevered out for cleaning and set up of the packaging machine between batches. 

Reading labels on bottles
100% inspection of labels on the curved surfaces of healthcare product bottles is possible using a line scan-based vision system. This eliminates the possibility of a product reaching a customer incorrectly coded or without any code at all. By rotating the bottle in front of the camera, the label is effectively ‘unwrapped’ to produce an image equivalent to imaging before it was stuck to the bottle. The vision system checks that the information on the label is correct as well as identifying missing or partial print. The system handles a variety of bottles at a variety of speeds. The bottles are randomly fed into a servo-controlled starwheel which rotates each bottle in turn to the camera position. The ‘unwrap’ process is achieved by using rollers to then rotate the bottle within the pocket of the starwheel in front of a line scan camera which builds up the image of the label line by line. A distortion-free image is produced so there are no false rejects and no possibility of any rogue products getting through. Since the bottles do not need to be oriented before entering the system, speeds up to 120 a minute can be achieved. The entire inspection cycle takes just 500 milliseconds. The rotation of the bottle is synchronised with the scanning speed of the camera to eliminate deformations in the image. By precisely matching the mechanics of the starwheel with the image acquisition, the system offers a stability that minimises both false rejects and further software processing steps.

Carton inspection
A 5-camera inspection system on a cartoning machine with two separate leaflet feeders can ensure that cartons not only contain a patient information leaflet but that it is also the correct one. The vision system, linked into the existing reject mechanism, also provides the option to ensure that the correct bottle has also been inserted into the carton. The cartoning machine cycles up to 70 times per minute and is used for two different product lines. Individual area scan cameras read the carton pharma code and check that the leaflet is in the carton as well as reading the leaflet pharma codes on two separate leaflet feeders. For one product range the leaflet is inserted into the carton and the product bottle pushed down on top of it. For the other product, the leaflet is inserted into a dedicated pouch inside the carton, next to the bottle. On receipt of a trigger command, each of the 5 networked cameras sends data back to the line PLC which is responsible for all of the critical data flow, error handling and interfacing with the existing cartoning machine. The cameras send a combination of two leaflet and one pharma-code verification strings and Pass/Fail result for the leaflet in carton checks. The PLC compares the data being read at each inspection point stored data and confirms it is as expected. It also manages all of the results and updates the signal for the existing reject gate.