An EMC gas component prediction method based on a V2C / V2O5 / SnO2 sensor and a GRU-Attention

Abstract

This invention belongs to the field of gas sensor detection technology, and particularly relates to an EMC gas component prediction method based on a V2C / V2O5 / SnO2 sensor and GRU-Attention. The EMC gas component prediction method includes: S1, preparing a multilayer V2C MXene; S2, preparing a V2C / V2O5 / SnO2 composite material; S3, preparing a gas sensor; S4, measuring the resistance signal; and S5, establishing a GRU-Attention gas detection model. This invention successfully synthesizes a V2C / V2O5 / SnO2 heterojunction gas-sensitive thin film. This heterojunction, through the synergistic effect of the high conductivity of V2C, the excellent gas-sensing properties of SnO2, and the catalytic activity of V2O5, provides a large number of adsorption and reaction sites for methyl ethyl carbonate molecules at a low temperature of 150℃. The sensor prepared by this invention exhibits significantly improved gas-sensing response, rapid response / recovery speed, and low detection limit, achieving efficient detection of EMC gases over a wide concentration range.

Description

Technical Field

[0001] This invention belongs to the field of gas sensor detection technology, and particularly relates to an EMC gas component prediction method based on V2C / V2O5 / SnO2 sensors and GRU-Attention. Background Technology

[0002] Ethyl methyl carbonate (EMC), an important organic solvent and chemical intermediate, is widely used in lithium-ion battery electrolytes and fine chemicals. However, EMC is volatile, and its vapor poses certain safety risks in high-concentration environments, potentially inhibiting the central nervous system and irritating the respiratory tract and mucous membranes. In industrial environments such as lithium battery production and quality control workshops, real-time and accurate monitoring of EMC gas concentrations is crucial for ensuring production safety, process control, and preventing environmental pollution. Therefore, developing gas sensors capable of rapid and sensitive detection of EMC gases is of significant practical importance.

[0003] Currently, commercial metal oxide semiconductor gas sensors are one of the mainstream technologies for detecting volatile organic compounds (VOCs). These sensors typically rely on a single gas-sensitive material (such as SnO2) undergoing a redox reaction with gas molecules at high temperatures (200-400 °C), causing a change in resistance to achieve detection. However, this approach has significant limitations. First, the high-temperature operating conditions lead to high power consumption, rapid material aging, and shortened lifespan. Second, their sensitivity to low-concentration target gases is limited, and they have poor ability to distinguish structurally similar VOC molecules (such as dimethyl carbonate and diethyl carbonate) in complex environments, resulting in unsatisfactory gas selectivity. Although nanotechnology and noble metal modification can improve performance to some extent, the fundamental contradictions mentioned above have not been effectively resolved. Electrochemical sensors, while capable of operating at room temperature, generally suffer from limited lifespan and severe cross-interference. Optical sensors, on the other hand, are limited in their widespread application due to their complexity and high cost.

[0004] Therefore, developing an EMC gas detection method that can operate at low temperatures or even room temperature and has high sensitivity, excellent selectivity, and good stability has become a key challenge that urgently needs to be overcome in this field. Summary of the Invention

[0005] The main objective of this invention is to provide a method for predicting EMC gas components based on V2C / V2O5 / SnO2 sensors and GRU-Attention.

[0006] To achieve the aforementioned objectives, the technical solution adopted by this invention includes: This invention provides a method for predicting EMC gas components based on V2C / V2O5 / SnO2 sensors and GRU-Attention, comprising the following steps: S1. Preparation of multilayer V2C MXene: Using aluminum vanadium carbide as raw material and hydrofluoric acid and hydrochloric acid solution as etching solution, multilayer V2C MXene is obtained by etching method; S2. Preparation of V2C / V2O5 / SnO2 composite material: Terephthalic acid and NaOH are dispersed in deionized water and stirred continuously to form a homogeneous solution. Then, SnSO4 and V2C MXene are added simultaneously and stirred continuously until a precipitate is formed. The precipitate is separated by centrifugation, washed with deionized water, and vacuum dried at room temperature to obtain Sn-MOF / MXene. Then, the Sn-MOF / MXene is heat-treated to obtain V2C / V2O5 / SnO2 composite material. S3. Preparation of gas sensor: Mix V2C / V2O5 / SnO2 composite material with terpineol, and drop-coat it onto gold and silver electrodes on the substrate surface to form a composite gas-sensitive film, thus obtaining a gas sensor based on V2C / V2O5 / SnO2 composite material. S4. Measuring Resistance Signals: When measuring pure EMC gas, only one signal from the gold electrode is used to measure the resistance signal of Au-V2C / V2O5 / SnO2-Au; when measuring a mixture of EMC and dimethyl carbonate, two signals are used to measure the resistance signals of Au-V2C / V2O5 / SnO2-Au and Au-V2C / V2O5 / SnO2-Ag. S5. Establish a GRU-Attention gas detection model: For the EMC and dimethyl carbonate mixture, a GRU-Attention gas detection model is established using the measured two resistance signals to detect the gas type and concentration of the EMC and dimethyl carbonate mixture.

[0007] Furthermore, the step of preparing multilayer V2C MXene in step S1 includes: adding concentrated hydrochloric acid, a hydrofluoric acid solution with a mass fraction of 38%-40%, and deionized water to a reaction vessel lined with polytetrafluoroethylene, adding aluminum vanadium carbide powder, stirring at 400-500 r / min for 25-30 min at room temperature, then heating to 35-40℃ and reacting for 45-48 h to obtain a black precipitate, washing the black precipitate with deionized water and anhydrous ethanol until the pH value is neutral, and vacuum drying at 60-70℃ for 12-14 h to obtain the multilayer V2C MXene.

[0008] Furthermore, the ratio of hydrofluoric acid, concentrated hydrochloric acid solution, deionized water and aluminum vanadium carbide powder is 10-20 mL: 10-20 mL: 10-20 mL: 1-2 g.

[0009] Furthermore, the step of preparing Sn-MOF in step S2 includes: dissolving terephthalic acid and NaOH in 250-300 ml of deionized water, continuously stirring to form a homogeneous solution, then adding SnSO4, stirring at 80-90℃ for 1-1.5 h, and then continuing to stir at room temperature for 4.5-5 h until a precipitate is formed, centrifuging the precipitate, washing it several times with deionized water, and then vacuum drying it at room temperature to obtain the Sn-MOF.

[0010] Furthermore, the ratio of terephthalic acid, NaOH, and SnSO4 is 2g:1g:6.5g.

[0011] Furthermore, the step of preparing the V2C / V2O5 / SnO2 composite material in step S2 includes: adding terephthalic acid, NaOH and SnSO4 into a reaction vessel, stirring continuously for 25-30 min at 20-25℃ and 400-500 r / min, dissolving 0.45 g of V2C MXene powder in the mixture, stirring continuously at 90℃ for 1-1.5 h, and then stirring continuously at room temperature for 4.5-5 h until a precipitate is formed. The precipitate is centrifuged, washed several times with deionized water, and then vacuum dried at room temperature to obtain Sn-MOF / V2C MXene powder. Then, Sn-MOF / MXene is calcined at 350-400℃ in air atmosphere for 3-4 h to obtain the V2C / V2O5 / SnO2 composite material.

[0012] Furthermore, the mass ratio of the V2C MXene powder to SnSO4 is 0.45g:6.5g.

[0013] Furthermore, the gas sensor prepared in step S3 includes an alumina substrate, a heating electrode, two gold electrodes, a silver electrode, and a V2C / V2O5 / SnO2 sensitive film.

[0014] Furthermore, the gas measurement method in step S4 includes: on the sensor surface, a gold electrode on one side serves as a common electrode, and the other gold and silver electrodes serve as two signal outputs; when measuring pure EMC gas, only the gold electrode signal is used to measure the resistance signal of Au-V2C / V2O5 / SnO2-Au; when measuring the EMC and dimethyl carbonate mixture, two signals are used to measure the resistance signals of Au-V2C / V2O5 / SnO2-Au and Au-V2C / V2O5 / SnO2-Ag.

[0015] Furthermore, the establishment of the GRU-Attention gas detection model in step S5 includes: using the measured resistance signals of Au-V2C / V2O5 / SnO2-Au and Au-V2C / V2O5 / SnO2-Ag as the input of the GRU-Attention network, and the concentrations of EMC gas and dimethyl carbonate gas as the output of the GRU-Attention network; the training of the GRU-Attention model includes the following steps: (1) Data preprocessing and dataset partitioning: The two resistance signals are used as the original input features and normalized. The sample data are randomly divided into training set, test set and validation set in a ratio of 60%:20%:20%. (2) Model initialization and construction: Initialize the GRU-Attention neural network model, take two resistance signals as input, set the loss function to mean absolute error, the optimizer to Adam optimizer, the batch size to 10, and the number of training cycles to 200; the model includes a bidirectional GRU layer, a self-attention mechanism layer and a fully connected layer. (3) Model training and optimization: The model is trained using the training set data. After each training cycle, the loss values ​​on the training set and validation set are calculated. The model is monitored for overfitting or underfitting by the validation set loss value. The model structure is adjusted, and the network structure and parameters with the smallest loss value on the validation set are selected as the final model.

[0016] Compared with the prior art, the advantages of the present invention include: 1. This invention successfully synthesized a V2C / V2O5 / SnO2 heterojunction gas-sensitive thin film and applied it to a semiconductor planar electrode gas sensor for detecting EMC gases. This heterojunction, through the synergistic effect of the high conductivity of V2C, the excellent gas-sensing properties of SnO2, and the catalytic activity of V2O5, provides a large number of adsorption and reaction sites for ethyl methyl carbonate (EMC) molecules at a low temperature of 150℃. Compared with traditional single SnO2-based sensors, the sensor prepared in this invention exhibits significantly improved gas-sensing response (216@500ppm), rapid response / recovery speed (17 / 38s), and a low detection limit (200 ppb), achieving efficient detection of EMC gases over a wide concentration range.

[0017] 2. This invention innovatively designs Au-V₂C / V₂O₅ / SnO₂-Au and Au-V₂C / V₂O₅ / SnO₂-Ag asymmetric electrode structures. Utilizing the asymmetric Schottky barrier formed when gold and silver metals contact the V₂C / V₂O₅ / SnO₂ gas-sensitive material, the same gas-sensitive membrane generates different resistance response signals under different electrode pairs. This "one-channel gas-sensing input, multiple-channel electrical output" mode provides a rich and multi-dimensional source of characteristic information for subsequent gas identification and concentration prediction, fundamentally enhancing the sensor's ability to distinguish different gases.

[0018] 3. This invention constructs a GRU-Attention gas prediction model based on a self-attention mechanism to process two resistance signals acquired by asymmetric electrodes. This model can deeply mine the temporal features in the two dynamic response-recovery curves and adaptively focus on the most critical signal segments and feature dimensions for EMC concentration prediction using an attention mechanism. By inputting these two signals as fusion features into the model, accurate prediction of EMC concentration in a two-component gas mixture is finally achieved, effectively solving the problems of poor selectivity and severe cross-interference of traditional sensors in complex environments. Attached Figure Description

[0019] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 This is a schematic diagram of the structure of the EMC sensor obtained by the present invention; Figure 2 SEM and TEM images of the V2C / V2O5 / SnO2 thin film; Figure 3 Elemental analysis diagram of V2C / V2O5 / SnO2 thin film; Figure 4 XRD and XPS characterization analysis results of V2C / V2O5 / SnO2 thin films; Figure 5 This is one of the sensor performance test diagrams in Embodiment 1 of the present invention; Figure 6 This is the second sensor performance test diagram in Embodiment 1 of the present invention; Figure 7 This is a schematic diagram illustrating the concentration prediction of EMC and DMC mixed gases using the sensor signal combined with the GRU-Attention model of this invention. Figure 8This is a graph showing the concentration prediction results of the GRU-Attention model of this invention for EMC and DMC mixed gases; The diagram is labeled as follows: Gold electrode 1, Silver electrode 2, Heating electrode 3, Sensitive thin film 4, Alumina substrate 5. Detailed Implementation

[0020] In view of the shortcomings of the prior art, the inventors of this invention, through long-term research and extensive practice, have proposed the technical solution of this invention. The following will further explain and illustrate this technical solution, its implementation process, and its principles.

[0021] The present invention will now be described in further detail with reference to specific embodiments. The given embodiments are merely illustrative of the invention and not intended to limit its scope. The embodiments provided below can serve as a guide for further improvements by those skilled in the art and do not constitute a limitation on the invention in any way.

[0022] Unless otherwise specified, the experimental methods used in the following embodiments are conventional methods, performed according to the techniques or conditions described in the literature in this field or according to the product instructions. Unless otherwise specified, the materials used in the following embodiments are commercially available.

[0023] This invention provides a method for predicting EMC gas components based on V2C / V2O5 / SnO2 sensors and GRU-Attention, comprising the following steps: S1. Concentrated hydrochloric acid, a 38%-40% hydrofluoric acid solution, and deionized water are added to a polytetrafluoroethylene-lined reactor. Aluminum vanadium carbide powder is added, and the mixture is stirred at 400-500 rpm for 25-30 minutes at room temperature. Then, it is heated to 35-40°C and reacted for 45-48 hours to obtain a black precipitate. The black precipitate is washed with deionized water and anhydrous ethanol until the pH value is neutral. It is then vacuum dried at 60-70°C for 12-14 hours to obtain the multilayer V2C MXene. The ratio of hydrofluoric acid, concentrated hydrochloric acid solution, deionized water, and aluminum vanadium carbide powder is 10-20 mL: 10-20 mL: 10-20 mL: 1-2 g.

[0024] S2. Dissolve terephthalic acid and NaOH in 250-300 ml of deionized water, stirring continuously to form a homogeneous solution. Then add SnSO4 and stir at 80-90℃ for 1-1.5 h, followed by stirring at room temperature for 4.5-5 h until a precipitate forms. Centrifuge the precipitate, wash it several times with deionized water, and then vacuum dry it at room temperature to obtain the Sn-MOF. The ratio of terephthalic acid, NaOH, and SnSO4 used is 2 g: 1 g: 6.5 g.

[0025] Terephthalic acid, NaOH, and SnSO4 were added to a reaction vessel and stirred continuously for 25-30 min at 20-25℃ and 400-500 r / min. 0.45 g of V2C MXene powder was dissolved in the mixture, and the mixture was stirred continuously at 90℃ for 1-1.5 h, then stirred for another 4.5-5 h at room temperature until a precipitate formed. The precipitate was centrifuged, washed several times with deionized water, and then vacuum dried at room temperature to obtain Sn-MOF / V2C MXene powder. The Sn-MOF / MXene was then calcined at 350-400℃ in air for 3-4 h to obtain the V2C / V2O5 / SnO2 composite material. The mass ratio of V2C MXene powder to SnSO4 was 0.45 g: 6.5 g.

[0026] S3. Preparation of gas sensor: Mix V2C / V2O5 / SnO2 composite material with terpineol, and drop-coat it onto gold and silver electrodes on the substrate surface to form a composite gas-sensitive film, thereby obtaining a gas sensor based on V2C / V2O5 / SnO2 composite material; The gas sensor prepared in step S3 includes an alumina substrate, a heating electrode, two gold electrodes, one silver electrode, and a V2C / V2O5 / SnO2 sensitive film.

[0027] S4. On the sensor surface, a single gold electrode on one side serves as a common electrode, while the other gold and silver electrodes serve as two signal outputs. When measuring pure EMC gas, only the gold electrode signal is used to measure the resistance signal of Au-V2C / V2O5 / SnO2-Au. When measuring a mixture of EMC and dimethyl carbonate, two signals are used to measure the resistance signals of Au-V2C / V2O5 / SnO2-Au and Au-V2C / V2O5 / SnO2-Ag.

[0028] S5. Establish a GRU-Attention gas detection model: For the EMC and dimethyl carbonate mixture, a GRU-Attention gas detection model is established using the measured two resistance signals to detect the gas type and concentration of the EMC and dimethyl carbonate mixture.

[0029] In some embodiments, establishing the GRU-Attention gas detection model in step S5 includes: using the measured resistance signals of Au-V2C / V2O5 / SnO2-Au and Au-V2C / V2O5 / SnO2-Ag as inputs to the GRU-Attention network, and using the concentrations of EMC gas and dimethyl carbonate gas as outputs of the GRU-Attention network; training the GRU-Attention model includes the following steps: (1) Data preprocessing and dataset partitioning: The two resistance signals collected by the asymmetric electrodes were used as the original input features. The two signal data were normalized, and the sample data were randomly divided into training set, test set and validation set in a ratio of 60%:20%:20%. The training set was used for learning model parameters, the validation set was used for hyperparameter optimization and model structure selection, and the test set was used to finally evaluate the generalization performance of the model.

[0030] (2) Model Initialization and Construction: Initialize the GRU-Attention neural network model. The model takes two resistance signals as input. The hyperparameters are set as follows: the loss function is set to mean absolute error (MAE), the optimizer is the adaptive moment estimator (Adam), the batch size is set to 10, and the number of training cycles is set to 200. The core of the model consists of a bidirectional GRU layer, which is used to extract deep features from the time series signal; then a self-attention mechanism layer is introduced, which can dynamically assign appropriate weights to the hidden state of the GRU at each time step, thereby focusing on the feature segments most critical to concentration prediction; finally, a fully connected layer is connected to map the weighted feature vector to the final EMC and DMC gas concentration prediction values.

[0031] (3) Model Training and Optimization: The training set data is input into the GRU-Attention network described above for training. After each training cycle, the model loss (MAE) on the training set and validation set is calculated simultaneously. The model is monitored for overfitting or underfitting by observing the changes in the validation set loss value, and the model structure is adjusted accordingly. Through multiple iterations of training and structure adjustment, the network structure and parameters corresponding to the minimum loss value on the validation set are finally selected, thus completing the construction of the optimal prediction model.

[0032] This invention synthesizes a V2C / V2O5 / SnO2 composite gas-sensitive film by heat-treating V2C MXene and Sn-MOF. The methyl ethyl carbonate (MEC) sensor prepared from this composite film exhibits advantages such as high sensitivity and fast response / recovery speed. It has a detection range of 200 ppb to 500 ppm for MEC gas concentration, effectively detecting even low concentrations of MEC gas at low detection temperatures. Furthermore, by utilizing asymmetric electrodes to acquire multiple signal outputs and establishing a GRU-Attention network model to predict EMC gas components in mixed gases, the gas selectivity of the sensor is significantly improved. This effectively addresses the shortcomings of traditional MEC gas sensors, such as poor selectivity, low sensitivity at low concentrations, and poor durability under high-temperature operating conditions.

[0033] To better understand the technical solution of the present invention, the following detailed discussion is provided in conjunction with specific embodiments.

[0034] Example 1 This embodiment provides a method for predicting EMC gas components based on V2C / V2O5 / SnO2 sensors and GRU-Attention, including the following steps: S1: Add 20 ml of concentrated hydrochloric acid, 20 ml of hydrofluoric acid solution and 20 ml of deionized water to a polytetrafluoroethylene-lined reactor, slowly add 2 g of aluminum vanadium carbide powder, stir at room temperature and 500 r / min for 30 min, then heat to 40 °C and react for 48 h. Wash the black precipitate with deionized water and anhydrous ethanol until the pH value is neutral, and dry it under vacuum at 70 °C for 14 h to obtain V2C MXene powder.

[0035] S2: 2g of terephthalic acid, 1g of NaOH, and 6.5g of SnSO4 were added to a reaction vessel and stirred continuously for 30 min at 20-25℃ and 400-500 r / min. 0.45g of V2C MXene powder was dissolved in the mixture, and the mixture was stirred continuously at 90℃ for 1 h, followed by stirring at room temperature for 5 h until a precipitate formed. The precipitate was centrifuged, washed several times with deionized water, and then vacuum dried at room temperature to obtain Sn-MOF / V2C MXene powder. The prepared Sn-MOF / MXene was then calcined at 400℃ for 3 h in air to obtain the V2C / V2O5 / SnO2 composite material.

[0036] S3: The V2C / V2O5 / SnO2 gas sensor mainly comprises an alumina substrate, a heating electrode, two gold electrodes, one silver electrode, and a V2C / V2O5 / SnO2 sensitive thin film. The V2C / V2O5 / SnO2 composite material is mixed with terpineol and drop-coated onto the substrate surface, ensuring close contact with the gold and silver electrodes, thus preparing the composite gas-sensitive thin film and obtaining a gas sensor based on the V2C / V2O5 / SnO2 composite thin film.

[0037] S4: During gas testing, the gold electrode on a single side of the sensor surface serves as the common electrode, while the other gold and silver electrodes provide two signal outputs. When measuring pure EMC gas, only the gold electrode signal is used to measure the resistance signal of Au-V₂C / V₂O₅ / SnO₂-Au. When measuring a mixture of EMC and dimethyl carbonate (DMC), two signals are used to measure the resistance signals of Au-V₂C / V₂O₅ / SnO₂-Au and Au-V₂C / V₂O₅ / SnO₂-Ag.

[0038] S5: The two resistance signals acquired by the asymmetric electrodes are used as the original input features. These two signal data are normalized, and the 600 sets of sample data are randomly divided into training, testing, and validation sets in a 60%:20%:20% ratio. The training set is used for learning model parameters, the validation set is used for hyperparameter optimization and model structure selection, and the testing set is used for the final evaluation of the model's generalization performance.

[0039] Initialize the GRU-Attention neural network model. This model takes two resistance signals as input. The hyperparameters are set as follows: loss function is set to Mean Absolute Error (MAE), the optimizer is the Adaptive Moment Estimator (Adam), batch size is set to 10, and training epochs are set to 200. The core of the model consists of a bidirectional GRU layer used to extract deep features from the time-series signal; subsequently, a self-attention mechanism layer is introduced, which dynamically assigns appropriate weights to the hidden states of the GRU at each time step, thus focusing on the feature segments most critical for concentration prediction; finally, a fully connected layer is connected to map the weighted feature vector to the final EMC and DMC gas concentration prediction values.

[0040] The training set data is input into the GRU-Attention network described above for training. After each training cycle, the loss value of the model on both the training and validation sets is calculated simultaneously. By observing the changes in the validation set loss value, the model is monitored for overfitting or underfitting, and the model structure is adjusted accordingly. Through multiple iterations of training and structural adjustments, the network structure and parameters corresponding to the minimum loss value on the validation set are finally selected, thus completing the construction of the optimal prediction model.

[0041] Figure 1 This is a schematic diagram of the EMC sensor fabricated according to the present invention, which mainly includes a gold electrode 1, a silver electrode 2, a heating electrode 3, a sensitive thin film 4, and an alumina substrate 5. The sensor size is 1.5 × 1.5 × 0.25 mm. The bottom layer is the heating electrode 3 module; above the heating electrode 3 module is the alumina substrate 5, which is used to isolate the heating electrode 3 and the sensitive thin film layer 4; the gold electrode 1 and the silver electrode 2 are deposited on the alumina substrate 5 by electron beam evaporation; the top layer is the sensitive thin film 4 layer of V2C / V2O5 / SnO2.

[0042] This invention also characterizes and analyzes the morphology of V2C / V2O5 / SnO2 materials, and the specific analysis results are shown in the figure: Figure 2 As shown in figure a, scanning electron microscopy (SEM) images of multilayer V2C MXene reveal that it possesses an accordion-like structure. Figure 2 b represents V2C MXene after heat treatment, indicating that heat treatment resulted in the formation of a large number of V2O5 particles on the surface of V2C MXene. Figure 2c represents the surface morphology of the V2C / V2O5 / SnO2 composite film, in which flocculent SnO2 is attached to multiple MXene layers, providing a larger contact area for gas molecules. Figure 2 d represents SnO2 derived from Sn-MOF. Figure 2 Images e and f are transmission electron microscopy (TEM) images of the V2C / V2O5 / SnO2 composite thin film. The lattice fringes with spacings of 0.87 nm and 0.41 nm correspond to the (002) crystal plane of V2C MXene and the (001) crystal plane of V2O5, respectively. The lattice fringes with a spacing of 0.33 nm correspond to the (110) crystal plane of SnO2.

[0043] This invention also performs elemental characterization analysis on the V2C / V2O5 / SnO2 composite thin film, and the specific analysis results are as follows: Figure 3 As shown, Sn elements mainly originate from SnO2, while V and C elements mainly originate from V2C MXene.

[0044] Figure 4 XRD patterns of the V2C / V2O5 / SnO2 composite films are shown in Figure 1. V2C MXene exhibits two significant characteristic peaks at 2θ = 13.2° and 41.2°, corresponding to the (002) and (113) crystal planes of V2C, respectively. The XRD pattern of V2C / V2O5 shows three significant characteristic peaks at 20.3°, 26.6°, and 31.4°. Simultaneously, due to heat treatment, the characteristic peaks of the (002) plane are significantly weakened, indicating that V2C was partially oxidized at high temperature to form V2O5. The XRD pattern of SnO2 shows five diffraction peaks at (110), (101), (200), (211), and (301). Figure 4 Figure 4(b) shows the XPS spectrum of the V₂C / V₂O₅ / SnO₂ composite material, which contains four major elements (Sn, O, V, and C). As shown in Figure 4(c), the V 2p spectrum can be fitted with four characteristic peaks. The peaks at 524.7 and 516.7 eV belong to V⁵⁺, while the peaks at 523.6 and 522.5 eV belong to V³⁺. Figure 4(d) shows the XPS spectrum of C 1s. The peaks at 288.4, 285.6, 284.4, and 282.9 eV correspond to OC=O, CO, CC, and CV, respectively. The Sn 3d spectrum is shown in Figure 4(e). The binding energies of Sn 3d³ / ² and Sn 3d⁵ / ² are 495.1 and 486.7 eV, respectively. The O 1s peak is shown in Figure 4(f). The peak energies of adsorbed oxygen (OA), oxygen vacancy (OV), and lattice oxygen (OL) are located at 533.7, 532.6, and 531.3 eV, respectively.

[0045] The present invention also tested the performance of the sensor, and the specific results are shown in the figure: Figure 5 a represents the gas-sensing response of V2C / V2O5, SnO2, and V2C / V2O5 / SnO2 sensors to EMC (500 ppm) at different temperatures. These three sensors exhibited the best gas-sensing responses at 120℃, 200℃, and 150℃, respectively, with V2CT showing the best response. x The / V2O5 / SnO2 sensor has the best performance. Figure 5 Figures b, 5c, and 5d illustrate the dynamic resistance changes of these three sensors when exposed to 50-500 ppm EMC gas at their respective optimal operating temperatures. The resistance of the V2C / V2O5 / SnO2 sensor increases with increasing EMC gas concentration, exhibiting n-type semiconductor characteristics. Figure 5 e shows the response and fitting curves of the three sensors. The V2C / V2O5 / SnO2 sensor not only showed a significant improvement in response value in the concentration range of 0-500 ppm, but also had a good linear relationship (R² = 0.9692). Figure 5 f shows the sensor response of composite materials with different mass ratios (V2C:SnO2). When the mass ratio is 1:5, the V2C / V2O5 / SnO2 sensor reaches its peak response to EMC.

[0046] Figure 6 a) is the sensor repeatability test. The V2C / V2O5 / SnO2 sensor has good repeatability for EMC gases of 500ppm and 200ppm. Figure 6 b represents the sensor response and recovery characteristic test. The response / recovery time of the V2C / V2O5 / SnO2 sensor is 17 / 38s. Figure 6 c represents the gas selectivity test of the sensor. The gas sensitivity response of the V2C / V2O5 / SnO2 sensor to 500 ppm CH4, C2H6, DMC, CO and EMC shows that the sensor exhibits excellent selectivity for EMC gas, but there is obvious cross-sensitivity for DMC gas. Figure 6 d indicates that the V2C / V2O5 / SnO2 sensor has a low detection limit and can produce a clear and regular response signal for low concentrations (200-2000 ppb) of EMC. Figure 6 e represents the test results of the effect of humidity on sensor performance; within the 30-60%RH range, the sensor response decreases by only 17%. Figure 6 f indicates that the V2C / V2O5 / SnO2 sensor exhibits excellent long-term stability.

[0047] like Figure 7 As shown, this invention also constructs a GRU-Attention model to predict the concentration of EMC and DMC mixed gases, and the specific results are as follows. Figure 8 As shown: Figure 8 'a' represents the gas ratio of EMC and DMC in the mixed gas. Figure 8 b represents the loss curve during the training of the GRU-Attention model. The loss value decreases rapidly during the first 50 iterations, and reaches its minimum after 60 iterations, indicating model convergence and peak concentration prediction accuracy. Figure 8 c and 8d represent the relative and absolute errors of the GRU-Attention model in predicting EMC gas concentrations, respectively. The prediction accuracy is 93.2%, and the mean absolute error is 11.58 ppm. Figure 8 e and 8f are the relative and absolute errors of the GRU-Attention model in predicting EMC gas concentration, respectively. The prediction accuracy is 91.5%, and the mean absolute error is 13.62 ppm.

[0048] The above descriptions are merely some embodiments of the present invention. It should be noted that those skilled in the art can make other modifications and improvements without departing from the inventive concept of the present invention, and these all fall within the protection scope of the present invention.

Claims

1. A method for predicting EMC gas components based on V2C / V2O5 / SnO2 sensors and GRU-Attention, characterized in that, Includes the following steps: S1. Preparation of multilayer V2C MXene: Using aluminum vanadium carbide as raw material and hydrofluoric acid and hydrochloric acid solution as etching solution, multilayer V2C MXene is obtained by etching method; S2. Preparation of V2C / V2O5 / SnO2 composite material: Terephthalic acid and NaOH are dispersed in deionized water and stirred continuously to form a homogeneous solution. Then, SnSO4 and V2C MXene are added simultaneously and stirred continuously until a precipitate is formed. The precipitate is separated by centrifugation, washed with deionized water, and vacuum dried at room temperature to obtain Sn-MOF / MXene. Then, the Sn-MOF / MXene is calcined at 350-400℃ for 3-4 h in air atmosphere to obtain V2C / V2O5 / SnO2 composite material. S3. Preparation of gas sensor: Mix V2C / V2O5 / SnO2 composite material with terpineol, and drop-coat it onto gold and silver electrodes on the substrate surface to form a composite gas-sensitive film, thus obtaining a gas sensor based on V2C / V2O5 / SnO2 composite material. S4. Measuring Resistance Signals: On the sensor surface, a single gold electrode on one side serves as the common electrode, while the other gold and silver electrodes provide two signal outputs. When measuring pure EMC gas, only the gold electrode signal is used to measure the resistance signal of Au-V2C / V2O5 / SnO2-Au. When measuring a mixture of EMC and dimethyl carbonate, two signals are used to measure the resistance signals of Au-V2C / V2O5 / SnO2-Au and Au-V2C / V2O5 / SnO2-Ag. S5. Establish a GRU-Attention gas detection model: For the EMC and dimethyl carbonate mixture, a GRU-Attention gas detection model is established using the measured two resistance signals to detect the gas type and concentration of the EMC and dimethyl carbonate mixture.

2. The EMC gas component prediction method based on V2C / V2O5 / SnO2 sensor and GRU-Attention according to claim 1, characterized in that, The step S1 for preparing multilayer V2C MXene includes: adding concentrated hydrochloric acid, a 38%-40% hydrofluoric acid solution, and deionized water to a polytetrafluoroethylene-lined reactor; adding aluminum vanadium carbide powder; stirring at 400-500 r / min for 25-30 min at room temperature; then heating to 35-40℃ and reacting for 45-48 h to obtain a black precipitate; washing the black precipitate with deionized water and anhydrous ethanol until the pH value is neutral; and vacuum drying at 60-70℃ for 12-14 h to obtain the multilayer V2C MXene.

3. The EMC gas component prediction method based on V2C / V2O5 / SnO2 sensor and GRU-Attention according to claim 2, characterized in that, The ratio of hydrofluoric acid, concentrated hydrochloric acid solution, deionized water and aluminum vanadium carbide powder is 10-20 mL: 10-20 mL: 10-20 mL: 1-2 g.

4. The EMC gas component prediction method based on V2C / V2O5 / SnO2 sensor and GRU-Attention according to claim 1, characterized in that, The ratio of terephthalic acid, NaOH, and SnSO4 used is 2g:1g:6.5g.

5. The EMC gas component prediction method based on V2C / V2O5 / SnO2 sensor and GRU-Attention according to claim 1, characterized in that, The gas sensor prepared in step S3 includes an alumina substrate, a heating electrode, two gold electrodes, a silver electrode, and a V2C / V2O5 / SnO2 sensitive film.

6. The EMC gas component prediction method based on V2C / V2O5 / SnO2 sensor and GRU-Attention according to claim 1, characterized in that, Step S5, establishing the GRU-Attention gas detection model, includes: using the measured resistance signals of Au-V2C / V2O5 / SnO2-Au and Au-V2C / V2O5 / SnO2-Ag as inputs to the GRU-Attention network, and using the concentrations of EMC gas and dimethyl carbonate gas as outputs of the GRU-Attention network; the training of the GRU-Attention model includes the following steps: (1) Data preprocessing and dataset partitioning: The two resistance signals are used as the original input features and normalized. The sample data are randomly divided into training set, test set and validation set in a ratio of 60%:20%:20%. (2) Model initialization and construction: Initialize the GRU-Attention neural network model, take two resistance signals as input, set the loss function to mean absolute error, the optimizer to Adam optimizer, the batch size to 10, and the number of training cycles to 200; the model includes a bidirectional GRU layer, a self-attention mechanism layer and a fully connected layer. (3) Model training and optimization: The model is trained using the training set data. After each training cycle, the loss values ​​on the training set and validation set are calculated. The model is monitored for overfitting or underfitting by the validation set loss value. The model structure is adjusted, and the network structure and parameters with the smallest loss value on the validation set are selected as the final model.

Images