Method and system for detecting concentration of separation gel in real time

Abstract

The invention relates to the technical field of concentration detection, in particular to a separation gel concentration real-time detection method and system.The method comprises the following steps that on the basis of input parameters of detection signals, multi-dimensional characteristic values of concentration detection signals are extracted, normalization processing is carried out on signal values in combination with gel matrix characteristics, and the concentration of the separation gel is obtained; after interference values are removed, linear variation and fitting degree calculation are carried out on the feature combination, and a concentration feature representation value is generated. According to the method, the weight is adjusted by combining abnormal deviation value extraction with a sparsity constraint rule, redundant features are optimized, and the influence of interference factors on a result is remarkably reduced. Through combination of nonlinear concentration interval weight distribution and dynamic weighting operation, concentration distribution fitting precision and adaptability in a complex environment are improved. By re-screening decoding parameters and iterative calculation, the accuracy and stability of a concentration distribution decoding result are ensured, a macroscopic trend value and a microcosmic difference value are integrated through multi-scale calculation, and the dynamic adaptability and data analysis depth of concentration prediction are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of concentration detection, and in particular to a method and system for real-time detection of separation gel concentration. Background Art

[0002] The technical field of separation gel concentration detection includes various methods and devices for detecting the concentration of chemical substances. The core content of this technical field is the real-time or offline analysis and measurement of the concentration of substances in different chemical solutions through physical, chemical or optical means, and is often used in the fields of pharmaceuticals, materials science, biomedicine, etc. Its overall technical system covers a variety of measurement methods, including the use of spectral analysis, electrochemical sensors, and detection methods based on density or viscosity changes. This technical field not only focuses on the accuracy and real-time nature of detection, but also has high requirements for the adaptability of the detection method, data acquisition and analysis capabilities, and the non-destructiveness of the sample, which constitutes an important part of concentration detection technology.

[0003] Among them, the real-time detection method of separation gel concentration refers to a method for real-time monitoring of the concentration of separation gel solution through specific detection means. This patent subject adopts a measurement method based on the change of optical refractive index in view of the characteristics of the change of separation gel concentration, and combines the sensor array to realize the rapid collection and analysis of concentration changes. The specific method includes using a laser light source to irradiate the separation gel solution, and establishing a quantitative correlation between the concentration and the light signal by receiving the change signal of the scattered light intensity or the reflected light wavelength, and realizing real-time concentration detection in combination with a pre-calibrated calibration model. This method is based on the principles of physical optics and quantitative models, covering a full set of solutions from the design of optical devices to the signal processing process to ensure the stability and timeliness of concentration detection.

[0004] Existing technologies rely on a single detection mode and lack a multi-dimensional feature value extraction and optimization mechanism, making it difficult to effectively eliminate interference factors, which affects detection stability. Feature processing methods are insufficient in identifying abnormal deviation values, resulting in redundant features significantly interfering with the results. Signal processing methods cannot adapt to nonlinear concentration change environments. The calibration model is based on static parameters and lacks dynamic adjustment capabilities, making it difficult to meet detection needs under complex conditions. The ability to integrate complex concentration distributions and multi-scale features is limited, which restricts the in-depth analysis of detection data, reduces its applicability in dynamic response scenarios, and may affect the accuracy and timeliness of key measurement results. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a real-time detection method and system for separation gel concentration.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: a real-time detection method for separation gel concentration, comprising the following steps: S1: Based on the input parameters of the detection signal, the multidimensional characteristic values ​​of the concentration detection signal are extracted, the signal values ​​are normalized in combination with the characteristics of the colloidal matrix, and after eliminating the interference values, the linear change and fitting degree of the characteristic combination are calculated to generate the concentration characteristic representation value; S2: Based on the concentration feature representation value, combined with the concentration distribution characteristics, extract the abnormal deviation value in the feature distribution, adjust the weight of the deviation value according to the sparsity constraint rule, optimize the redundant feature combination through multiple difference comparisons, and generate a potential feature constraint value; S3: Based on the potential characteristic constraint value, the characteristic deviation value is weighted in combination with the nonlinear concentration interval, a dynamic weighted operation is used to calculate the characteristic error distribution value, and the distribution value and the weight ratio of the dynamic adjustment parameter are integrated to generate a dynamic concentration fitting error value; S4: Based on the dynamic concentration fitting error value, the concentration distribution decoding parameters are re-screened in combination with the separation gel interface characteristics, the nonlinear prediction deviation is iteratively calculated, and a combination of the concentration distribution transformation value and the nonlinear characteristic distribution value is extracted to generate a concentration distribution decoding value; S5: Based on the concentration distribution decoding value, multi-scale calculation is performed on the macroscopic concentration characteristic trend value and the microscopic characteristic difference value, the difference between the trend value and the distribution interval is integrated, the fitting result of the distribution characteristic is calculated, and the characteristic integration after multiple processing of the concentration data is completed to generate the separation gel concentration prediction result value.

[0007] As a further solution of the present invention, the step of obtaining the concentration characteristic representation value is specifically as follows: S111: Based on the input parameters of the detection signal, extract the time domain eigenvalues ​​of the signal, obtain the time series parameters by calculating the signal changes in the differentiated time window, extract the amplitude domain eigenvalues ​​of the signal, calculate the signal peak value and distribution pattern to obtain the amplitude parameter, extract the frequency domain eigenvalues ​​of the signal, analyze the distribution and changes of the frequency components to obtain the frequency parameter, eliminate abnormal signals that exceed the data range, and generate a signal feature matrix; S112: performing normalization processing based on the signal feature matrix and in combination with the characteristics of the colloidal matrix, adjusting the eigenvalue range to match the colloidal matrix, calculating the eigenvalue deviation, eliminating unmatched data points, and generating a normalized feature matrix; S113: Based on the normalized feature matrix, linearly change the feature combination, calculate the weighted average to form the feature combination, analyze the fit of the combination and optimize the weight, integrate the fitting results of all features, and generate a concentration feature representation value.

[0008] As a further solution of the present invention, the weighted average is calculated to form a feature combination using the formula: ; Calculate weighted eigenvalues , analyze the combined fit and optimize the weight, integrate the fitting results of all features, and generate concentration feature representation values; in, Indicates the The weighted eigenvalues ​​of the features, Indicates the The sample in The absolute difference of the features, Indicates the The average value of the features, Indicates the The sample in The weight value on the feature, Express Sum of samples, Represents the square root of the sum of the squares of the weight values.

[0009] As a further solution of the present invention, the step of obtaining the potential feature constraint value is specifically as follows: S211: Based on the concentration characteristic representation value, analyze the concentration data in the characteristic distribution, calculate the mean and distribution range of the characteristic value, identify abnormal deviation values ​​by comparing the characteristic value with the upper and lower boundaries, extract characteristic deviation values ​​that exceed the boundaries, eliminate repeated and coupled deviation values, and generate an abnormal deviation characteristic value set; S212: Based on the abnormal deviation feature value set, combined with the sparsity constraint rule, the difference between the feature value and the adjacent feature value is calculated, the weight adjustment feature is assigned to adjust the impact of the feature on the overall distribution, the weight distribution is optimized according to the relationship between the deviation value and the overall proportion, and a sparsity weight adjustment result is generated; S213: Based on the sparsity weight adjustment result, multiple difference comparisons are performed on the adjusted deviation characteristic values ​​and the concentration characteristic representation values, the redundant relationship between the feature distributions is gradually calculated, the redundant feature combinations are removed, the optimized feature set is integrated, and the potential feature constraint value is generated.

[0010] As a further solution of the present invention, the step of obtaining the dynamic concentration fitting error value is specifically as follows: S311: Based on the potential characteristic constraint value and the nonlinear concentration range, the characteristic deviation values ​​are analyzed and grouped. During the analysis, the characteristic value range is selected based on the parameter type and distribution characteristics. The initial weight of each group of characteristic deviation values ​​is adjusted according to the adaptation characteristics to generate a group weight distribution table. S312: Based on the group weight allocation table and in combination with the dynamic parameter adjustment strategy, the group feature weight ratio is adjusted, and a weighted operation is performed on the group feature deviation value according to the weight ratio. During the operation, the feature data deviation values ​​are calibrated in sequence and distributed and classified to generate a feature error distribution table; S313: Based on the characteristic error distribution table and in combination with the dynamic adjustment parameters, the adaptability of the error distribution value and the weight ratio is analyzed, each set of concentration error data is calibrated according to the distribution weight and the concentration value is refitted to calculate the dynamic concentration fitting error value.

[0011] As a further solution of the present invention, the characteristic value range is screened by combining the parameter type and distribution characteristics during the analysis, using the formula: ; Calculate the standard deviation of the characteristic deviation ,During the analysis, the characteristic value range is selected by combining the parameter type and ,distribution characteristics, and the initial weight of each set of characteristic deviation values ​​,is adjusted based on the adaptation characteristics, and a grouping weight ,distribution table is generated; in, Representative Class features The standard deviation of the item, Representative Class features The eigenvalue of the term, Representative The mean of the class features, Representative The number of class features, Representative The item number of the class feature.

[0012] As a further solution of the present invention, the step of obtaining the concentration distribution decoding value is specifically as follows: S411: Based on the dynamic concentration fitting error value, combined with the separation gel interface characteristics, the concentration distribution decoding parameters are screened, the influence relationship between the parameters and the concentration fitting error is calculated, and the parameters with increased errors are eliminated by comparing the error variation range, and the optimized parameter set with reduced error is retained to generate an optimized concentration distribution decoding parameter set; S412: Based on the optimized concentration distribution decoding parameter set, iteratively calculate the nonlinear prediction deviation, adjust the influence weight of the decoding parameter, perform difference analysis based on the adjusted deviation value and the convergence standard, calculate the corrected weight ratio through each round of iterative calculation, gradually optimize the overall deviation distribution, and generate a nonlinear prediction deviation optimization result; S413: Based on the nonlinear prediction deviation optimization result, the concentration distribution transformation value and the nonlinear characteristic distribution value are extracted, and the distribution change range after the combination of the two is calculated. The overall dynamic characteristic distribution is analyzed by combining the characteristic values ​​and redundant feature combinations are eliminated to integrate and generate the concentration distribution decoding value.

[0013] As a further embodiment of the present invention, the step of obtaining the separation gel concentration prediction result value is specifically as follows: S511: Calculating a macroscopic concentration characteristic trend value based on the concentration distribution decoded value, analyzing the overall change of the concentration value within the distribution range, identifying the stable and fluctuating intervals of the concentration value by calculating the value change direction and boundary characteristics within the distribution range, extracting the trend characteristics of the stable interval and the boundary characteristics of the fluctuating interval, and generating a macroscopic concentration characteristic trend value; S512: Based on the macroscopic concentration characteristic trend value, a microscopic characteristic difference value is calculated. By comparing the trend value with the local concentration value within the distribution interval, the local variation amplitude in the concentration distribution is analyzed. By partitioning the characteristic fluctuation in the local distribution, characteristic fluctuation values ​​in multiple partitions are extracted and integrated into a difference feature set to generate a multi-scale concentration distribution difference value. S513: Based on the multi-scale concentration distribution difference value, the fitting result of the distribution feature is calculated, and the overall dynamic characteristics of the distribution are calculated by integrating the characteristic changes in the trend value and the difference value. The redundant feature combinations generated during the fitting process are removed, and the feature integration of the concentration data is completed to generate a separation gel concentration prediction result value.

[0014] As a further embodiment of the present invention, the overall change of the analytical concentration value within the distribution range is calculated using the formula: ; Calculate the concentration fluctuation characteristic value, the fluctuation characteristics of the concentration value within the distribution interval, and generate the macro concentration characteristic trend value; in, Represents the fluctuation characteristic value of the concentration value within the distribution interval, Represents the distribution range The concentration value of each sampling point, Represents the arithmetic mean of the concentration values ​​of all sampling points within the distribution interval, Represents the total number of sampling points within the distribution interval.

[0015] A real-time detection system for separation gel concentration, which is used to perform the above-mentioned real-time detection method for separation gel concentration, and comprises: The concentration feature extraction module extracts the multidimensional eigenvalues ​​of the signal based on the input parameters of the detection signal, normalizes the extracted signal eigenvalues ​​with the colloidal matrix characteristics, filters the interference values ​​of the normalized signal values, and readjusts the combination, calculates the linear change results and fitting degree values ​​of the eigenvalue set, and generates the concentration feature representation value; The feature deviation optimization module extracts abnormal deviation values ​​from the feature distribution based on the concentration feature representation value, combines the deviation values ​​with the concentration distribution characteristics, adjusts the sparsity parameter weights of the abnormal deviation values, performs multiple difference comparisons between the adjusted weight values ​​and the redundant parameter set of the feature distribution combination, optimizes the redundancy of the feature combination, and generates potential feature constraint values; The dynamic fitting calculation module assigns weights to the characteristic deviation values ​​based on the potential characteristic constraint values ​​in combination with the nonlinear concentration interval parameters, combines the assigned weight values ​​with the characteristic error distribution values, integrates the distribution values ​​with the weight ratios of the dynamic adjustment parameters, obtains the dynamic weighted calculation results of the concentration error distribution, and generates a dynamic concentration fitting error value; The concentration decoding distribution module extracts the concentration distribution parameters of the separation gel interface based on the dynamic concentration fitting error value, iteratively calculates the distribution parameters and the nonlinear prediction deviation value, combines the calculated distribution parameters with the characteristic distribution value, obtains the transformation result of the concentration distribution, and generates a concentration distribution decoding value; The feature integration prediction module extracts the macroscopic concentration feature trend value and the microscopic feature difference value based on the concentration distribution decoding value, combines the trend value with the difference value, calculates the fitting result through the difference amount of the distribution interval, integrates the fitting result with the distribution characteristics of multiple processing, and generates the separation gel concentration prediction result value.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, by extracting the multi-dimensional characteristic values ​​of the detection signal and normalizing them, the linear change and fitting calculation of the feature combination are realized after eliminating the interference value, thereby improving the detection accuracy and data stability. The abnormal deviation value extraction is combined with the sparsity constraint rule to adjust the weight, optimize the redundant features, and significantly reduce the impact of interference factors on the results. Combining nonlinear concentration interval weight allocation with dynamic weighted operations, the concentration distribution fitting accuracy and adaptability in complex environments are improved. By re-screening the decoding parameters and iterative calculations, the accuracy and stability of the concentration distribution decoding results are ensured. Multi-scale calculations integrate macro trend values ​​and micro difference values, significantly improving the dynamic adaptability of concentration prediction and the depth of data analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a schematic diagram of the workflow of the present invention; Figure 2 Flowchart of the steps for obtaining the concentration characteristic representation value of the present invention; Figure 3 Flowchart of the steps for obtaining potential feature constraint values ​​of the present invention; Figure 4 Flowchart of the steps for obtaining the dynamic concentration fitting error value of the present invention; Figure 5 Flow chart of the steps for obtaining the concentration distribution decoding value of the present invention; Figure 6 The figure is a flow chart of the steps for obtaining the predicted result value of separation gel concentration of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0019] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0020] Example 1 See also Figure 1 The present invention provides a technical solution: a method for real-time detection of separation gel concentration, comprising the following steps: S1: Based on the input parameters of the detection signal, the multidimensional characteristic values ​​of the concentration detection signal are extracted, the signal values ​​are normalized in combination with the characteristics of the colloidal matrix, and after eliminating the interference values, the linear change and fitting degree of the characteristic combination are calculated to generate the concentration characteristic representation value; S2: Based on the concentration feature representation value, combined with the concentration distribution characteristics, the abnormal deviation value in the feature distribution is extracted, the weight of the deviation value is adjusted according to the sparsity constraint rule, and the redundant feature combination is optimized through multiple difference comparisons to generate the potential feature constraint value; S3: Based on the potential characteristic constraint value, the characteristic deviation value is weighted in combination with the nonlinear concentration interval, and the characteristic error distribution value is calculated using dynamic weighting operation. The distribution value and the weight ratio of the dynamic adjustment parameter are integrated to generate the dynamic concentration fitting error value; S4: Based on the dynamic concentration fitting error value, the concentration distribution decoding parameters are re-screened in combination with the separation gel interface characteristics, the nonlinear prediction deviation is iteratively calculated, and the combination of the concentration distribution transformation value and the nonlinear characteristic distribution value is extracted to generate the concentration distribution decoding value; S5: Based on the concentration distribution decoding value, multi-scale calculations are performed on the macroscopic concentration characteristic trend value and the microscopic characteristic difference value. The difference between the trend value and the distribution interval is integrated, and the fitting results of the distribution characteristics are calculated. The characteristic integration after multiple processing of the concentration data is completed to generate the separation gel concentration prediction result value.

[0021] The concentration feature representation value specifically includes the multidimensional feature value, the normalized signal value, and the linear change fitting degree. The potential feature constraint value includes the abnormal deviation value weight, the sparsity constraint rule, and the redundant feature combination optimization value. The dynamic concentration fitting error value specifically includes the feature error distribution value, the dynamic adjustment parameter weight ratio, and the dynamic weighted operation value. The concentration distribution decoding value includes the concentration distribution transformation value, the nonlinear feature distribution value, and the separation gel interface decoding parameter. The separation gel concentration prediction result value specifically refers to the macroscopic concentration feature trend value, the microscopic feature difference value, and the distribution feature fitting result.

[0022] See also Figure 2 , the steps for obtaining the concentration feature representation value are as follows: S111: Based on the input parameters of the detection signal, extract the time domain eigenvalues ​​of the signal, obtain the time series parameters by calculating the signal changes in the differentiated time window, extract the amplitude domain eigenvalues ​​of the signal, calculate the signal peak value and distribution pattern to obtain the amplitude parameter, extract the frequency domain eigenvalues ​​of the signal, analyze the distribution and changes of the frequency components to obtain the frequency parameter, eliminate abnormal signals that exceed the data range, and generate a signal feature matrix; Based on the input parameters of the detection signal, the signal is first segmented and processed, and the time domain eigenvalues ​​are extracted based on the collected signal. Specifically, the signal amplitude sequence in a specific time window is obtained through the signal acquisition device. The time window length is set to 1 second, and the signal mean, variance, and peak value in each time window are calculated. The time series parameters are further obtained by statistically analyzing the eigenvalue change rate of these time windows. The signal is then subjected to eigenvalue extraction in the amplitude domain. Specifically, the peak value, average amplitude, skewness, and kurtosis of the amplitude distribution are statistically analyzed. The signal peak value is determined by the local maximum point in the detection signal. The amplitude distribution law is analyzed by constructing a signal histogram, and the probability distribution of the signal amplitude in different amplitude intervals is calculated. The amplitude domain parameters are calculated based on these distribution laws. In the frequency domain feature extraction, the signal is Fourier transformed to obtain the spectrum distribution. The frequency parameters are obtained by the changes in the spectrum energy distribution and frequency components. The abnormal signals outside the range are further eliminated by setting the frequency threshold range (such as 20Hz-2000Hz). The data is screened using software and the normal data is input into the feature extraction matrix to form the final signal feature matrix.

[0023] S112: performing normalization processing based on the signal feature matrix and in combination with the colloidal matrix characteristics, adjusting the eigenvalue range to match the colloidal matrix, calculating the eigenvalue deviation, eliminating unmatched data points, and generating a normalized feature matrix; Based on the signal feature matrix, the data is first normalized in combination with the colloidal matrix characteristics. Normalization is achieved by subtracting the mean from each column of the matrix data and dividing by the standard deviation to ensure that the distribution range of the matrix data matches the colloidal matrix characteristics. The deviation of the normalized eigenvalues ​​is then calculated. The specific method is to calculate the sum of the squares of the difference between each eigenvalue and the center of the preset range of the matrix characteristics, and to eliminate them by setting a deviation threshold. The deviation threshold can be set according to the actual experimental conditions. For example, based on the experimental data of the colloidal matrix, the deviation is set within the range of [0.05, 0.2]. Data points outside this range are directly eliminated. Finally, the eliminated normalized feature data are recombined to generate the normalized feature matrix.

[0024] S113: Based on the normalized feature matrix, linearly change the feature combination, calculate the weighted average to form the feature combination, analyze the fit of the combination and optimize the weight, integrate the fitting results of all features, and generate the concentration feature representation value.

[0025] Calculate the weighted average to form a feature combination using the formula: ; Calculate weighted eigenvalues , analyze the combined fit and optimize the weight, integrate the fitting results of all features, and generate concentration feature representation values; in, Indicates the The weighted eigenvalues ​​of the features, Indicates the The sample in The absolute difference of the features, Indicates the The average value of the features, Indicates the The sample in The weight value on the feature, Express Sum of samples, Represents the square root of the sum of the squares of the weight values.

[0026] formula: ; Detailed explanation of the formula and the process of formula calculation and derivation: Specific explanation and acquisition method of each parameter: Indicates the The weighted eigenvalues ​​of the features, Indicates the The sample in The absolute difference of the features, Indicates the The average value of the features is given by The sample value of each feature is obtained by calculating the arithmetic mean. Indicates the The sample in The weight value on each feature is obtained through feature importance analysis. Express Sum of samples, Represents the square root of the sum of the squares of the weight values.

[0027] Specific settings and sources of values: Monitoring The data of each feature comes from experimental data collection, and the five groups of sample values ​​are 3.1, 3.4, 3.7, 3.8, and 4.0 respectively. The absolute difference and weight of each sample are calculated. The weight values ​​are obtained through feature importance analysis, and the weights are 0.2, 0.25, 0.3, 0.15, and 0.1 respectively. The weight values ​​fluctuate with the degree of influence of the sample data on the fit.

[0028] Calculation process: Calculate the feature mean ; Calculate the absolute difference between each group of samples ; ; ; ; ; Calculates the numerator of the weighted absolute difference ; ; ; Calculate the square root of the sum of squared weights ; ; Substitute the numerator and denominator into the formula to calculate ; The result shows that the weighted eigenvalue of the first feature is 0.527, indicating that the weighted value for feature combination is calculated by comprehensively considering the absolute difference and weight distribution of the samples, which can be used as the input parameter for subsequent feature optimization weights and integrated fitting results.

[0029] See also Figure 3 , the steps for obtaining the potential feature constraint value are as follows: S211: Based on the concentration characteristic representation value, analyze the concentration data in the characteristic distribution, calculate the mean and distribution range of the characteristic value, identify abnormal deviation values ​​by comparing the characteristic value with the upper and lower boundaries, extract characteristic deviation values ​​that exceed the boundaries, eliminate repeated and coupled deviation values, and generate an abnormal deviation characteristic value set; Based on the concentration characteristic representation value, the concentration characteristic data is first analyzed, and the concentration data in the characteristic distribution is segmented by time or region. The distribution range of the characteristic value is analyzed by calculating the mean and standard deviation of each segment of data. Then, each characteristic value is compared with the set upper and lower boundary values ​​to determine whether there is an abnormal deviation value. The abnormal deviation value can be determined by the formula Calculate, where is the current eigenvalue, It is a boundary value. If the deviation value exceeds the boundary range, it is determined to be an outlier. All eigenvalues ​​exceeding the upper and lower boundary ranges are extracted to form a preliminary abnormal deviation eigenvalue set. Then, repeated or coupled outliers are eliminated through cluster analysis. The DBSCAN algorithm can be used here to cluster the outliers in the eigenvalue set, eliminate repeated data with a distance less than the set threshold, and finally generate an abnormal deviation eigenvalue set to ensure that the concentration feature data can be used for further calculation after eliminating the anomalies.

[0030] S212: Based on the abnormal deviation eigenvalue set and in combination with the sparsity constraint rule, the difference between the eigenvalue and the adjacent eigenvalues ​​is calculated, and the influence of the feature on the overall distribution is assigned a weight adjustment. The weight distribution is optimized according to the relationship between the deviation value and the overall proportion, and the sparsity weight adjustment result is generated. Based on the abnormal deviation eigenvalue set, the difference between the eigenvalue and the adjacent eigenvalue is first calculated in combination with the sparsity constraint rule, and the formula Calculate eigenvalues and neighboring eigenvalues The absolute difference between the eigenvalues ​​is then assigned a weight to each eigenvalue to adjust its impact on the overall distribution. The weight assignment is based on the degree of deviation between the eigenvalue and the outlier. The weight adjustment formula is: ,in is the eigenvalue weight, is the characteristic mean, In order to adjust the coefficient, the weight distribution is optimized according to the relationship between the deviation value and the overall proportion. By iteratively adjusting the weight distribution, the impact of abnormal deviation on the overall feature distribution is reduced, and finally the sparsity weight adjustment result is generated.

[0031] S213: Based on the sparsity weight adjustment result, multiple difference comparisons are performed on the adjusted deviation characteristic values ​​and the concentration characteristic representation values, the redundant relationship between the feature distributions is gradually calculated, the redundant feature combinations are removed, the optimized feature set is integrated, and the potential feature constraint value is generated.

[0032] Based on the sparsity weight adjustment results, multiple difference comparisons are performed on the adjusted deviation eigenvalues ​​and concentration characteristic values. First, the difference ratio between the adjusted eigenvalues ​​and the mean concentration characteristic is calculated, and the formula is used to calculate the difference ratio between the adjusted eigenvalues ​​and the mean concentration characteristic. Calculate, where is the adjusted eigenvalue, The concentration feature mean is calculated, and this process is repeated to analyze the redundant relationship between different feature values. Specifically, the correlation coefficient of each two groups of feature values ​​is calculated and a threshold is set. The redundant feature combinations with correlation higher than the set value (for example, 0.85) are removed, and then the optimized feature set is integrated. Through the iterative optimization process of gradually eliminating inefficient features and retaining significant features, the potential feature constraint value is finally generated to ensure that the feature set is sparse and representative.

[0033] See also Figure 4 , the specific steps for obtaining the dynamic concentration fitting error value are: S311: Based on the potential characteristic constraint value and the nonlinear concentration range, the characteristic deviation values ​​are analyzed and grouped. During the analysis, the characteristic value range is selected based on the parameter type and distribution characteristics. The initial weight of each group of characteristic deviation values ​​is adjusted according to the adaptation characteristics to generate a group weight distribution table. S312: Based on the group weight allocation table and in combination with the dynamic parameter adjustment strategy, the group feature weight ratio is adjusted, and the group feature deviation value is weighted according to the weight ratio. During the operation, the feature data deviation values ​​are calibrated in sequence and distributed and classified to generate a feature error distribution table; Based on the group weight allocation table, the group feature weight ratio is adjusted in combination with the dynamic parameter adjustment strategy, where the group weight allocation table can be constructed according to the group feature set. For example, the group data features include physical properties, chemical properties and experimental data of various sampling points. The dynamic parameter adjustment strategy realizes the optimization process of real-time data adaptation by introducing dynamic adjustment coefficients, and quantifies the group feature deviation value into a set of deviation values. For example, the degree of differentiation of a group of data from the standard value can be measured by calculating the deviation between its mean and median, and the standard deviation of each distribution point. Then, the group feature deviation value is weighted according to the weight ratio. The weighted calculation formula can be: ,in represents the weighted deviation value of group i, is the weight value of the jth feature, is the absolute value of the feature deviation. By performing step-by-step calculations on the above formula, the weighted deviation value after weight distribution is obtained. For example, for a certain feature weight value The interval is set from 0.1 to 0.9 through dynamic adjustment strategy. The calculation process first selects a group of feature deviation data with the highest weight value, and then re-optimizes the distribution of the remaining data groups according to the adjustment strategy. Finally, the feature data deviation value is calibrated in turn according to the size of the deviation value. By classifying and distributing the data, a set of error distribution tables is generated accordingly.

[0034] S313: Based on the characteristic error distribution table and in combination with the dynamic adjustment parameters, the adaptability of the error distribution value and the weight ratio is analyzed, each set of concentration error data is calibrated according to the distribution weight and the concentration value is refitted to calculate the dynamic concentration fitting error value.

[0035] Based on the feature error distribution table, combined with dynamic adjustment parameters, the adaptability of the error distribution value and the weight ratio is analyzed. The feature error distribution table uses the weighted deviation value generated in the previous section as input, and constructs the interval distribution according to the distribution range of the grouping feature. For example, the feature deviation distribution of a group of data can be obtained by calculating the quantiles of each feature group to obtain the adaptation interval, combined with dynamic adjustment parameters such as the adjustment coefficient , the calculation formula can be ,in is the difference in error distribution values, The difference after weight adjustment is used to analyze the adjusted adaptability, and each set of concentration error data is recalibrated according to the distribution weight. For example, the concentration data is used as input, and the deviation value is fitted by the distribution interval, such as using a nonlinear fitting function. , where the coefficient The fitting values ​​are extracted from the aforementioned distribution table and the dynamic concentration fitting error values ​​are calculated. During the concentration fitting process, the dynamic parameters of the grouped data need to be further decomposed and analyzed, and the concentration distribution values ​​are optimized by point-by-point adjustment. Finally, the dynamic fitting calculation is completed to obtain the final result value.

[0036] When parsing, the parameter type and distribution characteristics are combined to filter the eigenvalue range, using the formula: ; Calculate the standard deviation of the characteristic deviation ,During the analysis, the characteristic value range is selected by combining the parameter type and ,distribution characteristics, and the initial weight of each set of characteristic deviation values ​​,is adjusted based on the adaptation characteristics, and a grouping weight ,distribution table is generated; in, Representative Class features The standard deviation of the item, Representative Class features The eigenvalue of the term, Representative The mean of the class features, Representative The number of class features, Representative The item number of the class feature.

[0037] formula: ; Detailed explanation of the formula and the process of formula calculation and derivation: Parameter acquisition: :Acquired by collecting experimental data. For example, for monitoring water sample quality, collect specific values ​​of a certain type of indicator (such as dissolved oxygen concentration) in milligrams per liter, and collect data through monitoring equipment such as a portable dissolved oxygen meter. Assume that the dissolved oxygen concentration data is collected, and the results are 9.5, 9.2, 9.8, 10.1, 9.7, and 9.4 mg per liter, corresponding to 、 、 、 、 、 .

[0038] : The total number of data items collected. Based on the collected dissolved oxygen concentration data, the total number of items is 6.

[0039] : The average value of the data is obtained by calculating the arithmetic mean of the collected data. The calculation formula is: ; Bring in data: ; The average value was 9.6167 mg / L.

[0040] Formula derivation process: Enter the collected data and the calculated average value: ; Calculation steps: , the square is: ; , the square is: ; , the square is: ; , the square is: ; , the square is: ; , the square is: ; Add up all the squared results: ; Calculate the denominator: ; Substitute into the formula: ; The calculation results show that the standard deviation of the data is 0.3188 mg / L, which represents the fluctuation range of dissolved oxygen concentration in the collected data. A smaller standard deviation indicates a more concentrated distribution of the data. This calculation result is relevant to the subsequent parsing and grouping process and is used to analyze the distribution characteristics of each group of characteristic deviation values.

[0041] See also Figure 5 , the steps for obtaining the concentration distribution decoding value are as follows: S411: Based on the dynamic concentration fitting error value and combined with the separation gel interface characteristics, the concentration distribution decoding parameters are screened, the influence relationship between the parameters and the concentration fitting error is calculated, and the parameters with increased errors are eliminated by comparing the error change range, and the optimized parameter set with reduced error is retained to generate the optimized concentration distribution decoding parameter set; Based on the dynamic concentration fitting error value, the concentration distribution decoding parameters are screened in combination with the separation gel interface characteristics. First, the error variation range is extracted from the dynamic concentration fitting error value. The error value is obtained by the result of the nonlinear fitting formula. The extraction range can be obtained by calculating the mean and standard deviation of the error value. The calculation formula is: ,in is the mean, is the standard deviation, It is an adjustable factor. By analyzing the distribution characteristics of the error data within the range, the parameter set that matches the interface characteristics of the separation gel is screened out. The interface characteristics of the separation gel include interface viscosity, diffusion coefficient, etc. These characteristics can be obtained through experiments, such as measuring the interface viscosity through an interface rheometer and measuring the diffusion coefficient through spectrophotometry. These characteristic values ​​are regressed with the dynamic concentration fitting error value to determine the influence relationship between each parameter and the error. The regression model is used to eliminate parameters with increased errors. For example, through the significance test of the linear regression model, the parameters with low significance are eliminated, and the optimized parameter set with reduced error is retained, and finally the optimized concentration distribution decoding parameter set is generated.

[0042] S412: Based on the optimized concentration distribution decoding parameter set, iteratively calculate the nonlinear prediction deviation, adjust the influence weight of the decoding parameter, perform difference analysis based on the adjusted deviation value and the convergence standard, calculate the corrected weight ratio through each round of iterative calculation, gradually optimize the overall deviation distribution, and generate a nonlinear prediction deviation optimization result; Based on the optimized concentration distribution decoding parameter set, the nonlinear prediction deviation is iteratively calculated. The optimization process includes adjusting the influence weight of the decoding parameters and dynamically correcting the deviation value. First, an initial weight is assigned to each parameter in the optimized decoding parameter set. For example, the weight is initialized to ,in For each parameter error value, the weight value is dynamically adjusted by calculating its influence on the nonlinear prediction deviation. The adjusted weight The formula can be Calculate, where The error change of the i-th parameter in each round of iteration is calculated by calculating the correction weight ratio in each round of iteration, and the difference analysis is performed in combination with the convergence standard of the nonlinear fitting error, such as setting the convergence threshold , judge the error value after iteration Is it satisfied If it is not satisfied, the iteration continues until the convergence criteria are met, and finally the nonlinear prediction deviation optimization result is generated.

[0043] S413: Based on the nonlinear prediction deviation optimization results, the concentration distribution transformation value and the nonlinear characteristic distribution value are extracted, and the distribution change range after the combination of the two is calculated. The overall dynamic characteristic distribution is analyzed by combining the characteristic values ​​and redundant feature combinations are eliminated to integrate and generate the concentration distribution decoding value.

[0044] Based on the nonlinear prediction deviation optimization results, the concentration distribution transformation value and the nonlinear characteristic distribution value are extracted, wherein the concentration distribution transformation value can be calculated by applying the concentration distribution decoding parameter to the experimental data, for example, by the formula ,in is the original concentration value, To decode the parameters, the nonlinear characteristic distribution value is obtained by extracting the nonlinear characteristics of the concentration distribution fitting error. The main characteristic components are screened out using principal component analysis or support vector machine method. The concentration distribution transformation value is combined with the nonlinear characteristic distribution value to calculate the distribution change range of the combination of the two, such as calculating the maximum and minimum values ​​of each combination to form a distribution interval. , eliminate redundant feature combinations within the distribution variation range, use simple clustering algorithms, such as K-Means algorithm, to cluster the feature values ​​and screen effective feature combinations, and finally integrate to generate concentration distribution decoding values.

[0045] See also Figure 6 The specific steps for obtaining the predicted result value of separation gel concentration are as follows: S511: Calculate the macro concentration characteristic trend value based on the concentration distribution decoded value, analyze the overall change of the concentration value within the distribution range, identify the stable and fluctuating intervals of the concentration value by calculating the value change direction and boundary characteristics within the distribution range, extract the trend characteristics of the stable interval and the boundary characteristics of the fluctuating interval, and generate the macro concentration characteristic trend value; S512: Based on the macroscopic concentration characteristic trend value, the microscopic characteristic difference value is calculated. By comparing the trend value with the local concentration value within the distribution interval, the local variation amplitude in the concentration distribution is analyzed. By partitioning the characteristic fluctuation in the local distribution, the characteristic fluctuation values ​​in multiple partitions are extracted and integrated into a difference feature set to generate a multi-scale concentration distribution difference value. Based on the macroscopic concentration characteristic trend value, the microscopic characteristic difference value is calculated by comparing the trend value with the local concentration value within the distribution interval. First, it is necessary to define the calculation method of the macroscopic concentration characteristic trend value, which can usually be determined by the average value or moving average trend of the global concentration data. The calculation formula is ,in represents the i-th concentration value, Represents the total number of concentration data, and the microscopic feature difference value can be obtained by the formula Calculate, where For the local concentration value within the distribution interval, the concentration value of each interval is compared with the macro trend value to analyze the local variation range in the concentration distribution. The difference value of each local interval is calculated one by one using the above formula. The characteristic fluctuation in the local distribution is partitioned by combining the partition calculation strategy. For example, multiple intervals are divided according to the distribution range of the concentration value and the characteristic fluctuation value of each interval is calculated. The characteristic fluctuation value can be expressed by the standard deviation or deviation range of the concentration value within the interval. The formula is ,in is the mean concentration in this interval, is the number of data points in the interval. The characteristic fluctuation values ​​calculated for each partition are integrated into a difference feature set. Through cluster analysis or hierarchical screening, the multi-scale concentration distribution difference value is finally generated.

[0046] S513: Based on the multi-scale concentration distribution difference values, the fitting results of the distribution characteristics are calculated, and the overall dynamic characteristics of the distribution are calculated by integrating the characteristic changes in the trend value and the difference value. The redundant feature combinations generated during the fitting process are removed, and the feature integration of the concentration data is completed to generate the separation gel concentration prediction result value.

[0047] Based on the multi-scale concentration distribution difference value, the fitting results of the distribution characteristics are calculated. First, the multi-scale concentration distribution difference value needs to be used as the input parameter, and the concentration characteristic trend value is combined for overall fitting. The characteristic change amount in the integration trend value and difference value can be achieved through weighted calculation. For example, the formula is ,in is the integrated dynamic eigenvalue, and are weight coefficients, which respectively represent the contribution ratio of trend value and difference value in the fitting calculation. These two weight coefficients can be adjusted by the error minimization strategy to remove the redundant feature combinations generated in the fitting process. For example, by analyzing the feature contribution rate in the fitting results, the features with contribution rate lower than a certain threshold are screened and removed. This process can be achieved through the cumulative contribution rate calculation formula, such as setting the cumulative contribution rate threshold ,when When the specified ratio is reached, the corresponding feature combination is retained and the remaining redundant features are eliminated. Finally, the feature integration of the concentration data is completed and the separation gel concentration prediction result value is generated. The accuracy of this result can be verified by comparing it with the concentration range of the actual experimental data.

[0048] To analyze the overall change of concentration values ​​within the distribution range, the formula is used: ; Calculate the concentration fluctuation characteristic value, the fluctuation characteristics of the concentration value within the distribution interval, and generate the macro concentration characteristic trend value; in, Represents the fluctuation characteristic value of the concentration value within the distribution interval, Represents the distribution range The concentration value of each sampling point, Represents the arithmetic mean of the concentration values ​​of all sampling points within the distribution interval, Represents the total number of sampling points within the distribution interval.

[0049] formula: ; Detailed explanation of the formula and the process of formula calculation and derivation: The concentration data within the distribution interval is collected as input parameters. The concentration value of each sampling point in actual monitoring is obtained through professional concentration detection equipment. The value recorded by the detection equipment is in milligrams per cubic meter. The data collection frequency is set to record once per minute, the sampling interval is set to 60 minutes, and there are 60 sampling points in total. The following are the specific values ​​and calculation process: The concentration values ​​of the sampling points are as follows: ; Concentration mean calculation formula: ; Bring in the data: ; The total concentration value obtained by the actual sampling equipment is 651.2 mg per cubic meter, so: ; The absolute deviation calculation formula is: ; Bring in data: ; In actual calculations, the sum of the corresponding absolute deviations is 45.6, so: ; Variance calculation formula: ; Bring in data: ; The sampling device calculates the sum of the variances to be 36.72, so: ; Substitute the above calculation results into the main formula: ; Bring in data: ; The results show that the fluctuation characteristic value of the concentration value within the distribution range is 0.594 mg per cubic meter, which reflects the overall variation range of the concentration within the distribution range, indicating that the concentration value fluctuation in this range is relatively stable and close to its mean range.

[0050] Parameter explanation: Represents the fluctuation characteristic value of the concentration value within the distribution interval, Represents the distribution range The concentration value of each sampling point is in milligrams per cubic meter. Represents the arithmetic mean of the concentration values ​​of all sampling points within the distribution interval, in milligrams per cubic meter. Represents the total number of sampling points within the distribution interval, the value is 60, Represents the sum of the absolute deviations between the concentration values ​​of all sampling points and their mean, in milligrams per cubic meter. Represents the sum of the squared deviations between the concentration values ​​of all sampling points and their mean, in square milligrams per cubic meter; A real-time detection system for separation gel concentration is provided, which is used to perform the above-mentioned real-time detection method for separation gel concentration. The system comprises: The concentration feature extraction module extracts the multidimensional eigenvalues ​​of the signal based on the input parameters of the detection signal, normalizes the extracted signal eigenvalues ​​with the colloidal matrix characteristics, filters the interference values ​​of the normalized signal values, and readjusts the combination, calculates the linear change results and fitting degree values ​​of the eigenvalue set, and generates the concentration feature representation value; The feature deviation optimization module extracts abnormal deviation values ​​from the feature distribution based on the concentration feature representation value, combines the deviation value with the concentration distribution characteristics, adjusts the sparsity parameter weight of the abnormal deviation value, and compares the adjusted weight value with the redundant parameter set of the feature distribution combination multiple times to optimize the redundancy of the feature combination and generate potential feature constraint values; The dynamic fitting calculation module assigns weights to characteristic deviation values ​​based on potential characteristic constraint values ​​and nonlinear concentration interval parameters, combines the assigned weight values ​​with characteristic error distribution values, integrates the distribution values ​​with the weight ratios of the dynamic adjustment parameters, obtains the dynamic weighted calculation results of the concentration error distribution, and generates dynamic concentration fitting error values. The concentration decoding distribution module extracts the concentration distribution parameters of the separation gel interface based on the dynamic concentration fitting error value, iteratively calculates the distribution parameters and the nonlinear prediction deviation value, combines the calculated distribution parameters with the characteristic distribution value, obtains the concentration distribution transformation result, and generates the concentration distribution decoding value; The feature integration prediction module extracts the macroscopic concentration feature trend value and the microscopic feature difference value based on the concentration distribution decoding value, combines the trend value with the difference value, calculates the fitting result through the difference amount of the distribution interval, integrates the fitting result with the distribution characteristics of multiple treatments, and generates the separation gel concentration prediction result value.

[0051] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A real-time detection method for separation gel concentration, characterized in that, The following steps are involved: S1: Based on the input parameters of the detection signal, the multidimensional characteristic values ​​of the concentration detection signal are extracted, the signal values ​​are normalized in combination with the characteristics of the colloidal matrix, and after eliminating the interference values, the linear change and fitting degree of the characteristic combination are calculated to generate the concentration characteristic representation value; S2: Based on the concentration feature representation value, combined with the concentration distribution characteristics, extract the abnormal deviation value in the feature distribution, adjust the weight of the deviation value according to the sparsity constraint rule, optimize the redundant feature combination through multiple difference comparisons, and generate a potential feature constraint value; S3: Based on the potential characteristic constraint value, the characteristic deviation value is weighted in combination with the nonlinear concentration interval, a dynamic weighted operation is used to calculate the characteristic error distribution value, and the distribution value and the weight ratio of the dynamic adjustment parameter are integrated to generate a dynamic concentration fitting error value; S4: Based on the dynamic concentration fitting error value, the concentration distribution decoding parameters are re-screened in combination with the separation gel interface characteristics, the nonlinear prediction deviation is iteratively calculated, and a combination of the concentration distribution transformation value and the nonlinear characteristic distribution value is extracted to generate a concentration distribution decoding value; S5: Based on the concentration distribution decoding value, multi-scale calculation is performed on the macroscopic concentration characteristic trend value and the microscopic characteristic difference value, the difference between the trend value and the distribution interval is integrated, the fitting result of the distribution characteristic is calculated, and the characteristic integration after multiple processing of the concentration data is completed to generate the separation gel concentration prediction result value.

2. the separation gel concentration real-time detection method according to claim 1, is characterized in that, The steps for obtaining the concentration characteristic value are specifically as follows: S111: Based on the input parameters of the detection signal, extract the time domain eigenvalues ​​of the signal, obtain the time series parameters by calculating the signal changes in the differentiated time window, extract the amplitude domain eigenvalues ​​of the signal, calculate the signal peak value and distribution pattern to obtain the amplitude parameter, extract the frequency domain eigenvalues ​​of the signal, analyze the distribution and changes of the frequency components to obtain the frequency parameter, eliminate abnormal signals that exceed the data range, and generate a signal feature matrix; S112: performing normalization processing based on the signal feature matrix and in combination with the characteristics of the colloidal matrix, adjusting the eigenvalue range to match the colloidal matrix, calculating the eigenvalue deviation, eliminating unmatched data points, and generating a normalized feature matrix; S113: Based on the normalized feature matrix, linearly change the feature combination, calculate the weighted average to form the feature combination, analyze the fit of the combination and optimize the weight, integrate the fitting results of all features, and generate a concentration feature representation value.

3. the separation gel concentration real-time detection method according to claim 2, is characterized in that, The weighted average is calculated to form a feature combination using the formula: ； Calculate weighted eigenvalues , analyze the combined fit and optimize the weight, integrate the fitting results of all features, and generate concentration feature representation values; in, Indicates the The weighted eigenvalues ​​of the features, Indicates the The sample in The absolute difference of the features, Indicates the The average value of the features, Indicates the The sample in The weight value on the feature, Express Sum of samples, Represents the square root of the sum of the squares of the weight values.

4. the separation gel concentration real-time detection method according to claim 1, is characterized in that, The steps for obtaining the potential feature constraint value are specifically as follows: S211: Based on the concentration characteristic representation value, analyze the concentration data in the characteristic distribution, calculate the mean and distribution range of the characteristic value, identify abnormal deviation values ​​by comparing the characteristic value with the upper and lower boundaries, extract characteristic deviation values ​​that exceed the boundaries, eliminate repeated and coupled deviation values, and generate an abnormal deviation characteristic value set; S212: Based on the abnormal deviation feature value set, combined with the sparsity constraint rule, the difference between the feature value and the adjacent feature value is calculated, the weight adjustment feature is assigned to adjust the impact of the feature on the overall distribution, the weight distribution is optimized according to the relationship between the deviation value and the overall proportion, and a sparsity weight adjustment result is generated; S213: Based on the sparsity weight adjustment result, multiple difference comparisons are performed on the adjusted deviation characteristic values ​​and the concentration characteristic representation values, the redundant relationship between the feature distributions is gradually calculated, the redundant feature combinations are removed, the optimized feature set is integrated, and the potential feature constraint value is generated.

5. the separation gel concentration real-time detection method according to claim 1, is characterized in that, The steps for obtaining the dynamic concentration fitting error value are specifically as follows: S311: Based on the potential characteristic constraint value and the nonlinear concentration range, the characteristic deviation values ​​are analyzed and grouped. During the analysis, the characteristic value range is selected based on the parameter type and distribution characteristics. The initial weight of each group of characteristic deviation values ​​is adjusted according to the adaptation characteristics to generate a group weight distribution table. S312: Based on the group weight allocation table and in combination with the dynamic parameter adjustment strategy, the group feature weight ratio is adjusted, and a weighted operation is performed on the group feature deviation value according to the weight ratio. During the operation, the feature data deviation values ​​are calibrated in sequence and distributed and classified to generate a feature error distribution table; S313: Based on the characteristic error distribution table and in combination with the dynamic adjustment parameters, the adaptability of the error distribution value and the weight ratio is analyzed, each set of concentration error data is calibrated according to the distribution weight and the concentration value is refitted to calculate the dynamic concentration fitting error value.

6. The separation gel concentration real-time detection method according to claim 5, wherein During the analysis, the parameter type and distribution characteristics are combined to filter the characteristic value range, using the formula: ； Calculate the standard deviation of the characteristic deviation ,During the analysis, the characteristic value range is selected by combining the parameter type and ,distribution characteristics, and the initial weight of each set of characteristic deviation values ​​,is adjusted based on the adaptation characteristics, and a grouping weight ,distribution table is generated; in, Representative Class features The standard deviation of the item, Representative Class features The eigenvalue of the term, Representative The mean of the class features, Representative The number of class features, Representative The item number of the class feature.

7. The separation gel concentration real-time detection method according to claim 1, wherein The steps for obtaining the concentration distribution decoding value are specifically as follows: S411: Based on the dynamic concentration fitting error value, combined with the separation gel interface characteristics, the concentration distribution decoding parameters are screened, the influence relationship between the parameters and the concentration fitting error is calculated, and the parameters with increased errors are eliminated by comparing the error variation range, and the optimized parameter set with reduced error is retained to generate an optimized concentration distribution decoding parameter set; S412: Based on the optimized concentration distribution decoding parameter set, iteratively calculate the nonlinear prediction deviation, adjust the influence weight of the decoding parameter, perform difference analysis based on the adjusted deviation value and the convergence standard, calculate the corrected weight ratio through each round of iterative calculation, gradually optimize the overall deviation distribution, and generate a nonlinear prediction deviation optimization result; S413: Based on the nonlinear prediction deviation optimization result, the concentration distribution transformation value and the nonlinear characteristic distribution value are extracted, and the distribution change range after the combination of the two is calculated. The overall dynamic characteristic distribution is analyzed by combining the characteristic values ​​and redundant feature combinations are eliminated to integrate and generate the concentration distribution decoding value.

8. The separation gel concentration real-time detection method according to claim 1, wherein The steps for obtaining the separation gel concentration prediction result value are specifically as follows: S511: Calculating a macroscopic concentration characteristic trend value based on the concentration distribution decoded value, analyzing the overall change of the concentration value within the distribution range, identifying the stable and fluctuating intervals of the concentration value by calculating the value change direction and boundary characteristics within the distribution range, extracting the trend characteristics of the stable interval and the boundary characteristics of the fluctuating interval, and generating a macroscopic concentration characteristic trend value; S512: Based on the macroscopic concentration characteristic trend value, a microscopic characteristic difference value is calculated. By comparing the trend value with the local concentration value within the distribution interval, the local variation amplitude in the concentration distribution is analyzed. By partitioning the characteristic fluctuation in the local distribution, characteristic fluctuation values ​​in multiple partitions are extracted and integrated into a difference feature set to generate a multi-scale concentration distribution difference value. S513: Based on the multi-scale concentration distribution difference value, the fitting result of the distribution feature is calculated, and the overall dynamic characteristics of the distribution are calculated by integrating the characteristic changes in the trend value and the difference value. The redundant feature combinations generated during the fitting process are removed, and the feature integration of the concentration data is completed to generate a separation gel concentration prediction result value.

9. The separation gel concentration real-time detection method according to claim 8, wherein The overall change of the analytical concentration value within the distribution interval is calculated using the formula: ； Calculate the concentration fluctuation characteristic value, the fluctuation characteristics of the concentration value within the distribution interval, and generate the macro concentration characteristic trend value; in, Represents the fluctuation characteristic value of the concentration value within the distribution interval, Represents the distribution range The concentration value of each sampling point, Represents the arithmetic mean of the concentration values ​​of all sampling points within the distribution interval, Represents the total number of sampling points within the distribution interval.

10. A real-time detection system for separation gel concentration, characterized in that: The method for real-time detection of separation gel concentration according to any one of claims 1 to 9, wherein the system comprises: The concentration feature extraction module extracts the multidimensional eigenvalues ​​of the signal based on the input parameters of the detection signal, normalizes the extracted signal eigenvalues ​​with the colloidal matrix characteristics, filters the interference values ​​of the normalized signal values, and readjusts the combination, calculates the linear change results and fitting degree values ​​of the eigenvalue set, and generates the concentration feature representation value; The feature deviation optimization module extracts abnormal deviation values ​​from the feature distribution based on the concentration feature representation value, combines the deviation values ​​with the concentration distribution characteristics, adjusts the sparsity parameter weights of the abnormal deviation values, performs multiple difference comparisons between the adjusted weight values ​​and the redundant parameter set of the feature distribution combination, optimizes the redundancy of the feature combination, and generates potential feature constraint values; The dynamic fitting calculation module assigns weights to the characteristic deviation values ​​based on the potential characteristic constraint values ​​in combination with the nonlinear concentration interval parameters, combines the assigned weight values ​​with the characteristic error distribution values, integrates the distribution values ​​with the weight ratios of the dynamic adjustment parameters, obtains the dynamic weighted calculation results of the concentration error distribution, and generates a dynamic concentration fitting error value; The concentration decoding distribution module extracts the concentration distribution parameters of the separation gel interface based on the dynamic concentration fitting error value, iteratively calculates the distribution parameters and the nonlinear prediction deviation value, combines the calculated distribution parameters with the characteristic distribution value, obtains the transformation result of the concentration distribution, and generates a concentration distribution decoding value; The feature integration prediction module extracts the macroscopic concentration feature trend value and the microscopic feature difference value based on the concentration distribution decoding value, combines the trend value with the difference value, calculates the fitting result through the difference amount of the distribution interval, integrates the fitting result with the distribution characteristics of multiple processing, and generates the separation gel concentration prediction result value.

Images