TDLAS oxygen concentration prediction method and system based on fuzzy neural network

Abstract

The invention provides a TDLAS (Tunable Diode Laser Absorption Spectroscopy) oxygen concentration prediction method and system based on a fuzzy neural network. The method comprises the following steps: step 1, acquiring calibration data under different environmental conditions; 2, a fuzzy neural network model is constructed, the fuzzy neural network model comprises a fuzzification layer, a fuzzy reasoning layer and a defuzzification layer, the calibration data are used for completing training of the fuzzy neural network model, a loss function is obtained through calculation according to a predicted value and a true value of the model, and the fuzzy neural network model is obtained; optimizing network parameters of the fuzzy neural network model according to gradient information of the loss function; and 3, receiving real-time sensor data by using the trained fuzzy neural network model, and outputting the predicted oxygen concentration. The method has the advantages of high prediction precision, high reliability and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of gas sensors, and in particular to a TDLAS oxygen concentration prediction method and system based on a fuzzy neural network. Background Art

[0002] In many industries such as aerospace, petrochemicals, etc., accurate measurement of oxygen concentration is essential to ensure safe operation and optimize process. Traditional oxygen concentration measurement methods have many limitations, such as slow response speed, low accuracy or poor stability. In recent years, with the development of optical technology, TDLAS (tunable laser absorption spectroscopy) oxygen sensors based on Lambert-Beer law have been widely used due to their high accuracy, high stability and high durability.

[0003] However, the working principle of the TDLAS oxygen sensor determines that its output voltage is not only affected by the oxygen concentration, but also by environmental factors such as pressure and temperature. Specifically, the output voltage of the sensor With oxygen concentration Environmental pressure and temperature There is a complex nonlinear relationship between them. In order to solve this problem, it is usually necessary to establish the functional relationship between these parameters through calibration experiments, that is, , , Furthermore, in order to obtain a unique oxygen concentration value, it is necessary to invert and fit the concentration of central enterprises. in the form of.

[0004] In the prior art, the polynomial-least squares fitting method or the convolutional neural network (CNN) method is usually used to solve the above problem. Among them, the polynomial-least squares fitting method is to approximate the description of the polynomial function by using a polynomial function. and , , Since the nonlinear relationship between , , Due to the complexity and high nonlinearity of the function, ternary quartic polynomials are often used for fitting. Although this method can provide a certain degree of accuracy, the final equation often contains a large number of coefficients, which makes the model too complex to analyze and prone to overfitting. In view of the limitations of traditional mathematical statistics methods, a convolutional neural network (CNN) method is proposed. This method trains a deep learning model with a large amount of calibration data to replace the original unknown function relationship. Compared with the traditional polynomial fitting method, the CNN model can more effectively capture the complex relationship between input variables, thereby significantly improving the prediction accuracy. However, this method also faces challenges, especially when the amount of calibration data is limited, it may cause overfitting problems, thereby affecting the generalization ability of the model. By defining the membership function of the fuzzy neural network, the input variables (sensor output voltage, temperature, pressure, calibrated oxygen concentration) are converted into membership in the fuzzy set (this step is fuzzification), thereby increasing the robustness of the model, reducing the sensitivity to training data, and thus reducing the risk of overfitting. Summary of the invention The technical problem to be solved by the present invention is: In view of the above-mentioned problems in the prior art, a TDLAS oxygen concentration prediction method and system based on a fuzzy neural network with high prediction accuracy and high reliability is provided.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is: A TDLAS oxygen concentration prediction method based on fuzzy neural network comprises the following steps: Step 1: Data collection; Obtain calibration data under different environmental conditions; Step 2: Construction and training of fuzzy neural network model; Constructing a fuzzy neural network model, the fuzzy neural network model includes a fuzzification layer, a fuzzy reasoning layer and a defuzzification layer, the fuzzification layer uses a membership function to calculate the membership of the fuzzy language variable corresponding to each input component, the fuzzy reasoning layer is used to perform reasoning according to preset fuzzy rules to obtain a fuzzy result, and the defuzzification layer is used to convert the fuzzy result into a numerical value; using the calibration data to complete the training of the fuzzy neural network model, and calculating the loss function according to the predicted value and the true value of the model, and optimizing the network parameters of the fuzzy neural network model according to the gradient information of the loss function; Step 3: Use the trained fuzzy neural network model to receive real-time sensor data and output the predicted oxygen concentration.

[0006] As a further improvement of the method of the present invention: in the step 1, obtaining the calibration data under different environmental conditions is to obtain the voltage output value of each TDLAS oxygen sensor under a specified temperature, pressure, and oxygen concentration.

[0007] As a further improvement of the method of the present invention: in step 2, the input parameters of the fuzzy neural network model are defined as sensor output voltage, ambient pressure, and ambient temperature, and the output parameter is oxygen concentration.

[0008] As a further improvement of the method of the present invention: in step 2, the fuzzy reasoning layer includes setting a different number of neuron nodes for each input variable.

[0009] As a further improvement of the method of the present invention: in step 2, the fuzzy neural network model also includes a normalization layer, and the normalization layer uses linear fuzzy logic to map the input value to a predefined interval.

[0010] As a further improvement of the method of the present invention: in step 2, the method for optimizing the network parameters of the fuzzy neural network model is: Step 201: Setting the initial learning rate; Step 202: selecting different batches of samples, using the different batches of samples to train the fuzzy neural network model respectively, and recording the performance index of the fuzzy neural network model after each training, and obtaining a designated number of batches of samples with the best performance of the fuzzy neural network model according to the performance evaluation result; Step 203: further narrow the range of the designated batch of samples with the best performance obtained in step 202 to determine a new test interval, use the new test interval to train the fuzzy neural network model, and repeat step 202 until a stable hyperparameter combination is obtained.

[0011] As a further improvement of the method of the present invention: in the step 3, the output predicted oxygen concentration value is also fed back to the TDLAS oxygen sensor, and the system performs sliding window filtering on the oxygen concentration value output by the fuzzy neural network model each time to obtain the final oxygen prediction value.

[0012] The present invention also provides a TDLAS oxygen concentration prediction system based on a fuzzy neural network, comprising a microprocessor and a memory connected to each other, wherein the microprocessor is programmed or configured to execute the TDLAS oxygen concentration prediction method based on the fuzzy neural network.

[0013] The present invention also provides a computer-readable storage medium, in which a computer program / instruction is stored. The computer program / instruction is programmed or configured to execute the TDLAS oxygen concentration prediction method based on fuzzy neural network through a processor.

[0014] The present invention also provides a computer program product, comprising a computer program / instruction, wherein the computer program / instruction is programmed or configured to execute the TDLAS oxygen concentration prediction method based on fuzzy neural network through a processor.

[0015] Compared with the prior art, the advantages of the present invention are: The present invention uses membership functions to fuzzify input data, combines preset fuzzy rules for reasoning, and converts fuzzy results into precise values ​​through a defuzzification layer, thereby effectively capturing the intrinsic relationship between multidimensional variables such as sensor output voltage, temperature, and pressure. In addition, with the help of the parameter optimization process guided by the loss function, the model can continuously learn and adapt to new measurement conditions to ensure accuracy and reliability for long-term use. It can not only efficiently handle complex and nonlinear changes in environmental conditions, but also achieve high-precision and high-stability oxygen concentration prediction under different environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 Flow chart of the TDLAS oxygen concentration prediction method based on fuzzy neural network in an embodiment of the present invention.

[0017] Figure 2 This is a structural diagram of a fuzzy neural network for fitting oxygen concentration in an embodiment of the present invention.

[0018] Figure 3 Schematic diagram of training error in an embodiment of the present invention.

[0019] Figure 4 This is a graph showing the error between the model-predicted oxygen concentration and the actual oxygen concentration in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] like Figure 1 As shown, this embodiment is a TDLAS oxygen concentration prediction method based on a fuzzy neural network, comprising the following steps: Step 1: Data collection; Obtain calibration data under different environmental conditions.

[0022] In this embodiment, obtaining calibration data under different environmental conditions is to obtain the voltage output value of each TDLAS oxygen sensor under specified temperature, pressure, and oxygen concentration.

[0023] Specifically, each TDLAS oxygen sensor was subjected to five temperature points, 0°C, 10°C, 20°C, 30°C, and 40°C, to obtain the voltage output values ​​of the sensor under different oxygen concentrations and different pressures, as shown in Table 1.

[0024] Table 1 Oxygen sensor calibration data table

[0025] According to the data in Table 1, the precise ambient oxygen concentrations corresponding to different output voltages, different ambient pressures, and different ambient temperatures are extracted, and these calibration data are classified to ensure that each type of data has a set of clear oxygen concentrations as labels. The classified data is divided into two parts, including a training set and a test set. The training set is used to train the fuzzy neural network model, while the test set is used to evaluate the performance of the model.

[0026] Step 2: Construction and training of fuzzy neural network model; A fuzzy neural network model is constructed. The fuzzy neural network model includes a fuzzification layer, a fuzzy reasoning layer and a defuzzification layer. The fuzzification layer uses a membership function to calculate the membership of the fuzzy linguistic variables corresponding to each input component. The fuzzy reasoning layer is used to infer according to preset fuzzy rules to obtain fuzzy results. The defuzzification layer is used to convert the fuzzy results into numerical values. The fuzzy neural network model is trained using calibration data, and the loss function is calculated based on the predicted value and the true value of the model. The network parameters of the fuzzy neural network model are optimized based on the gradient information of the loss function.

[0027] In this embodiment, the input parameters of the fuzzy neural network model are defined as the sensor output voltage, ambient pressure, and ambient temperature, and the output parameter is the oxygen concentration.

[0028] This embodiment defines the membership function of the fuzzy neural network and converts the input variables (sensor output voltage, temperature, pressure, calibrated oxygen concentration) into membership in the fuzzy set, thereby increasing the robustness of the model, reducing the sensitivity to training data, and reducing the risk of overfitting.

[0029] In this embodiment, the fuzzy reasoning layer includes setting a different number of neuron nodes for each input variable.

[0030] Specifically, 3, 3, and 2 neuron nodes are set for each input variable, namely, output voltage, ambient pressure, and ambient temperature, and fuzzy results are inferred according to fuzzy rules. It can be understood that the number of neuron nodes for each input variable can be set according to actual application needs.

[0031] In this embodiment, the fuzzy neural network model further includes a normalization layer, and the normalization layer uses linear fuzzy logic to map the input value to a predefined interval.

[0032] In this embodiment, the method for optimizing the network parameters of the fuzzy neural network model is: Step 201: Setting the initial learning rate; Step 202: different batches of samples are selected, and the fuzzy neural network model is trained using the different batches of samples, and the performance index of the fuzzy neural network model after each training is recorded, and according to the performance evaluation result, a designated batch of samples with the best performance of the fuzzy neural network model is obtained; Step 203: Further narrow the range of the designated batch of samples with the best performance obtained in step 202 to determine a new test interval, use the new test interval to train the fuzzy neural network model, and repeat step 202 until a stable hyperparameter combination is obtained.

[0033] Specifically, set the initial learning rate and test the effect under different batch sample sizes in a small range; gradually narrow the range, repeat the adjustment and verification until a relatively stable hyperparameter combination is obtained. The method for confirming a stable hyperparameter combination is: after multiple adjustments within a certain range, when it is found that the model performance tends to be stable and no longer changes significantly, stop adjusting and confirm that the hyperparameter combination at this time is a relatively stable optimal configuration.

[0034] like Figure 2 As shown, in a specific application embodiment, constructing the structure of the fuzzy neural network includes the following steps: 1. Define the input parameters and output parameters of the fuzzy neural network. The input parameters are the sensor output voltage, ambient pressure, and ambient temperature, and the output parameter is the oxygen concentration.

[0035] 2. Determine the fuzzy neural network structure, including input layer, fuzzification layer, fuzzy reasoning layer, normalization layer, defuzzification layer and output layer.

[0036] 3. The input layer is used to transmit the input samples, namely the output voltage, ambient pressure, and ambient temperature, to the next layer.

[0037] 4. Fuzzification layer, using gbellmf (generalized bell-shaped membership function) to calculate the membership degree of the fuzzy variables corresponding to each input component.

[0038] 5. Fuzzy reasoning layer: set 3, 3, and 2 neuron nodes for each input variable, namely output voltage, ambient pressure, and ambient temperature, and infer the fuzzy results according to fuzzy rules.

[0039] 6. Normalization layer, using linear fuzzy logic, maps the input value to the interval [0,1] in preparation for defuzzification.

[0040] Specifically, the input variables (sensor voltage, temperature, pressure, oxygen concentration of standard gas) are calculated for the mean of the data ( μ ) and standard deviation ( σ ), so that the data is standard normally distributed with a mean of 0 and a standard deviation of 1.

[0041] 7. Defuzzification layer, using gbellmf membership function to realize defuzzification calculation.

[0042] Specifically, by calculating the centroid of the fuzzy output set To determine the control value, the function expression of the centroid of the fuzzy output set is: , in, μi It is i output membership values ​​(the membership values ​​calculated in the fuzzification layer according to gbellmf), xi is the corresponding output value.

[0043] 8. Output layer, get the output of oxygen concentration value.

[0044] The training data is input into the above constructed fuzzy neural network structure for training, so that the model can learn the relationship between input features (temperature, pressure, voltage) and output (oxygen concentration). Figure 3 As shown in the figure, the training error gradually decreases as the number of training iterations increases. In the early stages of training (0 to 20 iterations), the error decreases rapidly, indicating that the model is quickly learning patterns in the data. Between 20 and 40 iterations, the rate of error decrease slows, but it continues to decrease, indicating that the model is gradually optimizing its parameters. After 40 iterations, the error stabilizes and approaches a lower value, indicating that the model has basically converged and further training has little effect on reducing the error.

[0045] Step 3: Use the trained fuzzy neural network model to receive real-time sensor data and output the predicted oxygen concentration.

[0046] In step 3 of this embodiment, the output predicted oxygen concentration value is also fed back to the TDLAS oxygen sensor, and the system performs sliding window filtering on the oxygen concentration value output by the fuzzy neural network model each time to obtain the final oxygen prediction value.

[0047] Specifically, the system performs a sliding window filter of size 20 on the oxygen concentration value output by the model each time to obtain the final accurate oxygen prediction value, reducing the impact of occasional environmental factors and enhancing the stability and reliability of the system.

[0048] The implemented neural network is evaluated using 25°C test data: 1. Data preparation, the output of the sensor at 25°C is tested according to Table 1. The obtained data is used as the test data set of the fuzzy neural network.

[0049] 2. Test: send the above test data set into the fuzzy neural network model and compare the output oxygen concentration of the fuzzy neural network with the actual oxygen concentration. According to the error curve obtained by the test, Figure 4 As shown, the performance and effectiveness of the fuzzy neural network are explained and evaluated. The error between the oxygen concentration predicted by the model and the actual oxygen concentration is within ±1%. Therefore, the trained fuzzy neural network model can be applied to the TDLAS oxygen sensor to predict the oxygen concentration. In summary, the fuzzy neural network model of this embodiment only needs a small amount of calibration data to fit the output voltage value, pressure value, and temperature value of the oxygen sensor and accurately predict the current ambient oxygen concentration value.

[0050] This embodiment can accurately predict the current ambient oxygen concentration by fitting the output voltage value, pressure value, and temperature value of the oxygen sensor. It can be widely used in TDLAS oxygen sensors with an operating temperature range of 0-40°C, a pressure range of 0kPa-110kPa, and an oxygen concentration range of 0%-90%.

[0051] This embodiment also provides a TDLAS oxygen concentration prediction system based on a fuzzy neural network, comprising a microprocessor and a memory connected to each other, wherein the microprocessor is programmed or configured to execute a TDLAS oxygen concentration prediction method based on a fuzzy neural network.

[0052] It is understandable that the above method of this embodiment can be executed by a single device, such as a computer or server, etc., and can also be applied to a distributed scenario and completed by multiple devices in cooperation with each other. In the case of a distributed scenario, one of the multiple devices can only execute one or more steps in the above method of this embodiment, and multiple devices interact to complete the above method. The processor can be implemented in the form of a general-purpose CPU, a microprocessor, an application-specific integrated circuit, or one or more integrated circuits, etc., for executing related programs to implement the above method of this embodiment. The memory can be implemented in the form of a read-only memory ROM, a random access memory RAM, a static storage device, and a dynamic storage device. The memory can store an operating system and other applications. When the above method of this embodiment is implemented by software or firmware, the relevant program code is stored in the memory and called and executed by the processor.

[0053] This embodiment also provides a computer-readable storage medium, in which a computer program / instruction is stored. The computer program / instruction is programmed or configured to execute a TDLAS oxygen concentration prediction method based on a fuzzy neural network through a processor.

[0054] Those skilled in the art should understand that the above-mentioned embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application can take the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the process Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including an instruction device, which implements the functions specified in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The above are only preferred embodiments of the present invention and are not intended to limit the present invention in any form. Although the present invention has been disclosed as above with preferred embodiments, it is not intended to limit the present invention. Therefore, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solution of the present invention shall fall within the scope of protection of the technical solution of the present invention.

[0055] This embodiment also provides a computer program product, including a computer program / instruction, which is programmed or configured to execute a TDLAS oxygen concentration prediction method based on a fuzzy neural network through a processor.

[0056] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A TDLAS oxygen concentration prediction method based on fuzzy neural network, characterized in that: The following steps are involved: Step 1: Data collection; Obtain calibration data under different environmental conditions; Step 2: Construction and training of fuzzy neural network model; Constructing a fuzzy neural network model, the fuzzy neural network model includes a fuzzification layer, a fuzzy reasoning layer and a defuzzification layer, the fuzzification layer uses a membership function to calculate the membership of the fuzzy language variable corresponding to each input component, the fuzzy reasoning layer is used to perform reasoning according to preset fuzzy rules to obtain a fuzzy result, and the defuzzification layer is used to convert the fuzzy result into a numerical value; The fuzzy neural network model is trained using the calibration data, and a loss function is calculated based on the predicted value and the true value of the fuzzy neural network model, and the network parameters of the fuzzy neural network model are optimized based on the gradient information of the loss function; Step 3: Use the trained fuzzy neural network model to receive real-time sensor data and output the predicted oxygen concentration.

2. The TDLAS oxygen concentration prediction method based on fuzzy neural network according to claim 1 is characterized in that: In step 1, obtaining calibration data under different environmental conditions is to obtain the voltage output value of each TDLAS oxygen sensor under specified temperature, pressure, and oxygen concentration.

3. The TDLAS oxygen concentration prediction method based on fuzzy neural network according to claim 1 is characterized in that: In step 2, the input parameters of the fuzzy neural network model are defined as sensor output voltage, ambient pressure, and ambient temperature, and the output parameter is oxygen concentration.

4. The TDLAS oxygen concentration prediction method based on fuzzy neural network according to claim 1 is characterized in that: In step 2, the fuzzy reasoning layer includes setting a different number of neuron nodes for each input variable.

5. The TDLAS oxygen concentration prediction method based on fuzzy neural network according to claim 1 is characterized in that: In step 2, the fuzzy neural network model further includes a normalization layer, and the normalization layer uses linear fuzzy logic to map the input value to a predefined interval.

6. The TDLAS oxygen concentration prediction method based on fuzzy neural network according to claim 1, characterized in that: In step 2, the method for optimizing the network parameters of the fuzzy neural network model is: Step 201: Setting the initial learning rate; Step 202: selecting different batches of samples, using the different batches of samples to train the fuzzy neural network model respectively, and recording the performance index of the fuzzy neural network model after each training, and obtaining a designated number of batches of samples with the best performance of the fuzzy neural network model according to the performance evaluation result; Step 203: further narrowing the range of the best performing batches of samples obtained in step 202 to determine a new test interval, using the new test interval to train the fuzzy neural network model, and repeating step 202 until a stable hyperparameter combination is obtained.

7. The TDLAS oxygen concentration prediction method based on fuzzy neural network according to claim 1 is characterized in that: The step 3 also includes feeding back the output predicted oxygen concentration value to the TDLAS oxygen sensor, and the system performs sliding window filtering on the oxygen concentration value output by the fuzzy neural network model each time to obtain the final oxygen prediction value.

8. A TDLAS oxygen concentration prediction system based on fuzzy neural network, comprising a microprocessor and a memory connected to each other, characterized in that: The microprocessor is programmed or configured to execute the TDLAS oxygen concentration prediction method based on fuzzy neural network as described in any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program / instruction stored therein, characterized in that: The computer program / instruction is programmed or configured to execute the TDLAS oxygen concentration prediction method based on fuzzy neural network described in any one of claims 1 to 7 through a processor.

10. A computer program product comprising a computer program / instructions, characterized in that The computer program / instruction is programmed or configured to execute the TDLAS oxygen concentration prediction method based on fuzzy neural network described in any one of claims 1 to 7 through a processor.

Images