A method and system for particle size distribution testing of a laser particle size analyzer

Abstract

The present application relates to the technical field of laser particle size analysis, in particular to a laser particle size analyzer particle size distribution test method and system, the method comprises the following steps: extracting scattering characteristic splicing multi-dimensional matrix, combining theoretical model to calculate mismatch degree vector to divide form interference and stable section to generate binary mask vector, and constructing reduced dimension inversion equation set to solve basic particle size distribution accordingly, using weight reduction coefficient to establish weighted inversion equation set with penalty matrix to solve comprehensive distribution, calculating residual and combining confidence index to output correction result. In the present application, by isolating the scattering channel abnormally changed by non-spherical interference, the conversion path of optical difference to wrong components is blocked, the reduced dimension inversion is constructed relying on the mask mechanism and trend extraction, the weighted compensation constraint is implemented by introducing the weight reduction penalty, the false spread and false multi-peak phenomenon caused by misjudging abnormal scattering as additional components are eliminated, and the real size characteristics of the particles are restored in combination with evaluation.

Description

Technical Field

[0001] This invention relates to the field of laser particle size analysis technology, and in particular to a method and system for testing particle size distribution using a laser particle size analyzer. Background Technology

[0002] The field of laser particle size analysis technology mainly involves determining particle size and its distribution characteristics by utilizing the spatial scattering phenomenon of coherent light beams by particles. Its core aspects include light source emission, beam shaping, scattered light energy acquisition, and inversion calculations based on optical scattering theory. Overall, a monochromatic light beam is irradiated onto a group of moving particles, and a photodetector captures the spatial distribution signal of light energy at different deflection angles. The particle size information of the particle set is then calculated based on a physical and mathematical model of optical scattering. Traditional laser particle size analyzer particle distribution testing methods refer to the operational procedures for measuring particle size and its corresponding volume percentage.

[0003] Conventional testing of non-spherical particles always relies on a single spherical model to invert measured data. The morphological specificity of particles causes abnormal changes in the scattered light energy in specific angle ranges. Inherent optical differences are directly misinterpreted as additional spherical particle components of different sizes participating in the fitting. The forced introduction of false components causes the particle size distribution curve to be artificially broadened in the final output. Non-existent secondary peaks frequently appear in the test results, seriously masking the true planar size characteristics of non-spherical powders. This causes the measurement results to deviate from the actual physical properties of the material and lose the ability to accurately determine the particle size information of complex samples. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the existing technology and to propose a method and system for testing particle size distribution using a laser particle size analyzer.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a method for testing particle size distribution using a laser particle size analyzer, comprising the following steps:

[0006] S1: Calculate the deviation between the total light intensity sequence of the particle sample and the background light intensity sequence of the particle sample to obtain the full-angle scattered light energy distribution data. Extract the first-order derivative feature sequence and the second-order derivative feature sequence from the full-angle scattered light energy distribution data. Then, concatenate the full-angle scattered light energy distribution data, the first-order derivative feature sequence, and the second-order derivative feature sequence to generate a multi-dimensional feature matrix.

[0007] S2: Generate a theoretical scattered light energy distribution sequence and a mismatch vector based on the full-angle scattered light energy distribution data and the built-in spherical particle optical conversion matrix. Calculate a comprehensive evaluation value using the mismatch vector and the second derivative feature sequence in the multidimensional feature matrix. Based on the comprehensive evaluation value, divide the detector array into a morphological interference segment and a stable contribution segment, and generate a binary mask vector.

[0008] S3: Based on the difference between the full-angle scattered light energy distribution data and the theoretical scattered light energy distribution sequence in the morphological interference section, high-risk mapping areas are extracted to generate mapping trend information. The basic particle size distribution is obtained by establishing and solving the reduced-dimensional inversion equation set according to the binary mask vector.

[0009] S4: Calculate the weighting coefficient using the corresponding numerical value of the angle channel in the morphological interference section using the mismatch degree vector, and combine the weighting coefficient with the unit weight belonging to the stable contribution section to generate a global diagonal weight matrix. Combine the high-risk mapping area to establish a global weighted inversion equation system with a penalty matrix and solve it based on the basic particle size distribution to obtain the comprehensive particle size distribution.

[0010] S5: Calculate the residual values ​​of the angle channels belonging to the stable contribution segment based on the spherical particle optical transformation matrix, comprehensive particle size distribution, full-angle scattered light energy distribution data, and binary mask vector to obtain the first local residual. If the first local residual is determined to be lower than the preset convergence threshold, generate key particle size feature parameters and confidence index by combining the global diagonal weight matrix and mapping trend information. Calculate the morphological deviation index using the weighting coefficient and output the corrected particle size distribution result by combining the key particle size feature parameters and confidence index.

[0011] The present invention is improved in that the specific steps of S1 are as follows:

[0012] S111: Collect the total light intensity sequence of the particle sample within the set test time window, monitor the background light intensity sequence of the particle sample under the pure dispersion medium circulation state, calculate the light energy intensity deviation between the total light intensity sequence and the background light intensity sequence, and obtain the full-angle scattered light energy distribution data.

[0013] S112: Map the full-angle scattered light energy distribution data to a logarithmic angular coordinate system, perform a cubic spline interpolation algorithm to fit the full-angle scattered light energy distribution data, establish a continuous smooth energy curve, calculate the rate of change of light energy between the intervals formed by adjacent standardized angle sampling points on the continuous smooth energy curve, obtain the first derivative feature sequence, solve the second-order rate of change feature of the first derivative feature sequence in the angular dimension, perform an adaptive sliding window filtering algorithm to smooth the second-order rate of change feature, and obtain the second-order derivative feature sequence;

[0014] S113: Combine the full-angle scattered light energy distribution data, the first derivative feature sequence, and the second derivative feature sequence to generate a multi-dimensional feature matrix.

[0015] The present invention is improved in that the specific steps of S2 are as follows:

[0016] S211: Obtain the built-in spherical particle optical transformation matrix, execute the non-negative least squares algorithm to fit the mapping relationship between the full-angle scattered light energy distribution data and the spherical particle optical transformation matrix to solve for the equivalent spherical diameter distribution result, calculate the particle group scattering state based on the spherical particle optical transformation matrix and the equivalent spherical diameter distribution result, and obtain the theoretical scattered light energy distribution sequence;

[0017] S212: Calculate the absolute value of the deviation between the full-angle scattered light energy distribution data and the theoretical scattered light energy distribution sequence on each detector channel corresponding to the full-angle scattered light energy distribution data to obtain the mismatch vector. Convert the element values ​​contained in the mismatch vector to obtain the mismatch score. Extract the second derivative feature sequence contained in the multi-dimensional feature matrix. Calculate the curvature score corresponding to the second derivative feature sequence. Calculate the weighted sum between the mismatch score and the curvature score to obtain the comprehensive evaluation value.

[0018] S213: Extract the channel peak points that exceed the preset confidence threshold from the comprehensive evaluation values ​​as starting seed points, execute the bidirectional region growing algorithm to divide the detector array corresponding to each detector channel into morphological interference segments and stable contribution segments according to the starting seed points, and assign the values ​​one and zero to the detector array according to the distribution channel positions of the morphological interference segments and the stable contribution segments respectively to generate a binary mask vector.

[0019] The present invention is improved in that the process of extracting the channel peak points that exceed the preset confidence threshold from the comprehensive evaluation values ​​as the starting seed points is specifically as follows:

[0020] Obtain the blank background evaluation value of each detector channel under the pure medium circulation state, and aggregate the blank background evaluation values ​​of each detector channel to construct a blank background evaluation value set.

[0021] Calculate the global average value of the set of blank background evaluation values, and calculate the standard deviation value of all elements in the set of blank background evaluation values ​​relative to the global average value;

[0022] The global average value and the standard deviation value are summed, and the calculated value of the summation is configured as a preset reliability threshold.

[0023] By comparing the comprehensive evaluation value of each detector channel with the preset confidence threshold, candidate abnormal channels that are greater than the preset confidence threshold are selected from the detector array.

[0024] Extract the local maximum value channel within the candidate anomaly channel where the comprehensive evaluation value is greater than the comprehensive evaluation value corresponding to the channels of the two adjacent detectors, extract the local maximum value channel as the channel peak point, and use the channel peak point as the starting seed point.

[0025] The present invention is improved in that the specific steps of S3 are as follows:

[0026] S311: Calculate the energy difference between the full-angle scattered light energy distribution data and the theoretical scattered light energy distribution sequence in the morphological interference section to obtain the residual energy vector. Execute the Pearson correlation coefficient algorithm to calculate the linear correlation between the residual energy vector and the column vectors in the spherical particle optical transformation matrix. Compare the linear correlation with the correlation warning threshold. Based on the correlation warning threshold, select high-risk mapping areas from the particle size components corresponding to the spherical particle optical transformation matrix. Extract the component mapping trend data composed of the difference values ​​between the target column vector and adjacent column vectors in the high-risk mapping area to generate mapping trend information.

[0027] S312: Based on the spatial identifier bits contained in the binary mask vector, remove the full-angle scattered light energy distribution data and the angular dimension data belonging to the morphological interference segment in the optical transformation matrix of the spherical particle, and establish a set of reduced-dimensional inversion equations.

[0028] S313: For the reduced dimension inversion equation set, execute a non-negative least squares iterative algorithm with smoothing constraints to solve the residual minimum value of the reduced dimension inversion equation set and generate the basic particle size distribution.

[0029] The present invention is improved in that the process of extracting the component mapping trend data composed of the difference values ​​between the target column vector and the adjacent column vectors within the high-risk mapping region is specifically as follows:

[0030] Obtain the baseline residual energy sequence of a standard spherical particle sample;

[0031] Calculate the baseline correlation between the baseline residual energy sequence and the column vectors in the optical transformation matrix of the spherical particle;

[0032] The maximum value in the benchmark correlation is multiplied by the preset anti-false judgment ratio factor, and the output of the multiplication operation is configured as the correlation warning threshold.

[0033] Compare the linear correlation with the correlation warning threshold, and extract the target linear correlation that is greater than the correlation warning threshold;

[0034] Identify the target particle size components corresponding to the target linear correlation, and delineate the target particle size components as high-risk mapping regions;

[0035] Read the target column vector in the high-risk mapping region from the optical transformation matrix of the spherical particle;

[0036] Obtain the adjacent column vectors that are immediately adjacent to the boundary of the high-risk mapping region;

[0037] Calculate the difference between the target column vector and the adjacent column vectors, and use the difference as the component mapping trend data of the high-risk mapping area.

[0038] The present invention is improved in that the specific steps of S4 are as follows:

[0039] S411: Call the mismatch degree vector and the morphological interference segment, execute a nonlinear mapping algorithm to calculate the weighting coefficient of the angle channel of the mismatch degree vector in the morphological interference segment, combine the weighting coefficient and the unit weight belonging to the stable contribution segment to generate a global diagonal weight matrix;

[0040] S412: Based on the global diagonal weight matrix, the full-angle scattered light energy distribution data and the spherical particle optical transformation matrix are weighted to establish a global weighted inversion equation set, the high-risk mapping region contained in the mapping trend information is extracted, and the matrix elements of the particle size interval penalty term corresponding to the high-risk mapping region are configured according to the high-risk mapping region to obtain the penalty matrix;

[0041] S413: Set the basic particle size distribution as the initial iteration parameter, and based on the global weighted inversion equation set and the penalty matrix, solve for the minimum residual value of the global weighted inversion equation set containing the penalty matrix to obtain the comprehensive particle size distribution.

[0042] The present invention is improved in that the specific steps of S5 are as follows:

[0043] S511: Call the optical transformation matrix of the spherical particle and the comprehensive particle size distribution to calculate the updated theoretical scattering distribution, calculate the energy difference between the updated theoretical scattering distribution and the full-angle scattered light energy distribution data on each detector channel, obtain the full-angle residual vector, and filter the residual values ​​belonging to the angle channels in the stable contribution segment from the full-angle residual vector according to the binary mask vector to obtain the first local residual;

[0044] S512: Based on the first local residual, when it is determined that the first local residual is lower than the preset convergence threshold, extract the particle volume percentage corresponding to each particle size channel included in the comprehensive particle size distribution, calculate the sum of the distribution interval of the particle volume percentage, obtain the key particle size feature parameter, extract the weight coefficients included in the global diagonal weight matrix, and extract the risk warning index included in the mapping trend information. According to the weight coefficients and the risk warning index, calculate the confidence level value of the particle size component where the key particle size feature parameter is located, and obtain the confidence index.

[0045] S513: Based on the confidence index, calculate the percentage of the integral of the mismatch vector across the entire angular range, calculate the weighted score between the percentage of the integral and the weighting coefficient to obtain the morphological deviation index, and summarize and splice the confidence index, the morphological deviation index and the key granularity feature parameters to generate the corrected granularity distribution result.

[0046] The present invention is improved in that the process of obtaining the first local residual is specifically as follows:

[0047] The optical transformation matrix of the spherical particle is multiplied with the comprehensive particle size distribution to calculate the theoretical light energy distribution column vector corresponding to each detector channel;

[0048] The theoretical light energy distribution column vector is transformed into a row vector data structure by performing a matrix transpose operation, and the theoretical light energy distribution column vector transformed into a row vector structure is set as the updated theoretical scattering distribution;

[0049] Extract the corresponding values ​​of the updated theoretical scattering distribution and the full-angle scattered light energy distribution data on the same detector channel, and calculate the absolute value of the difference between the corresponding value of the updated theoretical scattering distribution and the corresponding value of the full-angle scattered light energy distribution data.

[0050] According to the spatial physical arrangement order of each detector channel, the absolute values ​​of the differences of each detector channel are sequentially spliced ​​to construct a full-angle residual vector;

[0051] Extract the target element with a value equal to one from the binary mask vector, and obtain the spatial index identifier of the target element in the binary mask vector;

[0052] Align and map the spatial index identifier to the full-angle residual vector, and extract the absolute value of the difference between the position corresponding to the spatial index identifier in the full-angle residual vector;

[0053] The extracted absolute value of the difference is set as the residual value belonging to the angle channel within the stable contribution segment, and all residual values ​​belonging to the angle channel within the stable contribution segment are aggregated to obtain the first local residual.

[0054] A laser particle size analyzer particle size distribution testing system is provided, the system being used to implement the aforementioned laser particle size analyzer particle size distribution testing method, the system comprising:

[0055] The scattering signal analysis module calculates the deviation between the total light intensity sequence of the particle sample and the background light intensity sequence of the particle sample to obtain the full-angle scattered light energy distribution data. It extracts the first-order derivative feature sequence and the second-order derivative feature sequence from the full-angle scattered light energy distribution data, and concatenates the full-angle scattered light energy distribution data, the first-order derivative feature sequence and the second-order derivative feature sequence to generate a multi-dimensional feature matrix.

[0056] The channel state assessment module generates a theoretical scattered light energy distribution sequence and a mismatch vector based on the full-angle scattered light energy distribution data and the built-in spherical particle optical conversion matrix. It calculates a comprehensive evaluation value using the mismatch vector and the second derivative feature sequence in the multi-dimensional feature matrix. Based on the comprehensive evaluation value, it divides the detector array into a morphological interference segment and a stable contribution segment and generates a binary mask vector.

[0057] The initial inversion solution module extracts high-risk mapping areas and generates mapping trend information based on the difference between the full-angle scattered light energy distribution data and the theoretical scattered light energy distribution sequence in the morphological interference section. It also establishes and solves the reduced-dimensional inversion equation system based on the binary mask vector to obtain the basic particle size distribution.

[0058] The weighted compensation iteration module calculates the weighting coefficient using the corresponding numerical values ​​of the angle channel in the morphological interference section of the mismatch degree vector, and combines the weighting coefficient with the unit weights belonging to the stable contribution section to generate a global diagonal weight matrix. It then establishes a global weighted inversion equation system with a penalty matrix in combination with the high-risk mapping area and solves the comprehensive particle size distribution based on the basic particle size distribution.

[0059] The parameter rating integration module calculates the residual values ​​of the angular channels belonging to the stable contribution segment based on the optical transformation matrix of the spherical particles, the comprehensive particle size distribution, the full-angle scattered light energy distribution data, and the binary mask vector to obtain the first local residual. When the first local residual is determined to be lower than the preset convergence threshold, the module generates key particle size feature parameters and confidence index by combining the global diagonal weight matrix and mapping trend information. The module calculates the morphological deviation index using the weighting coefficient and outputs the corrected particle size distribution result by combining the key particle size feature parameters and confidence index.

[0060] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0061] In this invention, by extracting the first- and second-order derivative feature sequences, splicing them into a multi-dimensional feature matrix, and linking the mismatch degree vector to calculate the comprehensive evaluation value, the scattering data channels that are abnormally altered due to the influence of non-spherical morphology are accurately located and isolated, blocking the transformation path from morphological optical differences to incorrect particle size components. The mapping trend is extracted based on the binary masking mechanism to construct a reduced-dimensional inversion equation set to obtain the basic distribution. Simultaneously, weighting coefficients and penalty matrices are introduced to establish a global weighted inversion solution process. Targeted weighted compensation and strict constraints are implemented on the interfered intervals to completely eliminate the false broadening and false multi-peak phenomenon caused by misjudging abnormal scattering as additional components. Combined with local residual evaluation, the true size characteristics of the sample are restored. Attached Figure Description

[0062] Figure 1 This is the main flowchart for particle size distribution testing in this invention;

[0063] Figure 2 This is a flowchart of the multidimensional feature matrix generation process of the present invention;

[0064] Figure 3 This is a flowchart illustrating the channel segmentation and risk mapping process of the present invention.

[0065] Figure 4 This is a flowchart of the weighted inversion and result output of the present invention. Detailed Implementation

[0066] To make the objectives, technical solutions, and advantages of this invention clearer, the 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 merely illustrative and not intended to limit the invention.

[0067] Please see Figures 1 to 4 This embodiment provides a method and system for testing particle size distribution using a laser particle size analyzer. In practical applications, such as during the online particle size distribution testing of particle samples using a laser particle size analyzer with wet circulating dispersion, the particle dispersion liquid in the sample cell enters the test optical path through the circulation pipeline. The background monitoring channel records the light intensity baseline corresponding to the pure dispersion medium. The detector array receives the scattered light energy at different angles. The particle size inversion processing program continuously processes the light intensity sequence, background sequence, optical transformation matrix, and channel status identifier to form a closed-loop data link from scattering signal acquisition to the output of corrected particle size distribution results. This includes a method for testing particle size distribution using a laser particle size analyzer, which includes the following steps:

[0068] S1: The deviation between the total light intensity sequence of the particle sample and the background light intensity sequence of the particle sample is calculated to obtain the full-angle scattered light energy distribution data. First-order and second-order derivative feature sequences are extracted from the full-angle scattered light energy distribution data. These three sequences are then concatenated to generate a multi-dimensional feature matrix. The total light intensity sequence of the particle sample is the light energy record formed by the detector array according to the angular channel order after the sample dispersion enters the test optical path. The background light intensity sequence of the particle sample is the baseline record formed by the same optical path under cyclic conditions in a pure dispersion medium. The full-angle scattered light energy distribution data consists of angular channel light energy fields saved according to the spatial arrangement order of the detectors after removing background influence. The first-order derivative feature sequence represents the light energy change trend field between adjacent angular channels, and the second-order derivative feature sequence represents the bending change field of the light energy change trend in the angular direction. The multi-dimensional feature matrix is ​​a data structure formed by column-wise concatenation of the light energy field, change trend field, and bending change field according to the same angular channel index, for subsequent channel state evaluation.

[0069] S111: Acquire the total light intensity sequence of the particle sample within the set test time window, monitor the background light intensity sequence of the particle sample under pure dispersion medium circulation, calculate the light energy intensity deviation between the total light intensity sequence and the background light intensity sequence, and obtain the full-angle scattered light energy distribution data. The test time window is given by the instrument test task configuration file and is synchronized with the sample pumping, light source illumination, detector sampling, and data buffer status. After the total light intensity sequence enters the original light intensity buffer, the channel integrity is first checked according to the detector channel identifier. The check includes whether the channel identifier is continuous, whether the acquisition status is valid, whether there are duplicate records in the same channel, whether the light intensity field conforms to the detector range and aperture, and whether the recording order is consistent with the array physical order. After the background light intensity sequence enters the background baseline buffer, it is aligned with the total light intensity sequence according to the same channel identifier. When a channel is missing, the background baseline status that has been verified within the same test task is used as a substitute input, and a background completion identifier is written to the channel status log. When the background baseline fails, the derivative extraction stage is stopped, and the background reacquisition status is output to avoid receiving misaligned angle data in subsequent steps.

[0070] The processing of light intensity deviation targets the total light intensity field and background light intensity field on the same detector channel. The processing order is as follows: first, channel alignment is completed, then background subtraction is performed, and finally, the physical arrangement order of the subtracted scattered light energy records is rearranged. The resulting full-angle scattered light energy distribution data is stored in the angular light energy data area, which stores the channel index, angular position identifier, background subtraction status, light energy validity identifier, and scattered light energy field. The validity identifier is jointly determined by the acquisition integrity, background availability, channel alignment results, and detector status. Abnormal channels do not participate in derivative feature generation and are recorded as pending compensation states by the trend of adjacent valid channels for reference in subsequent channel division.

[0071] S112: Map the full-angle scattered light energy distribution data to a logarithmic angular coordinate system, perform cubic spline interpolation to fit the full-angle scattered light energy distribution data, establish a continuous smooth energy curve, calculate the rate of change of light energy between intervals formed by adjacent standardized angle sampling points on the continuous smooth energy curve, obtain the first derivative feature sequence, solve the second-order rate of change feature of the first derivative feature sequence in the angular dimension, perform an adaptive sliding window filtering algorithm to smooth the second-order rate of change feature, and obtain the second-order derivative feature sequence. The logarithmic angular coordinate system is the angular expression space after the detector angular position is transformed according to the instrument angle calibration table. This space is used to allow the variation trends of small-angle and large-angle regions to be compared under the same standard. The standardized angle sampling points are jointly generated by the angle calibration table and the effective channel arrangement, carrying the original channel index, standardized angle position, and interpolation source identifier.

[0072] The cubic spline interpolation algorithm reads the scattered light energy field and standardized angle position corresponding to the effective channel, and establishes a continuous smooth energy curve between adjacent effective channels. If an isolated invalid channel exists, an interpolation marker is generated based on the curve trend of the effective channels on both sides, but this marker is not written back to the original light intensity buffer as the original acquired value. The first-order derivative feature sequence is formed by the direction and intensity of change of the continuous smooth energy curve in adjacent standardized angle intervals, and is saved as a trend field aligned with the detector channel index. The second-order rate of change feature is used to record the bending and turning state of the first-order derivative feature in the angular direction. The adaptive sliding window filtering algorithm adjusts the smoothing range according to the channel effectiveness, local noise state, and trend consistency of adjacent channels. The smoothed second-order derivative feature sequence enters the multi-dimensional feature matrix generation stage. If there are insufficient continuous effective channels during the filtering process, the smoothing state of the previous effective window is used and written into the window inheritance identifier. After the number of subsequent channels is restored, the current window state is switched.

[0073] S113: The full-angle scattered light energy distribution data, first-order derivative feature sequences, and second-order derivative feature sequences are stitched together to generate a multi-dimensional feature matrix. The stitching process uses the detector channel index as the unique alignment key, writing the scattered light energy field, light energy change trend field, and curvature change field corresponding to the same channel into the same row. The column fields of the multi-dimensional feature matrix sequentially carry the original scattered energy, first-order angular direction change features, second-order angular direction change features, channel validity identifier, and interpolation source identifier. After the matrix is ​​generated, it is stored in the feature cache and a matrix version identifier is output to the subsequent channel state evaluation process to ensure that the mismatch vector, curvature score, and comprehensive evaluation value call for the same batch of feature data.

[0074] S2: Based on the full-angle scattered light energy distribution data and the built-in spherical particle optical transformation matrix, a theoretical scattered light energy distribution sequence and a mismatch vector are generated. The comprehensive evaluation value is calculated using the mismatch vector and the second derivative feature sequence within the multi-dimensional feature matrix. Based on the comprehensive evaluation value, the detector array is divided into morphological interference segments and stable contribution segments, and a binary mask vector is generated. The spherical particle optical transformation matrix is ​​a data structure representing the conversion of standard spherical particle size components to the scattering response of the detector's angular channels, stored in the instrument's optical model library. The row direction is aligned with the detector's angular channels, and the column direction is aligned with the particle size component identifier. The theoretical scattered light energy distribution sequence is the angular channel light energy prediction field corresponding to the equivalent spherical diameter distribution under this transformation matrix. The mismatch vector is the deviation state field between the measured scattered light energy and the theoretical scattered light energy on each detector channel. The comprehensive evaluation value is the channel anomaly evaluation field formed by combining the mismatch state and the second derivative bending state.

[0075] S211: Obtain the built-in spherical particle optical transformation matrix, execute a non-negative least squares algorithm to fit the mapping relationship between the full-angle scattered light energy distribution data and the spherical particle optical transformation matrix to solve for the equivalent spherical diameter distribution result. Calculate the particle group scattering state based on the spherical particle optical transformation matrix and the equivalent spherical diameter distribution result to obtain the theoretical scattered light energy distribution sequence. The optical transformation matrix is ​​read from the instrument's optical model library and matched using the optical path configuration identifier, laser wavelength configuration identifier, detector array version identifier, and dispersion medium optical property identifier. If matching fails, the model configuration is entered into an abnormal state, and the output of the equivalent spherical diameter distribution result is stopped. The equivalent spherical diameter distribution result is a particle size component ratio field formed based on the assumption of spherical particles. This field is not used as the final correction result output, but rather as a data source for identifying non-spherical morphological interference and establishing a theoretical scattering benchmark.

[0076] In this step, the nonnegative least squares algorithm iteratively updates the particle size component proportion field according to nonnegativity constraints and residual convergence rules, making the scattering state after matrix mapping a comparable theoretical benchmark with the full-angle scattered light energy distribution data. The iterative input includes the full-angle scattered light energy field, channel validity identifiers, and optical transformation matrix. The output includes the equivalent spherical diameter distribution result, the theoretical scattered light energy distribution sequence, and the fitting status identifier. The fitting status identifier records whether the solution is complete, whether boundary constraints are triggered, and whether there are matrix column mismatch anomalies. If the fitting status is abnormal, the verified full-angle scattered light energy distribution data is retained, and the model mismatch status is output to the manual review interface.

[0077] S212: Calculate the absolute value of the deviation between the full-angle scattered light energy distribution data and the theoretical scattered light energy distribution sequence on each detector channel corresponding to the full-angle scattered light energy distribution data, obtaining the mismatch vector. Transform the element values ​​contained in the mismatch vector to obtain the mismatch score. Extract the second derivative feature sequence contained in the multi-dimensional feature matrix, calculate the curvature score corresponding to the second derivative feature sequence, and calculate the weighted sum between the mismatch score and the curvature score to obtain the comprehensive evaluation value. The elements of the mismatch vector come from the difference state between the measured scattered light energy field and the theoretical scattered light energy field under the same channel. This difference state is written into the mismatch cache after being filtered by the channel's valid identifier. The mismatch score is a dimensionless evaluation field after uniformly converting the mismatch state of each channel. The conversion caliber is given by the mismatch level rules in the instrument calibration library. The rule sources include blank background records, standard spherical particle records, and detector stability records.

[0078] The curvature score, derived from the second-order derivative feature sequence, describes the abrupt bending of the scattered energy curve in a local angular channel. During the conversion, the channel index corresponding to the second-order derivative feature is read first, and then the bending change field is mapped to the same evaluation caliber according to the curvature level rules. When the second-order derivative feature originates from a completion or inheritance window, the curvature score is appended with a source status identifier, and its trigger priority is reduced during the comprehensive evaluation phase. The comprehensive evaluation value is formed by the mismatch score and curvature score according to the weight field in the rule table. The weight field comes from the device calibration configuration, is updated with the optical path configuration and sample dispersion medium type, is stored in the evaluation rule table, and is called during the channel partitioning process. The evaluation process first processes the mismatch level, then the curvature level, then merges the effective channel states, and outputs a comprehensive evaluation record consistent with the spatial order of the detector array.

[0079] S213: Extract the channel peak points exceeding the preset confidence threshold from the comprehensive evaluation values ​​as starting seed points. Execute a bidirectional region growing algorithm to divide the detector array corresponding to each detector channel into morphological interference segments and stable contribution segments based on the starting seed points. According to the distribution channel positions of the morphological interference segments and stable contribution segments, assign interference state identifiers and stable state identifiers to the detector array respectively, generating a binary mask vector. The preset confidence threshold is a configurable field in the channel state evaluation rule table. This field is formed by the set of blank background evaluation values ​​for each detector channel under pure dispersed medium cyclic conditions. The set of blank background evaluation values ​​records the background fluctuation evaluation state when there is no particle scattering. When generating the threshold, first aggregate the blank background evaluation fields of each channel, then form the global background level and background discrete state. Subsequently, merge the two types of background states into a confidence judgment boundary and write it into the threshold rule table. After background record updates or detector maintenance, the threshold field is refreshed with the instrument calibration process.

[0080] The channel peak point screening process reads the comprehensive evaluation value and a preset confidence threshold. First, it filters candidate abnormal channels that exceed the threshold field's boundary. Then, it compares the evaluation status of adjacent channels within the candidate abnormal channels, extracting channels that are in a local peak state relative to their two adjacent channels as starting seed points. The bidirectional region growing algorithm expands from the starting seed point along the detector's physical arrangement direction to both sides. The expansion conditions read the comprehensive evaluation status of adjacent channels, the continuity of curvature changes, the continuity of mismatch states, and the channel's valid identifier. Region growing stops when adjacent channels no longer meet the same segment assignment rules. Morphological interference segments are used to identify the set of angular channels affected by particle morphological deviations, while stable contribution segments are used to identify the set of angular channels that still have a stable inversion contribution under the spherical particle optical transformation matrix. A binary mask vector stores the segment assignment identifier for each detector channel, and the identifier dictionary, channel index, and region boundary are written to the channel state cache for continuous calls during dimensionality reduction inversion, global weighted inversion, and local residual screening.

[0081] S3: Based on the difference between the full-angle scattered light energy distribution data and the theoretical scattered light energy distribution sequence in the morphological interference section, high-risk mapping areas are extracted to generate mapping trend information. The basic particle size distribution is obtained by establishing and solving a reduced-dimensional inversion equation set based on the binary mask vector. The high-risk mapping area is the set of particle size components in the optical transformation matrix of spherical particles that have a consistent change relationship with the morphological interference residual. The mapping trend information is the record of the change direction between the target column vector and the adjacent column vectors at the boundary of this set. The reduced-dimensional inversion equation set is the inversion data structure composed of the light energy field of the stable contribution section and the corresponding row field of the transformation matrix after removing the angular channels of the morphological interference section. The basic particle size distribution is the initial particle size component proportion field formed under the constraint of the stable contribution section.

[0082] S311: Calculate the energy difference between the full-angle scattered light energy distribution data and the theoretical scattered light energy distribution sequence within the morphological interference section to obtain the residual energy vector. Execute the Pearson correlation coefficient algorithm to calculate the linear correlation between the residual energy vector and the column vectors in the spherical particle optical transformation matrix. Compare the linear correlation with a correlation warning threshold. Based on the correlation warning threshold, select high-risk mapping areas from the particle size components corresponding to the spherical particle optical transformation matrix. Extract the component mapping trend data composed of the difference values ​​between the target column vector and adjacent column vectors within the high-risk mapping area to generate mapping trend information. The input channel of the residual energy vector comes from the morphological interference section identified by the binary mask vector. The vector content is the difference field between the measured scattered light energy and the theoretical scattered light energy in this section. In this step, the Pearson correlation coefficient algorithm is used to identify the degree of linear consistency between the residual change morphology and the particle size component scattering response column. The input objects are the residual energy vector and the particle size component column vectors of the optical transformation matrix, and the output object is the linear correlation field corresponding to the particle size component identifier.

[0083] The correlation warning threshold is formed by the baseline residual energy sequence of a standard spherical particle sample and a preset anti-misjudgment scaling factor. The baseline residual energy sequence of the standard spherical particle sample is stored in the calibration sample library and is used to record the residual forms allowed by the instrument in the spherical particle state. The preset anti-misjudgment scaling factor is a configurable field in the calibration rule table, derived from the instrument's factory calibration and subsequent maintenance calibration. When screening high-risk mapping areas, the baseline correlation state between the baseline residual energy sequence and the column vector of the optical transformation matrix is ​​read first, and then the correlation warning threshold field is generated. Subsequently, the current linear correlation field is compared with the threshold field according to rules. Particle size components exceeding the limit boundary of the threshold field are recorded as target particle size components and classified into the high-risk mapping area. The component mapping trend data consists of the difference state between the target column vector and the adjacent column vector with the immediate boundary in the high-risk mapping area. The difference state stores the direction of change, boundary side, particle size component identifier, and source column identifier. The mapping trend information is written to the risk mapping cache and called by the penalty matrix configuration and confidence index generation process.

[0084] S312: Based on the spatial identifier bits contained in the binary mask vector, remove the angular dimension data belonging to the morphological interference segment in the full-angle scattered light energy distribution data and the optical transformation matrix of the spherical particles, and establish a reduced-dimensional inversion equation set. The spatial identifier bits are the segment attribution fields in the binary mask vector that correspond one-to-one with the physical channels of the detector, carrying the interference state identifier, the stable state identifier, and the channel index. The removal process does not change the particle size distribution column structure, only removing the row fields belonging to the morphological interference segment in the angular dimension, while simultaneously retaining the scattered light energy field, the channel validity identifier, and the transformation matrix row fields corresponding to the stable contribution segment. If the channel index of the mask vector is inconsistent with that of the full-angle scattered light energy distribution data, an alignment check is first performed based on the channel index dictionary. If the alignment fails, the mask mismatch status is output and the basic particle size distribution solution is stopped.

[0085] The input to the reduced-dimensional inversion equations comes from the light energy field of the stable contribution segment and the transformation matrix field of the same segment. The constraint information comes from non-negativity constraints, smoothing constraints, and channel effective states. This equation set is stored in the inversion task cache as a data object for solving the basic particle size distribution. The cache also stores the list of eliminated channels and the list of stable channels for subsequent global weighted inversion to restore the full-angle weights.

[0086] S313: For the reduced-dimensional inversion equations, a non-negative least squares iterative algorithm with smoothing constraints is executed to solve for the residual minimum of the reduced-dimensional inversion equations, generating the basic particle size distribution. The non-negative least squares iterative algorithm with smoothing constraints reads the inversion task cache formed by the stable contribution segment, restricts the particle size component proportion field to a non-negative state, and suppresses isolated abrupt changes in the particle size distribution where there is no corresponding stable scattering evidence by the continuous change rule of adjacent particle size components. The smoothing constraint field comes from the inversion rule table, which is jointly indexed by the instrument optical model configuration and the sample dispersion state configuration. It does not write specific sample results, but only serves as the iterative boundary and result verification basis.

[0087] During the iterative solution process, the current particle size component proportion field generates a predicted scattering state through the stable segment row field of the optical transformation matrix. The predicted scattering state is then compared with the measured scattered light energy field of the stable contribution segment using residuals, and the residual state is fed back to the update direction of the particle size component proportion field. After the solution is completed, the basic particle size distribution, the basic inversion residual state, and the smoothing constraint trigger flag are output. If a non-converged state, insufficient effective stable channels, or abnormal matrix conditions occur during the iteration process, the basic particle size distribution is marked as pending verification, and the input snapshot of the reduced-dimensional inversion task is written to the traceability log. Subsequent global weighted inversions will only receive this result after the pending verification state is lifted.

[0088] S4: Calculate the weighting coefficients using the corresponding numerical values ​​of the angle channels within the morphological interference section using the mismatch degree vector. Combine these weighting coefficients with the unit weights belonging to the stable contribution section to generate a global diagonal weight matrix. Establish a global weighted inversion equation system containing a penalty matrix in conjunction with the high-risk mapping region, and solve for the comprehensive particle size distribution based on the basic particle size distribution. The weighting coefficients are weight fields used to suppress the contribution of the angle channels in the morphological interference section, derived from the mismatch degree vector and the nonlinear mapping rule table. The unit weights are the baseline weight fields used to maintain the conventional inversion contribution of the channels in the stable contribution section. The global diagonal weight matrix is ​​a channel weight data structure arranged in the order of the detector channels. The penalty matrix is ​​a constraint data structure configured for the particle size components in the high-risk mapping region. The comprehensive particle size distribution is a corrected particle size component proportion field formed under the combined effect of full-angle data, channel weights, and risk mapping constraints.

[0089] S411: The mismatch vector and morphological interference segment are retrieved. A nonlinear mapping algorithm is executed to calculate the weighting coefficients of the angle channels within the morphological interference segment. These weighting coefficients are combined with the unit weights assigned to the stable contribution segment to generate a global diagonal weight matrix. In this step, the nonlinear mapping algorithm refers to the process of reading the mismatch level rule table and mapping the channel mismatch state to the channel weight field. The rule table is derived from standard spherical particle records, non-spherical particle perturbation calibration records, and detector stability records. The more significantly the mismatch state deviates from the background baseline, the more the weight field of the corresponding channel is biased towards suppressing morphological interference contributions; when the mismatch state is close to the background baseline, the corresponding channel retains its contribution state for global inversion. The mapping process first reads the channel index of the morphological interference segment, then queries the mismatch level channel by channel, and finally outputs the weighting coefficients and writes them to the weight cache.

[0090] The channel weight field for stable contribution segments is read from the inversion rule table. This field is used to maintain the regular contribution of stable channels to particle size inversion. The global diagonal weight matrix combines the weighting coefficients of morphological interference segments and the unit weights of stable contribution segments according to the physical arrangement of the detectors. The matrix row and column indices are consistent with the angle channel indices. If a channel weight is missing, the weight status of adjacent valid channels within the same segment is read as a temporary weight source and a weight completion flag is written. The weight completion flag is read in the confidence index generation stage of the comprehensive particle size distribution and is used to reduce the confidence level of the corresponding particle size component.

[0091] S412: Based on the global diagonal weight matrix, a globally weighted inversion equation system is established by weighting the full-angle scattered light energy distribution data and the optical transformation matrix of spherical particles. High-risk mapping regions contained in the mapping trend information are extracted. Based on these high-risk mapping regions, matrix elements of the corresponding particle size interval penalty terms are configured to obtain the penalty matrix. The inputs to the globally weighted inversion equation system include the full-angle scattered light energy field, the spherical particle optical transformation matrix, the global diagonal weight matrix, and the initial state of the basic particle size distribution. Weighting is applied by applying channel weights to the measured scattered light energy field and the corresponding row field of the transformation matrix according to the angle channel index. This ensures that morphological interference segments are included in the global solution but are subject to weight constraints, while stable contribution segments maintain the normal constraint effect.

[0092] The penalty matrix is ​​configured based on high-risk mapping regions and component mapping trend data from the mapping trend information. The particle size intervals corresponding to high-risk mapping regions are written into the penalty configuration area, which records the particle size component identifier, risk source channel, mapping trend direction, and the status of adjacent components at the boundary. The matrix elements of the penalty term are configured according to the trend consistency between risk components and adjacent components. The trend indicates that when morphological interference residuals are concentrated and mapped to a certain particle size component, the change in the proportion of that particle size component in the global solution is constrained, preventing the residuals in morphological interference sections from being directly interpreted as true particle size peaks. If the high-risk mapping region is empty, the penalty matrix maintains a normal smoothing constraint state, and the risk-free mapping state is recorded in the inversion task cache for use in the confidence index generation stage.

[0093] S413: The basic particle size distribution is set as the initial iteration parameter. Based on the globally weighted inversion equations and the penalty matrix, the residual minimum of the globally weighted inversion equations, including the penalty matrix, is solved to obtain the comprehensive particle size distribution. The initial iteration parameter comes from the basic particle size distribution output by S313. This parameter carries the particle size component proportion field, the basic inversion residual state, and the smoothing constraint trigger flag. The global solution process starts from the basic particle size distribution, reads the scattered light energy field and the transformation matrix field after full-angle weighting, and reads the constraint state of the penalty matrix on the high-risk mapping region. Then, the particle size component proportion field is updated under the joint constraints of non-negativity constraints, smoothing constraints, and risk constraints.

[0094] Each updated particle size component proportion field is converted into a predicted scattering state and compared with the residual of the full-angle weighted scattered light energy field. The residual state is fed back to the next round of proportion field adjustment. The residual information of stable contribution segments maintains a dominant verification role, while the residual information of morphological interference segments participates in compensation through weighting coefficients. The penalty matrix limits the generation of sudden increases in high-risk mapping areas that are inconsistent with the mapping trend. The output comprehensive particle size distribution saves the particle size component proportion field, channel weight version, penalty matrix version, and global inversion state. If the basic particle size distribution is in a pending verification state, the weight cache is invalid, or the high-risk mapping area index is inconsistent with the particle size component index, the global weighted inversion task enters an abnormal suspension state, and the abnormality type is fed back to the data acquisition and model configuration check interface.

[0095] S5: Based on the optical transformation matrix of spherical particles, the comprehensive particle size distribution, the full-angle scattered light energy distribution data, and the binary mask vector, the residual values ​​of the angular channels belonging to the stable contribution segment are calculated to obtain the first local residual. If the first local residual is determined to be lower than the preset convergence threshold, key particle size feature parameters and a confidence index are generated by combining the global diagonal weight matrix and mapping trend information. The morphological deviation index is calculated using the weighting coefficient, and the corrected particle size distribution result is output by combining the key particle size feature parameters and the confidence index. The first local residual is the deviation field between the measured scattering state and the theoretical scattering state corresponding to the comprehensive particle size distribution within the stable contribution segment. The preset convergence threshold is the convergence judgment field in the inversion rule table, derived from instrument calibration records, standard spherical particle test records, and optical path stability maintenance records. The key particle size feature parameters are particle size feature fields extracted from the comprehensive particle size distribution. The confidence index is a reliable state field formed by combining channel weights and risk mapping states. The morphological deviation index is a morphological deviation rating field formed based on the mismatch ratio state and the weighting coefficient state.

[0096] S511: The spherical particle optical transformation matrix and the overall particle size distribution are used to calculate the updated theoretical scattering distribution. The energy difference between the updated theoretical scattering distribution and the full-angle scattered light energy distribution data at each detector channel is calculated to obtain the full-angle residual vector. Based on the binary mask vector, the residual values ​​belonging to the angle channels within the stable contribution segment are selected from the full-angle residual vector to obtain the first local residual. The updated theoretical scattering distribution is formed by matrix mapping of the spherical particle optical transformation matrix and the overall particle size distribution, and is converted into a data structure with the same direction as the full-angle scattered light energy distribution data according to the physical arrangement order of the detector channels. The matrix mapping process reads the particle size component proportion field from the overall particle size distribution and the particle size component column field from the optical transformation matrix, and outputs the theoretical light energy field corresponding to each angle channel; the fields are transposed and saved as the updated theoretical scattering distribution with an inversion version identifier.

[0097] The full-angle residual vector is composed of the absolute state of the difference between the updated theoretical scattering distribution and the full-angle scattered light energy distribution data on the same detector channel, and is spliced ​​according to the spatial physical arrangement order of the detector. When filtering the first local residual, the target element consistent with the stable contribution segment in the binary mask vector is read first, and then the spatial index identifier corresponding to the target element is obtained. Subsequently, the spatial index identifier is aligned and mapped to the full-angle residual vector, the residual value at the corresponding position is extracted and aggregated into the first local residual. If the stable identifier of the binary mask vector is inconsistent with the channel state dictionary, the segment attribution dictionary saved in S213 is read first for identifier conversion; if there is a missing channel in the residual vector, the channel enters the residual missing state and does not participate in the convergence determination, while the channel index and the source of the missing information are retained in the result traceability log.

[0098] S512: Based on the first local residual, when the first local residual is determined to be lower than the preset convergence threshold, the particle volume ratio field corresponding to each particle size channel in the comprehensive particle size distribution is extracted. The sum of the distribution interval integrals of the particle volume ratio field is calculated to obtain the key particle size feature parameters. The weight coefficients contained in the global diagonal weight matrix are extracted, and the risk warning indicators contained in the mapping trend information are extracted. Based on the weight coefficients and risk warning indicators, the confidence level value of the particle size component where the key particle size feature parameters are located is calculated to obtain the confidence index. Before the first local residual enters the convergence determination, it is compared with the preset convergence threshold field under the same rule table caliber. The determination result is divided into convergence state and non-convergence state. The convergence state triggers the generation of key particle size feature parameters, and the non-convergence state triggers the global weighted inversion parameter review, and the residual source channel is fed back to the weight cache and high-risk mapping cache.

[0099] Key granularity feature parameters are formed by the particle size channel, particle size interval identifier, and particle volume percentage field in the comprehensive particle size distribution. In engineering implementation, the distribution interval integral sum is represented by accumulating and truncating the volume percentage field along the particle size intervals, outputting the particle size feature field, interval source field, and percentage source field. The confidence index input includes the channel weight field in the global diagonal weight matrix, the risk warning indicator in the mapping trend information, the weight completion identifier, and the high-risk mapping area attribution identifier. The risk warning indicator is a status field composed of the high-risk mapping area, mapping trend direction, and risk source channel; when the particle size component corresponding to the key granularity feature parameter is within the risk mapping influence range, its confidence level is adjusted according to the risk warning indicator and the relevant channel weight status, and written to the result rating cache. After output, the confidence index is bound to the key granularity feature parameter for display, storage, and traceability of the corrected granularity distribution results.

[0100] S513: Based on the confidence index, calculate the integral proportion state of the mismatch vector across the entire angular range. Calculate the weighted score between the integral proportion state and the weighting coefficient to obtain the morphological deviation index. Summarize and concatenate the confidence index, morphological deviation index, and key granular feature parameters to generate a corrected granular distribution result. The integral proportion state is formed by the cumulative mismatch distribution of the mismatch vector across the entire angular range and the channel weight state, used to describe the coverage degree of morphological interference on the detector array. The weighting coefficient comes from the weight cache formed in S411. The weighted score is generated by the integral proportion state, weighting coefficient state, and channel attribution identifier according to the morphological deviation rating rule table, outputting the morphological deviation index. The morphological deviation index is saved as a level state field and associated with the global diagonal weight matrix version and the high-risk mapping area version.

[0101] The corrected particle size distribution result consists of the comprehensive particle size distribution, key particle size feature parameters, confidence index, morphological deviation index, channel status identifier, inversion status identifier, and anomaly traceability identifier. Before output, a field consistency check is performed, including whether the particle size component index is continuous, whether the channel weight version is consistent with the inversion version, whether the key particle size feature parameters originate from the convergence state, whether the confidence index is bound to the corresponding particle size component, and whether the morphological deviation index reads a valid reduction coefficient. If the check passes, the corrected particle size distribution result is entered into the instrument result cache, test report interface, and historical traceability log; if the check fails, the output result is placed in a pending review status, and version identifiers for the full-angle scattered light energy distribution data, theoretical scattered light energy distribution sequence, binary mask vector, global diagonal weight matrix, and mapping trend information are retained for re-inversion within the same test task or for manual review.

[0102] The laser particle size analyzer particle distribution testing system, implemented in conjunction with the aforementioned method, includes a scattering signal analysis module, a channel state evaluation module, an initial inversion solution module, a weighted compensation iteration module, and a parameter rating integration module. The scattering signal analysis module receives light intensity records from the detector array and background monitoring channel, performs background subtraction, angular coordinate mapping, derivative feature extraction, and multidimensional feature matrix generation, and transmits the full-angle scattered light energy distribution data and multidimensional feature matrix to the channel state evaluation module. The channel state evaluation module reads the optical transformation matrix of the spherical particles, generates a theoretical scattered light energy distribution sequence, a mismatch vector, a comprehensive evaluation value, and a binary mask vector, and writes the morphological interference segment, the stable contribution segment, and the channel state dictionary into the channel state cache.

[0103] The initial inversion solution module receives full-angle scattered light energy distribution data, theoretical scattered light energy distribution sequence, binary mask vector, and spherical particle optical transformation matrix. It extracts high-risk mapping regions and mapping trend information, establishes a reduced-dimensional inversion equation set, and outputs the basic particle size distribution. The weighted compensation iteration module reads the mismatch vector, morphological interference segment, basic particle size distribution, and mapping trend information, generates a global diagonal weight matrix and penalty matrix, establishes a global weighted inversion equation set, and outputs the comprehensive particle size distribution. The parameter rating integration module reads the comprehensive particle size distribution, full-angle scattered light energy distribution data, binary mask vector, global diagonal weight matrix, and mapping trend information. It forms the first local residual, key particle size feature parameters, confidence index, morphological deviation index, and corrected particle size distribution results, and writes the results to the display interface, report generation interface, and traceability log interface. Data alignment between modules is achieved through test task identifiers, channel indexes, matrix version identifiers, and result status identifiers. When any module outputs an abnormal status, subsequent modules read the abnormal status and stop using the corresponding invalid fields. The system retains the data cache and processing version that have been verified, maintaining the data traceability link within the same test task.

[0104] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for testing particle size distribution using a laser particle size analyzer, characterized in that, Includes the following steps: S1: Calculate the deviation between the total light intensity sequence of the particle sample and the background light intensity sequence of the particle sample to obtain the full-angle scattered light energy distribution data. Extract the first-order derivative feature sequence and the second-order derivative feature sequence from the full-angle scattered light energy distribution data. Then, concatenate the full-angle scattered light energy distribution data, the first-order derivative feature sequence, and the second-order derivative feature sequence to generate a multi-dimensional feature matrix. S2: Generate a theoretical scattered light energy distribution sequence and a mismatch vector based on the full-angle scattered light energy distribution data and the built-in spherical particle optical conversion matrix. Calculate a comprehensive evaluation value using the mismatch vector and the second derivative feature sequence in the multidimensional feature matrix. Based on the comprehensive evaluation value, divide the detector array into a morphological interference segment and a stable contribution segment, and generate a binary mask vector. S3: Based on the difference between the full-angle scattered light energy distribution data and the theoretical scattered light energy distribution sequence in the morphological interference section, high-risk mapping areas are extracted to generate mapping trend information. The basic particle size distribution is obtained by establishing and solving the reduced-dimensional inversion equation set according to the binary mask vector. S4: Calculate the weighting coefficient using the corresponding numerical value of the angle channel in the morphological interference section using the mismatch degree vector, and combine the weighting coefficient with the unit weight belonging to the stable contribution section to generate a global diagonal weight matrix. Combine the high-risk mapping area to establish a global weighted inversion equation system with a penalty matrix and solve it based on the basic particle size distribution to obtain the comprehensive particle size distribution. S5: Calculate the residual values ​​of the angle channels belonging to the stable contribution segment based on the spherical particle optical transformation matrix, comprehensive particle size distribution, full-angle scattered light energy distribution data and binary mask vector to obtain the first local residual. If the first local residual is determined to be lower than the preset convergence threshold, generate key particle size feature parameters and confidence index by combining the global diagonal weight matrix and mapping trend information. Calculate the morphological deviation index using the weighting coefficient and output the corrected particle size distribution result by combining the key particle size feature parameters and confidence index. The specific steps of S2 are as follows: S211: Obtain the built-in spherical particle optical transformation matrix, execute the non-negative least squares algorithm to fit the mapping relationship between the full-angle scattered light energy distribution data and the spherical particle optical transformation matrix to solve for the equivalent spherical diameter distribution result, calculate the particle group scattering state based on the spherical particle optical transformation matrix and the equivalent spherical diameter distribution result, and obtain the theoretical scattered light energy distribution sequence; S212: Calculate the absolute value of the deviation between the full-angle scattered light energy distribution data and the theoretical scattered light energy distribution sequence on each detector channel corresponding to the full-angle scattered light energy distribution data to obtain the mismatch vector. Convert the element values ​​contained in the mismatch vector to obtain the mismatch score. Extract the second derivative feature sequence contained in the multi-dimensional feature matrix. Calculate the curvature score corresponding to the second derivative feature sequence. Calculate the weighted sum between the mismatch score and the curvature score to obtain the comprehensive evaluation value. S213: Extract the channel peak points that exceed the preset confidence threshold from the comprehensive evaluation values ​​as starting seed points, execute the bidirectional region growing algorithm to divide the detector array corresponding to each detector channel into morphological interference segment and stable contribution segment according to the starting seed points, and assign the values ​​one and zero to the detector array according to the distribution channel positions of the morphological interference segment and the stable contribution segment respectively to generate a binary mask vector. The specific steps of S3 are as follows: S311: Calculate the energy difference between the full-angle scattered light energy distribution data and the theoretical scattered light energy distribution sequence in the morphological interference section to obtain the residual energy vector. Execute the Pearson correlation coefficient algorithm to calculate the linear correlation between the residual energy vector and the column vectors in the spherical particle optical transformation matrix. Compare the linear correlation with the correlation warning threshold. Based on the correlation warning threshold, select high-risk mapping areas from the particle size components corresponding to the spherical particle optical transformation matrix. Extract the component mapping trend data composed of the difference values ​​between the target column vector and adjacent column vectors in the high-risk mapping area to generate mapping trend information. S312: Based on the spatial identifier bits contained in the binary mask vector, remove the full-angle scattered light energy distribution data and the angular dimension data belonging to the morphological interference segment in the optical transformation matrix of the spherical particle, and establish a set of reduced-dimensional inversion equations. S313: For the reduced dimension inversion equation set, execute a non-negative least squares iterative algorithm with smoothing constraints to solve the residual minimum value of the reduced dimension inversion equation set and generate the basic particle size distribution. The process of extracting the component mapping trend data composed of the difference values ​​between the target column vector and its adjacent column vectors within the high-risk mapping region is specifically as follows: Obtain the baseline residual energy sequence of a standard spherical particle sample; Calculate the baseline correlation between the baseline residual energy sequence and the column vectors in the optical transformation matrix of the spherical particle; The maximum value in the benchmark correlation is multiplied by the preset anti-false judgment ratio factor, and the output of the multiplication operation is configured as the correlation warning threshold. Compare the linear correlation with the correlation warning threshold, and extract the target linear correlation that is greater than the correlation warning threshold; Identify the target particle size components corresponding to the target linear correlation, and delineate the target particle size components as high-risk mapping regions; Read the target column vector in the high-risk mapping region from the optical transformation matrix of the spherical particle; Obtain the adjacent column vectors that are immediately adjacent to the boundary of the high-risk mapping region; Calculate the difference between the target column vector and the adjacent column vectors, and use the difference as the component mapping trend data of the high-risk mapping area; The specific steps of S4 are as follows: S411: Call the mismatch degree vector and the morphological interference segment, execute a nonlinear mapping algorithm to calculate the weighting coefficient of the angle channel of the mismatch degree vector in the morphological interference segment, combine the weighting coefficient and the unit weight belonging to the stable contribution segment to generate a global diagonal weight matrix; S412: Based on the global diagonal weight matrix, the full-angle scattered light energy distribution data and the spherical particle optical transformation matrix are weighted to establish a global weighted inversion equation set, the high-risk mapping region contained in the mapping trend information is extracted, and the matrix elements of the particle size interval penalty term corresponding to the high-risk mapping region are configured according to the high-risk mapping region to obtain the penalty matrix; S413: Set the basic particle size distribution as the initial iteration parameter, and based on the global weighted inversion equation set and the penalty matrix, solve the residual minimum value of the global weighted inversion equation set containing the penalty matrix to obtain the comprehensive particle size distribution; The specific steps of S5 are as follows: S511: Call the optical transformation matrix of the spherical particle and the comprehensive particle size distribution to calculate the updated theoretical scattering distribution, calculate the energy difference between the updated theoretical scattering distribution and the full-angle scattered light energy distribution data on each detector channel, obtain the full-angle residual vector, and filter the residual values ​​belonging to the angle channels in the stable contribution segment from the full-angle residual vector according to the binary mask vector to obtain the first local residual; S512: Based on the first local residual, when the first local residual is determined to be lower than the preset convergence threshold, extract the particle volume percentage corresponding to each particle size channel included in the comprehensive particle size distribution, calculate the sum of the distribution interval of the particle volume percentage, obtain the key particle size feature parameter, extract the weight coefficients included in the global diagonal weight matrix, and extract the risk warning index included in the mapping trend information. Based on the weight coefficients and the risk warning index, calculate the confidence level value of the particle size component where the key particle size feature parameter is located, and obtain the confidence index. S513: Based on the confidence index, calculate the percentage of the integral of the mismatch vector across the entire angular range, calculate the weighted score between the percentage of the integral and the weighting coefficient to obtain the morphological deviation index, and summarize and splice the confidence index, the morphological deviation index and the key granularity feature parameters to generate the corrected granularity distribution result.

2. The particle size distribution testing method of the laser particle size analyzer according to claim 1, characterized in that, The specific steps of S1 are as follows: S111: Collect the total light intensity sequence of the particle sample within the set test time window, monitor the background light intensity sequence of the particle sample under the pure dispersion medium circulation state, calculate the light energy intensity deviation between the total light intensity sequence and the background light intensity sequence, and obtain the full-angle scattered light energy distribution data. S112: Map the full-angle scattered light energy distribution data to a logarithmic angular coordinate system, perform a cubic spline interpolation algorithm to fit the full-angle scattered light energy distribution data, establish a continuous smooth energy curve, calculate the rate of change of light energy between the intervals formed by adjacent standardized angle sampling points on the continuous smooth energy curve, obtain the first derivative feature sequence, solve the second-order rate of change feature of the first derivative feature sequence in the angular dimension, perform an adaptive sliding window filtering algorithm to smooth the second-order rate of change feature, and obtain the second-order derivative feature sequence; S113: Combine the full-angle scattered light energy distribution data, the first derivative feature sequence, and the second derivative feature sequence to generate a multi-dimensional feature matrix.

3. The particle size distribution testing method of the laser particle size analyzer according to claim 1, characterized in that, The process of extracting channel peak points that exceed a preset confidence threshold from the comprehensive evaluation values ​​as starting seed points is as follows: Obtain the blank background evaluation value of each detector channel under the pure medium circulation state, and aggregate the blank background evaluation values ​​of each detector channel to construct a blank background evaluation value set. Calculate the global average value of the set of blank background evaluation values, and calculate the standard deviation of all elements in the set of blank background evaluation values ​​relative to the global average value; The global average value and the standard deviation value are summed, and the calculated value of the summation is configured as a preset reliability threshold. By comparing the comprehensive evaluation value of each detector channel with the preset confidence threshold, candidate abnormal channels that are greater than the preset confidence threshold are selected from the detector array. Extract the local maximum value channel within the candidate anomaly channel where the comprehensive evaluation value is greater than the comprehensive evaluation value corresponding to the channels of the two adjacent detectors, extract the local maximum value channel as the channel peak point, and use the channel peak point as the starting seed point.

4. The particle size distribution testing method of the laser particle size analyzer according to claim 1, characterized in that, The process of obtaining the first local residual is as follows: The optical transformation matrix of the spherical particle is multiplied with the comprehensive particle size distribution to calculate the theoretical light energy distribution column vector corresponding to each detector channel; The theoretical light energy distribution column vector is transformed into a row vector data structure by performing a matrix transpose operation, and the theoretical light energy distribution column vector transformed into a row vector structure is set as the updated theoretical scattering distribution; Extract the corresponding values ​​of the updated theoretical scattering distribution and the full-angle scattered light energy distribution data on the same detector channel, and calculate the absolute value of the difference between the corresponding value of the updated theoretical scattering distribution and the corresponding value of the full-angle scattered light energy distribution data. According to the spatial physical arrangement order of each detector channel, the absolute values ​​of the differences of each detector channel are spliced ​​in sequence to construct a full-angle residual vector; Extract the target element with a value equal to one from the binary mask vector, and obtain the spatial index identifier of the target element in the binary mask vector; Align and map the spatial index identifier to the full-angle residual vector, and extract the absolute value of the difference between the position corresponding to the spatial index identifier in the full-angle residual vector; The extracted absolute value of the difference is set as the residual value belonging to the angle channel within the stable contribution segment, and all residual values ​​belonging to the angle channel within the stable contribution segment are aggregated to obtain the first local residual.

5. A laser particle size analyzer particle size distribution testing system, characterized in that, The system is used to implement the laser particle size analyzer particle distribution testing method as described in any one of claims 1-4, the system comprising: The scattering signal analysis module calculates the deviation between the total light intensity sequence of the particle sample and the background light intensity sequence of the particle sample to obtain the full-angle scattered light energy distribution data. It extracts the first-order derivative feature sequence and the second-order derivative feature sequence from the full-angle scattered light energy distribution data, and concatenates the full-angle scattered light energy distribution data, the first-order derivative feature sequence and the second-order derivative feature sequence to generate a multi-dimensional feature matrix. The channel state assessment module generates a theoretical scattered light energy distribution sequence and a mismatch vector based on the full-angle scattered light energy distribution data and the built-in spherical particle optical conversion matrix. It calculates a comprehensive evaluation value using the mismatch vector and the second derivative feature sequence in the multi-dimensional feature matrix. Based on the comprehensive evaluation value, it divides the detector array into a morphological interference segment and a stable contribution segment and generates a binary mask vector. The initial inversion solution module extracts high-risk mapping areas and generates mapping trend information based on the difference between the full-angle scattered light energy distribution data and the theoretical scattered light energy distribution sequence in the morphological interference section. It also establishes and solves the reduced-dimensional inversion equation system based on the binary mask vector to obtain the basic particle size distribution. The weighted compensation iteration module calculates the weighting coefficient using the corresponding numerical values ​​of the angle channel in the morphological interference section of the mismatch degree vector, and combines the weighting coefficient with the unit weights belonging to the stable contribution section to generate a global diagonal weight matrix. It then establishes a global weighted inversion equation system with a penalty matrix in combination with the high-risk mapping area and solves the comprehensive particle size distribution based on the basic particle size distribution. The parameter rating integration module calculates the residual values ​​of the angular channels belonging to the stable contribution segment based on the optical transformation matrix of the spherical particles, the comprehensive particle size distribution, the full-angle scattered light energy distribution data, and the binary mask vector to obtain the first local residual. When the first local residual is determined to be lower than the preset convergence threshold, the module generates key particle size feature parameters and confidence index by combining the global diagonal weight matrix and mapping trend information. The module calculates the morphological deviation index using the weighting coefficient and outputs the corrected particle size distribution result by combining the key particle size feature parameters and confidence index.

Images