ACOUSTIC DIAGNOSTIC OF OPERATION OF AN AEROSOL-GENERATING DEVICE

[0001]The present invention relates to an aerosol-generating device and a method of operating an aerosol-generating device.

[0002]Several types of aerosol-generating systems are currently commercially available. Such devices may heat aerosol-forming substrate to a temperature at which one or more components of the aerosol-forming substrate are volatilised without burning the aerosol-forming substrate. Other devices often referred to as nebulizers typically generate aerosol by mechanical action such as by a vibrating mesh module. Vibrating mesh nebulizers use mesh deformation or vibration to push a liquid through a mesh.

[0003]In a typical vibrating mesh nebulizer a piezo element, which is in contact with a mesh, is used to produce vibrations of the mesh. The mesh is adjacent and in direct contact with a liquid drug-containing substrate. The mesh deformation of the vibrating mesh generates a pressure field in the liquid, thus pumping and loading the holes with liquid. The liquid volume displaced through the holes breaks up into droplets, and are ejected into an aerosol-forming chamber. During inhalation, ambient air crosses the aerosol-forming chamber to transport the aerosol to the user.

[0004]Various causes could influence the aerosol nebulization throughput of an aerosol-generating device comprising a vibrating mesh. For example, non-optimal mechanical contact of the vibrating mesh with the exciting piezo elements or contact of the vibrating mesh with the cartridge holding the liquid to be volatilized, may cause changes of the resonance frequency of the membrane. Additional causes that may influence the aerosol nebulization throughput of an aerosol-generating device may include changes in the liquid viscosity, changes in transport material performance, aging of membrane, puffing airflow turbulences or the orientation of the device during aerosol generation. While changes of the resonance frequency of the membrane may be detected by monitoring impedance of the piezo-electric elements, most of the above causes of the variation of the aerosol nebulization throughput would not be detectable by measuring impedance, only.

[0005]Accordingly, it would be desirable to provide an aerosol-generating device that allows for detecting and identifying the causes for changes in aerosol nebulization throughput.

[0006]It would further be desirable to provide an aerosol-generating device that allows to generate and collect data regarding performance of the aerosol-generating device, which may be used during operation or in a subsequent analysis step to enhance the user experience.

[0007]According to an embodiment of the invention there is provided an aerosol-generating device comprising a controller for controlling operation of the aerosol-generating device, an acoustic source, which is connected to the controller and which is configured to generate an acoustic signal in use, an acoustic sensor, which is connected to the controller and which is configured to detect the acoustic signal generated by the acoustic source. The controller is configured to monitor an operational condition of the aerosol-generating device based on the detected acoustic signal.

[0008]The aerosol-generating device may comprise an aerosol-generating unit, configured to aerosolise an aerosol-forming substrate. The aerosol-generating unit may be configured to aerosolise an aerosol-forming substrate primarily by mechanical action. The aerosol-generating unit may comprise a vibrating element. The aerosol-generating unit may comprise a vibrating mesh.

[0009]As used herein the expression “vibrating mesh” refers to any vibrating element that may be used to generate an inhalable aerosol. The vibrating element may be formed from any kind of perforated membrane, which serves for aerosolisation purposes.

[0010]The vibrating mesh may be the acoustic source of the aerosol-generating device. The vibrating mesh may be used in the aerosol-generating device at the same time as part of the aerosol-generating unit and as the acoustic source. The vibrating mesh element may be operated in a large range of frequencies. The vibrating mesh element may be operated in the radio frequency range. The vibrating mesh element may be operated in a range from 20 kilohertz to 300 gigahertz. The vibrating mesh element may be operated in a range from 1 kilohertz to 1000 kilohertz. The vibrating mesh element may be operated in a range from 10 kilohertz to 500 kilohertz. The vibrating mesh element may be operated at a frequency of around 100 kilohertz.

[0011]During operation the vibrating mesh may generate an aerosol as described in the introductory portion of this specification. At the same time the vibrating mesh generates an acoustic signal corresponding to the mechanical vibration. Depending on the vibration frequency, the generated acoustic signal may be mainly in the ultrasound range. Such acoustic signals may be detected with a suitable acoustic sensor. In addition, the mechanical vibration of the vibrating mesh may generate sub-harmonics in the audible acoustic range of frequencies below 20 kilohertz.

[0012]During operation of the vibrating mesh the acoustic signal generated by the vibration of the mesh may be measured with a suitable acoustic sensor. The signature of the detected acoustic signal may be analyzed to verify the operating conditions of the device. In particular, the operating conditions of the vibrating mesh can be verified by an analysis of the acoustic signal generated by the vibrating mesh during generation of aerosol. In this way, it may be verified whether the device is operating in a normal or in a failure mode. Analysis of the acoustic signal may also allow to identify a depletion of the supply of aerosol-forming substrate. Such information may be used to notify a user of the operating condition of the aerosol-generating device. Such information may also be used to adjust the operating parameters aerosol-generating device. Thereby the overall user experience may be enhanced.

[0013]The aerosol-generating unit may comprise one or more electro-mechanical elements in mechanical contact with the vibrating element, wherein the electro-mechanical elements are controlled by the controller. The electro-mechanical element may be an annular piezo-mechanical element. The electro-mechanical element may comprise a plurality of piezo-mechanical elements. The electro-mechanical elements may be connected to the controller of the aerosol-generating device. The controller may be configured to control operation of the one or more electro-mechanical elements such that the vibrating mesh is excited to vibrate with a desired vibration frequency and with a desired amplitude.

[0014]The vibrating element may be integrally formed with the electro-mechanical, piezo element. For this purpose the vibrating element may be manufactured by making use of micro-electro-mechanical systems (MEMS) manufacturing techniques. Such manufacturing techniques include process technologies used in semiconductor device fabrication. These techniques include deposition of material layers, patterning by photolithography and etching of material in order to produce the required shapes. MEMS technologies allow precise manufacturing of vibrating mesh elements.

[0015]The acoustic sensor employed in the aerosol-generating device may be any suitable acoustic sensor known to the skilled person. The acoustic sensor may be a microphone. The acoustic sensor may be a MEMS microphone. A MEMS microphone may operate based on capacitive principle. A MEMS microphone is a micro-scale device that offers significant advantages. A MEMS microphone has a comparably high signal-to-noise ratio (SNR), low power consumption, good sensitivity and strong vibration resistance. In addition, MEMS microphones are small enough to be included in a tightly-integrated electronic product.

[0016]The acoustic sensor may be placed at any suitable position in the aerosol-generating device. The acoustic sensor may be placed in the airflow path defined in the aerosol-generating device. The acoustic sensor may be placed in the mouthpiece portion of the aerosol-generating device. Placing the acoustic sensor in the airflow path of the aerosol-generating device and placing the acoustic sensor close to the vibrating mesh may ensure good acoustic conditions for detecting the acoustic signal.

[0017]The acoustic sensor may be placed in contact with a surface of the aerosol-generating device. The acoustic sensor may be placed in contact with a surface of the housing of the aerosol-generating device. By placing the acoustic sensor in contact with a surface of the aerosol-generating device, acoustic signals transmitted through the material of the respective surface of the aerosol-generating device may be detected.

[0018]The acoustic sensor may also be placed distant from a surface of the aerosol-generating device. For example, the acoustic sensor may be placed hanging in open space within the aerosol-generating device. By placing the acoustic sensor remote from an internal surface of the aerosol-generating device, the acoustic sensor may be mechanically decoupled from the surfaces of the aerosol-generating device. In such configuration, the acoustic sensor may be particularly sensitive to acoustic signals transmitted from the vibrating mesh to the acoustic sensor via the air in the internal volume of the aerosol-generating device.

[0019]More than one acoustic sensors may be used for detecting the acoustic signal. Using a plurality of acoustic sensors may increase sensitivity of the acoustic detection. The acoustic sensors may be provided at multiple, different locations within the aerosol-generating device. The acoustic sensors may also be identical or may be of different types. Using different types of acoustic sensors and placing the acoustic sensors at different locations may allow to cover a larger dynamic range for detection of the acoustic signal. Using a plurality of acoustic sensors may also improve spatial resolution of the detection.

[0020]The acoustic signal may be transmitted to the one or more acoustic sensors via a rigid connector provided in contact with the vibrating element. The rigid connector may be in direct mechanical contact with both the vibrating element and an acoustic sensor in order to directly transmit the acoustic signal. This may further increase sensitivity of the acoustic detection.

[0021]The acoustic signal may also be transmitted via air contained in a tube. The tube may be a hollow flexible tube. One of the ends of the tube may be provided close or in contact with the vibrating element.

[0022]The acoustic sensor may be directly connected to an electronic board. This is particularly suitable if the acoustic sensor is a MEMS microphone. A MEMS microphone may be provided integral with electronic circuitry of an electronic board. Such system may allow for sufficient amplification and noise reduction and may therefore be beneficial for a subsequent signal processing.

[0023]The aerosol-generating device comprises electronic circuitry including a controller for processing the acoustic signal detected by the acoustic sensor. Processing and evaluation of the acoustic signal may include any suitable method for frequency analysis. Such methods may include frequency analysis by spectral analysis of the recorded acoustic signal. Such methods may also include amplitude analysis of the recorded acoustic signal. Spectral analysis may detect occurrence of specific frequencies such as sub-harmonics of the main vibration frequency of the vibrating mesh. Occurrence of such vibration frequencies may be indicative of mal-function or aging of the vibrating mesh. Low amplitudes of the acoustic signal may be indicative of increased damping of the vibrating mesh, which in turn may also indicate an The aerosol-generating device may further comprise a memory unit. The memory unit may be used to store previously recorded acoustic signals or signatures of previously recorded acoustic signals. In addition, or alternatively the memory unit may be used to store standard acoustic signals. Such standard acoustic signals may be calibrated acoustic signals. The standard acoustic signals are acoustic signals that are expected or desirable to be detected by the one or more acoustic sensors of the aerosol-generating device. In the memory unit a plurality of such standard acoustic signals may be stored.

[0024]The controller may be configured to evaluate the acoustic signal detected by the acoustic sensor, by comparing the detected signal to an expected acoustic signal stored in the memory unit.

[0025]Evaluation of the acoustic signal may be carried out using predefined diagnostic models. The classifier used in evaluation of the acoustic signal may be a pure statistical model. The classifier may be a simple thresholding algorithm. Evaluation of the acoustic signal may be carried out using a conventional machine learning model. Conventional machine learning models may include Gaussian Mixture Modelling (GMM) or Support Vector Machine (SVM) algorithms. Also, more recently developed models such as deep learning models may be used. The machine learning model may be an image-based algorithm using Convolutional Neural Network on acoustic spectral patterns.

[0026]Such predefined diagnostic models may be developed based on machine learning technologies. For this purpose, a model may be trained using a dataset of experimentally recorded acoustic signals. Such signals include desirable and undesirable acoustic signals. A portion of this dataset may be used as a training set to adjust the controller. Once the controller used is sufficiently adjusted, the model settings may be verified and validated by using further portion of the data set. The validation data set is used to further adjust parameters of the model and to repeat the training until a diagnostic model is obtained which performs well on the validation data set. A final test set may be used to finally evaluate performance of the diagnostic model.

[0027]With the diagnostic model it may not only be possible to distinguish between good and bad acoustic signals, but it may also be possible to identify the dominant cause of the malfunction of aerosol-generating device. Each cause of malfunction may lead to a distinct response detectable in the acoustic signal. For example, a change of the resonance frequency of the vibrating element may be indicative of a change in the clamping force on the membrane. If the device employs a replaceable cartridge, such change may be caused by the user during replacement of the cartridge. A low liquid flow onto the cartridge may also cause a change on the resulting vibrational state of the vibrating element. Such reduced liquid flow may be caused by a change in the viscosity of the liquid substrate or by a reduced transport material performance. The specific cause of the performance variation may be indicated to the user. If the liquid supply is depleted, the membrane becomes dry, which further affects its vibrational properties. By analysing the acoustic signal of the vibrating element, a depletion of the liquid supply may be identified. Thus, by using such predefined diagnostic models, which may be developed based on machine learning technologies, the aerosol-generating device may be trained to control and adjust its operation. The aerosol-generating device may thus be used to enhance overall user experience.

[0028]The aerosol-generating device may be configured to monitor an operational condition of the aerosol-generating device. Such operational conditions may include a puff detection, a characterization of airflow conditions, a dry mesh detection, a normal mode detection and a failure mode detection.

[0029]The controller of the aerosol-generating device may be configured to take at least one action, based on the result of the evaluation of the acoustic signal detected by the acoustic sensor. The controller may be configured to control the acoustic source based on the evaluation of the acoustic signal detected by the acoustic sensor. The controller may be configured to prevent aerosol generation by the aerosol-generating device, in case a malfunction is detected, which does not allow for reliable operation of the aerosol-generating device.

[0030]Monitoring the acoustic signal may be carried out during use of the aerosol-generating device, in particular during the nebulization process in which the aerosol is generated.

[0031]However, acoustic monitoring may not only be conducted during puffing, but may also be carried out between puffs. For example the vibrating element may be intentionally activated to monitor the general operational conditions of the device. In this way abnormal function of the device may be detected already in advance of a user experience.

[0032]The present invention further relates to an aerosol-generating system comprising the aerosol-generating device described herein and a cartridge comprising the aerosol-forming substrate. The cartridge may comprise the liquid storage portion.

[0033]As used herein, the term ‘aerosol-forming substrate’ may relate to a substrate capable of releasing volatile compounds that can form an aerosol or a vapor. The term ‘aerosol-forming substrate’ may also relate to a substrate capable to be mechanically nebulized. The aerosol-forming substrate may be in gel form or may be in liquid form. The terms ‘aerosol’ and ‘vapor’ are used synonymously.

[0034]The aerosol-forming substrate may comprise nicotine. The nicotine-containing aerosol-forming substrate may be a nicotine salt matrix.

[0035]The aerosol-forming substrate may comprise plant-based material. The aerosol-forming substrate may comprise tobacco.

[0036]The aerosol-generating device may comprise a housing. The housing may be elongate. The housing may comprise any suitable material or combination of materials. Examples of suitable materials include metals, alloys, plastics or composite materials containing one or more of those materials, or thermoplastics that are suitable for food or pharmaceutical applications, for example polypropylene, polyetheretherketone (PEEK) and polyethylene. Preferably, the material is light and non-brittle. The housing may include a user interface to activate the aerosol-generating device, for example a button to initiate heating of the aerosol-generating device or a display to indicate a state of the aerosol-generating device or of the aerosol-forming substrate.

[0037]The aerosol-generating device may comprise a power supply. The power supply may require recharging and may have a capacity that enables to store enough energy for one or more user experiences; for example, the power supply may have sufficient capacity to continuously generate aerosol for a period of around six minutes or for a period of a multiple of six minutes. In another example, the power supply may have sufficient capacity to provide a predetermined number of puffs. The aerosol-generating device may comprise a charging port for recharging the power supply.

[0038]The power supply may be a direct current (DC) power supply. In one embodiment, the power supply is a DC power supply having a DC supply voltage in the range of 2.5 Volts to 4.5 Volts and a DC supply current in the range of 1 Amp to 10 Amps (corresponding to a DC power supply in the range of 2.5 Watts to 45 Watts). The aerosol-generating device may advantageously comprise a direct current to alternating current (DC/AC) inverter for converting a DC current supplied by the DC power supply to an alternating current. The DC/AC converter may comprise a Class-D, Class-C or Class-E power amplifier. The AC power output of the DC/AC converter is supplied to the induction coil.

[0039]According to an embodiment of the invention there is provided a method of controlling an aerosol-generating device with a controller, an acoustic source, and an acoustic sensor. The method comprises controlling the operation of the aerosol-generating device by the controller, generating in use an acoustic signal with the acoustic source, detecting the acoustic signal generated by the acoustic source with the acoustic sensor, and monitoring by the controller an operational condition of the aerosol-generating device based on the detected acoustic signal.

[0040]Analysis of such acoustic signals offers the possibility to detect undesired operational conditions, which may result in uncontrolled variation of aerosol throughput. The proposed method may be used for development activities to enhance device control and handling. The method may also assist in manufacturing quality control and may offer integrated diagnostic tools in commercial products.

[0041]Evaluation of the acoustic signal detected by the acoustic sensor may be based on predefined diagnostic models. Such predefined diagnostic models may be developed based on machine learning technologies as described above.

[0042]The aerosol-generating device may comprise a memory unit. The method may include storing and reading data from the memory unit. For this purpose, the controller may be configured to carry out the evaluation of the acoustic signal detected by the acoustic sensor by comparing the detected signal to an expected acoustic signal signature stored in the memory unit.

[0043]The controller may further be configured for storing the detected acoustic signal in the microcontroller memory. Such stored acoustic signals may be used for subsequent data analysis. Such data analysis may include evaluation of operation of the aerosol-generating device. Data analysis may also be used to further develop data models that may be used for evaluation of the detected acoustic signal. The stored data sets may be used in machine learning context to develop evaluation strategies. The stored data sets may also be used in machine learning context as training data sets during training and testing the predefined diagnostic models for acoustic signal evaluation.

[0044]The method may include controlling the aerosol-generating device or more specifically controlling the acoustic source of the aerosol-generating device based on the evaluation of the acoustic signal detected by the acoustic sensor.

[0045]Further, evaluation of the generated acoustic signal may allow monitoring an operational condition of the aerosol-generating device. The monitored operational condition of the aerosol-generating device may include puff detection, characterization of airflow conditions, dry mesh detection, normal mode detection and failure mode detection.

[0046]Below, there is provided a non-exhaustive list of non-limiting examples. Any one or more of the features of these examples may be combined with any one or more features of another example, embodiment, or aspect described herein.

[0052]Example 2: The aerosol-generating device according to example 1, wherein the aerosol-generating device comprises an aerosol-generating unit, configured to aerosolise an aerosol-forming substrate.

[0053]Example 3: The aerosol-generating device according to example 2, wherein the aerosol-generating unit comprises a vibrating element, preferably a vibrating mesh.

[0054]Example 4: The aerosol-generating device according to example 2 or 3, wherein the aerosol-generating unit comprises electro-mechanical elements in mechanical contact with the vibrating element, wherein the electro-mechanical elements are controlled by the controller.

[0055]Example 5: The aerosol-generating device according to any one of the preceding examples, wherein the acoustic sensor is a microphone, preferably a MEMS microphone.

[0056]Example 6: The aerosol-generating device according to any one of the preceding examples, further comprising a mouthpiece portion, wherein the acoustic sensor is provided in the mouthpiece portion.

[0057]Example 7: The aerosol-generating device according to any one of the preceding examples, further comprising one or more additional acoustic sensors.

[0058]Example 8: The aerosol-generating device according to the preceding example, wherein the acoustic sensors are provided at different locations of the aerosol-generating device.

[0059]Example 9: The aerosol-generating device according to any one of the preceding examples, wherein the acoustic sensor is provided in contact with a surface of the aerosol-generating device.

[0060]Example 10: The aerosol-generating device according to any one of the preceding examples, wherein the acoustic signal is transmitted to the acoustic sensor via a rigid connector provided in contact with the vibrating element.

[0061]Example 11: The aerosol-generating device according to any one of the preceding examples, wherein the aerosol-generating device further comprises electronic circuitry for processing the acoustic signal detected by the acoustic sensor.

[0062]Example 12: The aerosol-generating device according to the preceding example, wherein the electronic circuitry is configured for amplification and noise reduction of the detected the acoustic signal.

[0063]Example 13: The aerosol-generating device according to the any one of examples 11 or 12, wherein the electronic circuitry is provided integral with the acoustic sensor.

[0064]Example 14: The aerosol-generating device according to any one of the preceding examples, wherein the aerosol-generating device comprises a memory unit, and wherein the controller is configured to evaluate the acoustic signal detected by the acoustic sensor, by comparing the signal to an expected acoustic signal signature stored in the memory unit.

[0065]Example 15: The aerosol-generating device according to any one of the preceding examples, wherein controller is configured to take at least one action, based on the result of the evaluation of the acoustic signal detected by the acoustic sensor.

[0066]Example 16: The aerosol-generating device according to any one of the preceding examples, wherein controller is configured to control the acoustic source based on the evaluation of the acoustic signal detected by the acoustic sensor.

[0067]Example 17: The aerosol-generating device according to any one of the preceding examples, wherein the monitoring an operational condition of the aerosol-generating device includes puff detection, characterization of airflow conditions, dry mesh detection, normal mode detection and failure mode detection.

[0073]Example 19: The method according to the preceding example, wherein the controller is configured to evaluate the acoustic signal detected by the acoustic sensor.

[0074]Example 20: The method according to any one of examples 18 to 19, wherein the controller is configured to evaluate the acoustic signal based on predefined diagnostic models.

[0075]Example 21: The method according to any one of examples 18 to 20, wherein the predefined diagnostic models are developed based on using machine learning technologies.

[0076]Example 22: The method according to any one of examples 18 to 21, wherein the aerosol-generating device comprises a memory unit, and wherein the controller is configured to evaluate the acoustic signal detected by the acoustic sensor by comparing the signal to an expected acoustic signal signature stored in the memory unit.

[0077]Example 23: The method according to any one of examples 18 to 22, wherein controller is configured to take at least one action, based on the result of the evaluation of the acoustic signal detected by the acoustic sensor.

[0078]Example 24: The method according to any one of examples 18 to 23, wherein controller is configured to control the acoustic source based on the evaluation of the acoustic signal detected by the acoustic sensor.

[0079]Example 25: The method according to any one of examples 18 to 24, wherein monitoring an operational condition of the aerosol-generating device includes puff detection, characterization of airflow conditions, dry mesh detection, normal mode detection and failure mode detection.

[0080]Example 26: The method according to any one of examples 18 to 25, wherein the controller is configured to store the detected acoustic signal in the microcontroller memory.

[0081]Example 27: The method according to the preceding example, wherein the stored acoustic signals are used for subsequent data analysis.

[0082]Features described in relation to one embodiment may equally be applied to other embodiments of the invention.

[0083]The invention will be further described, by way of example only, with reference to the accompanying drawings in which:

[0084]FIG. 1 shows a vibrating mesh nebulizer;

[0085]FIG. 2 shows a frequency spectrum of an acoustic;

[0086]FIG. 3 shows two acoustic signals of a vibrating mesh nebulizer;

[0087]FIG. 4 shows FFT spectra of the acoustic signals of FIG. 3;

[0088]FIG. 5 shows two acoustic signals of a vibrating mesh nebulizer;

[0089]FIG. 6 shows FFT spectra of the acoustic signals of FIG. 5;

[0090]FIG. 7 shows a scheme of training a controller to obtain a diagnostic model;

[0091]FIG. 8 shows an acoustic spectrogram of a regularly functioning vibrating element; and

[0092]FIG. 9 shows an acoustic spectrogram of an irregularly functioning vibrating element.

[0093]FIG. 1 shows a schematic illustration of an aerosol-generating device 10 comprising a housing 12 with a mouthpiece 14 at its proximal end. The aerosol-generating device 10 further comprises a liquid storage portion 16, a controller 18, a power supply 19, and a vibrating mesh element 20. The aerosol-generating device 10 may also be referred to as vibrating mesh nebulizer.

[0094]The vibrating mesh element 20 comprises a perforated, circular membrane 22 that is circumscribed by an annular, electro-mechanical piezo-element 24. An acoustic sensor in the form of a MEMS microphone 26 is provided within the mouthpiece 14. The power supply 19, the vibrating mesh element 20 and the MEMS microphone 26 are connected to the controller 18.

[0095]The perforated membrane 22 of the vibrating mesh element 20 is in contact with a liquid wicking material (not shown) which transports a liquid aerosol-forming substrate 17 from the liquid storage portion 16 to the vibrating mesh element 20. By use of the liquid wicking material, it is ensured that the perforated membrane 22 is always supplied with liquid aerosol-forming substrate 17.

[0096]In use the vibrating mesh element 20 is controlled by the controller 18. The controller 18 is configured to drive the piezo element 24 with an adjustable frequency signal such that the perforated membrane 22 is excited to vibrate.

[0097]The perforated membrane 22 is adjacent and in direct contact with the liquid aerosol-forming substrate 17. During vibration the perforations in the membrane 22 are loaded with the liquid substrate 17. The liquid substrate 17 is then displaced through the perforations and is released as droplets into an aerosol-forming chamber 28 of the mouthpiece 14. Accordingly, the vibrating mesh element 20 serves to aerosolize the liquid substrate 17. The vibrating mesh element 20 also generates an acoustic signal during vibration. Thus, the vibrating mesh element 20 also acts as an acoustic source of the aerosol-generating device 10.

[0098]During a user experience an airflow is generated from the air inlets 30 of the mouthpiece 14 via aerosol-forming chamber 28 adjacent the vibrating mesh element 20 towards the outlet 32. The aerosol droplets generated by the vibrating mesh element 20 are entrained in the airflow and are then inhaled by the user.

[0099]The MEMS microphone 26 is used during operation of the aerosol-generating device 10 to detect the acoustic signal generated by the vibrating mesh element 20. The detected acoustic signal is evaluated by the controller 18 and is used to verify an operational condition of the aerosol-generating device 10. In the event that the controller 18 detects an anomalous or undesired operational condition, the controller 18 is configured to take an appropriate action or to notify the user.

[0100]FIG. 2 shows a FFT frequency spectrum of a vibrational signal, wherein the x-axis represents the frequency and the y-axis shows the magnitude of the respective frequency signal in arbitrary units. The main peak in the centre of the spectrum relates to the main frequency signal which in this case corresponds to a frequency of about 80 kilohertz. In addition to the main frequency peak additional peaks are visible. These additional peaks relate to higher harmonics and subharmonics of the main frequency. The higher harmonics may occur at multiples of the main frequency. The subharmonics occur at frequencies obtained from integer division of the main frequency. Such frequency spectrum may be evaluated in order to derive information on the quality of the original vibrational signal. In an aerosol-generating device, the evaluation of the acoustic signal emitted from the vibrating mesh element allows to derive information on the vibration of the mesh element and therefore on the operational condition of the aerosol-generating device.

[0101]FIG. 3 shows acoustic signals VM_01 and VM_02 as recorded with two different devices over a period of 25 seconds. During these 25 seconds three puffs are taken with each aerosol-generating device. Each puff lasts for approximately 3 seconds. During the puffs the vibrating mesh element is activated, such that the amplitude of the recorded acoustic signal increases significantly. In between the puffs, activation of the vibrating mesh element is stopped, such that no acoustic signal is detectable between puffs. The quality of the signals is very similar, which allows to determine that the operational conditions of these aerosol-generating devices are similar.

[0102]The operational conditions of these aerosol-generating devices can be further evaluated by analysing the recorded acoustic signal using a FFT spectral analysis. The FFT spectrum of the acoustic signals of FIG. 3 is depicted in FIG. 4. As can be seen from the large scale FFT spectrum of FIG. 4, the acoustic signal has a main frequency of about 65 kilohertz. Some subharmonics are visible at the ½ and ⅓ of the main frequency. As can be seen from the enlarged view of the main FFT peak at about 65 kilohertz, the main FFT peak has a clearly defined single frequency with exponentially decaying shoulders. From these spectra it can be concluded that the vibrating mesh oscillates freely and that the operational conditions of the aerosol-generating device are nominal.

[0103]FIG. 5 shows a comparison of acoustic signals, which have been recorded under similar conditions as the acoustic signals depicted in FIG. 3. The signal VM_01 is identical to the corresponding signal of FIG. 3. The other acoustic signal VM_03 is recorded under dry mesh conditions, which means that the liquid storage portion in that case was depleted.

[0104]The acoustic signal VM_03 as recorded under dry mesh conditions is clearly distinguished from the acoustic signal VM_01 as recorded under normal conditions. Oscillation of the “dry mesh” signal starts and ends with a more intense overshoot than the in the acoustic signal VM_01. Furthermore, oscillation during the puff is less regular and there is even an acoustic signal detectable in between puffs.

[0105]This indistinct frequency response is also detectable by a comparison of the corresponding FFT spectra of the acoustic signals as depicted in FIG. 6. While signal VM_01 shows a clearly defined single frequency with exponentially decaying shoulders, the FFT spectrum of the dry mesh signal VM_03 shows at least two clearly distinguishable differences. Firstly, the main peak is shifted considerably towards higher frequencies and is now located exactly at 66 kilohertz. Secondly, additional peaks appear which are indicative of a diluted and somewhat indistinct frequency response of the vibrating mesh element.

[0106]Given these clearly distinguishable differences, the controller may be configured to interpret occurrence of such changes in a FFT spectrum as a depletion of the liquid supply portion. Under such operational conditions, further activation of the vibrating mesh element may be prevented. In addition, the user may be informed of the need to replace or replenish the liquid supply portion of the aerosol-generating device.

[0107]Evaluation of the acoustic signal may be carried out using a predefined diagnostic model. Such predefined diagnostic model may be developed based on machine learning technologies. For this purpose, a controller may be trained using a dataset 40 of experimentally recorded acoustic signals, as depicted in FIG. 7.

[0108]The dataset 40 of experimentally recorded acoustic signals is randomly divided into a training set 42, a validation set 44 and a test set 46. In the model of FIG. 7 the training set 42 comprises 60 percent of the entries of the initial dataset 40. The validation set 44 and the test set 46 each comprise 20 percent of the entries of the initial dataset 40.

[0109]The data of the training set 42 is used to train and establish a raw diagnostic model 48. The parameters are adjusted until the raw model 48 can correctly process the data of the training set 42. In a next step the raw model 48 is verified and validated by using the validation set 44. The validation set 44 is used to further adjust parameters of the controller and to repeat the training until a final diagnostic model 50 is obtained which performs well also on the data of the validation set 44.

[0110]In a last step the final model 50 is again checked using the data of the test set 46. This test set 46 may be used to finally evaluate performance of the final diagnostic model 50.

[0111]The diagnostic model 50 may allow the controller to verify the operational condition of an aerosol-generating device. In case of malfunction, the diagnostic model 50 may allow to identify the dominant cause thereof. The diagnostic model 50 may be used to detect a puff, to characterize airflow conditions, to detect a dry mesh, or to generally distinguish between normal or failure mode operation.

[0112]FIG. 8 shows an exemplary acoustic spectrogram as recorded with the acoustic sensor of a vibrating mesh nebulizer. The acoustic spectrogram shows the amplitude and the frequency of an acoustic signal generated by a vibrating element and recorded with an acoustic sensor over time. The acoustic spectrogram shows a sequence of three pulses of nebulization. The vibrating mesh nebulizer used for recording the spectrogram of FIG. 8 was used under regular operating conditions. In particular the vibrating mesh element used therein was regularly functioning. The signals recorded during the nebulization pulses, show pronounced maxima of the amplitude at about 66 kilohertz and a few less pronounced local maxima at lower frequencies between 10 and 50 kilohertz.

[0113]FIG. 9 shows a further acoustic spectrogram which is similar to the acoustic spectrogram of FIG. 8, but which has been recorded with an irregularly working vibrating mesh element. Particularly, the vibrating mesh used for recording the acoustic spectrogram of FIG. 9 produced an abnormally low nebulization. Such abnormally low nebulization may occur when the liquid storage portion is depleted. The spectrogram of FIG. 9 shows significant differences to the spectrogram of FIG. 8. The high frequency peaks at about 66 kilohertz are still visible. However, the peaks are less pronounced as compared with the peaks of the spectrogram of FIG. 8. In addition, the local maxima at lower frequencies are hardly recognizable anymore and the general amplitude level throughout all frequencies is increased. This diluted and somewhat indistinct frequency response of the vibrating mesh element again is indicative for abnormal operational conditions of the vibrating mesh.

[0114]Acoustic spectrograms as depicted in FIGS. 8 and 9 may be used in image-based machine learning algorithms as described herein. Spectral patterns of such spectrograms may be classified according to their dominant features such as the presence of pronounced frequency peaks, or the overall noise level recorded. Such machine learning algorithms may preferably use Convolutional Neural Network models. Such Convolutional Neural Network models are highly suitable to be to applied to image and video recognition, and in particular to image classification as required to evaluate acoustic spectrograms recorded with aerosol-generating devices as described herein.