AI Sensing Device for Wide-Range Gas and Vapor Detection

Abstract

A method and sensor device for determining the presence and concentration of each of a plurality of gases and / or volatile organic compounds, and a computerized method for training the sensor device.

Description

[Technical Field]

[0001] The present invention broadly relates to computerized methods and sensor devices for determining the presence and concentration of each of a plurality of gases and / or volatile organic compounds, and methods for training the sensor devices. [Background technology]

[0002] Any reference and / or description of prior art throughout this specification should not be taken in any way as an acknowledgement that this prior art is well known or forms part of the common general knowledge in the art.

[0003] There is a wealth of literature on the use of machine learning techniques, also known as artificial intelligence (AI), for gas and vapor sensing devices, sometimes called modern electronic noses (eNoses). For example, the article Towards a Chemiresistive Sensor-Integrated Electronic Nose: A Review, available at doi.org / 10.3390 / s131014214, provides a list of all commercially available eNose devices found in 2013.

[0004] Existing studies / devices either use highly selective materials that only react to certain types of gases / volatile organic compounds (VOCs) or perform chemical modifications, which increases the complexity and cost of such sensors.

[0005] Additionally, in existing research / products, the chemical fingerprint of each gas or VOC is derived from a unique array-like pattern formed in a sensor array by the combination of electrical responses obtained from individual sensing elements, each engineered to have different sensitivities, cross-sensitivities, and specificities, upon exposure to an analyte. For example, a chemical fingerprint may be the unique combination of variations in average current from an array of nine sensors arranged in a 3-by-3 format, where the three cells in the upper right corner are always higher than the three lower cells in the lower left corner when exposed to a particular type of analyte, but not other analytes. This also increases the complexity and cost of such sensors.

[0006] The method disclosed in "Selective Gas Sensing with a Single Pristine Graphene Transistor," Sergey Rumyantsev, Guanxiong Liu, Michael S. Shur, Radislav A. Potyrailo, and Alexander A. Balandin, Nano Lett. 2012, 12, 5, pp. 2294-2298, Publication Date: April 16, 2012, allows the use of a single 2DM-based chemiresistor to measure multiple sources of gases / vapors, but it cannot be used when two or more gas / VOC types are simultaneously present in the same environment (i.e., simultaneous fluctuations of two or more gas / vapor types). This is because, in this case, multiple unique Lorentzian frequencies are generated, each representing a specific gas / vapor type. As defined by the authors, a unique Lorentzian frequency is the frequency with the maximum power spectral density seen in a plot of the noise spectral density multiplied by frequency (i.e., the Lorentzian noise spectrum) versus frequency, and therefore, a user cannot identify all of them from the FFT spectrum analyzer used in the book. In this way, only one unique frequency is extracted from this multiple gas / VOC measurement, and the human user is unable to decipher the remaining information still hidden in the collected data.

[0007] Furthermore, the methods disclosed in the above-mentioned papers still rely on the overall incremental change in the conductivity of the chemiresistor when exposed to multiple gases / VOCs in the environment to measure concentration levels, and therefore are unable to identify the percentage change contributed by each individual gas / VOC to the overall incremental change detected in the conductivity of the chemiresistor.

[0008] Furthermore, the input features used for machine learning in existing studies / devices do not include the individual electrical responses from multiple sensors positioned closely together and exposed to the same analyte, which increases the complexity of such sensors.

[0009] SUMMARY OF THE INVENTION Embodiments of the present invention seek to address one or more of the above-mentioned needs. Summary of the Invention

[0010] According to a first aspect of the present invention, there is provided a computerized method for determining the presence and concentration of each of a plurality of gases and / or volatile organic compounds, the method comprising the steps of exposing one or more sensing elements having identical chemical and physical properties to the plurality of gases and / or volatile organic compounds, measuring electrical time series data of the one or more sensing elements during said exposure, analyzing the electrical time series data and Lorentzian noise information of the electrical time series data by an artificial intelligence (AI) system, and determining the presence and / or concentration of each of the plurality of gases and / or volatile organic compounds based on the analysis of the electrical time series data and Lorentzian noise information of the electrical time series data.

[0011] According to a second aspect of the present invention, there is provided a sensor device capable of determining the presence and concentration of each of a plurality of gases and / or volatile organic compounds, the sensor device comprising one or more sensing elements made from substantially identical sensing materials and an artificial intelligence (AI) system, wherein the AI ​​system is configured to analyze electrical time series data of the one or more sensing elements and Lorentzian noise information of the electrical time series data, and to determine the presence and / or concentration of each of the plurality of gases and / or volatile organic compounds based on the analysis of the electrical time series data and the Lorentzian noise information of the electrical time series data.

[0012] According to a third aspect of the present invention there is provided a method of training a sensor device of the second aspect so that it is capable of determining the presence and / or concentration of each of a plurality of gases and / or volatile organic compounds. [Brief explanation of the drawings]

[0013] The invention will be better understood by reference to the detailed description when considered in conjunction with the non-limiting examples and the accompanying drawings, in which:

[0014] [Figure 1] 1 shows a diagram illustrating a fabrication process for black phosphorus (bP), a two-dimensional material (2DM) chemiresistor, according to an exemplary embodiment.

[0015] [Figure 2] 1 illustrates the output characteristics (drain current ID vs. drain voltage VD) of a chemiresistor according to an exemplary embodiment.

[0016] [Figure 3] 3 illustrates an experimental setup of a wireless all-in-one gas sensor node 300 according to an exemplary embodiment for relative humidity (RH) measurement, according to an exemplary embodiment.

[0017] [Figure 4]FIG. 1 illustrates a block diagram of a sensor node in accordance with an illustrative embodiment.

[0018] [Figure 5] FIG. 1 shows a flow diagram illustrating the data collection process according to a preferred embodiment when training a single 2DM-based chemiresistor to classify and quantify multiple types of gases / vapors present in the same environment.

[0019] [Figure 6] 1 shows the statistics (i.e., mean, median, and standard deviation) of 68 samples collected over the first four days according to an exemplary embodiment.

[0020] [Figure 7] 10 illustrates the variation of measured current at similar RH according to an exemplary embodiment.

[0021] [Figure 8] 1 illustrates an experimental setup for training and testing an all-in-one gas sensor according to an exemplary embodiment for CO2, N2O, and RH.

[0022] [Figure 9] 1 illustrates an experimental setup for training and testing an all-in-one gas sensor according to an exemplary embodiment with no plants, plants without VOC emissions, and plants with VOC emissions.

[0023] [Figure 10] Illustrates the process of training a single 2DM-based chemiresistor via machine learning (ML) to classify and quantify multiple sources of gases / VOCs simultaneously present in the same environment, according to an exemplary embodiment.

[0024] [Figure 11] FIG. 1 shows a signal flow diagram illustrating the prediction of multiple gas / VOC sensing according to an exemplary embodiment.

[0025] [Figure 12] Briefly described is a system implementation block diagram of an all-in-one gas sensor according to an exemplary embodiment as an air quality sensor in a standalone consumer product for tracking health conditions, or as a climate monitoring sensing node in a wireless sensor network for production control in factory or industrial safety applications.

[0026] [Figure 13] Characteristic Lorentz frequencies extracted from three different ambient environments are shown, which can be seen as a first proof of principle of using characteristic Lorentz frequencies to distinguish multiple gases from a single 2DM-based chemiresistor, according to an exemplary embodiment.

[0027] [Figure 14] 1 shows characteristic Lorentzian frequencies extracted from three other different ambient environments, another ambient type 0 where there are no plants in the container, another ambient type 1 where there are plants in the container that are known not to emit any VOCs, and finally another ambient type 2 where there are plants in the container that are known to emit strong VOCs, according to an exemplary embodiment.

[0028] [Figure 15] 10 illustrates the predicted accuracy of an all-in-one sensor according to an exemplary embodiment for RH.

[0029] [Figure 16] 10 illustrates the predicted accuracy of an all-in-one sensor according to an exemplary embodiment for CO2 gas.

[0030] [Figure 17] 10 illustrates the predicted accuracy of an all-in-one sensor according to an example embodiment for N2O gas.

[0031] [Figure 18] We summarize the classification and regression results obtained from two different 2DM-based sensors according to an exemplary embodiment.

[0032] [Figure 19] 1 shows a flowchart illustrating a computerized method for determining the presence and concentration of each of a plurality of gases and / or volatile organic compounds, according to an exemplary embodiment.

[0033] [Figure 20] FIG. 1 shows a schematic diagram illustrating a sensor device capable of determining the presence and concentration of each of a plurality of gases and / or volatile organic compounds, according to an exemplary embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0034] Embodiments of the present invention provide an all-in-one smart sensor device that can simultaneously detect the presence and quantify the concentrations of multiple types of gases and VOCs using only a single physical sensor or an array of sensors with identical chemical and physical properties wirelessly. The all-in-one sensor device uses a two-dimensional layer material that is highly sensitive to various gases / VOCs and operationalizes its selectivity for different gases or VOCs through respective AI models (i.e., AI engines) developed by exhaustive machine learning (ML) in real-world settings. By providing software control over the selectivity of the sensor device according to exemplary embodiments, new sensing capabilities can be added on demand in real time without any changes to the physical hardware, which is not currently available in any conventional sensor system.

[0035] In general, wireless all-in-one smart sensors according to exemplary embodiments can be used in Internet of Things (IoT) platforms to track health, product, and manufacturing quality. Embodiments of the present invention have broad applications in environmental monitoring and emissions control, personal and military safety, agriculture, manufacturing, and medical diagnostics. Embodiments of the present invention can also be employed in heating, ventilation, and air conditioning (HVAC) systems to reduce energy consumption through better control of heating and air conditioning.

[0036] The sensing materials used by the exemplary embodiments include two-dimensional layer inorganic materials (2DMs) that, by their physical properties, are highly sensitive (i.e., non-selective) to a wide range of gases and VOCs, whereas existing research / devices either use highly selective materials that only respond to certain types of gases / VOCs, or perform chemical modifications or functionalizations on the sensing materials to create specificity for particular analytes.

[0037] Selectivity according to exemplary embodiments: Instead of chemically or physically functionalizing the sensing material as in existing studies / devices, gas / VOC type differentiation is achieved using machine learning (ML) techniques.

[0038] Chemical fingerprints of various gases or VOCs according to exemplary embodiments are derived from unique patterns detected using machine learning algorithms on information gathered from the low frequency noise spectrum of a single chemiresistor type via an ML training and deployment methodology.

[0039] Embodiments of the present invention advantageously extract information previously inaccessible from FFT spectrum analyzers used in existing research / devices via ML training and deployment methodologies.

[0040] Embodiments of the present invention use Lorentz frequencies associated with the kinetics of molecular adsorption and desorption as a result of exposure to vapors, which correspond to characteristic frequencies that are much lower compared to the Lorentz frequencies associated with charge traps created as a result of exposure to vapors as in existing studies / devices.

[0041] The input features used for machine learning according to the exemplary embodiments do not include individual electrical responses from each of multiple sensors positioned in close proximity and exposed to the same analyte. Instead, the input features used according to the exemplary embodiments include: Time-dependent electrical response (i.e., channel resistance) from a single type of 2DM-based chemiresistor, its derivatives, such as characteristic Lorentzian frequencies (i.e., peaks with maximum power spectral density) and their respective power spectral densities, from a plot of the noise spectral density multiplied by frequency (i.e., Lorentzian noise spectrum) versus frequency; the characteristic Lorentzian desorption time (i.e., the reciprocal of the characteristic frequency of the Lorentzian), Kurtosis, skewness, median of Lorentz noise spectrum The power spectral density ratio of the characteristic Lorentzian frequency to the median of the Lorentzian noise spectrum, the number of Lorentzian peaks found in the Lorentzian noise spectrum, the medians of these peaks, and the frequencies associated with these peaks; the full width at half maximum and full width at full maximum of the characteristic Lorentzian peak (i.e., the peak with the largest power spectral density), and Ambient temperature Includes information such as:

[0042] It should be noted that other operating conditions, such as runtime, may be used in different embodiments, and that operating conditions, such as ambient temperature, may not be used as input features in some exemplary embodiments, for example, if the sensor device is configured to be used only under specific operating conditions, such as room temperature.

[0043] The general physical mechanism of selective gas sensing using only a single 2D material as a chemiresistor according to exemplary embodiments is based on the recognition resulting from a systematic study that some gases change the electrical resistance of 2D devices, such as graphene, without changing their low-frequency noise spectrum, while other gases modify the noise spectrum by introducing a Lorentzian component with unique characteristics ["Selective Gas Sensing with a Single Pristine Graphene Transistor, Sergey Rumyantsev, Guanxiong Liu, Michael S. Shur, Radislav A. Potyrailo and Alexander A. Balandin", Nano Lett. 2012, 12, 5, 2294-2298 Publication Date: April 16, 2012]. According to the study, there are two reasons for the appearance of Lorentzian noise in graphene under gas exposure. First, gas molecules can create specific traps and scattering centers in graphene, leading to either fluctuations in the number of carriers due to fluctuations in trap occupancy or fluctuations in mobility due to fluctuations in scattering cross-section. Second, the kinetics of molecular adsorption and desorption upon exposure can also contribute to noise. The characteristic timescale for vapor adsorption and desorption has been reported to be hundreds of seconds. While the method disclosed in that study may enable the use of a single chemiresistor to measure multiple sources of gases / VOCs, it cannot be used for quantitative analysis when two or more gas / VOC types are present simultaneously in the same environment (e.g., simultaneous fluctuations of NO, CO, and relative humidity in the same environment). For example, if only ethanol is introduced into an environment, and then the FFT spectrum is analyzed to identify the Lorentzian frequency f at 399 Hz, CEthanol can be determined by observing the resistance of the channel measured from the multimeter, and the concentration of ethanol can be obtained by calculating dR / R (R = channel resistance measured from the multimeter, and dR = change in magnitude before and after exposure). However, if both ethanol and methanol are present in the environment, their respective concentration levels cannot be determined from the measured dR / R.

[0044] Instead, embodiments of the present invention advantageously provide a multi-gas / vapor sensor that uses a single sensing material configured to determine not only the types of gas / vapor species present, but also their respective concentration levels.

[0045] Fabrication of a Chemiresistor for a Sensor Device According to Exemplary Embodiments

[0046] FIG. 1 shows a diagram illustrating the fabrication process of a two-dimensional (2DM) black phosphorus (bP) chemiresistor for sensing relative humidity, according to an exemplary embodiment. A photoresist layer 100 is deposited on a substrate (here, a glass substrate 102), and a standard lift-off process is used to provide patterns 104, 106 for two electrical terminals 108, 110 (here, Au / Ni (30 nm / 1 nm)). In this embodiment, an AJA ATC-2200 UHV Sputter was used for the deposition of the metal 112, and a laser writer LW405B was used for patterning the photoresist layer 100. After lift-off of the metal 112, leaving the terminals 108, 110 on the substrate 102, the electrical channel of the chemiresistor 116 (here, an exfoliated bP piece 114) is then deposited across the two terminals 108, 110 by a dry transfer technique using polydimethylsiloxane (PDMS), in this exemplary embodiment. In a non-limiting example, the chemiresistor channel length and width is approximately 1 mm x 0.25 mm and its Raman spectrum 118 is -1 , 440cm -1 , and 467 cm -1 It shows the typical signature of black phosphorus.

[0047] Output characteristics of Chemiresistor 116 (drain current I D vs. drain voltage V D ) is shown in Figure 2. Because there is no significant hysteresis loop in the current measured in the double voltage sweep measurement, there are not a significant number of hidden traps that inhibit the current flow in the chemiresistor 116, which allows the bP chemiresistor 116 according to exemplary embodiments to operate in a low voltage range. It should be noted that for large-scale operation, a commercial printing process is preferably used to prepare chemiresistors according to various exemplary embodiments instead of dry transfer techniques using polydimethylsiloxane (PDMS).

[0048] Functional Description of Wireless Sensors According to Exemplary Embodiments

[0049] FIG. 3 illustrates an experimental setup of a wireless all-in-one gas sensor node 300 according to an exemplary embodiment for relative humidity (RH) measurement. In this exemplary embodiment, the sensor node 300 includes a Bluetooth-enabled tablet 302 for wirelessly receiving measured voltage across a chemiresistor 304 from a Bluetooth low energy microcontroller unit (TI CC2541) on a PCB 305. Relative humidity is controlled by a silica gel-filled desiccator 306, and a reference RH sensor 308 provides reference data used in machine learning according to the exemplary embodiment. Detailed functional circuits on the PCB 305, including an analog front-end unit (AFE), an analog-to-digital converter (ADC), a microcontroller unit (MCU), and a battery booster (here, a TI TPS61220 & LM4120), are described with reference to the block diagram shown in FIG. 4.

[0050] As can be seen from the diagram in Figure 4, a chemiresistor 400 is connected across the control electrode (CE) and working electrode (WE) terminals of an AFE 402 (here, a TI LM91000). This bias voltage for the chemiresistor 400 is supplied by a 3V coin-cell battery 404 (here, a CR-2032) and is preset to a minimum value of 25mV. A battery booster unit 406 regulates this battery power and provides a stable reference voltage of 2.5V to the control amplifier (A1) 407, transimpedance amplifier (TIA) 409, and analog-to-digital converter (ADC) found inside the microcontroller unit and radio frequency system-on-chip (here, an MCU+RF SoC CC2541) 412. To ensure that a constant 25mV is always maintained across the CE and WE, the reference electrode (RE) terminal of the AFE 402 is shorted to the CE. Under this configuration, AFE 402 acts like a potentiostat, and any change in the resistance of chemiresistor 400 is reflected in a current flowing through terminals 408, 410, which is converted to a voltage by TIA 409. The ADC in MCU 412 then converts this analog voltage to a digital signal, and Bluetooth transmitter 414 transmits this digital information to the receiver (Bluetooth-enabled tablet 302 in FIG. 3) periodically at a preset interval, e.g., 0.3 seconds (note: minimum available preset interval = 10 milliseconds). Once the digital voltage signal is received by tablet 302 (FIG. 3), a software application in tablet 302 (FIG. 3) converts the digital voltage signal back to its original analog current value based on the settings in MCU 412 and the ADC in TIA 409.

[0051] Data Collection for Machine Learning According to Exemplary Embodiments

[0052] FIG. 5 shows a flow diagram illustrating the data collection process according to a preferred embodiment when training a single 2DM-based chemiresistor to classify and quantify multiple types of gases / vapors present in the same environment. As shown in step 1, a single 2DM-based chemiresistor is exposed to a desired number N and types of gases / VOCs in a controlled manner in a closed environment to determine which gases / VOCs to measure and the range of concentration levels to adjust. As shown in step 2, the concentration of one gas / VOC is systematically varied and the corresponding channel resistance of the chemiresistor is measured. Preferably, a dataset of measurements at different concentration levels is collected, ensuring that the measurement points are evenly distributed within the determined adjustment range. This dataset is classified as ambient type 0 (increment this number by 1 for each new dataset for each different gas / VOC).

[0053] As shown in step 3a, repeat from step 1, selecting a different gas / VOC type until the concentrations of all gases / VOCs have been changed.

[0054] As shown as step 3b, repeat from step 1 to select two gas / VOC types for simultaneous variation of concentrations until the concentrations of all two gas / VOC combinations have been varied.

[0055] As shown as step 3c, repeat from step 1, increasing the number of gas / VOC types simultaneously varied each time (i.e., three gas / VOC combinations, four gas / VOC combinations, etc.) until the concentrations of all N possible gas / VOC combinations have been varied. Data collection is then completed and machine training begins.

[0056] It should be noted that in different embodiments, the names of ambient types need not be based on how many parameters changed during training, as described above with reference to FIG. 5. For example, the names of ambient types may be based on when training is performed in a chronological manner. In general, so long as all relevant combinations of gases are tested and trained with sensors according to example embodiments, the trained model is suitable for use with various example embodiments.

[0057] Data Collection for Machine Learning with N=3 [H2O / RH, CO2, and N2O] According to an Exemplary Embodiment

[0058] Ambient Type 0: Variations in relative humidity, RH, and CO2 only. The desiccator 306 shown in FIG. 3 is first filled with freshly calcined silica gel under a perforated plate using a battery-powered wireless bP sensor 304. The desiccator 306 is then closed to allow the silica gel to reduce RH within the desiccator 306. Once the reference RH sensor 308 has stabilized, data collection begins by allowing the current flowing across the bP chemiresistor 304 to be wirelessly recorded on the tablet 300 for three minutes at 0.3-second intervals. Thus, in this non-limiting example, each sample reading of the RH value will consist of 600 data points representing the measured current flowing across the bP chemiresistor 304 during this three-minute period. Data sets with any change in RH value within the three-minute recording period will be discarded. The experiment was conducted over a five-month period, with the only interruptions to the desiccator 306 environment occurring when the wireless sensor was powered on and off at the beginning and end of each day (to conserve battery energy). The number of sample files collected for RH is 945 (i.e., sample size = 945), with each sample containing 600 measurement points. The ambient concentration of N2O is assumed to remain unchanged for this ambient type 0, but the ambient CO2 concentration inside the dryer is seen to decrease due to diffusion throughout all experiments for this ambient type 0.

[0059] FIG. 6 shows the statistics (i.e., mean, median, and standard deviation) of 68 samples (note: for clarity, only 34 are shown in the figure) collected over the first four days. From this figure, a gradual decrease in the average current of each sample over time can be seen, independent of RH. This is an indication that the bP chemiresistor is affected by residual vapor adsorption on its surface. This is further illustrated in FIG. 7, where the average currents for the same RH taken on three different days are different, indicating that the bP chemiresistor has not recovered to its original state. Nevertheless, it will be shown later that these adverse effects have no bearing on the test score of the machine-trained gas sensing system according to the exemplary embodiment.

[0060] Ambient Type 1: Fluctuations in CO2 only. A photograph of the experimental setup for mechanically training the wireless 2DM-based gas sensing system for relative humidity and CO2 measurements is shown in Figure 8. (Note that the NO reference sensor 803 is not used; like oxygen and nitrogen, NO is part of the ambient air present, but its concentration level is very low, approximately 0.00003%.) In this setup, the CO2 concentration level is controlled by flowing CO2 gas (purity = 99.999%) through a mass flow controller 801 into the chamber 802 at a fixed rate of 5 SCCM for a fixed period ranging from 5 seconds to 60 seconds. Once primed, the gas inlet valve is closed, allowing CO2 to gradually diffuse through the edge of the setup door. The current flowing through the bP chemiresistor during this time is then measured and wirelessly recorded on a tablet at 0.3-second intervals for 3 minutes. In addition to the measured current, ambient temperature and concentration levels are also recorded from the reference CO2 and RH sensors 804, 806, respectively, once at the beginning of the three-minute current measurement and then again at the end of the three-minute measurement. The concentration levels, ambient temperature, and measured current fluctuations are then passed to a machine learning algorithm for training. Multiple sets of data are collected at ambient conditions across a wide range of CO2 concentration levels (e.g., 544 ppm to 1909 ppm). The experiment was conducted over a two-month period, and the number of sample files collected for CO2 and relative humidity was 642 (i.e., sample size = 642), with each sample containing 600 measurement points.

[0061] Note that the sensor is trained with this exemplary embodiment under ambient conditions, where ambient means that RH is present in the environment. Because HO / RH is one of the gases the sensor is trained to measure, its relatively constant concentration level is recorded to train the model to find differences in conductivity between situations where only CO is changing and where RH is changing. RH and CO can each change the conductivity of a chemiresistor. By changing only one component at a time and keeping the other components present but constant, a sensor according to an exemplary embodiment can be trained to make concentration predictions for multiple gases / VOCs. This advantageously overcomes the difficulty of predicting multiple gases from a single overall incremental change in the conductivity of a chemiresistor when exposed to multiple sources of gases / VOCs.

[0062] Ambient Type 2: CO and NO Fluctuations. The experimental setup is as shown in Figure 8. In this case, NO concentration (obtained from NO reference gas measurement sensor 803) is controlled by flowing NO gas (purity = 99.999%) into chamber 802 through mass flow controller 801 at a fixed rate of 5 SCCM for a fixed period ranging from 5 seconds to 60 seconds. Once primed, the gas inlet valve is closed, and CO is allowed to gradually diffuse through the edge of the setup door. The current flowing through the bP chemiresistor during this time is then measured and recorded wirelessly on a tablet at 0.3-second intervals for 3 minutes. In addition to the measured current, ambient temperature and concentration levels are also recorded from the reference NO, CO, and RH meters, respectively, once at the beginning of the 3-minute current measurement and then again at the end of the 3-minute measurement. The concentration levels on the CO2 sensor 804 and RH sensor 803 (the N2O sensor 803 does not have an ambient temperature readout), the ambient temperature fluctuations, and the measured current are then passed to a machine learning algorithm for training. Multiple sets of data are collected at ambient conditions across a wide range of CO2 concentration levels (e.g., 27 ppm to 874 ppm) and N2O concentration levels (e.g., 19 ppm to 945 ppm). The experiment was conducted over a two-month period, and the number of sample files collected for CO2 and N2O was 586 (i.e., sample size = 586), with each sample containing 600 measurement points.

[0063] Data Collection for VOC Detection via Machine Learning, According to an Exemplary Embodiment: N=2: CO2 and RH (e.g., ambient type 0 = empty container with no plants, type 1 = non-aromatic plants only, and type 2 = aromatic plants that emit VOCs).

[0064] Figure 9 shows the experimental setup for classifying the VOCs emitted by the plant. The aromatic plant is the basil plant, which is known in the literature to emit the following VOCs:

[0065] 1. α-pinene

[0066] 2. β-pinene

[0067] 3. Eucalyptol

[0068] 4. Linalyl acetate, phenylpropene

[0069] 5. Eugenol and sesquiterpenes

[0070] 6. α-bermagotene

[0071] 7. Germacrene-D

[0072] 8. γ-Murolene

[0073] 9. β-copaene

[0074] The only gases actively monitored for changes in this exemplary embodiment experiment were CO and RH. Note that if it were desired to predict the concentration level of each individual VOC emitted by basil plants, rather than simply classifying the ambient type, i.e., the presence of emitted VOCs, one could additionally obtain reference concentration values ​​for each VOC for regression training. However, this exemplary embodiment of VOC detection is intended to provide proof of concept that exemplary embodiments of the present invention may also be sensitive to VOCs.

[0075] The data collection and training methodology in embodiments in which the concentration level of each individual VOC is predicted is the same as that described herein with reference to Figures 5 and 10 for a multi-gas environment. Similar to, for example, CO or RH, VOCs are composed of molecules in a gaseous state, except one is organic in nature while the other is inorganic (i.e., CO). Having shown that the exemplary embodiment responds to this type of vapor, the same ML techniques can be applied and similar prediction accuracy can be expected, as will be understood by those skilled in the art.

[0076] Training a Wireless Chemiresistor for Multi-Gas Sensing via Machine Learning According to Exemplary Embodiments

[0077] In the data file collected during data collection, the current measured across the chemiresistor, the respective concentration levels and ambient temperatures taken from the reference sensor, the collection date and time, and the type of gas are available in time series. Using this information, a computer program (e.g., Python-based) can be written to create AI models for each gas / VOC type and use them for predictions. Figure 10 illustrates the process of training a single 2DM-based chemiresistor via machine learning (ML) to classify and quantify multiple sources of gases / VOCs simultaneously present in the same environment, according to an exemplary embodiment.

[0078] Specifically, as shown in step 1, training and testing data collection is performed as described above with reference to FIG. 5. As shown in step 2, a Fourier transform is applied to the various collected time-domain / time series electrical responses of the 2DM-based chemiresistor to obtain the low-frequency noise profile of the exposure and identify its associated Lorentzian component. From the noise profile, all necessary input features are calculated. In accordance with the exemplary embodiment described herein, a total of 15 features were used.

[0079] 1.Rch (ohms)

[0080] 2. Runtime (seconds)

[0081] 3. Ambient temperature (℃)

[0082] 4.RchSignatureFreq(Hz)

[0083] 5.RchMaxLorentzPSD(1 / Hz)

[0084] 6.RchSignatureDesorption(sec)

[0085] 7.RchKurtpLorentzPSD

[0086] 8.RchSkewpLorentzPSD

[0087] 9.RchMaxLorentzPSDmedian(1 / Hz)

[0088] 10.RchMaxLorentzPSDratio

[0089] 11.nLorentzPSDPeaks

[0090] 12.MedianPSDPeaks(1 / Hz)

[0091] 13.Median frequency peaks (Hz)

[0092] 14.FWHM max Hz (Hz)

[0093] 15.FWFMmaxHz(Hz)

[0094] As will be appreciated by those skilled in the art, the above characteristics include information such as:

[0095] Time-dependent electrical response (i.e., channel resistance) from a single 2DM-based chemiresistor, i.e., feature 1 above.

[0096] Operating conditions such as ambient temperature and runtime, i.e., features 2 and 3 above.

[0097] From the plot of the noise spectral density multiplied by frequency (i.e., the Lorentzian noise spectrum) versus frequency, characteristic Lorentzian frequencies (i.e., the peaks with the largest power spectral densities) and their derivatives, such as their respective power spectral densities (i.e., the Lorentzian noise spectrum), i.e., Features 4 and 5.

[0098] characteristic Lorentzian desorption time (i.e., the inverse of the characteristic Lorentzian frequency), i.e., feature 6.

[0099] Kurtosis, skewness and median of the Lorentzian noise spectrum, i.e., features 7, 8 and 9.

[0100] Power spectral density ratio of characteristic Lorentzian frequencies to the median of the Lorentzian noise spectrum, i.e., feature 10.

[0101] The number of Lorentzian peaks found in the Lorentzian noise spectrum, the median values ​​of those peaks, and the frequencies associated with those peaks, i.e., features 11, 12, and 13.

[0102] Full width at half maximum and full width at maximum of the characteristic Lorentzian peaks (i.e., the peaks with the largest power spectral density), i.e., features 14 and 15.

[0103] As shown in step 3a, using the measured time series of channel resistances, their Lorentzian derivatives, and operating conditions as input features, and all associated types of ambient environments as target labels, train a classification model with any established ML algorithm (e.g., from scikit-learn) to predict the type of ambient environment from the input features. If the accuracy is insufficient, repeat step 1.

[0104] As shown in step 3b, once classification modeling is complete, categorize each set of input features according to their classification label (i.e., surrounding type). Then, train a regression model for each surrounding type with any established ML algorithm using feature sets with the same surrounding type as inputs and their associated concentration levels as target values. The number of trained regression models should be equal to the number of surrounding types. Repeat step 1 if accuracy is insufficient.

[0105] It should be noted that while the data collection in the specific exemplary embodiment above does not include the complete spectrum of all ambient conditions specified by the preferred embodiment described above with reference to FIG. 5, it can be deduced from the results obtained from all specified ambient conditions (albeit incomplete), as will be understood by those skilled in the art, that for a three-gas sensor according to the preferred embodiment to cover all desired operating conditions, it would be preferable to train it as described above with reference to FIG. 5.

[0106] Similarly, it will be understood by those skilled in the art that it is not necessary to show sensors according to example embodiments that use a variety of different 2D materials in order to deduce that embodiments of the present invention will generally work for a variety of 2D materials that share the same general properties as those described in specific embodiments herein.

[0107] The training methodology according to an exemplary embodiment of the present invention is very versatile because it depends on how the sensor is used. For example, if a user knows with certainty that the operating conditions will not undergo any changes in one gas / VOC, such as RH, the sensor can still be used except for the fluctuation cycle for RH in the training loop. That is, the accuracy of the prediction will be considered along with the associated ambient conditions being trained. For example, as shown in FIG. 16, a particular prediction accuracy may be valid for ambient conditions where CO2 is changing but RH and NO are assumed constant, as will be described in more detail below.

[0108] If the user is not satisfied with such operating conditions, whether it is when one, two, or three gases are being changed simultaneously, the training loop described with reference to FIG. 5 can be continued / completed to obtain a three-gas sensor that can operate under all operating conditions according to the preferred embodiment.

[0109] Deployment of Trained Classification and Regression Models for Multiple Gas / VOC Sensing According to Exemplary Embodiments

[0110] FIG. 11 shows a signal flow diagram illustrating prediction for multiple gas / VOC sensing according to an exemplary embodiment. As shown at 1200, a 2DM chemiresistor is exposed to an environment with multiple sources of gases / VOCs. As shown at 1202, channel resistance is recorded at 0.3-second intervals over a three-minute period, and the measured time-series data is used to create the necessary input features (compare with the classification modeling described above with reference to FIG. 10). As shown at 1204, a trained classification model is applied to determine the corresponding ambient type from the input features and identify which types of gases / VOCs are present. As shown at 1206, for the determined ambient type, the concentration levels of each gas / VOC type are then predicted by the respective trained regression models (compare with the regression modeling described above with reference to FIG. 10).

[0111] System Implementation of a Gas Sensor or Gas Sensing Node According to Exemplary Embodiments

[0112] FIG. 12 briefly illustrates a system implementation block diagram of an all-in-one gas sensor 900 according to an exemplary embodiment, as an air quality sensor in a standalone consumer product for tracking health conditions or as a climate monitoring sensing node in a wireless sensor network for production control in factory or industrial safety applications. Output calculations (shown in block 902) can be performed either on a mobile device or via the cloud for further automation and complex data analysis calculations. Specifically, block 904 indicates that a gas takes / donates electrons, which in turn results in a change in resistance, as shown in block 906. In an exemplary embodiment, data representing the change in resistance is wirelessly transmitted to the mobile device / cloud, as shown in block 902. Calculated statistics and other data representing rates of change (shown in block 908) are transmitted (e.g., wirelessly) to an AI model block 910 of the all-in-one gas sensor 900 to make predictions of gas type and concentration level, as shown in block 912.

[0113] Results and Analysis from Illustrative Embodiments

[0114] The results in FIG. 13 show characteristic Lorentz frequencies and associated statistical information (compare also the feature list described above with reference to FIG. 10 ) extracted from three different ambient environments (i.e., Types 0, 1, and 2 described above for the exemplary embodiment), which can be viewed as a first proof of principle for using characteristic Lorentz frequencies to distinguish multiple gases, in this case, RH, CO, and NO, from a single 2DM-based chemiresistor according to the exemplary embodiment. Similarly, FIG. 14 shows characteristic Lorentz frequencies and associated statistical information (compare also the feature list described above with reference to FIG. 10 ) extracted from three other different ambient environments (compare FIG. 9 and the corresponding description above) according to the exemplary embodiment: another ambient Type 0 with no plants in the container; another ambient Type 1 with plants in the container that are known not to emit any VOCs; and finally, another ambient Type 2 with plants in the container that are known to emit strong VOCs. This result can be viewed as a first proof of principle for using characteristic Lorentz frequencies to distinguish affected plants based on their VOC emissions.

[0115] The results in FIGS. 15, 16, and 17 show the prediction accuracy of an all-in-one sensor according to an exemplary embodiment for RH (test score=0.998935), CO (test score=0.999999), and N2O (test score=0.999998) gases. As understood in the art, all test scores are based on support vector machine regression (SVR) and decision tree classification algorithms (test score=0.999243). The results advantageously demonstrate that the same physical sensor according to an exemplary embodiment can be simultaneously used to detect three types of gases, namely RH, CO2, and N2O gases, and their concentration levels in a dynamic manner (i.e., the gases do not need to be at a stable concentration) in a controlled environment. Specifically, FIG. 15 shows the RH predicted by support vector machine regression (data points) versus the actual RH (curve) with a response time of 1.5 minutes. RH training and testing were performed simultaneously while the concentration of CO2 gas was decreasing. Note that the sensors were trained under ambient conditions, including the presence of O, N, and N2O. No sources of O, N, and N2O were introduced and therefore assumed to be constant under experimental conditions (i.e., controlled environments). In Figure 16, support vector machine regression predicted CO (e.g., data points) versus actual CO (curve) is shown. Training and testing for CO was performed at relatively constant RH while simultaneously decreasing the concentration of N2O gas, albeit at a slower rate than CO2. Response time = 1.5 minutes. In Figure 17, support vector machine regression predicted N2O (e.g., data points) versus actual N2O (curve) is shown. Training and testing for N2O was performed at relatively constant RH while simultaneously decreasing the concentration of CO2 gas, albeit at a faster rate than N2O. Response time = 1.5 minutes.

[0116] An all-in-one 2DM-based sensor for relative humidity (RH), carbon dioxide (CO), and nitrogen oxides (NO) is provided according to exemplary embodiments. ML training according to exemplary embodiments achieves test scores of over 99.8% for all three gases with a response time of 1.5 minutes. This suggests that at least 99.8% of the variation in the dependent variables is explained by the dedicated AI engine. According to exemplary embodiments, a single physical sensor sensitive to a wide range of gases can be functionalized to selectively detect RH, CO, and NO through machine learning. The use of two-dimensional materials also advantageously achieves low power consumption (25 mV and <10 uA), a long shelf life, and a small physical size. Further training with other gases and vapors may provide an all-in-one portable sensor according to various embodiments for detecting air quality in the form of volatile gases, such as carbon dioxide, carbon monoxide, pollen, or toxins in air, in addition to relative humidity, oxygen, and nitrogen concentrations. Within a software application installed on a mobile device or fixed terminal, various embodiments can display the data, along with an actionable feed for a higher level of intelligence for the user to take action. By manufacturing different AI engines for different gases and vapors, new sensing capabilities can also be added to the all-in-one sensor according to various embodiments, on-demand and in real time, without any changes to the hardware. In various embodiments, reinforcement learning (another type of machine learning technique) can either eliminate or reduce the frequency of sensor calibration.

[0117] 18 summarizes classification and regression results obtained from two different 2DM-based sensors according to an exemplary embodiment, one based on black phosphorus (bP), Dev11-bP, and the other based on tellurene (Te), Dev01-Te. The bP-based sensor was shown to be capable of classifying three different ambient environments containing three gases, namely RH, CO2, and NO, and also quantifying their respective concentration levels within the environments. Meanwhile, the Te-based sensor was shown to be capable of classifying three different ambient environments: one with no plants inside but with elevated CO2 and reduced RH due to normal diffusion through a door gap; one with non-aromatic plants but with CO2 remaining between 595 and 615 ppm and RH remaining between 72 and 74%; and one with aromatic plants and with CO2 concentrations remaining around 605 and 615 ppm and RH rising from 65% to a maximum of 76% before settling back down to 72 and 73%.

[0118] 19 shows a flowchart 1900 illustrating a computerized method for determining the presence and concentration of each of a plurality of gases and / or volatile organic compounds, according to an exemplary embodiment. In step 1902, one or more sensing elements having identical chemical and physical properties are exposed to the plurality of gases and / or volatile organic compounds. In step 1904, electrical time series data of the one or more sensing elements are measured during the exposure. In step 1906, the electrical time series data and Lorentzian noise information of the electrical time series data are analyzed by an artificial intelligence (AI) system. In step 1908, the presence and / or concentration of each of the plurality of gases and / or volatile organic compounds is determined based on the analysis of the electrical time series data and the Lorentzian noise information of the electrical time series data.

[0119] The method further includes analyzing the ambient temperature with an artificial intelligence (AI) system and determining the presence and / or concentration of each of the plurality of gases and / or volatile organic compounds further based on the analysis of the ambient temperature.

[0120] The Lorentzian noise information may include features selected from the group consisting of characteristic Lorentzian peaks with maximum power spectral densities, their respective power spectral densities, characteristic Lorentzian desorption times, kurtosis (Kurt), skewness (Skew), median of the Lorentzian noise spectrum, power spectral density ratios of characteristic Lorentzian frequencies to the median of the Lorentzian noise spectrum, the number of Lorentzian peaks found in the Lorentzian noise spectrum, medians of the Lorentzian peaks, frequencies associated with the Lorentzian peaks, full width at half maximum and full width at maximum of the characteristic Lorentzian peaks.

[0121] The AI ​​system may be a classification or regression model AI system or a reinforcement learning AI system. Each sensing element may include a two-dimensional sensing material. The two-dimensional sensing material may be configured as a chemiresistor, and the electrical time series data may include resistance time series data. The two-dimensional sensing material may include one or a group consisting of black phosphorus (bP), tellurene, reduced graphene oxide, graphene, and transition metal dichalcogenides, or any two-dimensional allotrope of various elements or compounds that has carrier mobility comparable to black phosphorus for low-power operation.

[0122] 20 shows a schematic diagram illustrating a sensor device 2000 capable of determining the presence and concentration of each of a plurality of gases and / or volatile organic compounds, according to an exemplary embodiment. The sensor device 2000 includes one or more sensing elements 2002 having identical chemical and physical properties and an artificial intelligence (AI) system 2004 configured to analyze the electrical time series data and Lorentzian noise information of the electrical time series data and determine the presence and / or concentration of each of the plurality of gases and / or volatile organic compounds based on the analysis of the electrical time series data and Lorentzian noise information of the electrical time series data.

[0123] The AI ​​system 2004 may be further configured to analyze the ambient temperature and determine the presence and / or concentration of each of a plurality of gases and / or volatile organic compounds further based on the analysis of the ambient temperature.

[0124] The Lorentzian noise information may include features selected from the group consisting of characteristic Lorentzian peaks with maximum power spectral densities, their respective power spectral densities, characteristic Lorentzian desorption times, kurtosis (Kurt), skewness (Skew), median of the Lorentzian noise spectrum, power spectral density ratios of characteristic Lorentzian frequencies to the median of the Lorentzian noise spectrum, the number of Lorentzian peaks found in the Lorentzian noise spectrum, medians of the Lorentzian peaks, frequencies associated with the Lorentzian peaks, full width at half maximum and full width at maximum of the characteristic Lorentzian peaks.

[0125] The AI ​​system 2004 may be a classification or regression model AI system or a reinforcement learning AI system. Each sensing element may include a two-dimensional sensing material. The two-dimensional sensing material may be configured as a chemiresistor, and the electrical time series data may include resistance time series data. The two-dimensional sensing material may include one or a group consisting of black phosphorus (bP), tellurene, reduced graphene oxide, graphene, and transition metal dichalcogenides, or any two-dimensional allotrope of various elements or compounds that has carrier mobility comparable to black phosphorus for low-power operation.

[0126] In one embodiment, a method is provided for training a sensor device of the embodiment described above with reference to FIG. 20 to be able to determine the presence and concentration of each of a plurality of gases and / or volatile organic compounds.

[0127] The method may include a data collection step including: i) exposing one or more sensing elements to a desired number and types of gases and / or volatile organic compounds in a controlled environment and measuring a first data set of electrical time series data; ii) varying the concentration of one of the gases and / or volatile organic compounds and measuring a second data set of electrical time series data; and iii) repeating step ii) over a desired range of concentrations.

[0128] The method may further include a data collection step including: iv) varying the concentrations of each of two of said gases and / or volatile organic compounds and measuring further data sets of electrical time series data; and v) repeating step iv) over a desired range of combinations of the concentrations of each of the two gases and / or volatile organic compounds.

[0129] The method may further include a data collection step including vi) repeating steps iv) and v), where in each repetition an additional one of the gases and / or volatile organic compounds is added to step iv).

[0130] The method may further include performing machine training on the dataset collected in the data collection step. Performing machine learning may include training a classification model to predict the number and type of gases and / or volatile organic compounds, and training a regression model to predict the concentration of each of the gases and / or volatile organic compounds.

[0131] Embodiments of the invention may have one or more of the following features and associated benefits / advantages. [Table 1]

[0132] Embodiments of the present invention may have application as gas sensors, for example, air quality sensors in stand-alone consumer products for tracking health conditions, or as climate monitoring sensing nodes in wireless sensor networks for production control in factory or industrial safety applications. Gas sensors are devices that can detect the presence and quantify the concentration of specific gases in the atmosphere, such as water vapor (moisture), organic vapors, and hazardous gases. They are widely used in environmental monitoring and emissions control, personal and military safety, agriculture, industrial, and medical diagnostic production control.

[0133] However, conventional gas sensors are designed to detect only a single type of gas or vapor, so in a wireless sensor network that needs to monitor multiple gases or vapors, multiple unique gas sensors are required in the circuit, making the sensing node built with conventional sensors bulky and requiring multiple unique readout circuits and calibrations, resulting in high power consumption and maintenance costs.

[0134] Because embodiments of the present invention use only one physical sensor to obtain all gas footprints, the wireless sensor node is small (i.e., thumb-sized) and can last, for example, more than 12 months on a 3V coin-cell lithium battery. The AI ​​models developed for sensors according to exemplary embodiments can also be used to eliminate physical calibration requirements or reduce the frequency of calibration, for example, by using reinforcement learning.

[0135] The various functions or processes disclosed herein may be described as data and / or instructions embodied in various computer-readable media in terms of their behavior, register transfers, logical components, transistors, layout geometry, and / or other characteristics. Computer-readable media in which such formatted data and / or instructions may be embodied include, but are not limited to, various forms of non-volatile storage media (e.g., optical, magnetic, or semiconductor storage media) and carrier waves that may be used to transfer such formatted data and / or instructions through wireless, optical, or wired signal transmission media, or any combination thereof. Examples of transfer of such formatted data and / or instructions by a carrier wave include, but are not limited to, transfer (upload, download, email, etc.) over the Internet and / or other computer networks via one or more data transfer protocols (e.g., HTTP, FTP, SMTP, etc.). Once received within a computer system via one or more computer-readable media, such data and / or instruction-based representations of components and / or processes under the described system may be processed by a processing entity (e.g., one or more processors) within the computer system in conjunction with the execution of one or more other computer programs.

[0136] Aspects of the systems and methods described herein, such as data collection, machine learning, and AI sensor output generation, may be implemented as programmed functions in any of a variety of circuits, including programmable logic devices (PLDs), e.g., field programmable gate arrays (FPGAs), programmable array logic (PAL) devices, electrically programmable logic and memory devices, and standard cell-based devices, as well as application-specific integrated circuits (ASICs). Some other possibilities for implementing aspects of the system include a microcontroller with memory (e.g., electronically erasable programmable read-only memory (EEPROM)), embedded microprocessors, firmware, software, etc. Additionally, aspects of the system may be embodied in a microprocessor with software-based circuit emulation, discrete logic (sequential and combinatorial), custom devices, fuzzy (neural) logic, quantum devices, and hybrids of any of the above device types. Of course, the underlying device technology, e.g., metal-oxide-semiconductor field-effect transistor (MOSFET) technology such as complementary metal-oxide-semiconductor (CMOS), bipolar technology such as emitter-coupled logic (ECL), polymer technology (e.g., silicon-conjugated polymer and metal-conjugated polymer-metal structures), mixed analog and digital, etc., may be provided in a variety of component types.

[0137] The above description of illustrated embodiments of the systems and methods is not intended to be exhaustive or to limit the systems and methods to the precise forms disclosed. While specific embodiments and examples of system components and methods are described herein for illustrative purposes, various equivalent modifications are possible within the scope of the systems, components, and methods, as those skilled in the art will recognize. The teachings of the systems and methods provided herein may be applicable to other processing systems and methods, not just the systems and methods described above.

[0138] The elements and acts of the various embodiments described above can be combined to provide further embodiments. These and other changes can be made to the systems and methods in light of the above detailed description.

[0139] For example, although bP was used in the exemplary embodiments described herein, other materials may be used in different embodiments, including, but not limited to, tellurene, reduced graphene oxide, graphene, and transition metal dichalcogenides. Preferably, the material is a two-dimensional material with high carrier mobility and a large surface area to volume ratio, such as any two-dimensional allotrope of various elements or compounds with carrier mobility comparable to black phosphorus for low power operation.

[0140] In general, in the following claims, the terms used should not be construed to limit the systems and methods to the specific embodiments disclosed in the specification and claims, but should be construed to include all processing systems operating under the scope of the claims. Thus, the systems and methods are not limited by the disclosure; instead, the scope of the systems and methods is determined solely by the claims.

[0141] Unless otherwise clearly required by context, throughout the specification and claims, words such as "comprise," "comprising," and the like, should be construed in an inclusive sense, i.e., "including, but not limited to," as opposed to an exclusive or exhaustive sense. Words using the singular or plural form also include the plural and singular, respectively. Also, "as used herein, the words 'herein,' 'hereunder,' 'above,' 'below,' and words of similar import refer to this application as a whole and not to specific parts of this application. When the word 'or' is used in connection with a list of two or more items, the word covers all of the following interpretations of the word: any of the items in the list, all of the items in the list, and any combination of the items in the list.

Claims

1. 1. A computerized method for determining the presence and concentration of each of a plurality of gases and / or volatile organic compounds, said method comprising: exposing one or more sensing elements having identical chemical and physical properties to the plurality of gases and / or volatile organic compounds; measuring electrical time series data of the one or more sensing elements during the exposure; analyzing the electrical time series data and Lorentzian noise information of the electrical time series data together by an artificial intelligence (AI) system; determining the presence and concentration of each of the plurality of gases and / or volatile organic compounds based on the analysis of the electrical time series data and Lorentzian noise information of the electrical time series data; Equipped with analyzing the ambient temperature with the artificial intelligence (AI) system using the ambient temperature as an input; and determining the presence and / or concentration of each of the plurality of gases and / or volatile organic compounds further based on the analysis of the ambient temperature; the Lorentzian noise information has characteristics selected from the group consisting of distinctive Lorentzian peaks having maximum power spectral densities, their respective power spectral densities, distinctive Lorentzian desorption times, kurtosis (Kurt), skewness (Skew) and median of the Lorentzian noise spectrum, power spectral density ratios of distinctive Lorentzian frequencies to the median of the Lorentzian noise spectrum, the number of Lorentzian peaks found in the Lorentzian noise spectrum, medians of the Lorentzian peaks, frequencies associated with the Lorentzian peaks, full width at half maximum and full width at full maximum of the distinctive Lorentzian peaks.

2. 10. The method of claim 1, wherein the AI ​​system is a classification or regression model AI system or a reinforcement learning AI system.

3. The method of claim 1 or 2, wherein each sensor element comprises a two-dimensional sensing material.

4. The method of claim 3 , wherein the two-dimensional sensing material is configured as a chemiresistor, and the electrical time series data comprises resistance time series data.

5. 5. The method of claim 3 or 4, wherein the two-dimensional sensing material comprises one or the group consisting of black phosphorus (bP), tellurene, reduced graphene oxide, graphene, and transition metal dichalcogenides, or any two-dimensional allotrope of various elements or compounds that have carrier mobilities comparable to black phosphorus for low power operation.

6. 1. A sensor device capable of determining the presence and concentration of each of a plurality of gases and / or volatile organic compounds, comprising: one or more sensing elements made from substantially the same sensing material; an artificial intelligence (AI) system; the AI ​​system is configured to jointly analyze electrical time series data of the one or more sensing elements and Lorentzian noise information of the electrical time series data, and determine the presence and concentration of each of the plurality of gases and / or volatile organic compounds based on the analysis of the electrical time series data and the Lorentzian noise information of the electrical time series data; the AI ​​system is further configured to use an ambient temperature as an input, analyze the ambient temperature, and determine a presence and / or concentration of each of a plurality of gases and / or volatile organic compounds further based on the analysis of the ambient temperature; the Lorentzian noise information has characteristics selected from the group consisting of: distinctive Lorentzian peaks having maximum power spectral densities, their respective power spectral densities, distinctive Lorentzian desorption times, kurtosis (Kurt), skewness (Skew) and median of the Lorentzian noise spectrum, power spectral density ratios of distinctive Lorentzian frequencies to the median of the Lorentzian noise spectrum, the number of Lorentzian peaks found in the Lorentzian noise spectrum, medians of the Lorentzian peaks, frequencies associated with the Lorentzian peaks, full width at half maximum and full width at full maximum of the distinctive Lorentzian peaks; Sensor device.

7. The sensor device of claim 6 , wherein the AI ​​system is a classification or regression model AI system or a reinforcement learning AI system.

8. The sensor device of claim 6 or 7, wherein each sensing element comprises a two-dimensional sensing material.

9. The sensor device of claim 8 , wherein the two-dimensional sensing material is configured as a chemiresistor, and the electrical time series data comprises resistance time series data.

10. 10. The sensor device of claim 8 or 9, wherein the two-dimensional sensing material comprises one or the group consisting of black phosphorus (bP), tellurene, reduced graphene oxide, graphene, and transition metal dichalcogenides, or any two-dimensional allotrope of various elements or compounds that have carrier mobility comparable to black phosphorus for low power operation.

11. 11. A method of training a sensor device according to any one of claims 6 to 10 so as to be able to determine the presence and / or concentration of each of a plurality of gases and / or volatile organic compounds.

12. i) exposing the one or more sensing elements to a desired number and types of gases and / or volatile organic compounds in a controlled environment and measuring a first data set of the electrical time series data; ii) varying the concentration of one of the gases and / or volatile organic compounds and measuring a second data set of the electrical time series data; iii) repeating step ii) over the desired concentration range; The method of claim 11 , comprising a data collection step comprising:

13. iv) varying the concentration of each of two of said gases and / or volatile organic compounds and measuring further data sets of said electrical time series data; v) repeating step iv) over a desired range of combinations of concentrations of each of the two of the gases and / or volatile organic compounds; The method of claim 12 further comprising a data collection step comprising:

14. vi) repeating steps iv) and v), wherein in each repetition an additional one of said gases and / or volatile organic compounds is added to step iv). The method of claim 13 further comprising a data collection step comprising:

15. 15. The method of any one of claims 12 to 14, further comprising performing machine learning on the dataset collected during the data collection step.

16. 16. The method of claim 15, wherein performing the machine learning comprises training a classification model to predict the number and type of gases and / or volatile organic compounds, and training a regression model to predict the concentration of each of the gases and / or volatile organic compounds.

Images