An electronic controller signal calibration method based on a similar sample set

Abstract

The present invention provides a method for calibrating an electronic controller signal based on a similar sample set, including: Step 1: Determine the characteristic quantities that affect the change of the electronic controller signal; Step 2: Perform data preprocessing on the signal sample data collected by the electronic controller; Step 3: Classify the preprocessed sampling signals and their corresponding characteristic quantities to form a similar sample set of the sampling signals; Step 4: Train the similar sample set of the sampling signals and establish a calibration model; Step 5: Determine the similar sample set to which the sampling signal to be calibrated and its characteristic quantities belong; Step 6: Calibrate the signal to be calibrated using the calibration model trained by the similar sample set to which the sampling signal to be calibrated and its characteristic quantities belong. The calibration model established by the present invention can obtain high calibration accuracy and speed, has good generalization performance and network stability, is beneficial to the regulation and risk assessment of electronic controller signals, and has more practical application value.

Description

Technical Field

[0001] The present invention relates to the field of airborne computer system software, and particularly to an electronic controller signal calibration method based on a similar sample set. Background Art

[0002] Airborne computer systems are often applied to aspects such as engine control, engine status monitoring, and electromechanical management. To complete corresponding functions, signal regulation and conversion are often required through the sensor circuit in the aero-engine electronic controller. The output signals of the sensor interface circuit are generally small signals. Since the engine itself generates relatively large noise, the electronic controller also vibrates with the vibration of the engine. In addition, motors, relays, inertial navigation systems, communication devices, and other electronic devices on the aircraft, as well as lightning in the air, will all interfere with the aero-engine electronic controller and are interference sources for the sensor measurement system. When the interference signal is superimposed on the small signal output by the sensor interface circuit, the signal will be distorted, affecting the accuracy of the measurement result. Therefore, during the actual operation of the aero-engine, it is necessary to calibrate the signal of the sensor measurement system.

[0003] Currently, the more commonly used calibration methods mainly include polynomial linear calibration and neural network calibration. The polynomial linear calibration model has a simple structure, but there are also limitations: low-order polynomials cannot achieve good calibration effects for models with more influencing factors; the fitting curves of high-order polynomials are not smooth, usually showing large local waveforms with large fluctuations, and the calculation is more time-consuming. The neural network algorithm has the advantages of high fault tolerance for sample data, strong non-linear mapping ability, self-adaptation, and self-organization, and has been widely used in signal calibration. Currently, most of the electronic controller signal calibration methods based on neural networks usually take all influencing factors and the collected signal sample data as inputs, without performing feature subdivision on the collected signal sample training set for training the network model, which will result in a complex network model and a large number of iterations. In addition, due to the participation of some interference anomalies and sample data with large deviations in the establishment of the calibration model, the calibration accuracy of the calibration model will be affected. Summary of the Invention

[0004] In view of this, the embodiments of the present application provide an electronic controller signal calibration method based on a similar sample set. The calibration model established by this method can obtain high calibration accuracy and speed, has good generalization performance and network stability, is conducive to the regulation and risk assessment of electronic controller signals, and has more practical application value.

[0005] The embodiments of the present application provide the following technical solutions: An electronic controller signal calibration method based on a similar sample set, including:

[0006] Step 1: Determine the characteristic quantities that affect the change of the electronic controller signal;

[0007] Step 2: Perform data preprocessing on the signal sample data collected by the electronic controller;

[0008] Step 3: Classify the preprocessed sampling signals and their corresponding characteristic quantities to form a similar sample set of the sampling signals;

[0009] Step 4: Train the similar sample set of the sampling signals and establish a calibration model;

[0010] Step 5: Determine the similar sample set to which the sampling signal to be calibrated and its characteristic quantities belong;

[0011] Step 6: Calibrate the signal to be calibrated using the calibration model trained with the similar sample set to which the sampling signal to be calibrated and its characteristic quantities belong.

[0012] According to an embodiment of the present application, in Step 1, the factors affecting the signal change of the electronic controller are analyzed for correlation through the Pearson correlation coefficient, and the characteristic quantities affecting the signal change of the electronic controller are obtained.

[0013] According to an embodiment of the present application, in Step 2, the process of the data preprocessing includes: identifying and correcting the abnormal points and supplementing the missing data, smoothing the abnormal points, and reducing the interference of noise data or random fluctuations.

[0014] According to an embodiment of the present application, in Step 3, the Cannopy-optimized K-means clustering algorithm is used to classify the preprocessed sampling signals and their corresponding characteristic quantities, and the optimal number of clusters is determined by the change of the sum of squared errors, and a similar sample set with the number of the optimal number of clusters is formed from the preprocessed sampling signals and their corresponding characteristic quantities.

[0015] According to an embodiment of the present application, in Step 4, the similar sample set of each type of sampling signal and its corresponding characteristic quantity are used as the training input sample set of the model, and the calibration values corresponding to each sampling signal are used as the training output sample set of the model, and the model is trained and iterated to establish a PSO-RBF calibration model for each similar sample set.

[0016] According to an embodiment of the present application, in Step 5, the similar sample set to which the sampling signal to be calibrated and its characteristic quantities belong is determined through the similarity comprehensive index composed of the grey relational degree combined with the cosine similarity.

[0017] According to an embodiment of the present application, in Step 6, through the PSO-RBF calibration model of the similar sample set to which the sampling signal to be calibrated and its characteristic quantities belong, the signal to be calibrated is used as the input of the PSO-RBF calibration model to calibrate the signal to be calibrated, and the output of the PSO-RBF calibration model is the calibration value.

[0018] A method for calibrating electronic controller signals based on a similar sample set according to the present invention, after preprocessing the data of the signal samples collected by the electronic controller, through correlation analysis of the factors affecting the signal change of the electronic controller, the characteristic quantities affecting the signal of the electronic controller are obtained. Then, a similar sample set of the collected signal samples is constructed by the K-means algorithm optimized by Canopy to realize the characteristic subdivision of the collected samples, so that each type of sample set has similar characteristic quantities. Finally, by establishing a PSO-RBF calibration model for each type of similar sample set and its characteristic quantities, the PSO-RBF calibration model of the similar sample set to which the signal to be calibrated belongs is selected to effectively calibrate the signal to be calibrated, and the accuracy and speed of the electronic controller signal calibration are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0020] Figure 1 is a flowchart of the method for calibrating electronic controller signals based on a similar sample set according to an embodiment of the present invention;

[0021] Figure 2 is a flowchart of the K-means clustering algorithm based on Canopy according to an embodiment of the present invention;

[0022] Figure 3 is a flowchart of the PSO-RBF calibration model establishment according to an embodiment of the present invention;

[0023] Figure 4 is the relationship between SSE and the number of clusters k according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0024] The embodiments of the present application will be described in detail below with reference to the drawings.

[0025] It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the drawings and in combination with the embodiments. The technical solutions of the present invention are clearly and completely described. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0026] Such as Figure 1As shown in the figure, an electronic controller signal calibration method based on a similar sample set provided by an embodiment of the present invention includes:

[0027] S1: Determine the characteristic quantities that affect the change of the electronic controller signal.

[0028] Perform a correlation analysis on the factors that mainly affect the change of the electronic controller signal through the Pearson correlation coefficient r. The Pearson correlation coefficient r can measure the strength of the correlation between two variables, and its value range is [-1, +1]. If r > 0, it indicates that the two variables have a positive correlation; if r < 0, it is considered that the two variables have a negative correlation. The greater the absolute value of r, the stronger the correlation between the two variables is considered. The calculation of the Pearson correlation coefficient r is shown in the following formula:

[0029]

[0030] In the formula: r - Pearson correlation coefficient between the sampling signal and a certain influencing factor; n - number of influencing factors; x - specific value of a certain influencing factor; P - average value of the sampling signal. If the Pearson correlation coefficient between the sampling signal and a certain influencing factor satisfies |r| > 0.7, then this influencing factor can be regarded as one of the characteristic quantities of the sampling signal.

[0031] S2: Perform data preprocessing on the signal sample data collected by the electronic controller.

[0032] Perform data preprocessing on the signal sample data collected by the electronic controller, identify, correct, and supplement the abnormal points in the sampling data, smooth the abnormal points, reduce the interference of noise data or random fluctuations. If the sampling value and the calibration value satisfy the following formula, then discard this data.

[0033] max(|Y 采 -Y 标 |)>ε1

[0034] If the sampling value and the calibration value satisfy the following formula:

[0035] ε1〉max(|Y 采 -Y 标 |)>ε2

[0036] Then the data can be corrected according to the following formula.

[0037] Y 采 =Y 标 +(|Y 采 -Y 标 |) / n

[0038] In the formula: Y 采 is the sampling signal value; Y 标 is the sampling signal calibration value; ε1 is the discard threshold; ε2 is the correction threshold; n is the correction coefficient.

[0039] S3: Classify the preprocessed sampled signals and their corresponding characteristic quantities to form a similar sample set of the sampled signals.

[0040] Use the K-means clustering algorithm optimized by Cannopy to classify the preprocessed collected signal sample data and its corresponding characteristic quantities. This method can efficiently obtain the number of clustering clusters and select the center position of the clusters. For the specific process, please refer to Figure 2 . Determine the optimal number of clusters by observing the change of the sum of squared errors (SSE) curve: as the number of clusters gradually increases, the intra-cluster aggregation degree of each cluster will gradually rise. Before reaching the optimal number of clusters, the trend of the SSE curve is relatively steep; when reaching the optimal number of clusters, the trend of the SSE curve becomes relatively stable. Therefore, select the number of clusters where the SSE curve changes suddenly as the number of similar sample sets, so as to form the optimal number of similar sample sets for the overall collected signal samples and their corresponding characteristic quantities.

[0041] S4: Train the similar sample set of the sampled signal and establish a calibration model.

[0042] Establish a radial basis function neural network (PSO-RBF) calibration model combined with the particle swarm optimization algorithm for each type of sampled signal similar sample set and its corresponding characteristic quantity. Among them, the input sample set of the training model is each type of similar sample set and its corresponding characteristic quantity, and the output sample set of the training model is the calibration value corresponding to each sampled signal. Train and iterate the model to establish the PSO-RBF calibration model for each similar sample set and its corresponding characteristic quantity. In the process of establishing the PSO-RBF model, by integrating the RBF network parameters to be optimized: the center vector, the basis width, and the network connection weights into the same vector as the position vector for the particle swarm objective optimization, and by continuously iterating the particle swarm to change the fitness values of all particles, update the position vectors and velocity vectors of all particles until the global extreme value of the particle swarm meets the termination condition, and then decode it into the network parameters of the RBF neural network calibration model. The specific implementation process is as Figure 3 shown.

[0043] S5: Determine the similar sample set to which the sampled signal to be calibrated and its characteristic quantity belong.

[0044] Use the grey relational grade combined with the cosine similarity to form a similarity comprehensive index S to determine the similar sample set to which the sampled signal to be calibrated and its characteristic quantity belong. The grey relational grade R represents the overall correlation between the similar sample set and the characteristic quantity of the sampled signal to be calibrated. The closer the R value is to 1, the stronger the correlation. The cosine similarity D cos is used to describe the similarity of the change trend between the similar sample set and the characteristic quantity of the sampled signal to be calibrated. The closer D cos is to 1, the closer its change trend is. The specific relationship is shown in the following formula:

[0045] S i = αR i + (1 - α)D cos

[0046]

[0047]

[0048]

[0049] where: X i (k) is the value after normalization of the k-th influencing factor of the i-th clustering feature curve; X o (k) is the value after normalization of the k-th influencing factor of the signal to be calibrated; ρ is the discrimination coefficient, and its value range is 0 to 1, usually taking 0.5; α is the empirical weight coefficient, and when the influencing factor changes greatly, its value should be close to 0, otherwise close to 1.

[0050] Calculate the comprehensive similarity index S of the signal to be calibrated and its characteristic quantities with each similar sample set, and the similar sample set with the largest comprehensive similarity index S is the similar sample set to which the signal to be calibrated and its characteristic quantities belong.

[0051] S6: Calibrate the signal to be calibrated using the calibration model trained by the similar sample set to which the signal to be calibrated and its characteristic quantities belong.

[0052] Calibrate the signal to be calibrated through the PSO-RBF calibration model: The input is the signal to be calibrated and its characteristic quantities, the calibration model is the PSO-RBF calibration model of the similar sample set to which the signal to be calibrated and its characteristic quantities belong, and the output of the model is the calibrated signal value of the signal to be calibrated.

[0053] The following further describes the present invention in conjunction with the drawings and embodiments. As Figures 2 - 4 shown, where Figure 2 is the clustering flowchart of the Canopy-based K-means algorithm, Figure 3 is the modeling flowchart of the PSO-RBF calibration model, Figure 4 is the relationship between SSE and the number of clusters k.

[0054] In the first step, it is determined that the factors affecting the signal change of an electronic controller are: electromagnetic interference, acoustic wave interference, temperature, humidity, pressure, vibration intensity, etc. Calculate the Pearson correlation coefficient r between each influencing factor and the sampling signal as shown in Table 1, and it is obtained that the characteristic quantity affecting the signal change of this electronic controller is the temperature value.

[0055] Table 1

[0056]

[0057] In the second step, all the signal sample data collected by the electronic controller are preprocessed, and the abnormal points in the sampled data are identified, corrected and supplemented: the mutant signal sample values with large deviations are discarded, and the collected signal samples with slight offsets caused by noise are corrected.

[0058] In the third step, the Cannopy-optimized K-means clustering algorithm is used to classify all the sampled signal samples and their corresponding temperature values. For the specific implementation process, see Figure 2 。

[0059] Such as Figure 2As shown in the figure, in order to obtain the number of clustering clusters of the sampled signal samples and their corresponding characteristic quantities and select the positions of the cluster centers, the Cannopy-optimized K-means clustering algorithm is used to classify the preprocessed collected signal sample data and their corresponding characteristic quantities. Since the initial clustering centers of the K-means algorithm are selected artificially or randomly and have uncertainty, in order to accelerate the speed of constructing a similar sample set through the K-means algorithm, the Canopy algorithm can be used to accelerate the clustering speed. The Canopy algorithm uses a fast approximate distance metric and two distance thresholds T1 and T2 for calculation, and classifies each collected signal sample into different clusters. The Canopy clustering algorithm can efficiently obtain the number of clustering clusters, select the positions of the cluster centers, and improve the clustering efficiency of the K-means algorithm. During the Canopy clustering algorithm process, each sample object is represented by a multi-dimensional point. All the sampled signal samples and their corresponding characteristic quantity sets are used as the input of the Canopy algorithm, and two distance thresholds T1 and T2 (satisfying T1>T2) are set. Any data P is selected from the input sample set and used as the clustering center of the first cluster class. The Euclidean distance d between the remaining samples and the data Q is calculated. The sample data that meets the condition d<T1 is classified into the cluster class to which Q belongs. Referring to the Euclidean distance d calculated in the above steps, the data that meets the condition d<T2 is removed from the input sample set and is no longer classified into other Canopy clusters. The above steps are repeated until the input sample set is empty and the Canopy algorithm stops, completing the pre-clustering. The pre-clustering result contains a rough clustering result. The number of clusters and the clustering centers generated by the Canopy algorithm are used as the initial setting parameters of the K-means. According to the number of clusters and the clustering centers provided by the Canopy pre-clustering method, the distance between the remaining sample points of the input sample set excluding the clustering centers and the selected clustering centers is calculated. According to the minimum distance criterion, the sample points are classified into the set to which the designated clustering center belongs. The sum of the squares of the distances between the sample points included in different clusters and their corresponding cluster centers is calculated, and the mean value is obtained as the new clustering center. Check whether the clustering center has changed. If the clustering center no longer changes, the final clustering result is generated. If the new clustering center is different from the previous clustering center, the above steps are continued according to the new clustering center until the clustering center no longer changes.

[0060] During the clustering process, the optimal number of clusters is determined by observing the change of the sum of squared errors (SSE) curve. As Figure 4 can be seen, when the number of clusters is less than 3, the SSE curve shows a relatively steep trend; when the number of clusters is greater than 3, the SSE curve becomes relatively flat. Therefore, the optimal number of clusters k for this collected signal sample is 3, and the overall collected signal samples and their corresponding temperature values are divided into 3 similar sample sets.

[0061] Step 4: Establish PSO-RBF calibration models for the three types of similar sample sets of the sampling signals of the electronic controller respectively. By integrating the neural network parameters of the RBF into the target optimization position vector of the same particle swarm, the position and velocity vectors of the particles are updated through continuous iteration of the particle swarm until the population extreme value meets the termination condition. The particle individuals at the global optimal position during the iteration process are decoded and transformed into the network parameters of the RBF neural network calibration model. The input sample set for training the model is a type of similar sample set, and the output sample set for training the model is the calibration value corresponding to each sampling signal. Combining the RBF neural network parameters optimized by the PSO algorithm, establish PSO-RBF calibration models for the three types of similar sample sets respectively. The specific process can be referred to Figure 3 。

[0062] Such as Figure 3 shown, establish a radial basis function neural network (PSO-RBF) calibration model combined with the particle swarm algorithm for each type of sampling signal similar sample set and its corresponding characteristic quantity. First, use floating-point coding to integrate the RBF network parameters to be optimized: the center vector, the base width, and the network connection weights into the same vector as the target optimization position vector. Initialize the particle swarm parameters: set parameters such as the population size of the particles, the acceleration factor, the particle length, the inertia weight factor, and the maximum allowable number of iterations. Calculate the fitness of each particle, obtain the optimal position component of each particle, and update the individual extreme value. Select from the fitness values of all particles in the population, and select and save the population extreme value of the entire particle swarm. Select from the fitness values of all particles in the population, and select and save the population extreme value of the entire particle swarm. If the termination condition is not met, update the position vector and velocity vector of all particles. If the fitness value of a certain particle is better than the fitness value of its experienced optimal position, then save the current position of this particle as the individual historical optimal position, otherwise keep the historical individual optimal position unchanged. If the fitness value of a certain particle is better than the fitness value of the current optimal position in the population, then save the current position of this particle as the global optimal position, and then record the position of this particle, otherwise keep the historical global optimal position. If the termination condition is met, decode all the particle individuals in the particle swarm and transform them into the network parameters in the RBF neural network calibration model. Input the similar sample set to be trained and construct a calibration model based on PSO-RBF.

[0063] Step 5: Calculate the similarity comprehensive index S by combining the grey relational degree and the cosine similarity between the signal to be calibrated and its characteristic quantity temperature value and the three types of similar sample sets respectively. The similar sample set with the largest S value is the similar sample set to which the signal to be calibrated and its characteristic quantity temperature value belong.

[0064] In the sixth step, the signal to be calibrated and its characteristic quantity, the temperature value, are used as the inputs of the PSO-RBF calibration model trained by the similar sample set selected in the fifth step, and the output of the model is the calibration value corresponding to the calibration signal.

[0065] The method of the present invention has been applied to the signal calibration of an electronic controller in a certain project. By constructing a similar sample set to subdivide the signal acquisition samples, a similar corresponding relationship is established between each similar sample set, its characteristic quantity, and the corresponding calibration value. The radial basis neural network optimized by the particle swarm optimization can continuously update and adjust parameters such as the center vector, basis width, and network connection weights of the radial basis function in the calibration network, enabling the calibration network model to effectively calibrate the signal to be calibrated and improving the accuracy of the electronic controller signal calibration.

[0066] The above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed in the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. An electronic controller signal calibration method based on a similar sample set, characterized in that Including: Step 1: Determine the characteristic quantities that affect the signal change of the electronic controller; Step 2: Perform data preprocessing on the signal sample data collected by the electronic controller; Step 3: Use the K-means clustering algorithm optimized by Canopy to classify the preprocessed sampling signals and their corresponding characteristic quantities to form a similar sample set of the sampling signals; Among them, the Canopy algorithm uses a fast approximate distance metric and two distance thresholds T1 and T2 for calculation, and classifies each collected signal sample into different cluster sets; The Canopy clustering algorithm can efficiently obtain the number of clustering clusters and select the position of the cluster center, improving the clustering efficiency of the K-means algorithm. During the Canopy clustering algorithm, each sample object is represented by a multi-dimensional point. All sampling signal samples and their corresponding characteristic quantity sets are used as the input of the Canopy algorithm, and two distance thresholds T1 and T2 are set, where T1 > T2. Select any data P from the input sample set and use this data P as the clustering center of the first cluster class. Calculate the Euclidean distance d between the remaining samples and the data Q. Classify the sample data that meets the condition d < T1 into the cluster class to which Q belongs; According to the Euclidean distance d calculated in the above steps, remove the data that meets the condition d < T2 from the input sample set and no longer classify it into other Canopy clusters; Repeat the above steps until the input sample set is empty and stop the Canopy algorithm, so that the pre-clustering is completed; The pre-clustering result contains a rough cluster set result. Use the number of cluster sets and the clustering center generated by the Canopy algorithm as the initial setting parameters of the K-means; According to the number of cluster sets and the clustering center provided by the Canopy pre-clustering method, calculate the distance between the remaining sample points of the input sample set excluding the clustering center and the selected clustering center; According to the minimum distance criterion, classify the sample points into the set to which the specified clustering center belongs; Calculate the sum of the squares of the distances between the sample points included in different cluster sets and their corresponding cluster centers, and obtain the mean value as the new clustering center; Step 4: Train the similar sample set of the sampling signal and establish a calibration model; Among them, establish a radial basis neural network PSO-RBF calibration model combined with the particle swarm algorithm for each type of sampling signal similar sample set and its corresponding characteristic quantity; Among them, the input sample set of the training model is each type of similar sample set and its corresponding characteristic quantity, and the output sample set of the training model is the calibration value corresponding to each sampling signal. Train and iterate the model to establish the PSO-RBF calibration model of each similar sample set and its corresponding characteristic quantity; During the establishment of the PSO-RBF model, by integrating the RBF network parameters to be optimized: the center vector, the basis width, and the network connection weights into the same vector as the position vector for the particle swarm objective optimization. Through the continuous iteration of the particle swarm, change the fitness values of all particles, update the position vectors and velocity vectors of all particles until the population extreme value of the particle swarm meets the termination condition, and after decoding, convert it into the network parameters of the RBF neural network calibration model; Step 5: Determine the similar sample set to which the sampling signal to be calibrated and its characteristic quantities belong; Step 6: Calibrate the signal to be calibrated using the calibration model trained with the similar sample set to which the sampling signal to be calibrated and its characteristic quantities belong.

2. The method for calibrating an electronic controller signal based on a similar sample set according to claim 1, wherein In Step 1, the factors affecting the signal change of the electronic controller are analyzed for correlation through the Pearson correlation coefficient, and the characteristic quantities affecting the signal change of the electronic controller are obtained.

3. The method for calibrating an electronic controller signal based on a similar sample set according to claim 1, characterized in that, In Step 2, the process of the data preprocessing includes: by identifying and correcting the outliers and supplementing the missing data, smoothing the outliers, and reducing the interference of noise data or random fluctuations.

4. The method for calibrating an electronic controller signal based on a similar sample set according to claim 1, characterized in that In Step 3, the Cannopy-optimized K-means clustering algorithm is used to classify the preprocessed sampling signal and its corresponding characteristic quantities, and the optimal number of clusters is determined by the change of the sum of squared errors, and the preprocessed sampling signal and its corresponding characteristic quantities form a similar sample set with the number of the optimal number of clusters.

5. The method for calibrating an electronic controller signal based on a similar sample set according to claim 1, wherein In Step 5, the similar sample set to which the sampling signal to be calibrated and its characteristic quantities belong is determined through the comprehensive similarity index composed of the grey relational degree combined with the cosine similarity.

6. The method for calibrating an electronic controller signal based on a similar sample set according to claim 1, wherein In Step 6, through the PSO-RBF calibration model of the similar sample set to which the sampling signal to be calibrated and its characteristic quantities belong, the signal to be calibrated is used as the input of the PSO-RBF calibration model to calibrate the signal to be calibrated, and the output of the PSO-RBF calibration model is the calibration value.

Images