Intelligent baseline dynamic calibration data processing method for ion chromatography detection of soil gypsum

Abstract

The application discloses an intelligent baseline dynamic calibration data processing method for ion chromatography detection of soil gypsum, and belongs to the technical field of data processing of ion chromatography detection. The method comprises the following steps: acquiring original conductivity-time spectrum data collected by a conductivity detector during ion chromatography detection of a target soil sample; performing partition statistical analysis on the original data to identify a plurality of baseline segment data adjacent to a characteristic ion peak of gypsum; constructing a composite dynamic calibration coefficient based on the local density distribution characteristics of the baseline segment data; performing point-by-point sliding window baseline correction processing on the original data by using the coefficient to generate calibrated chromatography data; and delivering the calibrated data to a subsequent analysis module for ion concentration analysis. The application can adapt to complex baseline fluctuations and suppress chromatography peak distortion by constructing a dynamic calibration coefficient based on the local density distribution characteristics of the baseline segment and performing point-by-point sliding window correction.

Description

Technical Field

[0001] This invention relates to the field of ion chromatography detection data processing technology, specifically to an intelligent baseline dynamic calibration data processing method for soil gypsum ion chromatography detection. Background Technology

[0002] When detecting gypsum content in soil using ion chromatography, the raw conductivity-time spectrum data output by the conductivity detector is typically superimposed with a baseline signal that drifts and varies non-stationarily over time. This baseline fluctuation mainly originates from factors such as perturbations in the eluent composition, temperature gradients, contamination of the detection cell, and interference from the complex matrix of the soil leachate. Irregular changes in the baseline directly distort the peak shape and area of ​​gypsum characteristic ion peaks, leading to deviations in subsequent qualitative and quantitative analyses.

[0003] Existing baseline correction methods often employ fixed parameters or global fitting strategies, such as performing polynomial curve fitting or piecewise linear interpolation on the entire chromatogram and then subtracting it from the original signal as the baseline. These global strategies have limitations in their applicability to soil gypsum detection scenarios. Soil samples have complex compositions, and gypsum characteristic ion peaks often coexist with adjacent impurity peaks and shoulder peaks. Furthermore, the baseline exhibits differentiated fluctuation patterns and local density at different retention times. Global fitting struggles to capture the local density changes of the baseline in the vicinity of ion peaks, easily leading to overfitting or underfitting in peak boundary regions, resulting in peak asymmetry and errors in identifying peak start and end points. Another type of method relies on manually selecting several baseline points to construct the curve, resulting in poor reproducibility and an inability to meet the automated detection needs of large-scale soil sample testing.

[0004] For the detection of gypsum in soil by ion chromatography, it is necessary to solve how to accurately extract dynamic characterization parameters that reflect the local density distribution of the baseline in the region adjacent to the ion peak, and use these parameters to adaptively identify baseline candidate points within the sliding window to complete local correction, thereby reducing the distortion of the true ion peak signal by the baseline calibration results under the interference of dense clutter peaks. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent baseline dynamic calibration data processing method for soil gypsum ion chromatography detection, which solves the problems that global fitting and fixed parameters are difficult to adapt to local density changes in the baseline and are prone to introducing peak shape distortion in dense ion peak regions.

[0006] To achieve the above objectives, the present invention provides the following technical solution: The present invention discloses an intelligent baseline dynamic calibration data processing method for soil gypsum ion chromatography detection. This method eliminates the influence of baseline drift and noise interference on the identification of gypsum characteristic ion peaks by intelligently partitioning and dynamically calibrating the original conductivity-time spectrum data, thereby improving the accuracy and repeatability of low-concentration gypsum ion chromatography detection.

[0007] In the technical solution of this invention, after obtaining the raw conductivity-time spectrum data of the target soil sample collected by the conductivity detector during ion chromatography detection, noise floor assessment processing is performed on the raw conductivity-time spectrum data to determine the noise floor fluctuation of the detection data, and a dynamic signal response threshold is set based on the noise floor fluctuation. The raw conductivity-time spectrum data is traversed, and data points with values ​​lower than the dynamic signal response threshold are marked as invalid data points. The median of the valid data points in the neighborhood of the invalid data points is used to replace the original values ​​of the invalid data points, thus completing the data cleaning operation and suppressing the disturbance of background noise to subsequent baseline identification. Preferably, a continuous data sequence within the blank time period in the raw conductivity-time spectrum data where no chromatographic peaks appear is extracted. This continuous data sequence is sorted, and 5% of the data points at both ends of the sorted sequence are removed to obtain the main data segment. The difference between the maximum and minimum values ​​of the main data segment is used as the noise floor fluctuation, thereby making the noise assessment results more robust.

[0008] After data cleaning, the raw conductivity-time spectrum data is subjected to partitioned statistical analysis to identify multiple baseline data segments adjacent to the characteristic ion peaks of gypsum. Preferably, a second-order difference operation is performed on the raw conductivity-time spectrum data to obtain a difference signal sequence. Multiple data intervals are then divided based on the time positions of consecutive zero-crossing points in the difference signal sequence. From these intervals, data intervals located within a preset time window before and after the retention time of the characteristic ion peaks of gypsum are selected as baseline data segments. This method automatically avoids chromatographic peak regions and accurately selects signal segments reflecting baseline fluctuations.

[0009] A composite dynamic calibration coefficient is constructed based on the local density distribution characteristics of each baseline segment data point within multiple baseline data segments. As a preferred embodiment of this invention, the interquartile range of each baseline segment data point is calculated as a first discrete metric parameter, and the median absolute deviation between all data points in each baseline segment data point and its median is calculated as a second discrete metric parameter. The first and second discrete metric parameters are then weighted and multiplied to generate a local fluctuation intensity index for each baseline segment data point. The arithmetic mean of these local fluctuation intensity indices is then applied to obtain the composite dynamic calibration coefficient. This coefficient comprehensively reflects the overall intensity and statistical characteristics of baseline fluctuations during the detection process, providing an adaptive adjustment basis for subsequent calibration.

[0010] A point-by-point sliding window baseline correction process is performed on the raw conductivity-time spectrum data using a composite dynamic calibration coefficient to generate calibrated chromatographic data. Specifically, a sliding window centered on each data point in the raw conductivity-time spectrum data is defined. Within each sliding window, the composite dynamic calibration coefficient is compared with the local standard deviation of all data points within the window to determine the set of candidate data points belonging to the baseline region. Preferably, the arithmetic mean and local standard deviation of all data points within the sliding window are calculated. The value of each data point within the window is compared with the arithmetic mean to obtain the deviation magnitude. Data points with deviation magnitudes less than the product of the composite dynamic calibration coefficient and the local standard deviation are selected and included in the candidate data point set. A local polynomial fitting is performed on the candidate data point set to generate the local baseline value of the sliding window. The corrected value of the data point is obtained by subtracting the local baseline value from the raw conductivity-time spectrum data. As a further optimized approach, the time coordinates and conductivity values ​​of each data point in the candidate data point set are extracted. A first-order polynomial linear regression is then performed on the time coordinates and conductivity values ​​using the least squares criterion to obtain the slope and intercept parameters of the regression line. The conductivity value calculated by substituting the time coordinate of the center point of the sliding window into the regression line is then used as the local baseline value. Through this process, baseline correction can dynamically adapt to local signal fluctuation characteristics and effectively separate the true characteristic ion peaks of gypsum.

[0011] Before transmitting the calibrated chromatographic data to the subsequent qualitative and quantitative analysis modules, peak shape symmetry verification is performed on the calibrated chromatographic data. The skewness coefficient of each chromatographic peak is calculated, and abnormally deformed peaks are identified based on the comparison between the skewness coefficient and a preset peak shape symmetry threshold. For the data intervals corresponding to the abnormally deformed peaks, a point-by-point sliding window baseline correction is re-executed using a composite dynamic calibration coefficient, generating chromatographic data after secondary calibration. Preferably, the start and end times of each chromatographic peak are determined, the corresponding continuous data sequence is extracted, and the time position corresponding to the maximum peak height of the continuous data sequence is calculated as the peak peak time. The first time difference between the peak peak time and the start time, and the second time difference between the end time and the peak peak time are calculated, and the ratio of the first time difference to the second time difference is used as the skewness coefficient. This method can promptly detect peak shape distortion caused by improper baseline calibration and perform iterative correction, further ensuring the reliability of subsequent ion concentration analysis. Finally, the calibrated and verified chromatographic data is input into the subsequent qualitative and quantitative analysis modules to complete the concentration analysis of target ions in soil gypsum.

[0012] This invention avoids the problem of poor adaptability of fixed baseline models to complex soil matrices by using multi-stage adaptive baseline identification, dynamic calibration coefficient construction and sliding window correction. It effectively improves the accuracy of chromatographic peak identification and the precision of quantitative analysis, and is especially suitable for trace ion detection scenarios in low-concentration gypsum samples.

[0013] The technical effects and advantages provided by the present invention in the above technical solution are as follows: By extracting local density distribution features from multiple identified baseline segments, the interquartile range of each segment is calculated as the first discrete metric, and the median absolute deviation of all data points in each segment from its median is calculated as the second discrete metric. The weighted product of these two parameters yields the local fluctuation intensity index for each baseline segment. An arithmetic mean of these local fluctuation intensity indices from multiple baseline segments is then used to construct a composite dynamic calibration coefficient. This coefficient integrates robust discrete information regarding the dispersion and deviation trend of the baseline, characterizing the local density distribution near the gypsum characteristic ion peak. This allows the calibration coefficient to adaptively change with the actual baseline noise structure and density, avoiding misjudgments of baseline candidate points caused by the inability to track local density differences in conventional fixed thresholds.

[0014] In point-by-point sliding window baseline correction, a sliding window is set for each data point. The composite dynamic calibration coefficient is compared with the local standard deviation of all data points within the window. Data points with deviations less than the product of the composite dynamic calibration coefficient and the local standard deviation are selected as candidate data points. Local polynomial fitting is then performed on the candidate data point set to generate the local baseline value at the center of the window. By directly linking the selection criteria for data points within the window with the calibration coefficient that dynamically characterizes the local density distribution, the baseline candidate point determination for each window can adaptively adjust according to the actual dispersion and local density distribution of the data within that window. Within the narrow interval where gypsum characteristic ion peaks coexist with adjacent impurity peaks, the data points belonging to the baseline are accurately isolated for fitting, suppressing peak height reduction and peak shape asymmetry caused by the participation of misselected peak data points in baseline estimation. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0016] Figure 1 This is a flowchart of the intelligent baseline dynamic calibration data processing method for soil gypsum ion chromatography detection; Figure 2This is a flowchart of the baseline fragment data extraction method; Figure 3 This is a schematic curve illustrating the point-by-point sliding window baseline correction of the original conductivity-time spectrum data. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] See Figure 1 This invention provides an intelligent baseline dynamic calibration data processing method for ion chromatography detection of gypsum in soil. The method includes: acquiring raw conductivity-time spectrum data of a target soil sample collected by a conductivity detector during ion chromatography detection; performing partitioned statistical analysis on the raw conductivity-time spectrum data to identify multiple baseline fragments adjacent to the characteristic ion peaks of gypsum; constructing a composite dynamic calibration coefficient based on the local density distribution characteristics of each baseline fragment; performing point-by-point sliding window baseline correction processing on the raw conductivity-time spectrum data using the composite dynamic calibration coefficient to generate calibrated chromatographic data; and transmitting the calibrated chromatographic data as input to a subsequent qualitative and quantitative analysis module for ion concentration analysis.

[0019] Example 1: In specific implementation, noise floor assessment is performed on the raw conductivity-time spectrum data to determine the noise floor fluctuation of the detection data. A continuous data sequence is extracted from the blank time period in the raw conductivity-time spectrum data where no chromatographic peaks appear. The blank time period is identified by analyzing the change in conductivity values ​​over time in the raw conductivity-time spectrum data. Time periods where the conductivity values ​​of multiple consecutive data points are all below the first-order differential signal threshold and the absolute conductivity value is at the background level are determined as blank time periods without chromatographic peaks. The corresponding raw conductivity-time spectrum data point sequence within this blank time period is extracted to obtain the continuous data sequence.

[0020] The continuous data sequence is sorted, and 5% of the data points at both ends of the sorted sequence are removed to obtain the main data segment. Specifically, all conductivity values ​​in the continuous data sequence are arranged in ascending order to form an ordered sequence. The number of data points corresponding to 5% of the total number of data points in the ordered sequence is calculated. Data points corresponding to this number at the beginning and end of the ordered sequence are removed, and the remaining data points in the ordered sequence are retained as the main data segment.

[0021] Calculate the difference between the maximum and minimum values ​​of the main data segment, and use this difference as the noise floor fluctuation. The noise floor fluctuation is denoted as... .

[0022] A dynamic signal response threshold is set based on the noise floor fluctuation. Dynamic signal response threshold Calculate using the following formula: ; in, This is a preset proportional coefficient. The range of values ​​is to The preferred value is Select The basis for the value of 2 is that during the blank period without chromatographic peaks, the instantaneous amplitude fluctuation of conductivity noise is constrained to within... Within the peak-to-peak range characterized, the noise levels within this range fall within the normal fluctuation level of the instrument's background noise; when When =2, the dynamic signal response threshold expands to The peak-to-peak range can cover most of the numerical distribution of measured noise data within the blank time period (under actual ion chromatography noise distribution conditions, this threshold setting can cover more than...). Pure noise data points are correctly identified as invalid data points. At the same time, the threshold is significantly lower than the minimum response amplitude of the characteristic ion peak of gypsum, thus effectively filtering out background noise while avoiding mislabeling of the true chromatographic peak signal. The optimal value was determined by conducting preliminary experiments on the blank solvent under detection conditions to measure its noise amplitude distribution, ensuring that the noise mislabeling rate was lower than [a certain value]. To simplify parameter configuration, parameters can be directly selected when no prior experimental conditions are available. .

[0023] Iterate through the raw conductivity-time spectrum data, marking each data point with a value below the dynamic signal response threshold as an invalid data point. For each marked invalid data point, perform the following substitution operation: using the invalid data point's time index in the raw conductivity-time spectrum data as the center, extract the values ​​before and after it. Each data point is used as a neighborhood data point. For positive integers, the preferred value is [value]. Select data points from the neighborhood data points whose conductivity values ​​are greater than or equal to the dynamic signal response threshold as valid data points. If the number of valid data points is not zero, calculate the median conductivity value of these valid data points and replace the original conductivity value of the invalid data points with the median. If the number of valid data points is zero, then... The value increases The process involves reconstructing the neighborhood data points and filtering for valid data points until the number of valid data points is not zero. Then, the median conductivity value of the invalid data points is used to replace the original conductivity value of each valid data point. After performing this replacement operation on all invalid data points, the data cleaning operation is complete.

[0024] Example 2: In specific implementation, refer to Figure 2 A second-order difference operation is performed on the original conductivity-time spectrum data to obtain a differential signal sequence. Assume the original conductivity-time spectrum data contains... The data point, the first The time corresponding to each data point is The conductivity value is The second-order difference operation is performed as follows: ; in, Indicates the first Conductivity values ​​for each data point Indicates the first Conductivity values ​​for each data point Indicates the first Conductivity values ​​for each data point The range of values ​​is to The integer. The calculated result Arranged in sequential order, they form a differential signal sequence.

[0025] After obtaining the differential signal sequence, multiple data intervals are divided based on the time positions of consecutive zero-crossing points in the sequence. The method for identifying zero-crossing points is: traverse the differential signal sequence, and when... and The signs are different, that is At that time, it was determined that at the time point and There exists a zero-crossing point. The time position of the zero-crossing point. Determined by linear interpolation: ; All identified zero-crossings are sorted chronologically to obtain a zero-crossing sequence. A data interval is defined between every two adjacent zero-crossings, with the time boundary of the data interval jointly defined by the time positions of the preceding and following zero-crossings. All raw conductivity-time spectrum data points within this data interval constitute a candidate segment.

[0026] From the identified candidate fragments, data intervals within a preset time window before and after the retention time of the gypsum characteristic ion peak were selected, and these selected data intervals were used as multiple baseline fragments. The retention time of the gypsum characteristic ion peak was obtained by pre-determining the gypsum standard solution under the same ion chromatography conditions, and denoted as . The width of the preset time window is set based on the column efficiency of ion chromatography and the broadening of the gypsum peak. It is typically three to five times the half-peak width of the characteristic gypsum ion peak. Specific settings are as follows: minutes to Minutes. For each candidate segment, the average timestamps of all data points within that segment are calculated and used as the center time of that segment. If the center time of a candidate segment falls within a preset time window, the corresponding data interval is determined as a baseline segment. All eligible baseline segments are extracted for use in the subsequent construction of composite dynamic calibration coefficients.

[0027] Example 3: In specific implementation, local density distribution features are extracted from each of the multiple baseline data segments. The multiple baseline data segments were obtained in Example 2 through zero-crossing point division of the differential signal sequence and selection using the retention time window of gypsum characteristic ion peaks.

[0028] For a single baseline segment of data, the interquartile range of the baseline segment is calculated as the first discrete metric. The conductivity values ​​of all data points within the baseline segment are then sorted in ascending order to obtain a sorted sequence. The first quartile of the sorted sequence is then identified. That is, the position in the sorted sequence The conductivity value at the location; find the third quartile of the sorted sequence. That is, the position in the sorted sequence. The conductivity value at the location. Calculate the interquartile range. The interquartile range This is the first discrete metric parameter.

[0029] The median absolute deviation of all data points in the baseline segment data from its median is calculated and used as the second discrete metric. The median conductivity value of all data points within the baseline segment data is then calculated. For each data point in the baseline fragment data, calculate the conductivity value and median for that data point. The absolute value of the difference between the data points is used to obtain an absolute deviation value. An absolute deviation sequence is formed by organizing the absolute deviation values ​​corresponding to all data points. This sequence is then arranged in ascending order, and the median, located in the middle of the sorted sequence, is found. This median of the absolute deviation is the second discrete metric parameter, denoted as . .

[0030] For the first discrete metric parameter Second discrete metric parameter A weighted product operation is performed to generate a local fluctuation intensity index for the baseline segment data. The local fluctuation intensity index is calculated according to the following formula: ; in, Indicates the first The local fluctuation intensity index of each baseline segment data is expressed in units of electrical conductivity. Indicates the first The interquartile range of each baseline segment data, i.e., the first discrete metric parameter; Indicates the first The median absolute deviation of each baseline segment data, i.e., the second discrete metric parameter; This is the weighted fusion coefficient, and its value range is... to . Values to The preferred value is Select for The basis is: and The emphasis on characterizing the degree of data dispersion differs under different distribution patterns. It is insensitive to outliers but retains information about the overall span of the distribution. It provides a more detailed characterization of the dispersion of data near the distribution center, taking This approach ensures that the two measurement parameters contribute equally to the local fluctuation intensity index, achieving a balance between robustness and sensitivity, and is suitable for the typical distribution patterns of baseline fluctuations in soil gypsum ion chromatography detection. In application scenarios where the factory noise characteristics of the ion chromatography conductivity detector are known and sufficient historical baseline data has been accumulated, it can also be achieved by... to Within the range Perform a grid search for each candidate step size. The optimal selection criterion is to minimize the relative standard deviation of the characteristic ion peak area of ​​gypsum after calibration. The optimal value; when there is no historical data, directly take .

[0031] After arithmetically averaging the local fluctuation intensity indices of multiple baseline segments, and then normalizing them, dimensionless composite dynamic calibration coefficients are obtained. Let the number of baseline segments be... Each baseline segment of data corresponds to a local fluctuation intensity index. , The range of values ​​is to The integer. First, calculate the arithmetic mean of the local fluctuation intensity index. : ; After normalization, It is a dimensionless proportionality coefficient that characterizes the proportional relationship between the intensity of local fluctuations in the baseline and the overall level of dispersion. Its magnitude reflects the relative intensity of the baseline fluctuations and is used to dynamically adjust the sensitivity of candidate data point selection in subsequent sliding window correction.

[0032] The composite dynamic calibration coefficient The local density distribution characteristics of all baseline fragment data in the adjacent region of gypsum characteristic ion peaks are integrated for subsequent point-by-point sliding window baseline correction processing.

[0033] Example 4: In specific implementation, a composite dynamic calibration coefficient is used to perform point-by-point sliding window baseline correction processing on the original conductivity-time spectrum data. The composite dynamic calibration coefficient was obtained in Example 3 through the local density distribution characteristics of the baseline fragment data, and is denoted as... The raw conductivity-time spectrum data includes The data point, the first The time coordinates corresponding to each data point are The original conductivity value is .

[0034] For each data point in the original conductivity-time spectrum data, a sliding window centered on that data point is defined. For the ... For each data point, the sliding window is set up as follows: based on the time coordinate... Centered on a point, extend a fixed half-window width in both the decreasing and increasing time directions. This half-window width is denoted as . . The value is to The sampling interval corresponding to each data point is preferably set to a value of [value]. The sampling interval corresponding to each data point. The half-window width is taken as... The basis for this is: at typical ion chromatography sampling rates, The time width corresponding to each data point is approximately Instant The time width, measured in seconds, is sufficient to cover the local statistical characteristics of the baseline region while avoiding the inclusion of signals from adjacent chromatographic peaks in the calculation due to an excessively wide window, which could cause baseline estimation bias. The range of data point numbers covered by the sliding window is from... to The corresponding time interval is For boundary data points near the ends of the original conductivity-time spectrum data, when Less than When, the starting point of the sliding window is taken as the first... Data points; when Greater than When, the endpoint of the sliding window is taken as the first... Data points.

[0035] Within each sliding window, the set of candidate data points is determined. The arithmetic mean and local standard deviation of all data points within the sliding window are calculated. The sliding window contains... Data points, The value is determined as follows: In non-boundary regions, ; In the boundary area (i.e.) or hour), Take the number of data points actually covered by the sliding window, i.e. ; in ; ; No. The conductivity values ​​of the data points are denoted as follows: Arithmetic mean Calculate using the following formula: ; in, This represents the arithmetic mean of all data points within the sliding window. Indicates the total number of data points within the sliding window (the number of data points at the boundary is based on the actual number of data points contained within the window); Indicates the first [number]th ... Conductivity values ​​for each data point; The index number of the data point within the sliding window. The range of values ​​is to Integers.

[0036] Local standard deviation Calculate using the following formula: ; in, This represents the local standard deviation of all data points within the sliding window. In the boundary region, due to... Value less than The degrees of freedom in calculating local standard deviations are correspondingly reduced, but At the boundary, it is still not less than (That is, it contains at least half the width of the window of data points), which can ensure the stability of the statistical estimation.

[0037] After calculating the arithmetic mean and local standard deviation within the sliding window, the conductivity value of each data point within the sliding window is... with arithmetic mean Compare the data points and calculate the deviation for each data point. Deviation of data points Calculate using the following formula: ; in, Indicates the first [number]th ... The deviation of each data point.

[0038] For each data point within the sliding window, perform a candidate data point attribution determination. Then, determine the deviation magnitude. With composite dynamic calibration coefficient (Dimensionless) and Local Standard Deviation The products are compared. When When this happens, the data point is identified as a candidate data point belonging to the baseline region and added to the candidate data point set. It is a dimensionless proportionality coefficient. The dimension is the unit of electrical conductivity, and... The dimensions are consistent. When If the condition is not met, the data point will not be included in the candidate data point set. The process involves iterating through all data points within the sliding window to complete the filtering of the candidate data point set.

[0039] After obtaining the set of candidate data points within the sliding window, a local polynomial fitting is performed on the candidate data point set. The time coordinate and conductivity value of each candidate data point in the candidate data point set are extracted. Let the candidate data point set contain... The candidate data point, the first The time coordinates of the candidate data points are The conductivity value is , The range of values ​​is to Integers.

[0040] The least squares criterion was used to perform a first-order polynomial linear regression calculation on the time coordinates and conductivity values. The model expression for the first-order polynomial is: ; in, This represents the numerical value of electrical conductivity. Represents time coordinates; The slope parameter of the regression line; This represents the intercept parameter of the regression line.

[0041] slope parameter and intercept parameter This is obtained by solving the following system of normal equations: ; in, Indicates the first The time coordinates of the candidate data points; Indicates the first The conductivity values ​​of each candidate data point; This represents the total number of candidate data points in the candidate data point set.

[0042] The slope parameters are obtained by solving. and intercept parameter Then, the time coordinates of the center point of the sliding window are... Substitute the values ​​into the regression line expression to calculate the local baseline value for the sliding window. Local baseline value Calculate using the following formula: ; in, Indicates the time coordinate as The local baseline value corresponding to the data point; Represents the time coordinates of the center point of the sliding window; The slope parameter of the regression line; This represents the intercept parameter of the regression line.

[0043] The corrected value for a data point is obtained by subtracting the local baseline value from the original conductivity-time spectrum value. Calculate using the following formula: ; in, The time coordinate is The corrected values ​​of the data points; The time coordinate is The raw conductivity values ​​of the data points; The time coordinate is The local baseline value corresponding to the data point.

[0044] For each data point in the original conductivity-time spectrum data, the above-mentioned operations of setting the sliding window, determining the candidate data point set, calculating the local baseline value, and calculating the corrected value are repeated. After traversing all data points, the corrected values ​​of all data points are arranged in chronological order to generate the calibrated chromatographic data.

[0045] See Figure 3 The horizontal axis in the graph represents the detection time, in minutes. The vertical axis represents the conductivity value, in microsiemens per centimeter (µS / cm). In the figure, the blue solid curve represents the original conductivity-time spectrum data obtained in Example 4, the red dashed line represents the local baseline estimate obtained by point-by-point sliding window baseline correction within the same time range, and the green dotted line represents the corrected chromatographic data.

[0046] As can be seen from the figure, the original conductivity-time spectrum data is... to The minute interval shows two distinct peaks, located at approximately minutes and The retention time region of the characteristic ion peaks of gypsum is approximately [minutes]. The red dashed baseline is smooth overall, approximating a slowly changing curve, accurately reflecting the local baseline fluctuation characteristics of the original data. Furthermore, the baseline does not show obvious abnormal fluctuations in the two-peak region, demonstrating the effective fitting of the local baseline by the baseline correction process.

[0047] At the peaks of both peaks, the chromatographic data curve after green correction exhibits a sharp peak shape and is close to the zero baseline, indicating that the composite dynamic calibration coefficient has been successfully applied. With local standard deviation Using the product of the two factors as a filtering condition, the candidate data point set located by the sliding window accurately distinguishes the peak signal from the baseline, effectively stripping the baseline and enhancing the peak signal. This correction result conforms to the adjustment of the sliding window half-window width in Example 4. The description of the setting and local polynomial fitting fully demonstrates the advantages of combining local statistical features with dynamic calibration coefficients.

[0048] Example 5: In a specific implementation, peak shape symmetry verification is performed on the calibrated chromatographic data. The calibrated chromatographic data was generated in Example 4, and the calibrated chromatographic data includes... The data point, the first The time coordinates of the data points are The corrected conductivity value is , The range of values ​​is to Integers.

[0049] To identify all chromatographic peaks in the calibrated chromatographic data, the following steps are taken: First-order derivative calculations are performed on the calibrated chromatographic data to locate zero-crossing points where the first derivative changes from positive to negative. Each such zero-crossing point corresponds to a candidate peak position. From each candidate peak position, the calculation is extended in both the decreasing and increasing time directions until a first-order derivative zero-crossing point is encountered or the corrected conductivity value drops below the composite dynamic calibration coefficient. With noise floor fluctuation The product region boundaries determine the start and end time points of the chromatographic peaks. Each chromatographic peak corresponds to a continuous data sequence, which consists of all data points from the start to the end time point.

[0050] For each chromatographic peak, the skewness coefficient is calculated. The calculation process is as follows: extract the continuous data sequence corresponding to the chromatographic peak, search for the maximum corrected conductivity value in the continuous data sequence, and this maximum value is the peak height. The time coordinate corresponding to the peak height is recorded as the peak time. The starting time point of the chromatographic peak was recorded as follows: The termination time point is Calculate the first time difference between the peak time and the starting time. : ; in, This indicates the first time difference, which is the time span of the chromatographic peak rise phase; This indicates the peak time, i.e., the time coordinate corresponding to the maximum peak height of the chromatographic peak; Indicates the starting time point of the chromatographic peak.

[0051] Calculate the second time difference between the termination time and the peak time. : ; in, This indicates the second time difference, which is the time span of the chromatographic peak descent phase; Indicates the termination time point of the chromatographic peak; Indicates the peak time.

[0052] The ratio of the first time difference to the second time difference is used as the skewness coefficient. The skewness coefficient is calculated using the following formula: ; in, Indicates the skewness coefficient; Indicates the first time difference; Indicates the second time difference. Skewness coefficient. Greater than Indicates the chromatographic peak development and skewness coefficient. Less than Indicates peak tailing, skewness coefficient equal This indicates that the chromatographic peaks are completely symmetrical.

[0053] In practical implementation, a threshold range for peak shape symmetry is preset, and the lower limit of the peak shape symmetry threshold is denoted as... The upper limit of the peak shape symmetry threshold is denoted as . The value is , The value is .set up for and for The basis for this is that in ion chromatography analysis, normal chromatographic peaks, under ideal column efficiency conditions, approximate a Gaussian distribution, and their skewness coefficient is close to... Due to the physical inhomogeneity of the column packing and the influence of the mass transfer process, the skewness coefficient of the actual chromatographic peak is... to The peak shape can still be considered normal between these ranges; when the skewness coefficient exceeds this range, the chromatographic peak shows obvious asymmetric deformation, indicating that the baseline correction is undercorrected or overcorrected in this chromatographic peak region.

[0054] Based on the comparison between the skewness coefficient and the preset peak shape symmetry threshold, abnormally deformed peaks are identified. The skewness coefficient of each chromatographic peak is then calculated. respectively with and Comparison. When or When this occurs, the chromatographic peak is marked as an abnormally distorted peak. At that time, the chromatographic peak was marked as a normal peak.

[0055] For each chromatographic peak marked as an anomalous deformation peak, the starting time point of the anomalous deformation peak is extracted. and termination time point Determine the data interval corresponding to the abnormal deformation peak, and the time range of the data interval is... The composite dynamic calibration coefficient is applied to the data range corresponding to the abnormal deformation peak. Re-execute the point-by-point sliding window baseline correction process. The re-execution process is consistent with the steps described in Example 4: set a sliding window centered on each data point within the data interval, and apply the composite dynamic calibration coefficients within each sliding window. By comparing the data with the local standard deviation, a set of candidate data points belonging to the baseline region is determined. Local polynomial fitting is performed on the candidate data point set to generate local baseline values. The original conductivity value is subtracted from the local baseline value to obtain the value after secondary correction.

[0056] After recalibrating all abnormal peak data ranges, the original calibration values ​​for the corresponding data points in the calibrated chromatographic data are replaced with the recalibrated values, generating the recalibrated chromatographic data. This recalibrated chromatographic data is then used as input to the subsequent qualitative and quantitative analysis module for ion concentration analysis.

[0057] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for intelligent baseline dynamic calibration data processing in ion chromatography detection of soil gypsum, characterized in that, The method includes: Obtain raw conductivity-time spectrum data of the target soil sample collected by a conductivity detector during ion chromatography detection; The original conductivity-time spectrum data were subjected to partitioned statistical analysis to identify multiple baseline data segments adjacent to the characteristic ion peaks of gypsum; A composite dynamic calibration coefficient is constructed based on the local density distribution characteristics of each baseline segment data in the plurality of baseline segment data; The composite dynamic calibration coefficients are used to perform point-by-point sliding window baseline correction on the raw conductivity-time spectrum data to generate calibrated chromatographic data. The calibrated chromatographic data is then passed as input to the subsequent qualitative and quantitative analysis module for ion concentration analysis.

2. The intelligent baseline dynamic calibration data processing method for soil gypsum ion chromatography detection according to claim 1, characterized in that, The step of performing partitioned statistical analysis on the raw conductivity-time spectrum data to identify multiple baseline data segments adjacent to the characteristic ion peaks of gypsum specifically includes: Perform a second-order difference operation on the original conductivity-time spectrum data to obtain a difference signal sequence; Multiple data intervals are divided based on the time positions of consecutive zero-crossing points in the differential signal sequence; From the multiple data intervals, select each data interval that is within a preset time window before and after the retention time of the gypsum characteristic ion peak, and use each selected data interval as the multiple baseline segment data.

3. The intelligent baseline dynamic calibration data processing method for soil gypsum ion chromatography detection according to claim 1, characterized in that, The construction of composite dynamic calibration coefficients based on the local density distribution characteristics of each baseline segment data in the plurality of baseline segment data specifically includes: The interquartile range of each baseline segment data is calculated as the first discrete metric; the median absolute deviation between all data points in each baseline segment data and its median is calculated as the second discrete metric. The first discrete metric parameter and the second discrete metric parameter are weighted and multiplied to generate a local fluctuation intensity index for each baseline segment data. The local fluctuation intensity indices of the multiple baseline segment data are then arithmetically averaged to obtain the composite dynamic calibration coefficient.

4. The intelligent baseline dynamic calibration data processing method for soil gypsum ion chromatography detection according to claim 1, characterized in that, The step of performing point-by-point sliding window baseline correction on the original conductivity-time spectrum data using the composite dynamic calibration coefficient to generate calibrated chromatographic data specifically includes: For each data point in the original conductivity-time spectrum data, a sliding window centered on that data point is set; Within each sliding window, the composite dynamic calibration coefficient is compared and analyzed with the local standard deviation of all data points within the window to determine the set of candidate data points belonging to the baseline region within the window; A local polynomial fit is performed on the candidate data point set to generate the local baseline value of the sliding window; the corrected value of the data point is obtained by subtracting the local baseline value from the original conductivity-time spectrum data.

5. The intelligent baseline dynamic calibration data processing method for soil gypsum ion chromatography detection according to claim 4, characterized in that, Within each sliding window, the composite dynamic calibration coefficient is compared and analyzed with the local standard deviation of all data points within the window to determine the set of candidate data points belonging to the baseline region within the window. This specifically includes: Calculate the arithmetic mean and the local standard deviation of all data points within the sliding window; Compare the value of each data point in the window with the arithmetic mean, and calculate the deviation of each data point. Data points whose deviation is less than the product of the composite dynamic calibration coefficient and the local standard deviation are selected and included in the candidate data point set.

6. The intelligent baseline dynamic calibration data processing method for soil gypsum ion chromatography detection according to claim 4, characterized in that, The step of performing local polynomial fitting on the candidate data point set to generate the local baseline value of the sliding window specifically includes: Extract the time coordinates and conductivity values ​​of each data point in the candidate data point set; The least squares criterion is used to perform a first polynomial linear regression on the time coordinates and the conductivity values ​​to obtain the slope and intercept parameters of the regression line; Substitute the time coordinate of the center point of the sliding window into the regression line, and use the calculated conductivity value as the local baseline value.

7. The intelligent baseline dynamic calibration data processing method for soil gypsum ion chromatography detection according to claim 1, characterized in that, Before transmitting the calibrated chromatographic data as input to the subsequent qualitative and quantitative analysis module for ion concentration analysis, the method further includes: The calibrated chromatographic data is subjected to peak shape symmetry verification processing, and the skewness coefficient of each chromatographic peak is calculated; based on the comparison result of the skewness coefficient and the preset peak shape symmetry threshold, abnormally deformed peaks are identified. For the data range corresponding to the abnormally deformed peak, the composite dynamic calibration coefficient is called to re-execute the point-by-point sliding window baseline correction process to generate chromatographic data after secondary calibration.

8. The intelligent baseline dynamic calibration data processing method for soil gypsum ion chromatography detection according to claim 7, characterized in that, The step of performing peak shape symmetry verification on the calibrated chromatographic data and calculating the skewness coefficient of each chromatographic peak specifically includes: Determine the start and end time points of each chromatographic peak and extract the corresponding continuous data sequence. The time position corresponding to the maximum peak height of the continuous data sequence is calculated as the peak time; Calculate the first time difference between the peak time and the starting time, and the second time difference between the ending time and the peak time; The ratio of the first time difference to the second time difference is used as the skewness coefficient.

9. The intelligent baseline dynamic calibration data processing method for soil gypsum ion chromatography detection according to claim 1, characterized in that, After obtaining the raw conductivity-time spectrum data of the target soil sample collected by the conductivity detector during ion chromatography detection, the process further includes: The original conductivity-time spectrum data is subjected to noise floor assessment processing to determine the noise floor fluctuation of the detection data; a dynamic signal response threshold is set based on the noise floor fluctuation. The original conductivity-time spectrum data is traversed, and data points with values ​​lower than the dynamic signal response threshold are marked as invalid data points. The original value of the invalid data point is replaced with the median of the valid data points in the neighborhood of the invalid data point to complete the data cleaning operation.

10. The intelligent baseline dynamic calibration data processing method for soil gypsum ion chromatography detection according to claim 9, characterized in that, The step of performing noise floor assessment processing on the original conductivity-time spectrum data to determine the noise floor fluctuation of the detection data specifically includes: Extract the continuous data sequence within the blank time period in the original conductivity-time spectrum data where no chromatographic peaks appear; The continuous data sequence is sorted, and 5% of the data points at both ends of the sorted sequence are removed to obtain the main data segment. Calculate the difference between the maximum and minimum values ​​of the main data segment, and use the difference as the noise floor fluctuation.

Images