A gas sensor, its preparation method, its detection method, and its thermal runaway monitoring method.

Abstract

This application relates to a gas sensor, its preparation method, detection method, and thermal runaway monitoring method. The sensor includes multiple sensing units, each of which includes a bulk acoustic wave resonator substrate, interdigitated electrodes, a metal oxide sensitive layer, and a signal processing unit. The bulk acoustic wave resonator substrate emits high-frequency sound waves and enhances the activity of the metal oxide sensitive layer. The interdigitated electrodes are disposed on the surface of the bulk acoustic wave resonator substrate. The metal oxide sensitive layer is disposed on the surface of the interdigitated electrodes and adsorbs gas, causing a chemical reaction between the adsorbed gas and the metal oxide material, resulting in a change in the electrical signal of the interdigitated electrodes. The signal processing unit, connected to the interdigitated electrodes, receives the changes in the electrical signal of the interdigitated electrodes and determines the gas concentration and gas type based on these changes. The gas sensor provided by this application features high sensitivity and fast response speed, and can simultaneously detect multiple gases.

Description

Technical Field

[0001] This application relates to the field of gas sensor technology, and in particular to a gas sensor, a preparation method, a detection method, and a thermal runaway monitoring method. Background Technology

[0002] With the rapid development of new energy vehicles, lithium batteries have become an important energy source for electric vehicles, energy storage systems, and other fields due to their high energy density and long cycle life. However, lithium batteries are prone to thermal runaway under conditions such as overcharging and short circuits, which can rapidly release gases such as CO, H2, CH4, and C2H4. Changes in the concentration of these gases can serve as early warning signals of thermal runaway. Therefore, rapid and accurate detection of these gases is crucial for improving the safety of lithium batteries.

[0003] Currently, sensors for gas detection mainly include infrared gas sensors, electrochemical gas sensors, and semiconductor gas sensors. Infrared gas sensors primarily detect gases based on the absorption characteristics of gas molecules in specific infrared bands. However, these sensors rely on optical systems, which increase the overall size and cost of the sensor, making integration with lithium battery packs difficult, and they lack sensitivity for low-concentration gases. Electrochemical gas sensors measure gas concentration based on electrochemical reactions, but their electrolytes are prone to volatiles and their electrodes to passivation, resulting in a short lifespan (typically less than two years), making them unsuitable for long-term monitoring of battery gas leaks. Semiconductor gas sensors require temperatures between 200℃ and 400℃ to fully utilize their sensing characteristics, increasing energy consumption and potentially interfering with the battery's thermal management system, adding additional risks. Furthermore, existing gas sensors suffer from slow response times (>10 seconds) and insufficient multi-gas simultaneous detection capabilities. However, the thermal runaway process of lithium batteries is sudden and spreads rapidly. Existing gas sensors are unable to capture sudden changes in gas concentration in time, which can easily lead to delayed warnings and cause incalculable losses. Summary of the Invention

[0004] In view of the above-mentioned problems of the prior art, this application provides a gas sensor, a preparation method, a detection method, and a thermal runaway monitoring method. The gas sensor has the advantages of small size, long service life, high sensitivity, fast response speed, and support for simultaneous detection of multiple gases.

[0005] To achieve the above objectives, a first aspect of this application provides a gas sensor, comprising: multiple sensing units, each sensing unit comprising: a bulk acoustic wave resonator substrate, an interdigital electrode, a metal oxide sensitive layer, and a signal processing unit; the bulk acoustic wave resonator substrate is used to emit high-frequency sound waves and enhance the activity of the metal oxide sensitive layer through the high-frequency sound waves; the interdigital electrode is disposed on the surface of the bulk acoustic wave resonator substrate; the metal oxide sensitive layer is disposed on the surface of the interdigital electrode and is used to adsorb gas and react chemically with the adsorbed gas through the metal oxide material, thereby causing a change in the electrical signal of the interdigital electrode; the signal processing unit is connected to the interdigital electrode and is used to receive the change in the electrical signal of the interdigital electrode and determine the gas concentration and gas type based on the change in the electrical signal of the interdigital electrode.

[0006] As described above, when the gas sensor is placed in the environment to be detected, the metal oxide material in the metal oxide sensitive layer reacts chemically with the target gas, causing a change in the electrical signal of the interdigitated electrodes, thus achieving gas detection. This gas sensor features small size, long service life, high sensitivity, and fast response speed. Because the sensor includes multiple sensing units, each made of a different metal oxide material, it can support the simultaneous detection of multiple gases.

[0007] As one implementation of this aspect, the metal oxide sensitive layers of the plurality of sensing units are made of different metal oxide materials to adsorb different types of gases.

[0008] As a result, by using different materials to make different sensing units, multiple gases can be detected simultaneously.

[0009] As one implementation of this aspect, the bulk acoustic wave resonator substrate adopts a GHz-level solid-state assembled bulk acoustic wave resonator.

[0010] As shown above, the activity of the metal oxide sensitive layer can be better stimulated by using a GHz-level solid-state assembled bulk acoustic resonator.

[0011] As one implementation of this aspect, the metal oxide sensitive layer is coated onto the surface of the interdigitated electrode by a suspension spin coating method or an inkjet printing film formation method, and then subjected to high-temperature activation treatment.

[0012] As shown above, metal oxides can be uniformly coated on the surface of the interdigitated electrode by spin coating or inkjet printing. High-temperature activation treatment can improve the activity of the metal oxide sensitive layer and enhance the gas detection sensitivity.

[0013] As one implementation of this aspect, the metal oxide material of the metal oxide sensitive layer includes one or more of SnO2, WO3, and ZnO.

[0014] As one implementation of this aspect, the signal processing unit includes: a signal acquisition and preprocessing module, used to synchronously acquire time-series signals of interdigital electrode electrical signals corresponding to different sensing units, and to perform filtering, denoising, and normalization processing on each time-series signal to obtain each preprocessed signal; a feature extraction module, used to extract feature sequences from each preprocessed signal based on a convolutional neural network; a gas identification module, used to determine the concentration and type of gas corresponding to each feature sequence based on a long short-term memory model; wherein the model determines the concentration and type of gas by the response patterns of different materials in the sensitive layer to different gases; and an early warning module, used to issue early warning signals of different levels according to the concentration and type of gas.

[0015] Therefore, the detection accuracy and speed can be improved by using the above detection methods.

[0016] The second aspect of this application provides a method for fabricating a gas sensor, comprising: fabricating interdigitated electrodes on the surface of a bulk acoustic wave resonator substrate; uniformly coating a metal oxide material onto the surface of the interdigitated electrodes by a spin coating method or an inkjet printing method to form an initial state metal oxide sensitive layer; and placing the initial state metal oxide sensitive layer in an inert gas environment for high-temperature annealing to obtain the final metal oxide sensitive layer.

[0017] As one implementation of this aspect, the temperature range of the high-temperature annealing treatment is 300℃-600℃; the time range of the high-temperature annealing treatment is 2-5 hours.

[0018] A third aspect of this application provides a method for gas detection using the sensor described in any one of the first aspects, comprising: synchronously acquiring time-series signals of interdigital electrode electrical signals corresponding to different sensing units, and performing filtering, denoising, and normalization processing on each time-series signal to obtain pre-processed signals; extracting feature sequences from each pre-processed signal based on a convolutional neural network; and determining the gas concentration and gas type corresponding to each feature sequence based on a long short-term memory model; wherein the model determines the gas concentration and gas type by the response patterns of different materials in the metal oxide sensitive layer to different gases.

[0019] The beneficial effects in this regard can also be found in the descriptions of the beneficial effects in each part of the first aspect above.

[0020] The fourth aspect of this application provides a method for monitoring thermal runaway of a lithium battery, the method comprising: using a gas sensor as described in any of the first aspects above to monitor the type and concentration of gas released by the lithium battery, and determining whether the lithium battery has experienced thermal runaway based on the type and concentration of the gas.

[0021] The beneficial effects in this regard can also be found in the descriptions of the beneficial effects in each part of the first aspect above. Attached Figure Description

[0022] The various technical features of this application and their relationships will be further explained below with reference to the accompanying drawings. The drawings are exemplary; some technical features are not shown to scale, and some drawings may omit technical features commonly used in the art to which this application pertains that are not essential for understanding and implementing this application, or additionally show technical features that are not essential for understanding and implementing this application. In other words, the combination of various technical features shown in the drawings is not intended to limit this application. Furthermore, throughout this application, the same reference numerals refer to the same things. Specific descriptions of the drawings are as follows:

[0023] Figure 1 A three-dimensional structural schematic diagram of the gas sensor provided in the embodiments of this application;

[0024] Figure 2 A schematic diagram illustrating the fabrication of a metal oxide sensitive layer using a suspension spin coating method, provided in an embodiment of this application;

[0025] Figure 3 A schematic diagram illustrating the fabrication of a metal oxide sensitive layer by inkjet printing film formation method provided in an embodiment of this application;

[0026] Figure 4 A flowchart illustrating a method for gas detection using a gas sensor provided in an embodiment of this application. Detailed Implementation

[0027] The technical solutions provided in this application will be further described below with reference to the accompanying drawings and embodiments. It should be understood that the sensor structures and business scenarios provided in the embodiments of this application are mainly for illustrating possible implementations of the technical solutions of this application and should not be construed as the sole limitation on the technical solutions of this application. Those skilled in the art will recognize that, with the evolution of sensor structures and the emergence of new business scenarios, the technical solutions provided in this application are equally applicable to similar technical problems.

[0028] It should be understood that this application provides a gas sensor solution based on MOx-SMR. Since these technical solutions address the same or similar problems, some repetitive details may not be repeated in the following descriptions of specific embodiments. However, these specific embodiments should be considered as mutually referencing each other and can be combined with each other.

[0029] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. In case of any inconsistency, the meaning set forth in this specification or derived from the content described herein shall prevail. Furthermore, the terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit the scope of this application.

[0030] The gas sensor solution provided in this application is mainly used in the field of monitoring thermal runaway in lithium batteries. During thermal runaway, lithium batteries rapidly release gases such as CO, H2, CH4, and C2H4. The concentration changes of these gases precede temperature changes, making them crucial early warning signals for thermal runaway. Therefore, integrating the gas sensor provided in this application into a lithium battery module or battery management system, and using sensing units composed of different metal oxide materials within the sensor to detect different types of gases, not only enables simultaneous detection of multiple gases but also improves detection sensitivity and response speed, thereby providing timely and accurate monitoring of lithium battery thermal runaway.

[0031] The above application scenarios are merely illustrative examples. In other embodiments, the gas sensor solution provided in this application can also be applied to other scenarios, such as environmental monitoring and biomedical fields.

[0032] The embodiments of this application provide a gas sensor, which will be described below in conjunction with... Figure 1 This gas sensor will be described in detail.

[0033] like Figure 1 As shown, the gas sensor includes multiple sensing units ( Figure 1 Only one sensing unit is shown. Each sensing unit includes a bulk acoustic wave resonator substrate 110, interdigitated electrodes 120, a metal oxide sensitive layer 130, and a signal processing unit (not shown). As one implementation, the sensor may also include an encapsulation layer that allows gas passage to prevent the intrusion of external contaminants. Multiple sensing units may share the same bulk acoustic wave resonator substrate 110, or they may each have an independent bulk acoustic wave resonator substrate 110. Furthermore, multiple sensing units may share the same signal processing unit, or each sensing unit may be connected to an independent signal processing unit. The various structures of this gas sensor will be described in detail below.

[0034] The bulk acoustic wave resonator substrate 110 is used to emit high-frequency acoustic waves and activate and enhance the activity of the metal oxide sensitive layer through these high-frequency acoustic waves. It also provides support for each sensing unit. In this embodiment, the bulk acoustic wave resonator substrate 110 adopts a GHz-level solid-state assembled bulk acoustic wave resonator (SMR). Specifically, the GHz-level solid-state assembled bulk acoustic wave resonator (SMR) includes a bottom electrode 111, a piezoelectric layer (not shown), a top electrode 112, and a resonant region 113. A bulk acoustic wave resonator (SMR) may include one or more resonant regions 113, and the upper surface of the resonant region 113 is in contact with other components in the sensor.

[0035] Interdigitated electrodes 120 are disposed on the surface of the bulk acoustic wave resonator substrate 110. They amplify changes in electrical signals (e.g., resistivity, current, inductance) caused by the reaction of the metal oxide sensitive layer based on their own electric field distribution, thereby improving detection sensitivity. The electrode width of the interdigitated electrodes 140 can be set between 3 μm and 10 μm, and the spacing between two electrodes can be set between 2 μm and 8 μm. In this embodiment, the electrode width of the interdigitated electrodes 140 is set to 5 μm, the electrode spacing is set to 3 μm, the number of electrode pairs is 6, and the electrode material is platinum (Pt). In some embodiments, the interdigitated electrodes 120 can be configured with different shapes depending on the shape of the substrate.

[0036] A metal oxide sensitive layer 130 is disposed on the surface of the interdigital electrode 120 to adsorb gas and, through a chemical reaction between the adsorbed gas and the metal oxide material, cause a change in the electrical signal of the interdigital electrode 120. In this embodiment, the metal oxide sensitive layer 130 is fabricated using a spin-coating method (e.g., ...). Figure 2 (as shown) or inkjet printing to form a film (such as...) Figure 3 The coating (shown) is applied to the surface of the interdigital electrode 120 and subjected to high-temperature activation treatment to enhance the gas adsorption capacity. In this embodiment, the sensitive layers 130 of different sensing units in the sensor are made of different metal oxide (MOx) materials. For example, one or more of SnO2, WO3, and ZnO can be selected to make the sensitive layers to adsorb gases with different properties. In the sensor of this embodiment, the metal oxide sensitive layers 130 of different sensing units can be made of different metal oxide materials, thereby enabling a single sensor to simultaneously detect multiple gases.

[0037] The signal processing unit is connected to the interdigital electrode 120 and is used to receive changes in the electrical signal from the interdigital electrode 120, and determine the gas concentration and gas type based on these changes. The detection device includes a signal acquisition and preprocessing module, a feature extraction module, a gas identification unit, and an early warning unit. Specifically: the signal acquisition and preprocessing module is used to synchronously acquire time-series signals of the interdigital electrode electrical signals corresponding to different sensing units, and to perform filtering, denoising, and normalization processing on each time-series signal to obtain preprocessed signals. The feature extraction module is used to extract feature sequences from each preprocessed signal based on a convolutional neural network. The gas identification module is used to determine the gas concentration and gas type corresponding to each feature sequence based on a long short-term memory model; the model determines the gas concentration and gas type by observing the response patterns of different materials in the sensitive layer to different gases. The early warning module is used to issue different levels of early warning signals based on the gas concentration and gas type. The specific gas detection process can be found in the detailed description of the gas detection section below.

[0038] Based on the gas sensor provided in this application embodiment, in an environment containing a target gas, the bulk acoustic resonator substrate enhances the activity of the metal oxide sensitive layer by emitting high-frequency acoustic waves. The metal oxide material in the metal oxide sensitive layer reacts chemically with the target gas, thereby causing a change in the electrical signal of the interdigitated electrodes to achieve rapid gas detection. Furthermore, the sensor in this embodiment includes multiple sensing units, each made of a different metal oxide material, thus supporting the simultaneous detection of multiple gases.

[0039] The following describes the fabrication method of this gas sensor, which includes the following steps:

[0040] A. Fabrication of interdigitated electrodes 120: Interdigitated electrodes are fabricated on the surface of the bulk acoustic wave resonator substrate 110 (i.e., the top electrode side). Specifically, an interdigitated electrode 120 is fabricated by depositing a 100 nm platinum (Pt) thin film on the surface of the bulk acoustic wave resonator substrate 110, followed by electron beam evaporation and photolithography. Furthermore, the interdigitated electrode 120 requires gold wire bonding to extract electrode signals for connection to the signal processing unit.

[0041] B. Preparation of the metal oxide sensitive layer 130: by spin coating of suspension (see...) Figure 2 Or inkjet printing to form a film (see...) Figure 3A metal oxide material is uniformly coated onto the surface of the interdigitated electrode 120 to form an initial metal oxide sensitive layer. This initial metal oxide sensitive layer is then subjected to high-temperature annealing in an inert gas environment (e.g., nitrogen (N2), argon (Ar), or helium (He)) to obtain the final metal oxide sensitive layer. The high-temperature annealing temperature range is 300℃-600℃, and the annealing time ranges from 2 to 5 hours. In this embodiment, the temperature is 400℃ and the time is 2 hours.

[0042] Next, combine Figure 4 This paper introduces the method (i.e., detection principle) for gas detection using this gas sensor. The method mainly includes three steps: S410 data acquisition and preprocessing, S420 feature extraction, and S430 gas identification. Specifically:

[0043] In the S410 data acquisition and preprocessing steps: the sensor is placed in the environment to be tested. When the target gas is present in the environment, the metal oxide sensitive layer of the sensor comes into contact with the target gas and undergoes a chemical reaction, which causes the electrical signal of the interdigital electrode to change. The interdigital electrode amplifies the change in electrical signal through its own electric field distribution. Then, the signal processing unit connected to the interdigital electrode synchronously acquires the time series signals of the electrical signals transmitted from the corresponding interdigital electrodes of different sensing units based on a preset sampling frequency, and performs filtering, noise reduction and normalization processing on each time series signal to obtain each preprocessed signal.

[0044] As one approach, the raw real-time data (i.e., the time-series signals of electrical signals transmitted from the interdigital electrodes of different sensing units) can be stored for later use.

[0045] In the S420 feature extraction step: a convolutional neural network (CNN) is used to extract the respective feature sequences from each preprocessed signal. In this embodiment, the feature sequence includes the amount of change in the electrical signal.

[0046] In other embodiments, the feature sequence may include parameters such as response time and recovery time, in addition to the change in electrical signal. Response time represents the time difference between the time the sensor comes into contact with the target gas and the time when it reaches a first threshold (e.g., 90%) of its full-scale range. Recovery time represents the time difference between the time the sensor leaves the target gas and the time when it recovers to a second threshold (e.g., 10%) of its range. It should be understood that since the same metal oxide material exhibits different response magnitudes (i.e., the magnitude of the response value increase) and response rates (i.e., the slope of the response curve) to different gases, these two parameters can also be used as input parameters for subsequent algorithms, thereby enabling better evaluation of gas type and concentration.

[0047] In the S430 gas identification step: the concentration and type of gas corresponding to each feature sequence are determined based on the Long Short-Term Memory (LSTM) model; the model determines the gas concentration and type of gas by the response patterns of different materials in the metal oxide sensitive layer to different gases.

[0048] To better understand this detection method, a specific example will be used below.

[0049] First, a gas detection model is constructed based on a convolutional neural network (CNN) and a long short-term memory (LSTM) model. The training dataset for this gas detection model consists of sensor response data (or response patterns) of different gases at different concentrations. The gas detection model is trained using this training dataset.

[0050] The following describes the process of detecting gas types and concentrations using this gas detection model:

[0051] S410 Data Acquisition and Preprocessing:

[0052] (1) Data acquisition:

[0053] Input: Resistance change rate signal of each sensing unit (ΔR / R0, where R0 is the initial resistance value and ΔR is the real-time resistance change);

[0054] Data acquisition tool: 16-bit analog-to-digital converter (ADC), sampling frequency set to 100Hz, input voltage range 0-5V;

[0055] Output: Generates a time-series signal including each sensing unit (the time-series signal is the time sequence of the interdigital electrode electrical signal). Each sensing unit collects 100 data points per second, and each sensor collects a total of 600 data points in 6 seconds. In this embodiment, taking a sensor with 6 sensing units as an example, the 6 sensing units collect a total of 6*600 data points.

[0056] (2) Signal denoising:

[0057] Method: Wavelet transform, with the following parameters: Wavelet basis function: Daubechies wavelet; Decomposition level: 3 levels; Noise denoising threshold setting: Where σ is the noise standard deviation and N is the number of sampling points (N = 600 for each sensing unit's dataset);

[0058] Output: A smoothed time series after denoising, preserving gas response characteristics and suppressing environmental noise.

[0059] (3) Normalization:

[0060] Method: Min-Max normalization is used to linearly map the denoised signal to the [0,1] interval;

[0061] Output: Standardized time series, eliminating dimensional differences between sensor units and improving convergence efficiency.

[0062] S420 Feature Extraction:

[0063] (1) Data restructuring:

[0064] Input: Normalized time series of multiple sensing units (number of data points: 6*600);

[0065] Method: The normalized time series of multiple sensing units is converted into a two-dimensional matrix with the dimension being the number of sensing units * the number of sampling points per sensing unit (e.g., 6*600).

[0066] Output: Two-dimensional tensor, adapted to the input format of convolutional neural networks (CNN).

[0067] (2) Feature extraction using convolutional neural networks (CNN):

[0068] Network architecture: 3-layer convolutional neural network, with the following parameters:

[0069] First convolutional layer: 32 3×1 convolutional kernels, stride 1, activation function ReLU;

[0070] First pooling layer: 2×2 max pooling, stride 2;

[0071] Second convolutional layer: 64 3×1 convolutional kernels, stride 1, activation function ReLU;

[0072] Second pooling layer: 2×2 max pooling, stride 2;

[0073] The third convolutional layer: 128 3×1 convolutional kernels, stride 1, activation function ReLU;

[0074] Output: After convolution and pooling operations, a 3D spatial feature map is generated (dimensions: 1, 128, 146).

[0075] (3) Feature sequence tiling processing:

[0076] Method: The 3D feature map (dimensions: 1, 128, 146) output by the CNN is tiled according to the time step, with dimensions (1, 146, 128).

[0077] Output: A three-dimensional tensor with dimensions (1,146,128), suitable for the temporal modeling requirements of Long Short-Term Memory (LSTM) networks.

[0078] S430 Gas Detection (LSTM):

[0079] (1) LSTM network architecture:

[0080] A two-layer LSTM network with 64 units per layer is used. The activation function is tanh for the hidden state and sigmoid for the gated units. The regularization is Dropout ratio of 0.2 to prevent overfitting.

[0081] Input data: Feature sequences extracted by CNN;

[0082] Output: Time series hidden state sequence (dimensions: 1, 64);

[0083] (2) Fully connected layer:

[0084] Input: The time-series hidden state sequence output by LSTM (dimensions: 1, 64);

[0085] Output layer: The number of neurons equals the number of gas categories plus the number of concentration levels;

[0086] Activation functions: Softmax is used to output the probability distribution of gas types; Linear is used to output the predicted gas concentration.

[0087] Output results: probability distribution of gas types and predicted gas concentrations.

[0088] (3) Classification and Regression:

[0089] For gas types: the category with the highest output probability is taken as the gas type;

[0090] For gas concentration: output the gas concentration predicted by Linear as the gas concentration.

[0091] In this embodiment, a sliding window technique can also be used to update the input data. For example, the input data can be updated every 0.5 seconds to increase the real-time nature of the data and ensure the timeliness of the detection.

[0092] The detection methods for S410 to S430 in the above examples improve the speed of gas identification by using a CNN-LSTM hybrid model.

[0093] An embodiment of this application also provides a method for detecting thermal runaway of a lithium battery, comprising: integrating the aforementioned sensor with a battery management system, using the sensor to monitor the type and concentration of gases released by the lithium battery in real time, and determining whether the lithium battery has experienced thermal runaway based on the type and concentration of the gas. For example, when the concentration of a certain gas is detected to exceed a preset concentration, a graded alarm is automatically triggered.

[0094] Note that the above are merely preferred embodiments and the technical principles employed in this application. Those skilled in the art will understand that this application is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of this application. Therefore, although this application has been described in detail through the above embodiments, this application is not limited to the above embodiments. Many other equivalent embodiments may be included without departing from the concept of this application, all of which fall within the scope of protection of this application.

Claims

1. A gas sensor, characterized in that, include: Multiple sensing units, each sensing unit including: a bulk acoustic resonator substrate, interdigitated electrodes, a metal oxide sensitive layer and a signal processing unit; The bulk acoustic resonator substrate is used to emit high-frequency acoustic waves and enhance the activity of the metal oxide sensitive layer through the high-frequency acoustic waves. The interdigitated electrodes are disposed on the surface of the bulk acoustic resonator substrate; The metal oxide sensitive layer is disposed on the surface of the interdigitated electrode and is used to adsorb gas and react chemically with the adsorbed gas through the metal oxide material, so as to cause the interdigitated electrode to change the electrical signal. A signal processing unit, connected to the interdigitated electrodes, is used to receive changes in the electrical signals of the interdigitated electrodes and determine the gas concentration and gas type based on the changes in the electrical signals of the interdigitated electrodes; the metal oxide sensitive layers of the multiple sensing units are made of different metal oxide materials to adsorb different types of gases. The bulk acoustic wave resonator substrate adopts a GHz-level solid-state assembled bulk acoustic wave resonator. The metal oxide sensitive layer is coated onto the surface of the interdigitated electrode by a suspension spin coating method or an inkjet printing film formation method, and then subjected to high-temperature activation treatment; The signal processing unit includes: The signal acquisition and preprocessing module is used to synchronously acquire the time series signals of the interdigital electrode electrical signals corresponding to different sensing units, and to perform filtering, noise reduction and normalization on each time series signal to obtain the preprocessed signals. The feature extraction module is used to extract the respective feature sequences from the preprocessed signals based on a convolutional neural network; A gas identification module is used to determine the concentration and type of gas corresponding to each of the respective feature sequences based on a long short-term memory model; wherein, the model determines the concentration and type of gas by the response patterns of different materials in the sensitive layer to different gases; The early warning module is used to issue early warning signals of different levels based on the concentration and type of the gas.

2. The gas sensor according to claim 1, characterized in that, The metal oxide material of the metal oxide sensitive layer includes one or more of SnO2, WO3, and ZnO.

3. A method for preparing a gas sensor according to any one of claims 1-2, characterized in that, include: Interdigitated electrodes are fabricated on the surface of a bulk acoustic resonator substrate; The metal oxide material is uniformly coated onto the surface of the interdigitated electrode by a suspension spin coating method or an inkjet printing film formation method to form an initial state metal oxide sensitive layer. The initial state metal oxide sensitive layer is then placed in an inert gas environment for high-temperature annealing to obtain the final metal oxide sensitive layer.

4. The method according to claim 3, characterized in that, The temperature range of the high-temperature annealing treatment is 300℃-600℃; the time range of the high-temperature annealing treatment is 2-5 hours.

5. A method for gas detection using the gas sensor according to any one of claims 1-2, characterized in that, include: The time series signals of the interdigital electrode electrical signals of different sensing units are acquired synchronously, and each time series signal is filtered, denoised and normalized to obtain each preprocessed signal. The feature sequences of each preprocessed signal are extracted using a convolutional neural network. The concentration and type of gas corresponding to each of the respective feature sequences are determined based on a long short-term memory model; wherein, the model determines the concentration and type of gas by the response patterns of different materials in the metal oxide sensitive layer to different gases.

6. A method for monitoring thermal runaway in lithium batteries, characterized in that, The method includes: The gas sensor according to any one of claims 1-2 is used to monitor the type and concentration of gas released by the lithium battery, and the lithium battery is judged to have thermal runaway based on the type and concentration of gas.

Images