Analyzing aerosol releasing articles

Abstract

A device for analysing an aerosol-releasing article, such as a product of the tobacco industry or the e-cigarette industry, is disclosed. The device comprises means (22) for drawing an aerosol from the article, a plurality of sensors arranged to sense a characteristic of the aerosol to generate sensor data, means (60) for predicting at least one organoleptic response from the sensor data using a model (70) that relates the sensor data to the at least one organoleptic response, and means (72) for outputting the predicted organoleptic response to a user. This can allow an indication to be provided as to how the user will experience the article, which can help to reduce or avoid the need for a human organoleptic panel.

Description

[0001] This invention relates to techniques for analyzing aerosol-releasing articles, and more particularly to techniques for predicting sensory responses associated with aerosol delivery to the human mouth, lungs, and / or respiratory tract. This invention is particularly, but not exclusively, applicable to the analysis of products in the tobacco and e-cigarette industries.

[0002] In the tobacco industry, product formulations / blending are frequently changed. When making such changes, sensory testing may be necessary to ensure a similar or superior sensory experience for users. Other uses of sensory testing include: matching the sensory experience of a product to that of competitors; producing alternatives with similar characteristics to existing products, such as e-cigarettes or heated tobacco products; or adapting a product to a specific market.

[0003] Currently in the tobacco industry, smoking evaluation panels consist of trained and screened personnel who are able to smoke cigarettes / e-cigarettes and provide standardized responses to sensory experiences. Useful data can be obtained by "calibrating" the personnel in these smoking evaluation panels using products designed to represent specific sensory attributes such as "stimulation."

[0004] Evaluations conducted by panel members within the tobacco industry suffer from several significant flaws that translate into practical limitations. These limitations include cost, repeatability and reproducibility, ethical concerns, and a declining number of adult smokers willing to participate in such studies. Large tobacco companies invest substantial sums annually in consumer and taste evaluation panels. Alternatives that mitigate these significant shortcomings would offer substantial benefits.

[0005] It is known that smoking machines can be used to analyze smoking products. For example, the subject matter of WO 2006 / 100493 A1, incorporated herein by reference, discloses a smoking machine comprising one or more electrochemical cells for analyzing the composition of smoke streams from smoking products. This allows for direct, real-time analysis of gaseous or vapor phase chemicals. The concentration of a particular chemical can indicate the brand of the cigarette smoked.

[0006] Known mechanical smoking systems can detect the concentration of specific chemicals, but cannot provide any indication of how a user can experience the product.

[0007] Therefore, it is desirable to provide techniques for determining predictive sensory responses to aerosols that can reduce or avoid some of the shortcomings associated with the adoption of evaluation panels.

[0008] According to one aspect of the present invention, an apparatus for analyzing aerosol-emitting products is provided, the apparatus comprising: Apparatus for extracting aerosols from a product; Multiple sensors are arranged to sense the properties of aerosols to generate sensor data; A means for predicting at least one sensory response from sensor data using a model that associates sensor data with at least one sensory response; and A device for outputting a predicted sensory response to a user.

[0009] This invention offers the advantage that by providing multiple sensors arranged to generate sensor data, and means for predicting at least one sensory response from the sensor data using a model that correlates the sensor data with at least one sensory response, the device can provide indications about how a user will experience the article. This can help reduce the need for artificial sensory evaluation panels, thereby helping to address the costs, ethical issues, and other drawbacks associated with using such evaluation panels.

[0010] This invention is based on the understanding that sensor data from multiple sensors can provide a feature map or "fingerprint" that can be used to infer human sensory experience. The sensors do not necessarily target a specific sensor; rather, they can be non-targeted, meaning they do not search for a specific outcome. However, the combination of sensor data from multiple sensors can characterize the desired predicted sensory response. In this way, human sensory experience can be inferred using sensor data from multiple different sensors.

[0011] Therefore, the model can correlate sensor data from multiple sensors with the sensory response or each sensory response. In this way, sensor data from multiple sensors can be used to generate the predicted sensory response or each predicted sensory response. In other words, a single predicted sensory response can be generated using sensor data from two or more (different) sensors. This enables the generation of more accurate predictions about how the user will experience the product.

[0012] The aerosol-releasing article is preferably of the type intended for delivering aerosols to the human body. For example, the aerosol-releasing article may be of the type intended for delivering aerosols to the human mouth, lungs, and / or respiratory tract. In one embodiment, the aerosol-releasing article is a product of the tobacco or e-cigarette industry. For example, the aerosol-releasing article may be a cigarette, cigar, heated tobacco product, e-cigarette, e-cigarette device, or any similar article. However, embodiments of the invention may be used in conjunction with articles for delivering aerosols for other purposes, such as therapeutic applications.

[0013] The predicted sensory response can be a prediction of the response that will occur in the human body when the aerosol is inhaled. For example, the predicted sensory response can be a prediction of the response of one or more sensory receptors in the human body (e.g., in the mouth, tongue, throat, lips, nose, eyes, ears, and / or skin) that generate nerve impulses in the brain.

[0014] The predicted sensory response may include a quantitative description of at least one sensory experience reported by a (human) sensory evaluation panel. For example, the predicted sensory response may be in the form of a numerical value, or a classification such as "high," "medium," or "low," or any other suitable description of the sensory experience.

[0015] The predictive device is preferably arranged to predict multiple sensory responses based on sensor data. This provides a more complete picture of how a user might experience the product.

[0016] In one embodiment, the sensory response may include at least one of the following: a sensation of impact, throat irritation, oral tingling, aerosol filling, heat, oral coolness, draw resistance, oral dryness, or any equivalent thereof. In this context, "impact" may be defined as a sudden, intense, but brief sensory sensation felt when the aerosol contacts the chest during inhalation—sometimes described as a "chest hit"; "throat irritation" may be defined as a persistent burning or stinging sensation in the throat; oral tingling may be a stinging, numb, or peppery sensation in the mouth; aerosol filling may be the perceived amount of vapor entering and filling the mouth during inhalation; heat may be a combination of the temperature of the filter at the lips and the temperature of the aerosol in the mouth; oral coolness may be related to the perceived presence and amount of menthol or similar sensory sensations; draw resistance may be related to the perceived force required to inhale vapor from a lit product into the mouth; and oral dryness may be related to the dryness effect in the mouth after exhalation of the aerosol. Of course, these descriptive terms and definitions are merely examples, and other descriptive terms and definitions may be used alternatively or additionally. For example, in some cases, other terms may be used to describe similar, related, or equivalent sensory experiences.

[0017] It has been found that meaningful results can be obtained by using multiple different types of sensors to generate sensor data. Therefore, the multiple sensors can comprise a variety of different types of sensors. For example, the multiple sensors can be arranged to sense at least one (and preferably at least two) of the following: the physical properties of the aerosol; the chemical properties of the aerosol; and the size properties of the aerosol. The physical properties may include, for example, temperature, pressure drop, etc. The chemical properties may be related to the presence and / or concentration of chemical substances and can be obtained using, for example, electrochemical sensors such as tin oxide environmental sensors, spectroscopic methods, pH sensors, colorimetry, etc. The size properties may be related to the size of the aerosol particles and can be obtained, for example, using dynamic aerosol diameter measurement, light scattering, etc. Other types of measurement methods may also be used additionally or alternatively if desired. This can generate a characteristic pattern or fingerprint of the article.

[0018] In one embodiment of the invention, the aerosol properties include at least one of the following: aerosol density and / or extinction; aerosol particle size distribution; the presence and / or concentration of one or more chemical substances; aerosol temperature at one or more locations; aerosol humidity; and pressure and / or pressure drop. This combination of properties (or a subset thereof) has been found to be particularly effective in predicting human sensory experience. However, one or more other properties may also be used alternatively or additionally.

[0019] Suitable sensors can be configured to sense the aforementioned aerosol characteristics. Therefore, in one embodiment of the present invention, the plurality of sensors includes at least one of the following: an aerosol density and / or extinction sensor, a particle size distribution sensor, an electrochemical sensor, a temperature sensor, a humidity sensor, and a pressure sensor.

[0020] At least one of the plurality of sensors may be arranged to sense the physical properties of the aerosol in both the intensity and time domains. For example, samples of these physical properties may be collected at different time points (e.g., during the period the aerosol remains within the cavity). In this case, the model can correlate the sensor data in both the intensity and time domains with at least one sensory response. This allows the device to generate a predicted sensory response related to the duration of a specific sensory experience. This helps to provide a predicted sensory response corresponding to human sensory experience.

[0021] A device for predicting sensory responses may include a model that associates sensor data with at least one sensory response. The model may be a mathematical model, such as a statistical model, and may include algorithms for inferring human sensory responses from sensor data. By using a model that associates sensor data with at least one sensory response, human sensory experience can be inferred from sensor data.

[0022] In one embodiment, the model includes a machine learning algorithm. The machine learning algorithm may use, for example, neural networks, linear regression, logistic regression, clustering, decision trees, random forests, or any other suitable method. The machine learning algorithm is preferably pre-trained. For example, the machine learning algorithm may be pre-trained using sensor data and reported sensations from a sensory evaluation panel. The sensor data is preferably obtained from multiple test artifacts (which may be "standard" artifacts), and the reported sensations are preferably obtained from the same or corresponding test artifacts, preferably under the same or corresponding conditions.

[0023] In one embodiment, the model uses a decision tree algorithm to predict sensory responses. A decision tree uses a sequence of branches representing associated decisions, which can be represented by a tree diagram. The decision tree algorithm can use tree models such as decision trees, random forests, or gradient boosting tree models such as XGBoost. This method has been found to be effective in predicting sensory responses based on sensor data.

[0024] However, in alternative embodiments, the model may alternatively or additionally employ other machine learning methods, such as artificial neural networks (ANNs), support vector machines (SVMs), regression analysis, Bayesian networks, Gaussian processes, etc. Furthermore, in some embodiments, the sensory response model may be in the form of one or more mathematical formulas, such as polynomial equations, provided by a linear mixed-effects model.

[0025] If necessary, multiple models can be provided, each corresponding to a specific type of product and / or a set of test variables. In this case, the device can be configured to select the appropriate model, for example, automatically or in response to user input.

[0026] The device may also include means for preprocessing the sensor data before it is input into the model. The means for preprocessing the sensor data may be arranged, for example, to reduce the amount of data input into the model. This may be desirable, for example, when sampling the properties of the aerosol at a relatively high sampling rate (compared to aspiration), in order to reduce the amount of data that needs to be processed.

[0027] In one embodiment, the device may include means for flattening sensor data to obtain a set of features for each sensor. In this case, the model may be arranged to predict at least one sensory response based on the flattened sensor data. The flattening process may, for example, involve converting the data into another form and / or reducing the amount of data. The features may be, for example, synthetic information such as mean, standard deviation, linearity, entropy, curvature, etc. This can help reduce processing load and facilitate matching sensor data with the responses of evaluation panel members.

[0028] Devices for extracting aerosols from an article of work may include a suction engine. This allows for the controlled extraction of aerosols from the article of work. The suction engine may include any device suitable for applying negative pressure to the article of work. The suction engine may be programmed with suitable suction parameter profiles, such as the volume and / or shape of the suction.

[0029] The device may further include a cavity, and means for extracting aerosols from the article may be arranged to extract the aerosols into the cavity. In this case, at least one of a plurality of sensors may be arranged to sense the physical properties of the aerosols within the cavity. The cavity may, for example, simulate physical features of the human body, such as the mouth, throat, or lungs. This allows the device to simulate the conditions that aerosols undergo within the human body.

[0030] If needed, the volume of the chamber can be adjustable. For example, this allows testing to be adapted to different products, conditions, and intended end uses.

[0031] If needed, some or all of the sensors can be housed in the same cavity. However, it has been found that this can lead to space constraints or potentially impair the arrangement of the sensors. To address this issue, the device may include multiple cavities and / or multiple suction engines. In this case, each cavity may include at least one sensor. This allows for the deployment of a greater number of sensors without compromising sensor arrangement due to space constraints. Each cavity may have its own suction engine, or a suction engine may be shared by one or more cavities.

[0032] In one embodiment, the device is arranged to extract aerosols from a plurality of aerosol-releasing articles, and the plurality of sensors are arranged to sense at least one characteristic of the aerosol from each of the plurality of articles. Sensory responses can then be predicted based on sensor data from the plurality of articles. Each article can be a different sample of the same product. This allows for the deployment of a larger number of sensors without compromising the sensor arrangement. Each article can be aspirated using its own suction engine, or a suction engine can be shared by one or more articles. The aerosol from each article can be extracted into a separate chamber, or alternatively, the aerosol from two or more articles can be extracted into the same chamber.

[0033] In one embodiment, the plurality of sensors includes a plurality of temperature sensors arranged to sense the temperature of the aerosol at multiple different locations. For example, when the aerosol is drawn into a cavity, the plurality of temperature sensors can sense the temperature at different locations within the same cavity and / or at different locations within different cavities (or the temperature at other locations, such as the temperature of the smoking article itself at one or more locations, such as the filter or mouthpiece). This has been found to help predict one or more reported sensations, such as impact and / or flavor intensity, and / or reported sensations that can be assessed at different sites, such as sensations in different areas of the mouth, throat, or nose.

[0034] In one embodiment, the plurality of sensors includes a humidity sensor arranged to sense the humidity of the aerosol. For example, when the aerosol is drawn into a cavity, the humidity sensor can sense the humidity of the aerosol within the cavity. The humidity sensor can sense relative humidity, preferably the ratio of the amount of water present to the amount of moisture that would be present when the air is saturated. This can help predict one or more reported sensations, such as irritation, a feeling of impact, and / or a dry mouth.

[0035] In one embodiment, the plurality of sensors includes pressure sensors arranged to measure pressure changes during aspiration. For example, when aerosol is aspirated into a cavity, the pressure sensors can sense the pressure drop within the cavity during aspiration. This can help predict one or more sensations, such as aspiration force.

[0036] In one embodiment, the plurality of sensors include a weighing device arranged to determine the mass of particle deposition. For example, in the case of aerosol being extracted into a cavity, the weighing device may be arranged to determine the mass of particle deposition within the cavity. This can help predict one or more sensations, such as oral coverage and / or taste intensity. The weighing device may be, for example, a quartz microbalance, or any other device suitable for measuring mass.

[0037] In one embodiment, the plurality of sensors includes optical sensors arranged to measure particle scattering of the aerosol. This can help predict one or more sensations, such as aerosol fullness. The optical sensors may be positioned, for example, in a conduit through which the aerosol passes, or in other locations, such as within a cavity.

[0038] In one embodiment, the plurality of sensors include at least one electrochemical cell. The electrochemical cell may be arranged to sense the concentration of one or more substances, such as compounds or compound types. For example, the electrochemical sensor may be arranged to sense the concentration of volatile organic compounds (VOCs), oxygen (O2), carbon monoxide (CO), carbon dioxide (CO2), nicotine, and / or any other compound or element. This can help predict one or more sensations, such as impact and taste intensity.

[0039] At least one electrochemical cell can be a solid-state sensor, such as a tin oxide sensor. This allows for the use of relatively inexpensive and readily available sensors.

[0040] In one embodiment, the plurality of sensors comprises a plurality of electrochemical cells. In this case, each electrochemical cell may be arranged to sense a substance (such as a different compound or compound type) that is different from at least one other electrochemical cell. This can help predict one or more sensations, such as impact and taste intensity. Each electrochemical cell may be located in the same position as the other electrochemical cells, or in a different position from the other electrochemical cells.

[0041] In one embodiment, the plurality of sensors includes a particle size measuring device. This can help predict one or more sensations, such as a feeling of mouth coverage, aerosol fullness, and mouth dryness. The particle size measuring device may be located, for example, on the exhaust port side of a suction engine, or in other locations.

[0042] In one embodiment, the plurality of sensors includes a spectral sensor. This can help predict one or more sensations, such as impact and taste intensity. The spectral sensor may be located, for example, on the exhaust port side of a suction engine, or in other locations.

[0043] The means for outputting the predicted sensory response to a user may include a display, or any other suitable means for presenting information to a user, such as a printer, light fixture, speaker, or other visual or audio indicator. The predicted sensory response may be presented to the user in any suitable manner, such as a series of heatmaps, radar charts, or pseudo-videos of data changes. Alternatively or additionally, the means for outputting the predicted sensory response may be arranged to store the predicted sensory response and / or transmit the predicted sensory response to another device.

[0044] According to another aspect of the present invention, an apparatus arranged for analyzing aerosol release articles is provided, the apparatus comprising: Devices (such as suction engines) are arranged to extract aerosols from the product. Multiple sensors are arranged to sense the properties of aerosols to generate sensor data; A processor configured to predict at least one sensor response from the sensor data using a model that correlates sensor data with sensor responses; and The user interface is designed to output the predicted sensory response to the user.

[0045] Corresponding methods can also be provided. Therefore, according to another aspect of the present invention, a method for analyzing aerosol-emitting products using an analytical device is provided, the method comprising: Extracting aerosols from the aerosol-releasing product; Multiple sensors are used to sense the properties of the aerosol to generate sensor data; Predict sensory responses from sensor data using a model that associates sensor data with at least one sensory response; and Output the predicted sensory response to the user.

[0046] The method may include using the model to convert sensor data into at least one predicted sensory response. The method may also include training the model using training data. For example, the model may be trained in advance using sensor data obtained from multiple test products and reported sensations obtained from the same or corresponding test products. Therefore, the method may further include the step of training the model using sensor data and reported sensations from a sensory evaluation panel.

[0047] A feature of one aspect can be combined with any other aspect. Any device feature can be provided as a method feature, and vice versa.

[0048] Preferred features of the invention will now be described purely by way of example, with reference to the accompanying drawings, in which: Figure 1The invention illustrates components of a device for collecting sensor data in one embodiment; Figure 2 Components of a sensory testing device are shown in one embodiment of the present invention; Figure 3 The components of the monitoring unit are shown in more detail; Figure 4 This illustrates how sensory descriptors can be presented to a user in an exemplary embodiment; and Figure 5 The steps performed by a sensory testing device in one embodiment are shown.

[0049] Many industries rely on human sensory responses to ensure the success of their products. The tobacco industry is one such industry. Changes in tobacco or product formulations / blending occur frequently in this sector. Furthermore, sensory expectations can differ across different markets. Therefore, sensory testing can be used to ensure that sensory experiences match market expectations.

[0050] It should be noted that the "sensory experience" considered by the tobacco industry goes beyond what can be defined as "taste" or "smell." It describes the experience throughout the mouth, lungs, and nasal cavity when the aerosol is inhaled. Terms such as "mouthfeel," "fullness," "aftertaste," and "dryness" are very common, but these terms do not constitute relevant characteristics within the realm of taste; rather, they can be described as the "tactile" experience of the product's aerosol. It is important to note that a single sensory response is the result of the combined effects of multiple physical reactions produced by the human body in response to smoke.

[0051] Currently in the tobacco industry, smoking evaluation panels consist of trained and screened personnel who are able to smoke and provide standardized responses to sensory experiences. These sensory experiences are inherently qualitative, although they may contain approximate quantitative elements (e.g., high, medium, low, or numbers from 1 to 5). Useful data can be obtained by "calibrating" the evaluation panel members using products designed to embody characteristics such as "stimulation."

[0052] Table 1 lists a variety of potential sensory descriptors that evaluation panel members can use to describe the product.

[0053] Table 1

[0054] It should be understood that the sensory descriptors shown in Table 1 are merely illustrative and may be used in place of or additionally to describe other descriptors and sensations.

[0055] Large tobacco companies invest heavily in consumer and taste evaluation groups every year. Alternatives that can reduce these high costs would bring significant benefits to manufacturers.

[0056] Reproducibility and repeatability are often challenging when assessing subjective effects or impact using human perception. This problem can be partially mitigated through standardized procedures. This involves training participants, establishing a common language of descriptive terms to explain sensory experiences (although the language used may vary between companies), and “calibrating” individual responses using standards or controls. However, even with these precautions, variability in results still frequently occurs.

[0057] The ethical issues surrounding the use of human evaluation panel members remain significant. Panel members are exposed to cigarette smoke, and although existing smokers are selected for this task, this is still considered an ethical gray area. Furthermore, the use of human smoking evaluation panels is inconsistent with tobacco companies' strategies for transitioning to a smoke-free world and launching a range of low-harm products.

[0058] Even if used only as part of a screening tool, further minimizing the variability of results obtained by human evaluation panel members, reducing the cost of a single test, and reducing or eliminating ethical dilemmas would be a major advancement in the field of tobacco aerosol sensory response research.

[0059] Embodiments of the present invention aim to replace human evaluation, at least in part, with objective assessments using electronic or other sensing devices. It should be recognized that electronic systems are unlikely to fully simulate a sensory evaluation panel, and therefore the breadth of experience / data obtained may be limited. Furthermore, the final decision-maker regarding issues such as product acceptability is usually still a human. However, electronic systems can provide an early screening mechanism that eliminates some routine testing in product development, allowing only the most promising candidates to enter the full panel evaluation phase, or can support multiple product iterations to find the sensory optimality.

[0060] Another use is to determine whether intentional or unintentional changes to the tobacco formulation in a product will have any noticeable impact on the consumer experience, such as changes in formulation (changes in the ratio of Virginia tobacco / spiced tobacco / burley tobacco) or changes in added ingredients (e.g., excessive licorice). This method can then become part of product quality assurance, rather than just a research tool.

[0061] A fundamental problem that this invention aims to solve is to replace or partially replace human sensory evaluation panels with a screening device based on a sensor with electronic output that can output valid results (at least one predicted sensory response).

[0062] According to the present invention, it has been found that by deploying multiple sensors to jointly form the characteristic response pattern of the aerosol being measured, it may be possible to infer human sensory experience through pattern matching.

[0063] In some embodiments, a suction engine or a series of suction engines is provided, which can be programmed to inhale tobacco products such as cigarettes, heated tobacco products, e-cigarettes, and e-cigarette devices via standardized suction parameter curves (which can be defined mathematically or reproduced from records of human inhalation patterns). Aerosol from the product enters one or more chambers. The chambers are adjustable in size to simulate the volume of the mouth, a single inhalation, or the lungs. The aerosol remains in the chamber for analysis until the next inhalation is performed, or is expelled after a set residence time. The chambers are equipped with sensors and may also include facilities for introducing aliquots of smoke into a more sensitive device that may require dilution. The aerosol is held for a set time—and a bypass pathway for flushing with clean air may also be provided—before being expelled, or inhaled into another chamber equipped with a second sensor array, a process that may be accomplished using a filter that mimics the human body's filtration mechanism.

[0064] A series of chambers can be used, each capable of simultaneously drawing different cigarettes of the same product using the same draw parameter curve. This allows for the deployment of large sensor arrays without having to reduce the number of sensors or change their placement due to space constraints. In this scenario, it is assumed that drawing the same product will produce the same sensory experience.

[0065] During inhalation and smoke residence, the electronic output of the sensors is recorded (timestamped to ensure synchronization), generating a set of data (electronic sensor profile) that represents the characteristics of the sensory experience as perceived by the sensory evaluation team members. The intensity of the sensor response and the timing of that response are correlated (the response profile is not just a single reading).

[0066] By combining the response maps of individual sensors within the time domain, a sensory "fingerprint" of the product under test can be generated. This sensory fingerprint can be expanded by considering each sip in the sequence as part of the overall user experience fingerprint rather than the fingerprint of a single sip.

[0067] In order to provide predictable sensory responses, the system is first calibrated based on the responses of the sensory evaluation panel members.

[0068] Currently, there are "standard" products that have been characterized by members of the sensory evaluation panel. These products have known sensory evaluations under the test specifications, including sensations and the intensity of said sensations. Using such products, it is possible to generate spectra characterizing the smoking mechanism, smoking sensations, taste sensations, and residual sensations. By using a series of standards, specific spectra associated with the reported sensations can be identified. In this way, a set of analytical measurements directly related to the sensory evaluation panel members' responses in terms of intensity and presence can be obtained.

[0069] Alternatively, a series of head-to-head comparisons can be made between the sensor evaluation panel data and the sensor array response specific to a particular product to generate a set of standardized data. Based on this data, a sensor / sensory response model can be created that can be universally applied to the new system under test.

[0070] A direct correlation between a single sensor measurement and the sensory evaluation panel's assessment is unlikely. More likely, a series of sensor outputs correlate with the sensory evaluation panel's data points. This panel response does not always simply or linearly "proportionately" to the sensor matrix response—some responses dominate, some have a counteracting effect, and some only have a measurable effect when considered in conjunction with other sensor responses. Based on the array of sensor data from the sensory evaluation panel and the sensor data, sensors that can be used to define the sensory response, as well as those that are almost uncorrelated or uncorrelated with the sensory response under investigation, can be identified.

[0071] In addition, the continuous sensitivity of the system (which may be poisoned by chemicals in aerosols or physically coated to form an aerosol barrier layer) can be determined and confirmed by periodic inspections using the aforementioned standard products.

[0072] The sensor does not need to target a specific sensation; instead, it can be "non-targeted," meaning it does not detect a specific outcome, but when combined with other sensors, it can form an array that characterizes the identified response. This array can have a strong, weak, or even negative correlation with the reported sensation. For example, particle size can have a negative or positive correlation with taste; aerosol density has a strong correlation with taste; aerosol temperature has a weak correlation; humidity has a strong correlation with taste; pH has a negative correlation with taste; and so on.

[0073] The "non-targeted" nature of this method (e.g., its purpose is not necessarily to measure a specific chemical substance, but to map human responses to multiple associated measurement results) integrates physical measurements (weight, pressure drop, etc.), chemical measurements (tin oxide environmental sensors, spectrometry, pH, potential colorimetry, etc.) and traditional particle size measurements (aerosol dynamic diameter measurement, light scattering, etc.) into a product experience profile that changes over time (inhalation), forming a characteristic profile or fingerprint of the product in the intensity domain and time domain.

[0074] Data arrays can be processed in various ways and presented to system users, such as as a series of heatmaps, radar charts, or pseudo-videos of data changes. Pre-defined algorithms can be used to "convert" sensor data arrays into a set of sensory evaluation terms, such as "high throat hit."

[0075] This data transformation is a major challenge in the development of sensory substitution technologies, as is the cross-correlation with the responses of a human sensory evaluation panel. The sensing system needs to be correlated with the sensory evaluation panel. For example, it may be necessary to map each sensory evaluation panel member's "rating" to a sensor, involving different orders of magnitude, and then validate it against "unknown samples."

[0076] The process becomes more complex when the time domain is taken into account. Some sensory evaluation panel responses may involve the duration of a scent or experience. This means that sensor data should be collected in the time domain, and in some cases, the spatial domain, as well as the intensity dimension.

[0077] Figure 1 Components of a device for collecting sensor data, as shown in one embodiment of the present invention, are illustrated. (Refer to...) Figure 1 The device includes a clamp 10 for holding the aerosol-generated article 12, a first conduit 14, an analysis chamber 16, a particulate filter 18, a second conduit 20, a suction engine 22, and a regulating gas source 24.

[0078] The aerosol-generating article 12 can be, for example, a cigarette, cigar, heated tobacco product (HTP), electronic cigarette, electronic cigarette device, or any similar article. A holder 10 is arranged to hold the end of the article 12 that abuts against a consumer, such as a filter or mouthpiece. The holder 10 includes at least one seal, such as a lip seal, which forms a seal with the article and mimics the consumer's lips. The holder 10 is in fluid communication with an analysis chamber 16 via a first conduit 14. The analysis chamber 16 is in fluid communication with a suction engine 22 via a second conduit 20.

[0079] The dimensions of the analysis chamber 16 can be adjustable, for example, to simulate the volume of the oral cavity, a single aspiration, or the lungs (although this feature is optional). This can be achieved, for example, by extending a bellows, an extension ring, or a sliding seal, and can be automated. For example, a stepper motor can be provided that can be actuated to increase and decrease the size of the chamber. The chamber 16 is equipped with sensors and may also include mechanisms for drawing aliquots of flue gas into more sensitive instruments that may require dilution.

[0080] A suction engine 22 is used to perform suction from the article 12 to generate an aerosol stream. The suction engine 22 is programmed with standardized suction parameter curves, which may be mathematically defined or reproduce records of human suction patterns. A particulate filter 18 is used to collect particles from the aerosol stream. The particulate filter 18 is an optional component and can be removed or replaced.

[0081] During operation, the suction engine 22 draws smoke from the aerosol-generating article 12 to generate an aerosol stream. The aerosol stream is drawn from the article 12 and enters the analysis chamber 16 via the first conduit 14. The aerosol remains in the chamber 16 for a set residence time. The aerosol stream then passes through the particulate filter 18 and the second conduit 20 to reach the suction engine 22. The aerosol is then discharged from the suction engine 22 through the exhaust port 44.

[0082] exist Figure 1 In the arrangement, the device includes an optical sensor 30, a first temperature sensor 32, a second temperature sensor 33, a third temperature sensor 49, a humidity probe 34, a quartz microbalance 36, a pressure drop sensor 38, an electrochemical sensor array 40, an impact sampler or other capture device 42, a particle size measuring device 46, and a spectral sensor 48. An optional dilution valve 50 is provided for the particle size measuring device 46 and the spectral sensor 48.

[0083] An optical sensor 30 is disposed within the first conduit 14 to measure the "thickness" of the particle scattering or aerosol. A first temperature sensor 32 is disposed towards the front of the cavity 16 to measure the aerosol temperature near the point where the aerosol enters the cavity. A second temperature sensor 33 is disposed towards the rear of the cavity 16 to measure the aerosol temperature near the point where the aerosol exits the cavity. A third temperature sensor 49 is disposed towards the aerosol-generating article 12 to measure the temperature at one or more points on the aerosol-generating article, such as the filter temperature. A humidity probe 34 is disposed within the cavity 16 to measure the humidity of the aerosol within the cavity 16. A quartz microbalance 36 is used to measure the weight of the particles within the cavity 16. A pressure drop sensor 38 is used to measure the pressure change within the cavity 16.

[0084] An electrochemical sensor array 40, disposed within the second conduit 20, comprises multiple electrochemical cells. Each electrochemical cell is used to measure the concentration of a specific gas or vapor in the aerosol stream. For example, the electrochemical sensor array 40 may include sensors for detecting the concentrations of substances such as volatile organic compounds (VOCs), oxygen, carbon monoxide (CO), carbon dioxide (CO2), and nicotine. An amplifier may be included to amplify the sensor output. In this example, the electrochemical sensor is shown located within the second conduit 20, downstream of the particulate filter 18, to reduce the risk of contamination. However, some or all of the electrochemical cells may be located in other locations, such as within the cavity 16 or at the exhaust port 44.

[0085] An impact sampler 42 may be positioned at the inlet of the suction engine 22 to collect particulate samples from the aerosol stream. Without the filter 18 installed, a particle size measuring device 46 measures the particle size in the exhaust stream from the suction engine 22. A spectroscopic sensor 48 is positioned at the outlet of the particle size measuring device 46. The spectroscopic sensor 48 is a device such as a non-dispersive infrared (NDIR) instrument or a mass spectrometer (MS). The spectroscopic sensor 48 is tuned for real-time quantitative detection of various chemical substances. The spectroscopic sensor 48 may use, for example, time-of-flight mass spectrometry (TOFMS), ion mobility spectrometry (IMS), quadrupole mass spectrometry, or any other applicable technique.

[0086] Figure 1 Two shut-off valves 26 and 28 are also shown. Shut-off valves 26 and 28 are in fluid communication with the analysis chamber 16. The first shut-off valve 26 is open to the atmosphere, and the second shut-off valve 28 is connected to the regulating gas source 24. Shut-off valves 26 and 28 can be actuated to allow purging of the analysis chamber 16 between suction and aspiration cycles.

[0087] It should be understood that Figure 1 The arrangement shown is merely exemplary and should not be considered a limitation on the type or arrangement of sensors or other components. For example, some sensors may be omitted, some sensors may be placed in other locations, or other sensors may be provided alternatively or additionally.

[0088] During aspiration and aerosol residence, data from optical sensor 30, temperature sensors 32, 33, and 49, humidity probe 34, quartz microbalance 36, pressure drop sensor 38, electrochemical sensor array 40, and impact sampler 42 are recorded and timestamped to ensure synchronization. After the aerosol is exhausted, data from particle size measurement device 46 and spectral sensor 48 are also recorded. This generates a set of sensor data arrays (electronic sensor mode), which represents the sensory experience experienced by the evaluation panel members. The intensity of the sensor response and the timing of the response are recorded (the response mode is not limited to a single reading). Aspiration control and data acquisition are handled by a computer system ( Figure 1 (Not shown in the image) Execute.

[0089] In this embodiment, the target is not subjected to specific analysis, or even high-precision analysis. Instead, a "pattern" of response is generated using a sensor array; therefore, relative sensitivity and the stability of these sensitivities are desired characteristics / results.

[0090] Table 2 summarizes the results of studies conducted to determine at what stage each sensor contributes to a sensory response that the evaluation panel members can assess.

[0091] Table 2

[0092] Sensor data will take on multiple forms, a mixture of the following: continuous data generated in the time domain (e.g., data from an electrochemical array), readings acquired at discrete time intervals throughout the pumping cycle (e.g., data from an optical sensor), or "spectrums" acquired at discrete time intervals that encompass multiple regions or areas of interest (e.g., data from a spectroscopic sensor or particle size measurement device). Therefore, the data will have both temporal and numerical elements at the schematic level.

[0093] Subsequently, data analysis was used to correlate sensor data with specific sensory experiences. The analytical problem to be solved was to predict unknown responses from new evaluation panel members using sensor data. To achieve this goal, the system was first calibrated based on the responses from the evaluation panel members.

[0094] For example, data from a sensory evaluation panel is collected for a range of representative products under different environmental conditions. In each test, panel members score a set of sensations using a 1-10 scale for each puff. Simultaneously, data is collected using an electronic sensory testing station for the same test products under identical conditions and at the same time. Based on this, a sensory prediction model can be developed.

[0095] Feature engineering is a prerequisite for training a prediction model. Two problems need to be addressed: Data misalignment. Sensor evaluation panel data and sensor data are not naturally correlated at a single line level, which means that there is a potential N×M correlation between the data, where N is the number of evaluation panel members and M is the number of times the given device is run on the sensory test bench under specific environmental conditions.

[0096] Representation of time series. Sensor data consists of a set of time-dependent variables, with a sampling rate of, for example, 1 second or 0.1 seconds. This means that 30 or 300 data points are collected for each sampling, while the evaluation panel members' response corresponds to 1 data point for each sampling.

[0097] This ensures that the responses of the evaluation panel members and the operation of the sensory testing station are independent of each other. Based on this, the data can be "flattened," representing the response of each evaluation panel member for each aspiration as the mean, median, and quantile, thus obtaining a single value for each device, condition, and aspiration.

[0098] Table 3 shows examples of how the evaluation panel members' responses can be flattened into the mean, median, and quantiles for each sample.

[0099] Table 3

[0100] Once sensor data is flattened over time, it generates a set of features for each device, condition, operation, and extraction (in this context, "features" refers to synthesized information such as mean, standard deviation, linearity, entropy, and curvature).

[0101] Table 4 shows examples of how sensor data can be flattened to correspond to a set of features for each device, condition, operation, and suction.

[0102] Table 4

[0103] Standards (known products that have sensory responses from an evaluation panel at range endpoints such as 1 or 9) are sets of information proven to effectively improve prediction quality and help models calibrate under new equipment or conditions.

[0104] A set of models can be tested in different scenarios. The above information is summarized in Table 5.

[0105] Table 5

[0106] All possible combinations of scenarios can be evaluated using the minimum absolute error (MAE) (for regression) and accuracy (for classification). MAE is the average of the absolute differences between predicted and true values. Accuracy is the ratio of correct predictions to all predictions. Furthermore, the expected inference capability under new equipment and conditions was also considered. The following results and conclusions can be drawn from this: To provide reliable forecasting results, it is advisable to have new equipment or standards corresponding to new conditions.

[0107] The classification model is accurate.

[0108] The aspiration results are an important predictor variable.

[0109] The linear mixed effects model has been found to be an applicable model, but its performance is not superior to random forests or decision trees.

[0110] The calibration process generally follows the same procedure as data acquisition, with the main difference being the use of multiple calibration pieces with known sensory properties. The sensor array responds to these test pieces and sets characteristic levels so that the generated "pattern" matches an ideal pattern to be stored in a database. Therefore, sensor-by-sensor calibration adjustments can be obtained by stretching or compressing this response to align it with the expected response.

[0111] The experiments conducted by the applicant demonstrate that meaningful results (in terms of predicting sensory responses) can be obtained using the following sensor data: Aerosol density / extinction Particle size distribution Data from at least one chemical sensor (organic / volatile) Aerosol relative humidity (RH) Aerosol temperature at two or more locations Temperature of smoking products at one or more locations pressure drop However, it is not necessary to have all of the above sensor data; other sensor data can be used or substituted.

[0112] It has been found that the above sensor data can be used to predict one or more of the following sensory responses: Impact sensation (sound of a blow to the chest - sudden and brief) Throat irritation (persistent burning or stinging sensation) A tingling or numb sensation in the mouth (which is also part of the stimulation). aerosol fullness The sensation of heat (the combined sensation of the temperature of the lip filter and the aerosol in the mouth). Cooling sensation in the mouth (menthol or similar sensation) Extraction resistance Dry mouth sensation (the drying effect in the mouth after exhaling aerosols) Therefore, in a preferred embodiment, at least a portion of the sensor data is used to predict one or more sensory responses (or their equivalents).

[0113] In situations requiring the use of multiple sensors, a series of chambers can be configured, each capable of simultaneously performing suction on the same product using the same suction parameter curve. This allows for the deployment of large sensor arrays without being constrained by space limitations in terms of the number of sensors used or their placement.

[0114] Figure 2 Components of a sensory testing device according to one embodiment of the present invention are shown. (Refer to...) Figure 2 The device includes multiple grippers 101-10 NMultiple analysis chambers 161 – 16 N Multiple suction engines 221-22 N The device includes a data recording unit 56, a suction control unit 58, a monitoring unit 60, a display unit 62, and a user interface 64. Each gripper has 101-10 grippers. N Arranged to maintain the corresponding aerosol-generating articles 121-12 N Each gripper has 101-10 grippers. N Corresponding analysis chamber 161 – 16 N Fluid connectivity. Analysis chamber 161 – 16 N It can be size-adjustable, however this feature is optional. Each analytical chamber 161 – 16N can be the same size or different from the rest of the analytical chambers. Each analytical chamber 161 – 16 N With corresponding suction engine 221-22 N Fluid communication. Clamp 101 – 10 N Analysis chamber 161 – 16 N And suction engine 221 – 22 N For example, the above reference can be used. Figure 1 The aforementioned format can also be set. Figure 1 Any other components shown (though some components may be omitted).

[0115] exist Figure 2 In the arrangement, the first sensor array 521-52 N With each analysis chamber 161 – 16 N Related. The sensors in each of the first sensor arrays may, for example, be disposed within the analytical chamber, or within a conduit leading to or from the analytical chamber. Second sensor arrays 541 – 54 N With each suction engine 221 – 22 N Associated. Each second sensor array has 541 – 54 N Located on the exhaust side of the corresponding suction engine. Each sensor array has 521-52 sensors. N 541 – 54 N It includes one or more sensors, such as those referred to above. Figure 1 One or more of the sensors described. For example, each of the first sensor arrays is 521-52. N It may include one or more of the following: optical sensors, temperature sensors, humidity probes, quartz microbalances, pressure drop sensors, electrochemical sensor arrays, and impact samplers. Each second sensor array has 541-54... N It may include one or more of the following: particle size measuring device and spectral sensor.

[0116] Each sensor array has 521-52 N 541 – 54 N It may include sensors different from other sensor arrays, or at least partially the same sensors, or any suitable combination thereof. One or more sensor arrays may be omitted if needed. Any other suitable sensors may also be added or replaced if required. For simplicity, Figure 2 Sensor arrays 521-52 are shown in the diagram. N 541 – 54 N It is associated with the analysis chamber and the suction engine. However, in practical applications, at least some of the sensors themselves can, for example, Figure 1 The device is located in a conduit between devices or at an exhaust port, or in another location within the system.

[0117] In a non-limiting embodiment, sensor arrays 521-52 N 541 – 54 N Includes multiple (two or more) temperature sensors, multiple (two or more) tin oxide-based chemical sensors, optical extinction sensors, pressure drop sensors, humidity sensors, particle size sensors, pH sensors and / or nicotine sensors.

[0118] During operation, each suction engine 221-22 N Used to generate articles 121-12 from each aerosol N A suction engine is used to generate an aerosol stream. In some arrangements, one suction engine can be used to suction two or more articles. Suction engine 221 – 22 N It is controlled by a suction control unit 58. The suction control unit 58 is programmed with suction parameter curves, which can be defined mathematically or can replicate the recorded shape of the human body during suction. The suction parameter curves corresponding to each suction engine may be the same or different.

[0119] From products 121-12 N The aerosol stream was drawn into analysis chamber 161-16 N Aerosols in cavity 161–16 N It remains inside for a set time before being discharged. Sensor array 521 – 52 N Each sensor in the array senses one or more parameters of the aerosol. The output of each sensor is sent to the data logging unit 56. After the aerosol is discharged, the sensor array 541-54... N Parameters used to sense emitted aerosols. From sensor array 541-54N The data was also transmitted to data recording unit 56. This generated a data array characterizing the sensory experiences of the evaluation team members. The intensity of the sensor response and the time of occurrence of the response were recorded.

[0120] Overall system control and data analysis are performed by monitoring unit 60. Data logging unit 56, suction control unit 58, and monitoring unit 60 can be implemented, for example, as software routines running on one or more processors. Analysis results are displayed on display 62. User interface 64 is used to receive input from the user and / or output data to the user. User interface may include, for example, a keyboard, mouse, touchscreen, display, etc.

[0121] Figure 3 The components of the monitoring unit 60 are shown in more detail. (Refer to...) Figure 3 The monitoring unit 60 includes a system control unit 66, a data flattening unit 68, a sensory response model 70, and a reporting unit 72. The monitoring unit 60 can be implemented as software running on a processor. It should be understood that other hardware and / or software devices may also exist, but these are not shown for simplicity. Furthermore, some of the units shown may be combined.

[0122] The system control unit 66 provides overall control of suction and data acquisition. It receives input from the user interface 64 and instructs the suction control unit 58 to execute an appropriate suction procedure. The system control unit 66 can also control other system variables, such as cavity volume, valve setpoints, and / or sensor setpoints.

[0123] The data flattening unit 68 receives sensor data from the data recording unit 56 and flattens the sensor data into a set of features for each aspiration of the test product. The flattened sensor data is then provided to the sensory response model 70.

[0124] Sensory response model 70 is a statistical model that correlates sensor data with a sensory profile. Sensory response model 70 is obtained by pre-mapping sensor evaluation panel data to sensor data from a sensor array. For this purpose, sensory response model 70 is trained using machine learning based on training data. This training is preferably performed using a set of "standard" products. The sensory evaluation panel's responses to these products are mapped to sensor data obtained using the same products under identical conditions.

[0125] In one embodiment, the sensory response model 70 is a tree model, such as a decision tree, random forest, or gradient boosting tree model like XGBoost. A decision tree is a hierarchical decision support model that employs a tree-like structure of decisions and their possible consequences. A random forest is a machine learning algorithm that combines the outputs of multiple decision trees to obtain a single result. Gradient boosting is a machine learning technique based on boosting methods in a function space. It yields a predictive model in the form of multiple weak predictive models (typically simple decision trees). Such tree-based models are known in the art and will not be described further.

[0126] In alternative embodiments, the sensory response model 70 may alternatively or additionally employ other machine learning methods, such as artificial neural networks (ANNs), support vector machines (SVMs), regression analysis, Bayesian networks, Gaussian processes, etc. Furthermore, in some embodiments, the sensory response model may be in the form of one or more mathematical formulas, such as polynomial equations provided by a linear mixed-effects model.

[0127] Multiple sensory response models can be provided when needed, each corresponding to a specific type of product and / or set of test variables. In this case, the system control unit 66 can select the appropriate sensory response model to use, for example, based on test conditions or user input.

[0128] During operation, the data flattening unit 68 receives sensor data from the data recording unit 56. The sensor data can be data from any of the aforementioned sensors. For example, in one embodiment, the sensor data may include at least one of the following: aerosol density / extinction; particle size distribution; chemical sensor data (organic matter / volatiles); aerosol residual humidity (RH); aerosol temperature at two or more locations; and pressure drop. The sensor data includes samples collected from each sensor at different time points (i.e., in the intensity and time domains). For each aspiration of the test product, the data flattening unit 68 flattens the sensor data into a set of features (such as mean, standard deviation, linearity, entropy, curvature, etc.). The flattened data is provided to the sensory response model 70.

[0129] Sensory response model 70 receives flattened sensor data as input and uses the sensor data to predict the evaluation panel members' responses to the test product. The output of sensory response model 70 is a set of sensory descriptors describing the predicted response of the evaluation panel members to the test product. For example, in one embodiment, the output of sensory response model 70 may include sensory descriptors such as: impact; throat irritation; oral tingling; aerosol filling; heat; oral coolness; extraction resistance; and / or oral dryness. The sensory descriptors are provided to reporting unit 72.

[0130] Reporting unit 72 receives the set of sensory descriptors and organizes them into a format that can be presented to the operator. The sensory descriptors are then presented to the user on display 62. The sensory descriptors can be presented to the user in various ways, such as as a series of heatmaps, radar charts, or pseudo-videos of data changes. The sensory descriptors can also be provided to communication unit 74 for transmission to another device and / or storage for later retrieval.

[0131] Figure 4 This illustration shows how sensory descriptors can be presented to a user in an exemplary embodiment. In this example, the sensory descriptors are presented in the form of a radar chart. The radar chart has five axes, each corresponding to a specific sensory descriptor (in this example, "aerosol filling," "throat stimulation," "impact," "heat," and "pulsation resistance"). Each sensory descriptor has a score between 0 and 5. Figure 4 The results of three different aspirations are shown. This allows for real-time visual comparison of different aspirations and different products.

[0132] It should be understood that Figure 4 The visualizations shown are merely examples; data can be presented in any suitable manner, such as tables, bar charts, heatmaps, videos, etc. Additionally, other sensory descriptive terms may be shown as supplementary or alternative methods.

[0133] Figure 5 The steps performed by the sensory testing equipment for testing a product are illustrated in one embodiment. (Refer to...) Figure 5 The test begins at step 100. The test can be initiated, for example, by a user entering a command to the system control unit 66 via the user interface 64. Alternatively, the test can be initiated automatically, or upon receiving a command from another device.

[0134] In step 102, the system control unit 66 selects the aspiration parameter curve to be used during the test. The aspiration parameter curve may include, for example, the volume of each aspiration and the shape of each aspiration (volume change over time). The aspiration parameter curve may be selected, for example, based on the type of test product or the type of sensory response to be predicted. The aspiration parameter curve may be a standardized volume and shape, or based on data collected from human aspiration shapes (e.g., the mouth or lungs or a single aspiration) or other arbitrary data. The selected aspiration parameter curve is output from the system control unit 66 to the aspiration control unit 58 for controlling the aspiration engine.

[0135] In step 104, the system control unit 66 selects the volume of the analytical chambers 161–16N to be used during the test. The chamber volume can be selected, for example, based on the type of the test product or the type of sensory response to be predicted. The chamber volume can be a standardized volume or based on human morphology or other arbitrary criteria. The selected volume is output to the analytical chambers 161–16N. N This is used to set the volume of each cavity.

[0136] In step 106, the system control unit 66 selects process variables to be used during the test. Process variables may include, for example, purge / non-purge between two aspirations, residence time within the chamber, and the total number of aspirations to be performed. Process variables can be selected based on any suitable criteria, such as the type of test product or the type of sensory response. Process variables are used to control the aspiration engine and / or other devices, such as shut-off valves, during the aspiration process.

[0137] In step 108, initial suction is performed using a suction engine (before the work-in-process is activated). Subsequently, in step 110, background clean air is established within the analytical chamber. This helps ensure that no residual aerosols or other potential contaminants are present at the start of the test.

[0138] In step 112, one or more of the tested articles are ignited or activated. For cigarettes or similar smoking articles, the article may be ignited manually or automatically. For example, the article may be ignited using an ignition system such as that disclosed in WO 2006 / 056747 (the subject matter of which is incorporated herein by reference) or any other suitable method. For heated tobacco articles, electronic cigarettes, or electronic cigarette devices, the article may be activated manually or automatically, depending on the article's design.

[0139] In step 114, the data recording unit 56 begins recording data from the sensor. In step 116, the test article is suctioned once using a suction engine. In step 118, it is determined whether the number of suctions NP performed is greater than or equal to the total number of suctions NT to be performed during the test. If the number of suctions performed is less than the total number of suctions to be performed, the process returns to step 116. If the number of suctions performed is greater than or equal to the total number of suctions to be performed, the process continues to step 120. In step 120, the data recording unit 56 stops recording data from the sensor.

[0140] In step 122, the sensor data is flattened into a set of features for each aspiration of the test article using the data flattening unit 68. In step 124, the sensory response of the test article is modeled using the sensory response model 70. The input to the sensory response model is the flattened sensor data from the data flattening unit 68. The output of the sensory response model is one or more sensory descriptors that predict the evaluation panel members' responses to the test article. In step 126, the reporting unit 72 uses the sensory descriptors to generate a report of the article's sensory profile in a standard format. In step 128, the sensory profile report is displayed on the display 62. The testing process then ends in step 130.

[0141] It should be understood that Figure 5 The steps shown are given as examples only and can be modified in many ways. For example, some steps may be omitted, and other steps not shown may also be present. In addition, the order of the steps may be changed as appropriate.

[0142] The above embodiments provide a method for measuring and describing sensory and tactile sensations associated with aerosol delivery to the human mouth, lungs, and respiratory tract via machine analysis. Specifically, a sensory association and measurement method is provided, deploying an array of multiple sensors that collectively form characteristic response patterns of the tested aerosol or flue gas. Patterns acquired from sensing elements deployed in the aerosol path can be used to correlate the evaluation results of a sensory evaluation panel using similar or identical artifacts, achieving a correlation between the analytical measurement results and human sensory experience through pattern matching. The method can utilize response patterns from the sensor array, which vary in both the time and intensity domains and are characteristic of the target sensory experience. The method can use non-targeted or non-specific sensors, which, in combination with other sensing elements, can be used to form characteristic response patterns of specific tactile or sensory judgments made by evaluation panel members. The method may involve the use of sensors that generate electronic signals. The method may involve the use of standard artifacts whose characteristics are known based on patterns obtained from a test chamber, ensuring that calibration is maintained and adjusted to obtain consistent results. The correlation between the measured response in the test chamber and the judgment results of the evaluation panel members of the control product can be used to construct a set of patterns that can be used to generate the same or similar judgment results as the evaluation panel members (predictive ability) when the product is no longer the selected control product. The method can use a chamber to hold a certain volume of aerosol for a set time to simulate the retention and exhalation process of the human body when smoking or using e-cigarettes. The method can use test chambers of different volumes, which can be matched to the volume of inhaled aerosol. The method can use a suction engine that can be programmed for any desired suction pattern and volume to simulate human suction parameter curves in inhalation and exhalation functions. The method can use multiple chambers and suction engines that operate synchronously, with different chambers equipped with different sensors. The data from these sensors are collected synchronously to form a single picture of the product aerosol. The method can perform sensing on the inhalation side and / or the exhalation side of the suction engine. The method can use sensing elements at the exhaust port or exhalation section of the suction engine. The method can use sensing elements that provide time-domain information. The method may include control over the rate and / or volume of exhalation. The aerosol being tested may be generated from tobacco products that can be heated, burned, or formed from a liquid or gel. The method may include storing discrete sensor data in different formats and converting it into pattern characteristics of sensory parameter curves for evaluation panel members, and matching the acquired data with an "example" pattern to generate sensory parameter curves equivalent to the evaluation panel's parameter curves via a device. A corresponding device configured to perform the method may also be provided.

[0143] The embodiments of the present invention have been described above by way of example, and various modifications are possible. For example, while the described embodiments are applicable to tobacco products, other embodiments may be applicable to electronic cigarette products and / or products for other purposes such as therapeutic applications that deliver aerosols to the human body. Features of one embodiment may be used in combination with features of any other embodiment. Various other modifications and variations in detail will be apparent to those skilled in the art within the scope of the appended claims.

Claims

1. An apparatus for analyzing aerosol release products, the apparatus comprising: Apparatus for extracting aerosols from a product; Multiple sensors are arranged to sense the properties of aerosols to generate sensor data; A means for predicting a sensory response from sensor data using a model that associates sensor data with at least one sensory response; as well as A device for outputting a predicted sensory response to a user.

2. The device according to claim 1, wherein, The model associates sensor data from multiple sensors with the sensory response or each sensory response.

3. The device according to claim 1 or 2, wherein, The aerosol-releasing product is of the type intended for delivery of aerosols to the human body.

4. The device according to any of the preceding claims, wherein, The aerosol-releasing products are products of the tobacco industry or the e-cigarette industry.

5. The device according to any of the preceding claims, wherein, The predicted sensory response is a prediction of the response that will occur in the human body when the aerosol is inhaled.

6. The device according to any of the preceding claims, wherein, The predicted sensory response includes a quantitative description of at least one sensation reported by the sensory evaluation panel.

7. The device according to any of the preceding claims, wherein, The sensory responses include at least one of the following: impact, throat stimulation, oral tingling, aerosol filling, heat, oral coolness, extraction resistance, and oral dryness.

8. The device according to any of the preceding claims, wherein, The multiple sensors include various different types of sensors.

9. The device according to any of the preceding claims, characterized in that, The plurality of sensors are arranged to sense at least one of the following: the physical properties of the aerosol; the chemical properties of the aerosol; and the size properties of the aerosol.

10. The device according to any of the preceding claims, characterized in that, The characteristics of the aerosol include at least one of the following: aerosol density and / or extinction; aerosol particle size distribution; presence and / or concentration of one or more chemical substances; aerosol temperature at one or more locations; aerosol humidity; and pressure and / or pressure drop.

11. The device according to any of the preceding claims, characterized in that, The plurality of sensors include at least one of the following: an aerosol density and / or extinction sensor; a particle size distribution sensor; an electrochemical sensor; a temperature sensor; a humidity sensor; and a pressure sensor.

12. The device according to any of the preceding claims, characterized in that, At least one of the plurality of sensors is arranged to sense the characteristics of the aerosol in the intensity domain and the time domain.

13. The device according to any of the preceding claims, characterized in that, The model includes a machine learning algorithm, and the machine learning algorithm is preferably trained in advance using sensor data and reported sensations from a sensory evaluation panel.

14. The device according to any of the preceding claims, characterized in that, The model includes at least a tree-based model and / or a linear mixed-effects model.

15. The device according to any of the preceding claims, characterized in that, The device includes means for flattening the sensor data to obtain a set of features for each sensor, and the model is arranged to predict at least one sensor response based on the flattened sensor data.

16. The device according to any of the preceding claims, characterized in that, Devices used to extract aerosols from a product include a suction engine.

17. The device according to any of the preceding claims, characterized in that, It also includes a cavity, wherein means for extracting aerosol from the article is arranged to extract the aerosol into the cavity, and at least one of the plurality of sensors is arranged to sense the aerosol characteristics within the cavity.

18. The device according to claim 17, characterized in that, The volume of the cavity is adjustable.

19. The device according to claim 17 or 18, characterized in that, The device includes multiple chambers and / or multiple suction engines.

20. The device according to any of the preceding claims, characterized in that, The device is arranged to extract aerosols from a plurality of aerosol release articles, and the plurality of sensors are arranged to sense at least one characteristic of the aerosols from each of the plurality of articles.

21. The device according to any of the preceding claims, characterized in that, The plurality of sensors includes at least one of the following: Multiple temperature sensors are arranged to sense the temperature of aerosols at multiple different locations; A humidity sensor is configured to sense the humidity of aerosols. A pressure sensor is configured to measure pressure changes during the suction process; A weighing device arranged to determine the mass of particle deposition; An optical sensor is arranged to measure particle scattering of the aerosol; At least one electrochemical cell arranged to sense one or more organic compounds and / or volatile compounds; At least one temperature sensor is arranged to sense the temperature of the workpiece being measured; Particle size measuring device; as well as Spectral sensor.

22. The device according to any of the preceding claims, wherein the plurality of sensors comprises at least one electrochemical cell arranged to sense the concentration of at least one of the following: volatile organic compounds, oxygen, ammonia, carbon monoxide, carbon dioxide, and nicotine.

23. The device according to any of the preceding claims, wherein the plurality of sensors comprises a plurality of electrochemical cells, each electrochemical cell being arranged to sense different substances from at least one of the remaining electrochemical cells.

24. A method for analyzing aerosol-releasing products using analytical equipment, the method comprising: Extracting aerosols from the aerosol-releasing product; Multiple sensors are used to sense the properties of the aerosol to generate sensor data; The sensory response is predicted from the sensor data using a model that associates sensor data with at least one sensory response; as well as Output the predicted sensory response to the user.

25. The method of claim 24, further comprising the step of: The model was trained using sensor data and reports of feelings from a sensory evaluation panel.

Images