Feature
The future of shelf-life testing
Tucked away among the bustling stretch of Tower Bridge Road is the small but efficient London offices of BlakBear.
Despite its humble surroundings, this building is home to some significant kit which has the potential to disrupt the way shelf-life dates are established in years to come.
Founded in 2017 by a team of scientists and engineers who met at Imperial College London, the deep tech start-up has created an innovative sensor that can tell the user how fresh their food is, which are now used by major food manufacturers such as Cranswick.
Traditional shelf-life testing
Conventional shelf-life testing works on clever but rather rudimentary principles, with multiple packs opened and tested with sensory and microbiological methods over the course of several days.
The issue with this is that it creates a lot of waste, both in packaging terms and food, not to mention it’s slow, labour intensive and expensive.
Moreover, the nature of the test only enables the tester to obtain one data point which means the shelf-life test might not to wholly accurate.
“To determine a use by date for a typical pack of chicken, they’ll [labs] get 20 or so sample packs, store them in a fridge at controlled temperatures, and open one or more every few days. They’ll smell it and taste it and count the bacteria on it – and once it goes beyond a threshold, i.e. smells bad or has too many bacteria on it, that will establish the use by, within industry guidelines” co-founder of BlakBear, Dr Max Grell explained.
“A lot of shelf-life dates are probably inaccurate but by law food businesses cannot sell it. Printed dates don’t actually know the freshness of a food package, they’re chosen to cover the worst-case scenario and there’s a huge variation between packs. That means there’s a lot of unnecessary waste because of a fixed date,” he added.
“I think in five years we’ll look back and say it was crazy we were trashing food because of a fixed printed date.”
The digital sniff test: how it works
BlakBear replaces this with real-time monitoring using low-cost sensors.
“We think food should tell you how fresh it is,” Grell told Food Manufacture.
The core IP (published/peer-reviewed and patented) of BlakBear is a gas sensor that sits inside the food packaging measuring the gases the commodity (e.g. raw chicken) emits.
“The result is 80-90% lower cost than a traditional shelf-life test, with 75% less waste and real-time data rather than waiting 3-5 days for each micro culture.”
One particularly impressive attribute of the electrical gas sensor is its ability to thrive in the moist and corrosive environment of food packaging, where ordinary electrical sensors would fail.
But in this sensor the humidity is used to its advantage, with an aqueous layer built into the device. The gases dissolve into that layer and the level of resistance within it indicates the freshness of the product.
The sensor takes measurements every second, assessing how much is in the layer and how that gas has changed. That gas change correlates to how the food smells and therefore how much bacteria there are and how fresh the product is.
The two types of sensors
The team have designed two different sensors– one which is for ‘virtual labs’ and re-used on sites (i.e. in-house testing on a manufacturing site) and another smaller ‘smart label’ version which has the potential for retail and at-home use on individual packages
The first and larger of the two sensors sits loose within the packaging of the food product. As the gases emit, the data is shared via Bluetooth, powered by a battery pack in the box, and received by a gateway device which shares the readings to the Cloud.
On the Cloud the data is stored and assessed by the accompanying software which has been trained to read the measurements and translate them into meaningful graphs.
The idea behind this larger device is that a manufacturer could use it to monitor its processes and flag any potential issues. And, because the data is individual to each pack, it’s more likely that outliers will be spotted – for example, temperature differences in certain areas of an industrial fridge or a leaking pack.
After a few days of taking readings, the software will also be able to forecast the shelf life of each pack, spotting quality issues before they become complaints.
The second version of the sensor is much smaller as it has no battery. This not only makes it cheaper and slight enough to be embedded within the packaging, but it also opens up potential future uses for the technology within the retail and consumer space.
Based on the same technology used widely in the fashion industry – the sensors are equipped with RFID tags – small electronic devices that store and transmit information to a reader.
Like all RFID, these are powered wirelessly, for example through a phone. Once triggered, the sensors will take a reading of the gases, which can be viewed on the software as well as via a consumer-friendly App – the latter of which will be a particularly useful for at-home testing.
These sensors are also small enough to fit into vacuum packs, with the tech slightly tweaked.
“Whilst there isn’t headspace for gases, there is purge – the juices that come off meat – and using the same mechanisms, we can measure the spoilage from that,” Grell elaborated.
Training the tech
Of course, taking measurements is only one part of the puzzle – and a lot of effort went into both the design and machine learning capabilities of the BlakBear software.
Alongside the black bear paraphernalia that adorns the London HQ – a fun nod to the company’s name which honours this particular animal’s keen sense of smell – the facility also comprises an array of different sized incubators (fridges).
These incubators are set at varying temperatures and allow the team to test how certain products react to specific temperatures and temperature spikes. This information is then used to train the software on the gases and the levels that indicate food degradation.
The varying degrees of temperature is an important element of training the model as it enables the team to establish the boundaries for different food types and how they react in different extremes. This information can then be used to teach the model to recognise these patterns and make predictions.
“You might not store your food at 5°C, you might store it at 12°C. We want to make sure we have a good understanding of what will happen in those temperatures,” Grell said.
Each sensor has an ID, so when logging onto the software the user can choose the corresponding device. The software also enables the operator to toggle between sites and product type (e.g. whole or diced chicken).
The data is recorded into a graph – one shows ‘digital micro’ (bacteria counts) and the other ‘digital odour’.
“We measure a wide of spoilage organisms. As the food spoils, the bacteria are competing and typically a single species will win. But it doesn’t matter what the bacteria is, the data is just showing when it grows exponentially.”
The software also has the capability of translating this into a map, tracking a sensor’s journey through the supply chain. This has significant benefits in establishing where, for example, temperature abuse may have occurred specifically.
“We have done work with retailers in the UK to measure freshness through the supply chain to identify the temperature problems and what impact that is having on shelf life.”
Moreover, with the ability to embed small sensors into labels, companies could also use this to track the recycling of their products.
So…how accurate is the tech?
During the business’s early stages, BlakBear had its sensors tested out by Campden BRI to check how it stood up against its own shelf-life testing. Microbiology testing was conducted on four organisms: Aerobic Plate Count (APC), Enterobacteriaceae, Lactic acid bacteria and Pseudomonas.
In this study, raw chicken breast samples, supplied by 2 Sisters Food Group, were sealed with BlakBear sensors in a modified atmosphere packaging (MAP) and refrigerated for up to 15 days.
“Campden BRI carried out a traditional shelf-life test using the sensor in around 50 packs. These were stored at 4°C [for 7 days], spiked at 23°C for a few hours to simulate someone taking their products home from the shops, and then stored at 9°C to replicate a fridge at home.”
On scheduled sampling dates, several packs were removed from chilled storage and transferred for gas, microbiological and sensory analysis by Campden. Meanwhile, BlakBear sensors continuously monitored the spoilage profile of the samples from the point of packaging.
The results determined a 90% correlation between Campden’s own micro/sensory data and BlakBear’s.
The future of use-by dates
For Grell, the future of manufacturing is a more automated system, with the technology feeding back information and recommendations for safer and best practice.
“If you make a change to your product in manufacturing, it generally counts as a new product and often you have to do quite extensive shelf life testing again. This could be the same product, but just in a different volume package, for example. You could definitely make a strong argument that that doesn’t need extensive testing – but it’s often required and it is expensive,” he explained.
“If you can make testing easier, you could test lots of different formulations and products and therefore what is best for shelf life and waste. [So you’d be able to] rapidly understand what ingredients and packaging makes sense for you and what temperatures you need.”
For retail the picture is similar, with Grell envisioning supermarket fridges equipped with antenna to power up the RFID sensors (which BlakBear is currently doing in some of its own incubators) – enabling stores to keep track on freshness.
This would mean that if, for example, a supermarket fridge door was left open by a consumer, the sensors would be able to determine if a product had been affected. The same notion could also apply to meal kit services, which can often be left on doorsteps for hours at a time.
Going one step further, he ventured that fridges could then be automated to lower or increase in temperature as needed. For example, a particularly hot day could see fridges automatically tweak their settings.
This kind of technology could also be used to track inventory in store and keep tabs on where food is located – we’ve all seen misplaced stock, with ready meals finding their way to ambient shelves.
“You know where all your food is all the time, what temperature it is, how fresh it is in real time. As a result of that you don’t have to send people around to count stock, you don’t have to over or under order,” Grell suggested. “If the food is not fresh, the fridge should be cooler and vice versa, you’ll be able to tell if the fridge is too cold and where you can save energy. There should be these feedback loops in there…that’s not impossible to imagine at all.
All in all, the technology has the potential to be pretty groundbreaking in terms of waste reduction, cost savings and safety.
In related news, a team from the University of Arkansas System Division of Agriculture believes its made significant headway in improving AI quality inspection systems.
