Electronic nose based on multi-task model and gas detection method
By adopting a multi-task model and a detachable gas sensor in the electronic nose, combined with backpropagation neural network and LoRa wireless transmission, the problem of electronic nose being unable to predict gas concentration and sensor vulnerability is solved, achieving higher versatility and accuracy sex.
Patent Information
- Application Number
- CN202411554782.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-04
- Publication Date
- 2025-01-03
AI Technical Summary
The existing electronic nose has problems in gas detection that cannot predict gas concentration and sensor vulnerability to damage, resulting in equipment failure, reducing its versatility.
An electronic nose based on a multi-task model is designed, using a detachable gas sensor and backpropagation neural network model, and simultaneously detects gas species and concentration by sharing convolutional layer and task branch layer, and uses LoRa wireless transmission technology to achieve it. Data transmission.
It improves the versatility of the electronic nose, can be replaced when the gas sensor is damaged, and the multi-task model can accurately predict gas types and concentrations, making up for the shortcomings of the prior art.
Smart Images
Figure CN119246776A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of gas detection, and in particular to an electronic nose and a gas detection method based on a multi-task model. Background Art
[0002] With the rapid development of bionic tools, human sensory capabilities have been greatly expanded, and the electronic nose (E-nose) is one of the most important of these tools. The electronic nose imitates the human olfactory system and overcomes many defects of human smell, such as the inability to detect toxic and colorless gases. In related technologies, many gas detection models in electronic noses are only trained on public data sets, which often only contain information about gas types, so gas concentrations cannot be predicted. At the same time, the sensors in the electronic nose are usually welded together with the analysis system. When a single sensor is damaged, it often causes the entire electronic nose to fail, which greatly reduces the versatility of the electronic nose. Summary of the invention
[0003] The purpose of the present application is to provide an electronic nose and a gas detection method based on a multi-task model. The gas sensor on the electronic nose is replaceable and can detect the type and concentration of gas at the same time, thereby improving the versatility of the electronic nose.
[0004] To achieve the above objectives, this application provides the following solutions:
[0005] In a first aspect, the present application provides an electronic nose based on a multi-task model, comprising:
[0006] A gas detection module, used to detect the gas-sensitive response value of the gas to be detected;
[0007] A microcontroller module, connected to the gas detection module, and used to determine the type and concentration of the gas to be detected according to the gas-sensitive response value of the gas to be detected;
[0008] Wherein, the gas detection module is a gas sensor array; the gas sensor array includes a plurality of detachable gas sensors, and the gas sensors are used to detect the gas-sensitive response value of the gas to be detected;
[0009] A multi-task model is provided in the microcontroller module, and the multi-task model is used to determine the type and concentration of the gas to be detected according to the gas-sensitive response value of the gas to be detected; the multi-task model is obtained by training a back-propagation neural network model with training samples, and the training samples include gas-sensitive response value sample data of multiple gases and corresponding labels, and the labels are the type and concentration of the gas.
[0010] Furthermore, the back propagation neural network model includes a shared convolutional layer and a task branching layer.
[0011] Furthermore, the gas sensor adopts a micro-electromechanical system gas sensor.
[0012] Furthermore, the microcontroller module comprises:
[0013] A heater, used to heat the gas sensor so that the sensitive material in the gas sensor reacts with the gas to be detected and adsorbs or desorbs the gas to be detected;
[0014] a voltage module, connected to the heater and used to provide a DC voltage to the heater;
[0015] A cache module, connected to the gas sensor, and used to store the gas-sensitive response value of the gas to be detected;
[0016] The controller is connected to the cache module and the voltage module respectively, and is used to send control instructions to the voltage module so that the voltage module generates the DC voltage according to the control instructions, and stores the acquired gas-sensitive response value of the gas to be detected in the cache module, and is also used to determine the type and concentration of the gas to be detected through a set multi-task model.
[0017] Furthermore, the electronic nose based on the multi-task model further includes:
[0018] The wireless transmission module is connected to the controller and is used to send the type and concentration of the gas to be detected to the host computer.
[0019] Furthermore, the wireless transmission module adopts LoRa wireless transmission technology.
[0020] Furthermore, the gas detection module and the microcontroller module are integrated on a circuit board.
[0021] Furthermore, the multi-task model electronic nose also includes a shell, and the circuit board is arranged inside the shell.
[0022] In a second aspect, the present application provides a gas detection method, which is applied to the electronic nose based on the multi-task model described above, comprising:
[0023] Obtaining the gas-sensitive response value of the gas to be detected;
[0024] The gas-sensitive response value of the gas to be detected is input into the multi-task model for detection to obtain the type and concentration of the gas to be detected; wherein the multi-task model is obtained by training a back propagation neural network model with training samples, and the training samples include gas-sensitive response value sample data of multiple gases and corresponding labels, and the labels are the type and concentration of the gas.
[0025] Furthermore, the multi-task model is a trained back-propagation neural network model, and the back-propagation neural network model includes a shared convolution layer and a task branching layer;
[0026] The gas-sensitive response value of the gas to be detected is input into the multi-task model for detection to obtain the type and concentration of the gas to be detected, specifically including:
[0027] Inputting the gas-sensitive response value of the gas to be detected into the shared convolution layer for feature extraction to obtain target features;
[0028] The target feature is input into the task branch layer for detection to obtain the type and concentration of the gas to be detected.
[0029] According to the specific embodiments provided in this application, this application discloses the following technical effects:
[0030] The present application provides an electronic nose and a gas detection method based on a multi-task model. The electronic nose gas detection module includes a plurality of detachable gas sensors, which can be replaced when the gas sensors are damaged, thereby improving the versatility of the electronic nose. At the same time, the multi-task model set in the microcontroller module can simultaneously determine the type and concentration of the gas to be detected according to the gas-sensitive response value of the gas to be detected, thereby making up for the defect that the electronic nose cannot predict the gas concentration. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0032] Figure 1 This is a schematic diagram of functional modules of an electronic nose based on a multi-task model in one embodiment of the present application;
[0033] Figure 2 Schematic diagram of the original resistance curve and gas sensitive response value curve of the gas sensor array detecting different gases in one embodiment of the present application, wherein (a) is the original resistance curve of eight gas sensors to BZ, (b) is the gas sensitive response value curve of eight gas sensors to EA, and (c) is the gas sensitive response value curve of eight gas sensors to IBA;
[0034] Figure 3 A schematic diagram of the structure of a multi-task model provided in another embodiment of the present application;
[0035] Figure 4Schematic diagram of the evaluation results and real-time prediction results of the multi-task model provided in an embodiment of the present application, wherein (a) is the accuracy and loss curve of the 5-fold cross-validation in the classification task, and (b) is the R of the 5-fold cross-validation in the regression task 2 Score and MSE curve, (c) is the confusion matrix of the classification task in the multi-task model, (d) is the violin plot of the regression task in the multi-task model, (e) is the real-time prediction result of the multi-task model for EAC, and (f) is the real-time prediction result of the multi-task model for EA. DETAILED DESCRIPTION
[0036] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0037] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.
[0038] The electronic nose based on the multi-task model provided in the embodiment of the present application is as follows: Figure 1 As shown, the electronic nose based on the multi-task model includes a gas detection module 1 and a microcontroller module 2. The gas detection module 1 is a gas sensor array, which includes a plurality of detachable gas sensors, and the gas sensors are used to detect the gas-sensitive response value of the gas to be detected.
[0039] In this embodiment, the gas sensor adopts a micro-electromechanical system gas sensor, and the gas sensitive response value is the ratio of Ra / Rg, where Ra is the baseline resistance value when the gas sensor does not react with the gas, that is, the resistance of the sensitive material of the gas sensor in the air, and Rg is the real-time resistance change value of the gas sensor for different types and concentrations of gas, that is, the resistance of the sensitive material of the gas sensor in the gas to be tested. When the sensitive material of the gas sensor encounters different types and concentrations of gas, its resistance tends to decrease, and Rg is often less than Ra, so the gas sensitive response value curve obtained by dividing the two is always greater than 1.
[0040] In an exemplary embodiment, the gas sensor array includes eight detachable gas sensors. The number of gas sensors can be set according to specific circumstances. The eight detachable gas sensors in this embodiment are GM-102B, GM-202B, GM-302B, GM-402B, GM-502B, GM-602B, GM-702B and GM-512B. The gas sensor array is used to conduct multiple detection experiments on the gas to be detected. The gas to be detected has different volatile organic compounds: acetone (CP), benzaldehyde (BZ), butyl acetate (BAC), ethanol (EA), ethyl acetate (EAC), isobutyl alcohol (IBA), isopropyl alcohol (IPA), toluene (TL), n-butanol (NBA) and n-propanol (NPA). In each experiment, the gas sensor array is exposed to a specific type of volatile organic compound with a concentration of 10ppm to 100ppm, and the sampling rate is 2Hz. The test process for each gas is repeated five times. The response of the gas sensor is quantified by the gas sensitive response value. Figure 2 (a) shows the original resistance curves of eight gas sensors to benzaldehyde, i.e., the Rg curves. Figure 2 (b) and Figure 2 (c) in the figure shows the gas sensitive response value curves for ethanol and isobutanol, respectively, where the horizontal axis is time and the vertical axis is the sensing response value of each gas sensor to each gas at 10 to 100 ppm, i.e., the gas sensitive response value.
[0041] Continue to see Figure 1 The microcontroller module 2 includes a heater 21, a voltage module 22, a cache module 23 and a controller 24. The heater 21 in the microcontroller module 2 is used to heat the gas sensor so that the sensitive material in the gas sensor reacts with the gas to be detected and adsorbs or desorbs the gas to be detected.
[0042] The voltage module 22 is connected to the heater 21 , and provides a DC voltage to the heater 21 .
[0043] The controller 24 is connected to the voltage module 22 and the buffer module 23 respectively. The controller 24 stores the acquired gas-sensitive response value of the gas to be detected in the buffer module 23. The controller 24 sends a control instruction to the voltage module 22, so that the voltage module 22 generates a DC voltage according to the control instruction.
[0044] In actual use, the voltage module 22 includes a low-pass filter and an emitter follower. Specifically, the host computer and the controller 24 are connected and communicate wirelessly through a wireless serial port. The host computer sends an instruction to specify the duty cycle of the pulse width modulation (PWM) wave. The controller 24 sends a PWM wave with a specified duty cycle according to the received instruction. The PWM wave is converted into a DC voltage through a low-pass filter. The voltage value of the DC voltage is determined by the duty cycle of the PWM wave, and the duty cycle of the PWM wave is set by the host computer; the emitter follower amplifies the input very small current to improve the load capacity of the circuit. In this embodiment, the duty cycle is set to 70%, and the corresponding heating voltage of the heater 21 is 2.31V.
[0045] The buffer module 23 is connected to the gas sensor 1 , and the buffer module 23 stores the gas-sensitive response value of the gas to be detected.
[0046] The controller 24 is also used to obtain the gas-sensitive response value of the gas to be detected from the cache module 23 and determine the type and concentration of the gas to be detected through the set multi-task model. The multi-task model is used to determine the type and concentration of the gas to be detected according to the gas-sensitive response value of the gas to be detected. The multi-task model is obtained by training the back propagation neural network model with training samples, and the training samples include gas-sensitive response value sample data of multiple gases and corresponding labels, and the labels are the type and concentration of the gas.
[0047] like Figure 3 As shown, the back propagation neural network model includes a shared convolution layer 41 and a task branch layer 42. The shared convolution layer 41 includes three layers of neurons, wherein the number of nodes in each layer of neurons is 128, 64, and 32 respectively; the task branch layer 42 includes a classification task and a regression task. The classification task includes a layer of neurons, wherein the number of nodes in each layer of neurons is 10; the regression task includes two layers of neurons, wherein the number of nodes in each layer of neurons is 16 and 1 respectively. In addition, the softmax activation function is applied in the classification task to convert the output of the classification task into the predicted probability of the gas, i.e., the type of gas; the output of the regression task is the concentration of the gas.
[0048] In this embodiment, the SJTU-G919 test system is used to establish a data set. The system uses a self-reconfiguration adjustment circuit and an automatic control script to generate a large-scale gas-sensitive response value data set for different types of volatile organic compounds. The gas-sensitive response value data set is divided proportionally to obtain training samples. The training samples include gas-sensitive response value sample data of multiple gases and corresponding labels, and the labels are the types and concentrations of the gases. The back-propagation neural network model is trained using the training samples, and the trained back-propagation neural network model is used as a multi-task model.
[0049] To verify the generalization performance of the multi-task model, a 5-fold cross validation was performed. The evaluation index curves of the classification task and regression task are as follows: Figure 4 As shown in (a) and (b) in Figure 4 (a) The blue color represents the accuracy and loss curves of the classification task, the light color represents the 5 loss curves corresponding to the 5-fold cross validation, and the dark color represents the average loss curve of the 5 curves; the yellow color represents the accuracy curve of the classification task, the light color represents the 5 accuracy curves corresponding to the 5-fold cross validation, and the dark color represents the average accuracy curve of the 5 curves.
[0050] exist Figure 4 In (b), the blue color represents the determination coefficient R of the regression task. 2 Score and Mean Squared Error (MSE). The light color represents the 5 MSE curves corresponding to the 5-fold cross validation, and the dark color represents the average of the 5 MSE curves, i.e., the average MSE. The yellow color represents the R of the regression task. 2 Score curves, light colors are the 5 determination coefficient curves corresponding to the 5-fold cross validation, and dark colors are the 5 R 2 The average value of the score curve is the average coefficient of determination.
[0051] Table 1 shows the evaluation indicators of the multi-task model in this application tested on edge devices and host computers. Figure 4 (a) Figure 4 As shown in (b) and Table 1, using the multi-task model on the MCU, the accuracy and R 2 The scores are 98.22% and 0.9583 respectively. The confusion matrix and violin plot are shown in Figure 4 (c) and (d) in Figure 4 In (c), the stronger the blue diagonal line, the higher the classification task accuracy. Figure 4 In (d), the closer the center line of the violin plot is to the red dashed line, the more accurate the regression task is. For comparison, the same multi-task model was tested on the host computer, and the accuracy of the classification task reached 98.50%, and the R of the regression task was 2.34. 2 The score is 0.9631, as shown in Table 1. The performance difference between the edge device and the host computer is minimal, and there is no significant difference in the time required for the two methods to make a prediction, which further highlights the potential of multi-task models for fast prediction on edge devices.
[0052] Table 1
[0053] Multi-task model Classification Accuracy <![CDATA[R 2 Score]]> Prediction time Deployed on edge devices 98.22％ 0.9583 2.0ms Deploy on computer 98.50％ 0.9631 1.6ms
[0054] like Figure 4As shown in (e) and (f) in the figure, the multi-task model in this application is used to detect EAC and EA in real time. The actual gas concentrations are 10ppm, 20ppm, 30ppm and 40ppm, and the gas concentration changes every 7.5 seconds. The results show that the multi-task model in this application can effectively perform real-time prediction tasks. The classification accuracy of EAC and EA is 95% and 93.33% respectively, and the R 2 The scores are 0.9450 and 0.9140 respectively.
[0055] The controller 24 uses a micro control unit (MCU), and the MCU model is STM32L431 to reduce power consumption.
[0056] The gas detection module 1 and the microcontroller module 2 in this embodiment are integrated on a circuit board. The size of each side of the circuit board is only 5 cm, and the small size makes the electronic nose highly portable.
[0057] In actual use, the gas sensor is set on the sensor board, and the microcontroller module is set on the main board. The sensor board and the main board are connected by using a board-to-board connector to achieve a pluggable design between boards, which facilitates the replacement of the gas sensor and reduces the failure rate of the electronic nose.
[0058] In this embodiment, the electronic nose based on the multi-task model further includes a shell, and the circuit board is arranged inside the shell. The shell can protect the gas detection module 1 and the microcontroller module 2 from being affected by external dust.
[0059] The electronic nose based on the multi-task model in this embodiment also includes a front-end circuit, which includes a Wheatstone bridge and a 12-bit analog-to-digital converter. The front-end circuit is used to convert the gas-sensitive response value of the gas to be detected obtained by the gas sensor into a digital signal that can be processed by the MCU.
[0060] The electronic nose based on the multi-task model also includes a wireless transmission module, which is connected to the controller 24 and is used to send the type and concentration of the gas to be detected to the host computer. Specifically, the wireless transmission module adopts LoRa wireless transmission technology, which stands for long-distance wireless transmission technology and is a wireless network protocol that uses radio communication to transmit data over a wide range and bandwidth.
[0061] Based on the same inventive concept, the embodiment of the present application also provides a gas detection method for the electronic nose based on the multi-task model involved above. The implementation scheme for solving the problem provided by the method is similar to the implementation scheme recorded in the above-mentioned device, so the specific limitations in one or more gas detection method embodiments provided below can be found above and will not be repeated here.
[0062] In an exemplary embodiment, a gas detection method is provided, comprising steps 101 and 102:
[0063] Step 101, obtaining the gas-sensitive response value of the gas to be detected.
[0064] Step 102, inputting the gas-sensitive response value of the gas to be detected into the multi-task model for detection to obtain the type and concentration of the gas to be detected; wherein the multi-task model is obtained by training a back propagation neural network model with training samples, and the training samples include gas-sensitive response value sample data of multiple gases and corresponding labels, and the labels are the type and concentration of the gas.
[0065] In an exemplary embodiment, the multi-task model is a trained back-propagation neural network model, and the back-propagation neural network model includes a shared convolution layer and a task branch layer. Step 102 specifically includes: inputting the gas-sensitive response value of the gas to be detected into the shared convolution layer for feature extraction to obtain the target feature; inputting the target feature into the task branch layer for detection to obtain the type and concentration of the gas to be detected.
[0066] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.
[0067] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).
[0068] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.
[0069] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0070] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.
Claims
1. An electronic nose based on a multi-task model, characterized in that: The electronic nose based on the multi-task model includes: A gas detection module, used to detect the gas-sensitive response value of the gas to be detected; A microcontroller module, connected to the gas detection module, and used to determine the type and concentration of the gas to be detected according to the gas-sensitive response value of the gas to be detected; Wherein, the gas detection module is a gas sensor array; the gas sensor array includes a plurality of detachable gas sensors, and the gas sensors are used to detect the gas-sensitive response value of the gas to be detected; A multi-task model is provided in the microcontroller module, and the multi-task model is used to determine the type and concentration of the gas to be detected according to the gas-sensitive response value of the gas to be detected; the multi-task model is obtained by training a back-propagation neural network model with training samples, and the training samples include gas-sensitive response value sample data of multiple gases and corresponding labels, and the labels are the type and concentration of the gas.
2. The electronic nose based on the multi-task model according to claim 1, characterized in that: The back propagation neural network model includes a shared convolutional layer and a task branching layer.
3. The electronic nose based on the multi-task model according to claim 2, characterized in that: The gas sensor adopts a micro-electromechanical system gas sensor.
4. The electronic nose based on the multi-task model according to claim 2, characterized in that: The microcontroller module comprises: A heater, used to heat the gas sensor so that the sensitive material in the gas sensor reacts with the gas to be detected and adsorbs or desorbs the gas to be detected; a voltage module, connected to the heater and used to provide a DC voltage to the heater; A cache module, connected to the gas sensor, and used to store the gas-sensitive response value of the gas to be detected; The controller is connected to the cache module and the voltage module respectively, and is used to send control instructions to the voltage module so that the voltage module generates the DC voltage according to the control instructions, and stores the acquired gas-sensitive response value of the gas to be detected in the cache module, and is also used to determine the type and concentration of the gas to be detected through a set multi-task model.
5. The electronic nose based on the multi-task model according to claim 4, characterized in that: The electronic nose based on the multi-task model further includes: The wireless transmission module is connected to the controller and is used to send the type and concentration of the gas to be detected to the host computer.
6. The electronic nose based on the multi-task model according to claim 5, characterized in that: The wireless transmission module adopts LoRa wireless transmission technology.
7. The electronic nose based on the multi-task model according to claim 1, characterized in that: The gas detection module and the microcontroller module are integrated on a circuit board.
8. The electronic nose based on the multi-task model according to claim 7, characterized in that: The electronic nose of the multi-task model further comprises a shell, and the circuit board is arranged inside the shell.
9. A gas detection method, applied to the electronic nose based on the multi-task model according to any one of claims 1 to 8, the gas detection method comprising: Obtaining the gas-sensitive response value of the gas to be detected; The gas-sensitive response value of the gas to be detected is input into the multi-task model for detection to obtain the type and concentration of the gas to be detected; wherein the multi-task model is obtained by training a back propagation neural network model with training samples, and the training samples include gas-sensitive response value sample data of multiple gases and corresponding labels, and the labels are the type and concentration of the gas.
10. The gas detection method according to claim 9, characterized in that: The multi-task model is a trained back-propagation neural network model, and the back-propagation neural network model includes a shared convolution layer and a task branching layer; The gas-sensitive response value of the gas to be detected is input into the multi-task model for detection to obtain the type and concentration of the gas to be detected, specifically including: Inputting the gas-sensitive response value of the gas to be detected into the shared convolution layer for feature extraction to obtain target features; The target feature is input into the task branch layer for detection to obtain the type and concentration of the gas to be detected.