In today’s business climate, food and beverage (F&B) manufacturers face a number of challenges. In many Asian countries, regulatory oversight is becoming increasingly stringent, especially in the light of food scandals that have perpetuated over recent years. As producers move beyond their borders to export their products to international markets, legislative standards in those countries also have to be adhered to.
With the advancement of internet technologies and the proliferation of social media, consumers too, have a greater awareness of food safety issues. Mistakes made by companies that affect consumer health have made the headlines and spread like wildfire, quickly bringing down reputations that took years to build.
Substandard InspectionOn the production side, manufacturers are often perplexed by the difficulties in ensuring process efficiency while maintaining product quality. Quality control (QC) usually becomes an issue when manual processes are involved—such as employing operators to visually check for product defects along the production line.
A manufacturer of powdered milk deployed six operators on each work shift to check for defects on milk tins. The operators had to ensure that the manufacturing dates, lot codes and expiry dates were present on each tin that passed by.
The problem with this method is that it was impossible to inspect every single tin that came through. Moreover, such a manual process was largely dependent on the ability, experience and mood of each operator on duty. This means that the effectiveness of the QC process was directly linked to the operator’s skill and level of alertness for that particular shift. The issue was further exacerbated when the facility ramped up production to meet increased market demand.
The manual system also means that while operators could verify the presence of the printed information on a product, it was extremely difficult for the human eye to ensure that the dates and lot codes are correct. On top of that, it is impossible for the operators to read any printed barcodes.
After the installation of a machine vision system, the facility’s QC processes saw dramatic improvement. The cameras are able to read every can that passes by, ensuring a 100 percent thorough inspection. Since the cans move along the line and can present themselves at random orientations or even upside-down, the cameras have to be able to check and verify the printed information under such circumstances.
Error-Free QCInspections are no longer dependent on subjective decision making by human operators. Another benefit is that only one operator is now required on each shift (instead of six), allowing the facility to streamline its workforce and re-deploy human resources into areas that require greater human intuition and strategic thinking.
Within this production environment, machine vision can also be applied to check the contents of each tin. Since each tin must contain a scoop, the system ensures the presence of one in each tin and checks that it is of the right colour and size.
When comes to bottling, be it ketchup, beer or soft drinks, a common challenge is in ensuring that the caps have been properly placed or screwed onto their corresponding bottles. For plants that rely on human vision to perform this task, the same production problem rears its head—how to ensure that a 100 percent thorough inspection has been made on 100 percent of the products.
Machine vision however, can fulfil this task. With the help of software, vision cameras ensure that bottle caps are correctly positioned. Any bottles with crooked or missing caps are immediately identified by the system. In certain facilities that manufacture a range of products (for example, ketchup and chilli sauce) or that make products for different brand names, mix-ups on the production line can occur.
For example, a cap that reads ‘Ketchup’ may be placed on a chilli sauce bottle, or beer brand A’s cap may be used to seal brand B’s bottle. Such mix-ups can be a source of potential embarrassment for the manufacturer and could result in a loss of reputation for the brands involved.
Machine vision however, is able to differentiate between the different colours of caps and the logos or brand names that are printed on them. If the system detects any violations in the capping process, it then alerts the operator via the software interface. To ensure uninterrupted production, the system can be set up to divert the unacceptable bottles into a separate rejection bin. The operator then proceeds to collect these bottles after the production shift.
Trail Of AccountabilityIn a manual inspection system, it can be quite difficult for the operators on duty to record down any errors that are discovered during that shift—especially when they are busy weeding out errors on the production line at the same time. With an automated vision system, images of each bottle are stored and the system is able to generate reports on the number of errors that are detected during each shift. This provides managers with the necessary information to fine-tune and improve their production processes.
An automated system ensures accountability and creates a reliable audit trail that facilitates investigations. These mechanisms are especially useful for management to identify problems on the production line—for example, if a particular batch has a higher-than-usual number of rejections—and trace them to the source. In the unfortunate event of a product recall, these images may also help investigators to track down the cause of the problems.
Needless to say, the technology can aid a contract manufacturer in convincing potential brand name clients that its production QC and traceability processes are sufficiently robust and reliable for the job. Furthermore, it usually takes just about six to 12 months for a facility to recoup its investment for such a system.
Allergen ManagementAllergen management is another aspect of packaged food production that cannot be neglected. In most countries, government regulations require that manufacturers list down the ingredients that are contained in a product.
This means that it is the manufacturer’s responsibility to ensure that the food label accurately reflects the contents. The problem here is that labels for different products can sometimes become mixed up within a facility (for example, an operator loads the wrong set of labels into the labelling machine).
The task of ensuring that such problems do not leave the warehouse, then falls on the QC process. Having a human-eye driven process again has certain limitations, as not all packages can be thoroughly inspected. In this context, the manufacturer runs the risk of an incorrectly described product reaching the shelves of supermarkets.
In milder cases, this could result in public embarrassment. In serious cases, however, it could be consumed by an unsuspecting customer who has an allergic reaction to certain ingredients (for example, peanuts) contained in the product.
This could lead to severe physical symptoms or even death. Such an incident can quickly tarnish the company’s reputation that was painstakingly built up over many years, not to mention the expensive litigation involved. In most situations, mismatches between labels and their contents will end up in a recall of products—often a costly and logistically intensive affair.
Yet, such incidents can be effectively avoided with the help of machine vision. Using cameras that capture images of the labels, a software can perform pattern matching on unique features in artwork. Cameras are positioned to look at the packaging/container from various faces where the labels are located, for example, top and sides.
Since the software has already been trained to recognise and understand specific images, it now looks for specific areas of text and characteristics of the label. If a match is found, the product is considered to have passed inspection and the system moves on to the next item on the line. In the case of a mismatch, the package/container is automatically diverted to a reject station downstream. In addition, the software also ensures that the label is correctly positioned on the lid.
This provides a fail-safe method of verifying the correct packaging for the appropriate product. The software is able to tolerate variations in product presentation angles, and can cope with perspective distortion. This ability to acquire the required detailed information ensures not only zero defects but also a negligible rate of false rejects.
Implementation ConsiderationsWhen it comes to upgrading technology, a common worry of facility owners is the amount of production downtime that is required to install the new hardware. Another is the complexity of the installation and whether it involves significant modifications to the current production line.
One advantage of the vision system is the ease of setup—a revamp of the existing production line hardware is usually not required. The cameras simply need to be mounted along the line and hooked up to a computer. From there, the advanced software is trained to recognise images of the product that it needs to check, and to flag out any areas that do not meet the mark.
This lowers the cost of the capital investment and also reduces the implementation time that is required. More importantly, the latter translates into less downtime for the production line. Due to the low-installation complexity nature of such vision systems, installation and implementation can be typically performed during scheduled production downtimes, such as during breaks and maintenance timeslots. Another advantage is that the solution is fully scalable and can handle higher production speeds by simply adding more cameras to the system.
Track & TraceGiven the numerous food scandals that have hit certain countries in Asia, manufacturers are becoming increasingly concerned with protecting their supply chains. With the aid of 1D barcodes and 2D codes, image-based technology is gaining traction across many industries for supply chain management.
Codes that are printed onto animals for slaughter for example, can help to track meat products throughout the processing stages and back to the source. In a particular facility, pig carcasses have to be put through a series of baths besides removing their hair and having them disembowelled. Under such harsh conditions, 2D codes are used which can still be reliably read by image-based readers.
With 2D codes, it is possible that shoppers in the future could use a smart phone or kiosk to determine whether the meat came from grain-fed, grass-fed or hormone free animals. Information on the animal—and how it was processed—can also be stored on the 2D codes.
A manufacturer of cheese adopts labels that are made of casein (a milk protein that is biocompatible with cheese). This marking method is indelible and cannot be removed or falsified. It is the cheese's ‘passport’ that provides reliable identification, which allows it to defend against imitations while ensuring traceability.
Since cheese ‘evolves’ as it matures, reading the label can become a challenging task. This means that the image-based reader has to be capable of maintaining consistent performance, despite changes in the reading surface such as dimensional degradation, incorrectly positioned labels, or codes that are partially damaged during the cheese production processes.
Clarity Of SightLighting plays a major role in the ability of the system to perform successful reads. In production environments where the lighting is too dim, the vendor may have to install a different set of lights to suit the application. Where the light is overly bright, reflections on the read surface may result and could affect the reading ability of the system. In this situation, one possibility is to put up tinted glass or plastic panels to filter off the excessive glare.
Another concern that facility owners have is the complexity of operating a newly implemented system. Systems, which are difficult to learn and operate, require operators to spend large amounts of their time to attend training sessions. This problem is also likely to repeat itself in future if there is any staff turnover and incoming personnel need to be trained to use the system. Moreover, the chances of mistakes being made during production are also higher as the complexity of operation increases.
Fortunately, a robust image-based system is easy to use and there is no need for any complex parameter configurations to be performed by the operators. Whenever new products are introduced, a software wizard assists the maintenance staff. The latter are guided through the steps that are required to teach the system to understand the new pattern template. Once completed, the settings are stored on the system. Operators simply have to select that product at the beginning of a production run, via a software interface.
Many advantages are offered by image-based systems for product tracking and quality control. It pays for manufacturers to take the time to find out how such technology can aid them in creating a production environment that is efficient, profitable and error-free.