1 of 3
Enose prototype analytical dept chemical faculty GUT Gdansk.
2 of 3
The nose is packed with multiple chips and circuits, each focused on the complex task of converting an odor to a data set.
3 of 3
Human versus computer “smelling”.
Electronic noses are devices that, as the name implies, mimic human smelling and evaluation of aromas. For close to a decade, they have produced striking results in more and more areas of tea processing and quality management.
All teas, regardless of their type, have a distinct and unique aroma. It’s a major determinant of quality and a differentiator for both high end loose leaf and commodity machine-processed crops. For premium teas, there is a very strong correlation between aroma and price. Electronic noses offer an increasing degree of understanding of aromas and how to use this to authenticate teas and determine their precise grade and quality dimensions.
Tea bag and iced teas compete in a brutal, competitive export market, marked by overcapacity, margin erosion, and crop yield pressures. Here, a small advantage in quality can be decisive for the economic health of regions and even countries. Electronic noses provide insight and tools to improve the processing of categories of tea, especially the timing of oxidation in black tea, by far the largest market segment. In all these situations, aroma is one of the most critical elements of quality.
Aroma detection and analysis
Aroma is extraordinarily complex in so many ways. It is created by the interaction among molecular volatile compounds that evaporate. This generates responses in the human brain through electrical pulses that act at the cellular level. These are affected by ambiance, psychochemical evaluation, mood, and other contextual factors. The typical tea contains around 200 volatiles, which form molecular pathways connecting, creating and releasing classes of chemical compound – theaflavins, leafy alcohols, nitrous sulfides, etc. A few have 600 identified volatiles.
Electronic noses play a major role in applied research in identifying the full range of volatiles for a given target category of tea, and uncovering more and more through the analytic, statistical, and artificial intelligence tools built into them. They also provide higher levels of assessment than the skilled panels of human tasters who rate quality but must rely on subjective 1-10 rating scales and words like “oaky” or “floral.” Alexander Graham Bell posed the core question of aroma assessment just over a century ago. “Did you ever try to measure a smell? Can you tell whether one smell is just twice as strong as another?... until you can measure their likenesses and differences you can have no science of odor. If you are ambitious to found a new science, measure a smell.”
Electronic noses are a cornerstone in that science. They share a common set of design principles with a wide range of technology, analytic, and application developments. The market is edging towards making the noses portable, standardizing components, and adding new chips that speed up or extend analysis. They can detect counterfeit teas and track the impacts of storage on aroma loss. Most promisingly, they are providing reliable and detailed recommendations for improving manufacturing processes.
Here are a few examples:
Identifying the optimum timing in oxidizing black tea in the factory, taking into account individual characteristics of batches and the process stages that include withering, rolling, and heating, that obviously change the structure and characteristic of the leaf. The aroma is a relative indicator of quality. Levels of key volatiles can vary by 50% within a few minutes. Controlling oxidation is essential for black and oolong teas. Greens are pan-fried or steamed to halt this process which is rather like letting the slices of an apple wither and brown. Rust is the result of oxidation of metals.
Differentiating aged raw and ripe puehr samples with around 98% accuracy. Given the variety, complexity, and counterfeiting of this most expensive of teas, this ability to discriminate is a powerful tool for assessing quality. The electronic nose does not affect the tea – no breaking it up, diluting it, or adding chemicals.
Precisely discriminating raw puehr by different storage years. The age of a puehr is a key selling point and the market is noted for short cuts, misrepresentation, counterfeits, and unreliable heritage. One study meticulously and accurately traced the differences of the same puehr for each of 10 years of age via an electronic nose.
Classifying the storage age of Longjing Dragonwell green tea, one the highest rated teas in the world. The range of grades is extensive with no formal metrics, methods, and verification. The rating gap between the scarce highest premium Dragonwell, little of which is exported or easy to access, and the good can translate to large price disparities. Storage is critical and loss of aroma signals poor initial quality, degradation, and poor handling in the supply chain pipeline.
Capturing the quality differences of Chinese green and black teas. The nose builds statistical profiles of the high, medium and low grades of the teas and provides baseline data for comparative analysis and pattern recognition. The results showed significant differences in aroma compounds that mark the grades and these correlate closely with prices. The electronic nose offers more certification than the label and more objectivity than tea taster evaluations.
Tracking the changes in compounds that are core to a tea’s flavors and postulated health benefits in high end teas, including Margaret’s Hope (India), Keemum (China), sencha, hojicha (Japan), and Uva, Dimbula (Sri Lanka.)
Other examples are assessing the severity of damage to bushes through climate stress, season, or disease; analyzing the sensory impacts of brewing of green and black teas, particularly infusion time; broader “precision agriculture” monitoring of soil conditions, comparison among oolongs of pan-frying versus steaming and oven-heating that has detected previously unknown compounds that affect sweetness, smokiness, and fruit taste flavors; aroma formation in white teas; quality analysis of Indonesian flavored green teas – ginger, lemon, and local herbs; detecting nonvisible pesticide levels.
These are impressive and it’s easy to see their value. Commercialization has been slow and almost all the cases are in the lab, not the field or factory. A mass market for affordable, standardized, and portable tools is likely to emerge within the next two to five years, driven by one of more of the sectors that has a stronger financial base and technology-receptive demand: food safety, medical diagnosis, police work – e-noses can detect spoiled meat early and quickly and smell breast cancer and it isn’t just dogs sniffing for drugs and explosives at airports. These are helping reduce the size of the decidedly non-portable services that sit on desks in the lab and not in the hands of field workers.
The market for devices and software is fragmented, with many options. There is a clear increase in firms competing to build market niches. More and more research studies demonstrate the value of electronic noses. As so often in the deployment of new technology, buyers are taking the lead in turning technical and scientific invention into market and product innovation. One striking instance is the Port of Rotterdam, Europe’s largest. It has installed around 250 electronic noses that detect minute changes in the air immediately. The port provides key customers with noses and is gradually building an inter-port platform. It has the infrastructures needed to integrate and communicate the electronic nose data. Organizationally, it has the needed skill base, buying power, and leadership drive. It also has the money.
That’s beyond the financial capital resources of most tea firms. It seems likely that precision agriculture platform providers will lead in easing many of the operational problems: Microsoft Farmbeats, Sentara, CropX, Bayer, and IBM are a few instances.
The technology
Here’s a summary of the core technology of electronic noses. It has many variations in every element but the basic design principles and methods are relatively generic. The nose “sniffs” a sample of the complex of scents and passes them through a set of sensors. The most common enabler here is gas chromatography. This breaks the sample into individual compounds via sensitivity to material coating, electrical responses, and, increasingly, specialized chips. Detection triggers electrical pulses. Some noses capture the changes down to the individual molecule, providing a unique signal for the nose’s software to interpret.
The sensors are moving towards a shared mainstream of metal oxide semiconductors (MOS), transistors for amplifying and transferring electrical signals. A review of research experiments showed over a dozen sensor types used by the “most known” noses, offered by close to 20 vendors. An individual nose may contain an array of 4-20 sensors. Their input is the “headspace” sample, with all its collection of compounds; the output is the equivalent of a digitized database of the separated elements.
The core of the nose is pattern recognition that organizes these sensor outputs to meet the many needs of the application. This includes profiles of teas, benchmarks, machine learning AI neural models, statistical classification, and expert tea tasters’ evaluations. A main tool is PCA (principal component analysis) that organizes and displays the information in the large dataset in a more meaningful and focused format. Each method has its limits, features, and data requirements and the nose may include a variety of software tools. AI is becoming a major area of development through neural models that are the core of machine learning.
The data aspect of electronic noses is where tea poses problems in moving from lab to field. There is so much variety of combinations of tea plant, terrain, altitude agroforestry, and seasonality plus the regional and local nature of tea categories such as puehrs versus whites. Tea is not standardized even in the basics. “White” tea, for instance, has no formal definition.
Electronic noses are a major and valuable line of applied research. Commercialization lags, as it does in AI application, blockchain, genetic DNA fingerprinting, drones, fertigation, IoT (internet of things) platforms, multispectral imaging, and other areas of technology-centered innovation. Much of this reflects the shrinking of sources of financial capital for long-term investment with delayed though substantial payoffs. Expertise is scarce – how many electronic nose specialists have a sound understanding of tea production and vice versa?
Technology in and of itself rarely provides systemic improvements – it has to be meshed into management methods, production processes, and supply chains to gain full leverage. Electronic noses are high on the payoff list and it’s worth tracking news and developments: once the industry bandwagons get moving, pick a few and jump on.
Back to Bell’s comment: “until you can measure their likenesses and differences you can have no science of odor.” We now have that science.