A gas classification system and method based on an array of MEMS sensors

Abstract

A gas classification system and method based on a MEMS sensor array, the system comprising: a MEMS sensor array including a bearing substrate, a plurality of heating elements installed in an array on the bearing substrate, and a plurality of MEMS gas sensors corresponding to the plurality of heating elements; a plurality of voltage followers corresponding to the plurality of MEMS gas sensors; an input end of a microcontroller connected to the plurality of voltage followers through an ADC module, and an output end of the microcontroller connected to the plurality of heating elements through a DAC module and a power amplification module in sequence. The method comprises: changing the heating conditions of the MEMS gas sensors to obtain more gas response data and construct a gas data set; training a gas classification model using the gas data set; testing a gas using the MEMS gas sensors to obtain response data, extracting data features of the response data, and classifying the gas using the gas classification model. The system and method can more accurately identify different gases or gas mixtures at a low cost.

Description

Technical Field

[0001] This invention belongs to the field of intelligent gas identification technology, specifically a gas classification system and method based on a MEMS sensor array. Background Technology

[0002] Currently, most electronic nose systems based on MEMS gas sensors adopt a hardware design scheme that integrates multiple sensors. The purpose of this design is to provide as much raw data as possible for subsequent identification algorithms. However, when the number of physical sensors increases significantly, the system as a whole loses its characteristics of miniaturization, low power consumption, and low cost, thus limiting the application scope of the gas classification system.

[0003] In existing technologies, to enhance the selectivity of MEMS gas sensors, pulse heating modulation is commonly used. This modulation scheme has two main advantages: firstly, high-frequency pulse modulation reduces power consumption while maintaining sensor stability; secondly, low-frequency pulse modulation obtains a "gas dynamic response," thereby improving classification accuracy. Clearly, low-frequency pulse modulation also reduces power consumption, and verification has shown that low-frequency pulses have a beneficial impact on improving classification accuracy. For high-frequency pulse modulation, the modulation scheme is usually relatively simple, often requiring fixed-amplitude square waves, triangular waves, or sine waves. While heating with these waveforms can increase the amount of effective data, it does not significantly reduce the complexity of the network model. Addressing the problem of limited gas response data sources and inaccurate gas classification in existing technologies, there is an urgent need for a novel intelligent gas classification system and method. Summary of the Invention

[0004] To address the problems existing in the prior art, this invention provides a gas classification system and method based on a MEMS sensor array. This system has a simple structure and low manufacturing cost. It can achieve the response effect of a massive number of MEMS gas sensors even with a limited number of sensors, significantly enhancing the source response data and greatly increasing its information content, thus contributing to accurate gas classification. The method is simple to implement and highly intelligent. Under limited resource conditions, it can obtain more data on gas responses by varying the heating conditions of the gas sensors, enabling more accurate identification of different gases or gas mixtures.

[0005] To achieve the above objectives, the present invention provides a gas classification system based on a MEMS sensor array, including a MEMS sensor array, a voltage follower, an ADC module, a DAC module, a power amplifier module, and a microcontroller.

[0006] The MEMS sensor array includes a support substrate, multiple heating elements mounted in an array on the support substrate, and multiple MEMS gas sensors correspondingly mounted on the multiple heating elements;

[0007] Multiple voltage followers are connected to multiple MEMS gas sensors accordingly;

[0008] The input terminal of the microcontroller is connected to multiple voltage followers through an ADC module, and the output terminal of the microcontroller is connected to multiple heating elements in sequence through a DAC module and a power amplifier module.

[0009] As a preferred embodiment, the microcontroller is a ZYNQ chip, which includes a PS section and a PL section. The PS section includes an ARM dual-core processor and a DDR memory module, and the PL section includes an FPGA. The ARM dual-core processor and the FPGA are connected and interact with each other via an AXI bus.

[0010] As a preferred embodiment, the ZYNQ chip is a Xilinx Zynq7020, the ARM dual-core processor module is a Cortex-A9 ARM core, and the DDR memory module is DDR3.

[0011] As a preferred embodiment, the MEMS gas sensor is a carbon monoxide sensor or a hydrogen sensor.

[0012] As a preferred embodiment, the carbon monoxide sensor is model GM-702B.

[0013] As a preferred embodiment, the hydrogen sensor is model GMV-2021B.

[0014] Furthermore, to facilitate the use of low-frequency signals to drive the MEMS gas sensor, a low-pass filter module is also included, which is positioned between the DAC module and the power amplifier module. Since the complete gas response of the MEMS gas sensor occurs on a timescale of several minutes, the amplitude modulation drive signal of the sensor should be a low-frequency signal.

[0015] In this invention, the MEMS gas sensor is mounted on a heating element on a supporting substrate. This allows for easy control of the heating element to alter the operating temperature of the MEMS gas sensor. Due to the flexibility of the MEMS gas sensor under different temperature conditions, its sensitivity can be easily adjusted. By varying the sensor's heating conditions, more information about the gas response can be obtained, leading to response data under different temperature conditions. This increases the diversity and richness of the sample dataset, enabling the training of a gas classification model with higher accuracy. This facilitates more accurate identification of different gases or gas mixtures using this technique. By placing a power amplification module between the DAC module and the heating element, the voltage drive signal from the microprocessor can be amplified, enhancing the signal energy. This allows for control of the amplitude and voltage of the signals received by each sensor, enabling more efficient driving of the heating element to change the operating temperature of the MEMS gas sensor. Equipping each MEMS gas sensor with a voltage follower provides each sensor with an independent dedicated data acquisition channel. Furthermore, connecting the microcontroller to the voltage follower via the ADC module facilitates real-time acquisition of the MEMS gas sensor's response data. This system has a simple structure and low manufacturing cost. It can easily change the operating temperature of multiple MEMS gas sensors, thereby changing the sensitivity of multiple MEMS gas sensors. As a result, it can achieve the response effect of a large number of MEMS gas sensors with a limited number of MEMS gas sensors, significantly enhancing the source response data and greatly increasing the information content of the source response data. This is conducive to obtaining a gas classification model with higher classification accuracy under low cost conditions, and helps to achieve accurate gas classification.

[0016] The present invention also provides a gas classification method based on a MEMS sensor array, which employs a gas classification system based on a MEMS sensor array and includes the following steps;

[0017] Step 1: Obtain the gas dataset;

[0018] S11: Place the MEMS sensor array in the test chamber and prepare a mixed gas sample according to the set component concentration information;

[0019] S12: The microcontroller uses multi-waveform modulation technology to generate k voltage drive signals based on k pre-set waveform drive signals. These voltage drive signals are amplitude-adjustable periodic signals, and each voltage drive signal is assigned a unique drive status identifier signal before being sent to the DAC module. The DAC module then sends the received k voltage drive signals to the low-pass filter module. The low-pass filter module filters the voltage drive signals and sends them to the power amplifier module. The power amplifier module amplifies the input voltage drive signals and outputs them to the heating element, using different voltage drive signals to drive the corresponding heating element for heating operations, enabling the corresponding MEMS gas sensor to operate under different temperature conditions. Once the operating temperature of the MEMS gas sensor stabilizes, a mixed gas with pre-prepared component concentration data is injected into the measuring chamber.

[0020] S13: Enables multiple MEMS gas sensors to respond under mixed gas conditions with set component concentration data. After the multiple MEMS gas sensors reach a steady state, the voltage signal of the corresponding MEMS gas sensor is collected in real time using a voltage follower and sent to the ADC module. The ADC module converts the received analog voltage signal into a digital voltage signal and sends the digital voltage signal to the microcontroller. The microcontroller obtains the response data of the MEMS gas sensors based on the received digital voltage signal. The microcontroller synchronously records the drive status indicator signal and the corresponding response data as the original gas sample data, adds the corresponding gas tag, and stores it in the storage module.

[0021] S14: Inject clean air into the measuring chamber and displace the mixed gas to the outside until the measuring chamber is filled with clean air;

[0022] S15: Repeat S12 to S14 multiple times to obtain a large amount of raw gas sample data;

[0023] S16: Denoise the raw gas sample data, then separate the raw gas data according to the driving state identifier signals of different voltage driving signals to obtain response data under each voltage driving signal; for the separated response data, use the bisection peak finding algorithm and the local inflection point algorithm to extract data features, then use the sample interpolation algorithm to expand the sample dimension, and then use feature point fitting to obtain the feature curve; merge different feature curves belonging to the same response process to form the feature sample data of a single gas, and generate a set of feature sample data corresponding to multiple gas labels to form a gas dataset, thereby completing the construction of the gas dataset;

[0024] S17: Divide the gas dataset into training set, test set and validation set according to the set ratio;

[0025] Step 2: Construct a gas classification model;

[0026] S21: Construct an MLP-based classification model in the microcontroller, wherein the weights and biases of the classification model are set in the storage module of the PS part, and the parts of the classification model other than the weights and biases are set in the PL part, and the PL part communicates with the storage module through DMA.

[0027] S22: The classification model is trained using the training set. During the training process, the weights and biases are continuously updated through the backpropagation algorithm to reduce the value of the loss function and achieve the process of model parameter tuning. Then, the trained classification model is tested using the test set and validated using the validation set to finally obtain the gas classification model.

[0028] Step 3: Classify the gases;

[0029] S31: Place the MEMS gas sensor in the test environment and obtain the response data of the MEMS gas sensor to the gas in the test environment.

[0030] S32: First, the response data is denoised, and then the bisection peak finding algorithm and the local inflection point algorithm are used to extract data features;

[0031] S33: Input the data features as input data into the gas classification model. The gas classification model compares the current data features with the known response feature data, and then identifies and outputs gas classification information.

[0032] As a preferred embodiment, in step S15 of step one, 70% of the gas dataset is divided into a training set, 15% into a test set, and 15% into a validation set.

[0033] Furthermore, in order to better avoid the loss of important feature information, in step S16 of step one, during the process of expanding the sample dimension using the sample interpolation algorithm, the dimension is expanded to 2000.

[0034] This invention proposes a gas classification method based on hybrid waveform modulation technology. By employing multi-waveform modulation technology to drive the heating element of the gas sensor in a MEMS sensor array, the temperature sensitivity of semiconductors can be fully utilized, enabling the MEMS gas sensor to operate under different temperature conditions. This results in a sufficiently rich set of data features for distinguishing gas types during the gas sensor's response, effectively increasing the feature dimension of the gas response data and significantly reducing the complexity of the network model. The voltage driving signal used is an amplitude-adjustable periodic signal, and the MEMS gas sensor response data is acquired under each voltage driving signal. This effectively captures the sensor's response to gas under specific heating conditions, providing gas sample data. By analyzing this data, different types of gases can be detected and classified. This achieves the effect of increasing the number of sensors without increasing the actual number of sensors, enhancing the source data at the data generation source, greatly increasing the information content of the gas data, and obtaining more accurate gas classification results. In each data acquisition process, this invention uses multiple waveform driving signals to generate multiple voltage driving signals, thereby facilitating the use of multiple waveform superposition heating methods to maximize the data features. In the feature extraction process, a binary peak-finding algorithm and a local inflection point algorithm are employed to extract data features. By setting appropriate thresholds, peaks, troughs, and inflection points can be identified, enhancing the feature extraction process and providing a more comprehensive representation of sensor responses. Compared to traditional techniques, this invention further increases the absorption peaks of the gas, improves the gas reaction rate, and provides more data features for the subsequent classification model, thereby significantly reducing the complexity of the classification model and achieving the effect of reducing the complexity of the subsequent MLP neural network structure. Simultaneously, by employing a sample interpolation algorithm to expand the sample dimension, this invention effectively prevents the actual peak values ​​from being filtered out as noise in the subsequent filtering process, thus avoiding the loss of important feature information and further ensuring the accuracy of the classification model. Furthermore, in the process of constructing the classification model, this invention stores the weights and biases in the storage module of the PS part, while enabling the PL part to call DMA for high-speed data transmission, thereby significantly reducing the utilization rate of the FPGA and thus reducing power consumption. This technique effectively balances the complexity of the hardware structure and the algorithm, which is beneficial to improving the target accuracy of the classification task.

[0035] This method is simple to implement and highly intelligent. Under limited resource conditions, it can obtain more data on gas response by changing the heating conditions of the gas sensor, thereby increasing the diversity of the dataset. Richer data can lead to a classification model with higher classification accuracy, thus enabling more accurate identification of different gases or gas mixtures. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the gas classification system in this invention;

[0037] Figure 2 This is a magnified view of a feature point in the response curve in one embodiment of the present invention;

[0038] Figure 3 This is a waveform diagram of the driving signal used in one embodiment of the present invention;

[0039] Figure 4 This is a schematic diagram of the ZYNQ chip structure in this invention;

[0040] Figure 5 This is a schematic diagram of the original gas sample data in this invention;

[0041] Figure 6 The data feature map is extracted using the bisection peak-finding algorithm and the local inflection point algorithm in this invention.

[0042] In the diagram: 1. Microcontroller, 2. DAC module, 3. Power amplifier module, 4. ADC module, 5. MEMS sensor array, 6. MEMS gas sensor, 7. Support substrate. Detailed Implementation

[0043] The invention will now be further described with reference to the accompanying drawings.

[0044] like Figures 1 to 6 As shown, the present invention provides a gas classification system based on a MEMS sensor array, including a MEMS sensor array 5, a voltage follower, an ADC module 4, a DAC module 2, a power amplifier module 3, and a microcontroller 1;

[0045] The MEMS sensor array 5 includes a support substrate 7, multiple heating elements mounted in an array on the support substrate 7, and multiple MEMS gas sensors 6 correspondingly mounted on the multiple heating elements;

[0046] Multiple voltage followers are correspondingly connected to multiple MEMS gas sensors 6;

[0047] The number of heating elements, MEMS gas sensors 6, and voltage followers can be determined based on the expected goals of the classification task and the performance indicators of the optional sensors. As a preferred option, the number of heating elements, MEMS gas sensors 6, and voltage followers are all four.

[0048] The input terminal of the microcontroller 1 is connected to multiple voltage followers through the ADC module 4, and the output terminal of the microcontroller 1 is connected to multiple heating elements in sequence through the DAC module 2 and the power amplifier module 3.

[0049] As a preferred embodiment, the microcontroller 1 is a ZYNQ chip, which includes a PS section and a PL section. The PS section includes an ARM dual-core processor and a DDR memory module, and the PL section includes an FPGA. The ARM dual-core processor and the FPGA are connected and interact with each other via an AXI bus. Figure 4 As shown.

[0050] As a preferred embodiment, the ZYNQ chip is a Xilinx Zynq7020, the ARM dual-core processor module is a Cortex-A9 ARM core, and the DDR memory module is DDR3.

[0051] As a preferred embodiment, the MEMS gas sensor 6 is a carbon monoxide sensor or a hydrogen sensor.

[0052] As a preferred embodiment, the carbon monoxide sensor is model GM-702B.

[0053] As a preferred embodiment, the hydrogen sensor is model GMV-2021B.

[0054] To facilitate driving the MEMS gas sensor using low-frequency signals, a low-pass filter module is also included, which is positioned between the DAC module 2 and the power amplifier module 3. Since the complete gas response of the MEMS gas sensor occurs on a timescale of several minutes, the amplitude modulation drive signal of the sensor should be a low-frequency signal. Therefore, the amplitude modulation drive signal of the MEMS gas sensor should be a low-frequency signal (typically effective at frequencies ≤ 1 Hz). Preferably, the low-pass filter module is a 7th-order Butterworth low-pass filter.

[0055] In this invention, the MEMS gas sensor is mounted on a heating element on a supporting substrate. This allows for easy control of the heating element to alter the operating temperature of the MEMS gas sensor. Due to the flexibility of the MEMS gas sensor under different temperature conditions, its sensitivity can be easily adjusted. By varying the sensor's heating conditions, more information about the gas response can be obtained, leading to response data under different temperature conditions. This increases the diversity and richness of the sample dataset, enabling the training of a gas classification model with higher accuracy. This facilitates more accurate identification of different gases or gas mixtures using this technique. By placing a power amplification module between the DAC module and the heating element, the voltage drive signal from the microprocessor can be amplified, enhancing the signal energy. This allows for control of the amplitude and voltage of the signals received by each sensor, enabling more efficient driving of the heating element to change the operating temperature of the MEMS gas sensor. Equipping each MEMS gas sensor with a voltage follower provides each sensor with an independent dedicated data acquisition channel. Furthermore, connecting the microcontroller to the voltage follower via the ADC module facilitates real-time acquisition of the MEMS gas sensor's response data. This system has a simple structure and low manufacturing cost. It can easily change the operating temperature of multiple MEMS gas sensors, thereby changing the sensitivity of multiple MEMS gas sensors. As a result, it can achieve the response effect of a large number of MEMS gas sensors with a limited number of MEMS gas sensors, significantly enhancing the source response data and greatly increasing the information content of the source response data. This is conducive to obtaining a gas classification model with higher classification accuracy under low cost conditions, and helps to achieve accurate gas classification.

[0056] The present invention also provides a gas classification method based on a MEMS sensor array, which employs a gas classification system based on a MEMS sensor array and includes the following steps;

[0057] Step 1: Obtain the gas dataset;

[0058] S11: Place the MEMS sensor array 5 in the test chamber and prepare a mixed gas sample according to the set component concentration information;

[0059] As a preferred embodiment of the present invention, the mixed gas samples can be prepared according to the component concentration information table in Table 1, and 47 mixed gas samples can be prepared.

[0060] Table 1: Component Concentration Information Table

[0061]

[0062]

[0063]

[0064] S12: Microcontroller 1 employs multi-waveform modulation technology to generate k voltage drive signals based on pre-set k waveform drive signals, and assigns a unique drive status identifier signal to each voltage drive signal. When the voltage drive signal is changed, a corresponding drive status identifier signal is generated synchronously, such as... Figure 3 As shown;

[0065] The k voltage drive signals represent the driving voltages of the gas sensor's heating element under different operating conditions. Each different waveform drive signal corresponds to a different voltage change. The voltage drive signals are adjustable amplitude periodic signals, meaning their amplitude (maximum voltage value) and period (frequency of signal repetition) can be adjusted as needed. These voltage drive signals are used to activate the gas sensor's heating element, which can generate different temperature conditions under different voltage drives, thereby altering the sensor's response characteristics to the gas.

[0066] Then, k voltage drive signals are sent to DAC module 2. DAC module 2 sends the received k voltage drive signals to low-pass filter module. Low-pass filter module filters the voltage drive signals to obtain smooth waveform control signals, which are then sent to power amplifier module 3. Power amplifier module 3 amplifies the input voltage drive signals and outputs them to the heating element. In this way, sufficient energy is generated to drive the corresponding heating element to perform heating operations using different voltage drive signals, so that the corresponding MEMS gas sensor 6 can work under different temperature conditions. After the operating temperature conditions of MEMS gas sensor 6 stabilize, a mixed gas with the prepared component concentration data is injected into the measuring chamber.

[0067] As a preferred embodiment, the frequency of the voltage drive signal is 50Hz;

[0068] S13: Multiple MEMS gas sensors 6 are enabled to respond under mixed gas conditions with set component concentration data. After the multiple MEMS gas sensors 6 reach a steady state, the voltage signal of the corresponding MEMS gas sensor 6 is collected in real time using a voltage follower and sent to the ADC module 4. The ADC module 4 converts the received analog voltage signal into a digital voltage signal and sends the digital voltage signal to the microcontroller 1. The microcontroller 1 obtains the response data of the MEMS gas sensor 6 based on the received digital voltage signal. The microcontroller 1 synchronously records the drive status indicator signal and the corresponding response data as the original gas sample data, adds the corresponding gas tag, and stores it in the storage module.

[0069] S14: injecting clean air into the measurement chamber and displacing the mixed gas to the outside until the measurement chamber is filled with clean air;

[0070] S15: repeating S12 to S14 multiple times to obtain massive original gas sample data;

[0071] S16: performing denoising processing on the original gas sample data to make the original gas data sample more accurate, then separating the original gas data according to driving state identification signals of different voltage driving signals, and obtaining response data under each voltage driving signal after separation;

[0072] For the separated response data, a binary peak-finding algorithm and a local inflection point algorithm are used to extract data features, as Figure 6 shown, points highlighted in color can be used to represent feature points of a response curve. These color-coded points represent key features such as peaks, valleys and inflection points, providing a clear visual representation of important features in the data.

[0073] The foregoing feature extraction algorithm is specifically as follows: for a set of discrete data y1,y2,...,y n , for each data point y i (2≤i≤n-1): if y i >y i-1 +w and y i >y i+1 +b, then y i is a peak. If y i <y i-1 -w ′ and y i <y i+1 -b ′ , then y i is a valley. If |k i-1 |<E and |k i+1 |>E ′ , then y i is an inflection point. Where w, b, w ′ , b ′ represent determination parameters for peaks and valleys, E represents a determination parameter for inflection points, k i-1 represents the slope between y i and y i-1 , k i+1 represents the slope between y i and y i+1 The size and step size of the sliding window will affect the recognition accuracy of the algorithm. From the perspective of the features and variation trend of the data itself, for a peak point, the data interval on the left side is smaller, and the data interval on the right side is larger. Therefore, when selecting parameters, it is necessary to ensure w < b. For a valley point, the situation is exactly the opposite, it must be ensured that w′ >b ′ A grid search method was used to determine the window length and step size. After multiple rounds of experiments, the parameter values ​​were determined as follows: w = 0.01305, b = 0.01633, w ′ =0.00954, b ′ =0.00538, E=0.03, E ′ =0.1, window length is 300, step size is 10;

[0074] Then, a sample interpolation algorithm is used to expand the sample dimensions, and then feature curves are obtained through feature point fitting. Different feature curves belonging to the same response process are merged to form feature sample data for a single gas. Multiple gas labels correspond to a set of feature sample data, forming a gas dataset, thus completing the construction of the gas dataset. Figure 5 As shown;

[0075] S17: Divide the gas dataset into training set, test set and validation set according to the set ratio;

[0076] Step 2: Construct a gas classification model;

[0077] S21: Build an MLP-based classification model in microcontroller 1. The weights and biases of the classification model are set in the storage module of the PS part, and the parts of the classification model other than the weights and biases are set in the PL part. The PL part communicates with the storage module through DMA. In this way, the use of hardware logic can be significantly reduced.

[0078] As a preferred approach, based on the principles of MLP, the weight matrices are defined as follows: the first layer weight matrix is ​​weights1

[50]

[2000] , the second layer weight matrix is ​​weights2

[50]

[50] , and the third layer (output layer) weight matrix is ​​weights3

[50] [8]. The bias matrices are defined as biases1

[50] , biases2

[50] , and biases3[8]. Each array element is a signed floating-point number. If these weights are implemented in FPGA hardware, they will consume a significant portion of BRAM (Block Random Access Memory) resources.

[0079] S22: The classification model is trained using the training set. During the training process, the weights and biases are continuously updated through the backpropagation algorithm to reduce the value of the loss function and achieve the process of model parameter tuning. Then, the trained classification model is tested using the test set and validated using the validation set. The accuracy of the gas classification model is shown in Table 2. Finally, the gas classification model is obtained.

[0080] Table 2: Accuracy Ranking of Gas Classification Models

[0081]

[0082] Step 3: Classify the gases;

[0083] S31: Place the MEMS gas sensor 6 in the environment to be tested and obtain the response data of the MEMS gas sensor 6 to the gas in the environment to be tested;

[0084] S32: First, the response data is denoised, and then the bisection peak finding algorithm and the local inflection point algorithm are used to extract data features;

[0085] S33: Input the data features as input data into the gas classification model. The gas classification model compares the current data features with the known response feature data, and then identifies and outputs gas classification information.

[0086] As a preferred embodiment, in step S15 of step one, 70% of the gas dataset is divided into a training set, 15% into a test set, and 15% into a validation set.

[0087] To better avoid the loss of important feature information, in step S16 of step one, the dimension is expanded to 2000 during the process of expanding the sample dimension using the sample interpolation algorithm.

[0088] This invention proposes a gas classification method based on hybrid waveform modulation technology. By employing multi-waveform modulation technology to drive the heating element of the gas sensor in a MEMS sensor array, the temperature sensitivity of semiconductors can be fully utilized, enabling the MEMS gas sensor to operate under different temperature conditions. This results in a sufficiently rich set of data features for distinguishing gas types during the gas sensor's response, effectively increasing the feature dimension of the gas response data and significantly reducing the complexity of the network model. The voltage driving signal used is an amplitude-adjustable periodic signal, and the MEMS gas sensor response data is acquired under each voltage driving signal. This effectively captures the sensor's response to gas under specific heating conditions, providing gas sample data. By analyzing this data, different types of gases can be detected and classified. This achieves the effect of increasing the number of sensors without increasing the actual number of sensors, enhancing the source data at the data generation source, greatly increasing the information content of the gas data, and obtaining more accurate gas classification results. In each data acquisition process, this invention uses multiple waveform driving signals to generate multiple voltage driving signals, thereby facilitating the use of multiple waveform superposition heating methods to maximize the data features. In the feature extraction process, a binary peak-finding algorithm and a local inflection point algorithm are employed to extract data features. By setting appropriate thresholds, peaks, troughs, and inflection points can be identified, enhancing the feature extraction process and providing a more comprehensive representation of sensor responses. Compared to traditional techniques, this invention further increases the absorption peaks of the gas, improves the gas reaction rate, and provides more data features for the subsequent classification model, thereby significantly reducing the complexity of the classification model and achieving the effect of reducing the complexity of the subsequent MLP neural network structure. Simultaneously, by employing a sample interpolation algorithm to expand the sample dimension, this invention effectively prevents the actual peak values ​​from being filtered out as noise in the subsequent filtering process, thus avoiding the loss of important feature information and further ensuring the accuracy of the classification model. Furthermore, in the process of constructing the classification model, this invention stores the weights and biases in the storage module of the PS part, while enabling the PL part to call DMA for high-speed data transmission, thereby significantly reducing the utilization rate of the FPGA and thus reducing power consumption. This technique effectively balances the complexity of the hardware structure and the algorithm, which is beneficial to improving the target accuracy of the classification task.

[0089] This method is simple to implement and highly intelligent. Under limited resource conditions, it can obtain more data on gas response by changing the heating conditions of the gas sensor, thereby increasing the diversity of the dataset. Richer data can lead to a classification model with higher classification accuracy, thus enabling more accurate identification of different gases or gas mixtures.

Claims

1. A gas classification system based on a MEMS sensor array, comprising a MEMS sensor array (5), and further comprising a voltage follower, an ADC module (4), a DAC module (2), a power amplifier module (3), and a microcontroller (1). The MEMS sensor array (5) includes a support substrate (7), multiple heating elements mounted in an array on the support substrate (7), and multiple MEMS gas sensors (6) correspondingly mounted on the multiple heating elements. Multiple voltage followers are correspondingly connected to multiple MEMS gas sensors (6); The input terminal of the microcontroller (1) is connected to multiple voltage followers through the ADC module (4), and the output terminal of the microcontroller (1) is connected to multiple heating elements through the DAC module (2) and the power amplifier module (3) in sequence. The microcontroller (1) is a ZYNQ chip, which includes a PS part and a PL part. The PS part includes an ARM dual-core processor and a DDR memory module, and the PL part includes an FPGA. The ARM dual-core processor and the FPGA are connected and interact with each other via an AXI bus.

2. The gas classification system based on a MEMS sensor array according to claim 1, characterized in that, The ZYNQ chip is a Xilinx Zynq7020, the ARM dual-core processor is a Cortex-A9 ARM core, and the DDR memory module is DDR3.

3. The gas classification system based on a MEMS sensor array according to claim 1, characterized in that, The MEMS gas sensor (6) is a carbon monoxide sensor or a hydrogen sensor.

4. A gas classification system based on a MEMS sensor array according to claim 3, characterized in that, The carbon monoxide sensor is model GM-702B.

5. A gas classification system based on a MEMS sensor array according to claim 3, characterized in that, The hydrogen sensor is model GMV-2021B.

6. A gas classification system based on a MEMS sensor array according to claim 1, characterized in that, It also includes a low-pass filter module, which is located between the DAC module (2) and the power amplifier module (3).

7. A gas classification method based on a MEMS sensor array, employing the gas classification system based on a MEMS sensor array as described in claim 2, characterized in that, Includes the following steps; Step 1: Obtain the gas dataset; S11: Place the MEMS sensor array (5) in the test chamber and prepare a mixed gas sample according to the set component concentration information; S12: The microcontroller (1) uses multi-waveform modulation technology to generate k voltage drive signals based on k pre-set waveform drive signals. The voltage drive signals are amplitude-adjustable periodic signals, and a unique drive status identifier signal is set for each voltage drive signal. The signal is then sent to the DAC module (2). The DAC module (2) sends the received k voltage drive signals to the low-pass filter module. The low-pass filter module filters the voltage drive signals and then sends them to the power amplifier module (3). The power amplifier module (3) amplifies the input voltage drive signals and outputs them to the heating element. Different voltage drive signals are used to drive the corresponding heating element to perform heating operations, so that the corresponding MEMS gas sensor (6) works under different temperature conditions. When the working temperature conditions of the MEMS gas sensor (6) are stable, a mixed gas with the prepared component concentration data is injected into the test chamber. S13: Enable multiple MEMS gas sensors (6) to respond under mixed gas conditions with set component concentration data. When multiple MEMS gas sensors (6) reach steady state, use a voltage follower to collect the voltage signal of the corresponding MEMS gas sensor (6) in real time and send it to the ADC module (4). After the ADC module (4) converts the received analog voltage signal into a digital voltage signal, it sends the digital voltage signal to the microcontroller (1). The microcontroller (1) obtains the response data of the MEMS gas sensor (6) according to the received digital voltage signal. The microcontroller (1) synchronously records the drive status indicator signal and the corresponding response data as the original gas sample data, adds the corresponding gas label, and stores it in the storage module. S14: Inject clean air into the test chamber and displace the mixed gas to the outside until the test chamber is filled with clean air; S15: Repeat S12 to S14 multiple times to obtain a large amount of raw gas sample data; S16: Denoise the raw gas sample data, then separate the raw gas sample data according to the driving state identifier signals of different voltage driving signals to obtain response data under each voltage driving signal; for the separated response data, use the bisection peak finding algorithm and the local inflection point algorithm to extract data features, then use the sample interpolation algorithm to expand the sample dimension, and then use feature point fitting to obtain the feature curve; merge different feature curves belonging to the same response process to form the feature sample data of a single gas, and generate a set of feature sample data corresponding to multiple gas labels to form a gas dataset, thereby completing the construction of the gas dataset; S17: Divide the gas dataset into training set, test set and validation set according to the set ratio; Step 2: Construct a gas classification model; S21: Construct an MLP-based classification model in the microcontroller (1), wherein the weights and biases of the classification model are set in the storage module of the PS part, and the parts of the classification model other than the weights and biases are set in the PL part, and the PL part communicates with the storage module through DMA. S22: The classification model is trained using the training set. During the training process, the weights and biases are continuously updated through the backpropagation algorithm to reduce the value of the loss function and achieve the process of model parameter tuning. Then, the trained classification model is tested using the test set and validated using the validation set to finally obtain the gas classification model. Step 3: Classify the gases; S31: Place the MEMS gas sensor (6) in the test environment and obtain the response data of the MEMS gas sensor (6) to the gas in the test environment; S32: First, the response data is denoised, and then the bisection peak finding algorithm and the local inflection point algorithm are used to extract data features; S33: Input the data features as input data into the gas classification model. The gas classification model compares the current data features with the known response feature data, and then identifies and outputs gas classification information.

8. A gas classification method based on a MEMS sensor array according to claim 7, characterized in that, In step S15 of step one, 70% of the gas dataset is divided into the training set, 15% into the test set, and 15% into the validation set.

9. A gas classification method based on a MEMS sensor array according to claim 7, characterized in that, In step S16 of step one, during the process of expanding the sample dimension using the sample interpolation algorithm, the dimension is expanded to 2000.

Images