An inkjet printing wastewater treatment data online analysis method and system

Abstract

The present application relates to the field of wastewater treatment, and more particularly to a method and system for online analysis of inkjet printing wastewater treatment data, the method comprising: obtaining water quality, process and equipment state parameter data of inkjet printing wastewater based on each sensor, and constructing an original data matrix after preprocessing; decomposing the matrix, selecting a preset principal component number for reconstruction, and generating a reconstructed data matrix; calculating reconstruction error by analyzing the difference between the original and reconstructed matrices, analyzing the correlation distortion index based on row vector similarity, constructing a target function to determine the optimal principal component number; and using reconstruction error and distance for clustering, filtering abnormal data points and adjusting process parameters. The present application effectively solves the noise interference problem under complex water quality changes by optimizing the principal component number and improving the clustering analysis, significantly improves the abnormal detection accuracy and system stability, and enhances the robustness and adaptability of the inkjet printing wastewater treatment system.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of wastewater treatment. In particular, it relates to an online analysis method and system for inkjet printing wastewater treatment data. BACKGROUND

[0002] The inkjet printing wastewater contains high chroma dyes, organic auxiliaries and heavy metals, and the water quality and quantity fluctuate dramatically with order changes and cleaning frequencies, becoming a typical wastewater difficult to treat in the printing and dyeing industry. If not effectively monitored and directly discharged, it will cause chroma accumulation, COD exceeding standard and ecological toxicity risk in water bodies. Therefore, establishing a real-time and accurate monitoring system is the prerequisite for evaluating treatment effect and process control.

[0003] Accurate and continuous detection of inkjet printing wastewater is a key means to timely capture its water quantity and quality fluctuations under the influence of dye batches, cleaning cycles and process switching, and prevent chroma, COD and heavy metal emissions from exceeding standards. Only by obtaining reliable data can decision-making basis be provided for dynamic optimization of process parameters, real-time evaluation of treatment effect and rapid warning of abnormal events, and ultimately ensure that the wastewater treatment system is stable, meets standards and reduces environmental risks.

[0004] However, when facing such mutant water quality, the existing monitoring means is severely disturbed by noise such as flow pulse, dye batch difference, chemical residue, etc., resulting in concentration peak distortion, baseline drift and frequent abnormal values, which leads to lag or even misjudgment of conventional feedback control. SUMMARY

[0005] To solve the problem that noise such as flow pulse, dye difference and chemical residue superimposed under mutant water quality makes monitoring data peak and drift value chaotic, and causes conventional feedback control to lag or misjudge, the present application provides solutions in the following aspects.

[0006] In a first aspect, an online analysis method for inkjet printing wastewater treatment data includes: obtaining parameter data of inkjet printing wastewater based on each sensor and preprocessing to construct an original data matrix, wherein the parameter data includes water quality parameters, process parameters and equipment state parameters; decomposing the preprocessed original data matrix to obtain an orthogonal eigenvector matrix, and selecting parameter data of a preset number of retained principal components for reconstruction to generate a reconstructed data matrix; calculating the difference of all corresponding elements in the original data matrix and the reconstructed data matrix, quantitatively analyzing the reconstruction error before and after reconstruction, analyzing the correlation distortion index based on the similarity of corresponding elements between row vectors in the original data matrix and the reconstructed data matrix, constructing an objective function with the reconstruction error and the correlation distortion index as variables, and determining the optimal solution of the number of retained principal components; clustering based on the error and distance between the original data matrix and the reconstructed data matrix of the optimal solution of the number of retained principal components, screening abnormal data points, and adjusting process parameters during inkjet printing wastewater treatment according to the abnormal data points.

[0007] By comprehensively considering the reconstruction error and correlation distortion index to optimize the principal component number, the relationship between noise reduction and feature preservation is effectively balanced, so as to accurately identify the abnormal data points in the complex and variable inkjet printing wastewater treatment scene, and analyze the abnormal reasons according to the sensor data sequence corresponding to the abnormal data points, and then realize the accurate regulation of the wastewater treatment process parameters. Not only improve the accuracy and reliability of the abnormal detection, avoid the misjudgment caused by noise interference, but also enhance the adaptability of the system to the complex working conditions such as water quality mutation, ensure the stability and efficiency of the wastewater treatment process, and significantly improve the overall performance and quality of the inkjet printing wastewater treatment.

[0008] Preferably, the step of constructing the original data matrix comprises:

[0009] The missing values of each parameter data are filled by linear interpolation, and the supplemented parameter data are time-sequenced aligned according to the same time sequence as the reference, and each parameter sequence after alignment is spliced into an original data matrix in time sequence, wherein each row of the original data matrix corresponds to the same sampling time, and each column corresponds to the same sensor parameter.

[0010] Preferably, the orthogonal eigenvector matrix comprises the steps of:

[0011] The mean and standard deviation of the elements in each column vector in the original data matrix are calculated, the corresponding column vector is standardized based on the mean and standard deviation, to obtain a standardized matrix, the covariance matrix of the standardized matrix is calculated and eigenvalue decomposition is performed to obtain an orthogonal eigenvector matrix.

[0012] The differences in the dimensions and numerical ranges of different parameters are eliminated by standardization, so that each column vector is comparable. The covariance matrix is calculated based on the standardized matrix and eigenvalue decomposition is performed to obtain an orthogonal eigenvector matrix which can accurately reflect the direction of each principal component in the data, providing a key basis for dimensionality reduction and reconstruction, so as to remove noise while preserving the main structure and internal correlation of the data, providing a high-quality feature space for subsequent anomaly detection and process parameter regulation.

[0013] Preferably, obtaining the reconstruction data matrix comprises the steps of:

[0014] Setting the number of principal components to be retained as , selecting the eigenvectors corresponding to the first characteristic values to construct a dimensionality reduction projection matrix; multiplying the standardized original data matrix by the dimensionality reduction projection matrix to obtain a principal component score matrix projected into dimensional principal component space; multiplying the principal component score matrix by the transpose of the dimensionality reduction projection matrix to obtain a reconstruction data matrix back to the original space.

[0015] By selecting the eigenvector corresponding to the characteristic value The dimensionality reduction projection matrix is constructed by selecting the eigenvector corresponding to the characteristic value, and the high-dimensional data is effectively projected into the low-dimensional principal component space, thereby removing noise and retaining the main structure of the data. The principal component score matrix is mapped back to the original space through the inverse projection operation to generate the reconstructed data matrix. Not only does it reduce the dimensionality of the data and reduce the computational complexity, but it also removes the noise component through dimensionality reduction, making the reconstructed data more clearly reflect the true pattern of the data, providing a more accurate and reliable data basis for subsequent anomaly detection and process parameter control.

[0016] Preferably, the calculation method of the reconstruction error comprises:

[0017] The square errors of all corresponding elements in the original data matrix and the reconstructed data matrix are summed and then square rooted to obtain the reconstruction error of the data before and after reconstruction.

[0018] Preferably, the calculation method of the correlation distortion index comprises:

[0019] The correlation coefficients between each two column vectors in the original data matrix and the reconstructed data matrix are calculated respectively to obtain the correlation coefficient matrix of the original data matrix and the correlation coefficient matrix of the reconstructed data matrix. The correlation coefficients between each pair of rows in the standardized matrix are calculated, and the correlation coefficient matrix is constructed. The square errors between each pair of rows in the correlation coefficient matrix and the reconstructed correlation coefficient matrix are summed as the total error, and the square root of the total error is taken to obtain the correlation distortion index.

[0020] By calculating the correlation coefficient matrices of the original data matrix and the reconstructed data matrix respectively, and comparing the differences between the two, the correlation distortion in the reconstruction process can be quantified. The degree of preservation of the intrinsic correlation of the data in the dimensionality reduction and reconstruction process can be accurately evaluated, ensuring that important correlation structures in the data are not excessively damaged while removing noise. The calculation of the correlation distortion index provides a key basis for optimizing the number of principal components, thereby balancing between noise reduction and data structure preservation, and improving the accuracy and reliability of subsequent anomaly detection and process parameter control.

[0021] Preferably, the calculation method of the optimal solution of the number of retained principal components comprises:

[0022] The minimum value of the number of retained principal components in the objective function is set, the reconstruction error when retaining principal components is calculated, the correlation distortion index of the data matrix reconstruction when retaining principal components is calculated, the mapping result is multiplied by the reconstruction error using an exponential function, and the number of retained principal components is obtained. The target function in the case of one principal component, and the optimal principal component number is selected based on minimizing the target function.

[0023] Preferably, the step of obtaining the abnormal data points comprises:

[0024] Based on the optimal solution of the number of principal components reserved, the data matrix is reconstructed, and the error between the standardized vector of the original data matrix and the corresponding element of the standardized vector of the reconstructed data matrix is calculated.

[0025] The error is used as the influence weight of the distance between each point data and other point data in the clustering process between the original data matrix and the reconstructed data matrix, to obtain a modified distance.

[0026] Based on the modified distance, all corresponding data points at all times are clustered to obtain a plurality of clustering clusters, and isolated points not belonging to any clustering cluster are screened out and marked as abnormal data points.

[0027] The data sequence collected by each sensor at the time corresponding to the abnormal data points is extracted, and the abnormal reason is analyzed and the process parameters are adjusted according to the performance of the data collected by each sensor.

[0028] In a second aspect, an online analysis system for inkjet printing wastewater treatment data is provided, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the online analysis method for inkjet printing wastewater treatment data is realized.

[0029] The present application has the following effects:

[0030] 1. The present application constructs a target function by combining reconstruction error and correlation distortion index, and determines the optimal solution of the number of principal components reserved based on the target function, so that the internal correlation of the data is retained while the noise is removed. The improved distance is used for clustering analysis, which can more accurately identify abnormal data points, effectively solve the problem of abnormal peak drift value caused by the superposition of flow pulse, dye difference and residual reagent noise under sudden water quality, avoid the lag or misjudgment of conventional feedback control, and significantly improve the accuracy and reliability of abnormal detection.

[0031] 2. The present application can adapt to the complex water quality changes and process fluctuations in inkjet printing wastewater, and through online analysis and real-time control, the robustness and adaptability of the wastewater treatment system are enhanced. Even in the face of complex working conditions such as sudden changes in water quality, it can quickly respond and make reasonable adjustments to ensure the stable operation of the wastewater treatment system, and improve the controllability and stability of the entire inkjet printing wastewater treatment process. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1is a method flow chart of steps S1-step S4 in a method for online analysis of inkjet printing wastewater treatment data according to an embodiment of the application.

[0033] Figure 2 is a structural block diagram of an online analysis system for inkjet printing wastewater treatment data according to an embodiment of the application. DETAILED DESCRIPTION

[0034] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are some but not all of the embodiments of the application.

[0035] Reference Figure 1 An online analysis method for inkjet printing wastewater treatment data includes steps S1-S4, as follows:

[0036] S1: Based on each sensor, parameter data of inkjet printing wastewater is obtained and preprocessed to construct an original data matrix, wherein the parameter data includes water quality parameters, process parameters and equipment state parameters.

[0037] The missing values of each parameter data are filled by linear interpolation, and the supplemented parameter data are time-aligned based on the same time sequence as the reference, and each parameter sequence after alignment is spliced into an original data matrix in time sequence, wherein each row of the original data matrix corresponds to the same sampling time, and each column corresponds to the same sensing parameter.

[0038] The sequence break caused by missing sampling is eliminated by interpolation, and the time misalignment caused by sampling clock deviation of each sensor is eliminated by alignment, so that each row of the matrix accurately corresponds to the same sampling time, and each column corresponds to the same sensing parameter, so that the subsequent standardization, covariance calculation and feature decomposition can accurately reveal the true correlation between parameters on the basis of time-consistent and variable-comparable data.

[0039] S2: The preprocessed original data matrix is decomposed to obtain an orthogonal eigenvector matrix, and the parameter data of a preset number of principal components are selected for reconstruction to generate a reconstructed data matrix.

[0040] The mean and standard deviation of the elements in each column vector in the original data matrix are calculated, the corresponding column vector is standardized based on the mean and standard deviation to obtain a standardized matrix, the covariance matrix of the standardized matrix is calculated and decomposed to obtain an orthogonal eigenvector matrix.

[0041] The dimension and numerical range differences between different sensing parameters are eliminated by column-wise mean removal and variance normalization, ensuring that the subsequent covariance only reflects the true linear correlation between parameters; the orthogonal eigenvectors are obtained by eigen decomposition of the covariance matrix, so as to extract mutually uncorrelated principal component directions while retaining the maximum variance information, thereby providing an interpretable and statistically robust feature space for subsequent dimension reduction, denoising and anomaly detection.

[0042] Due to the zero drift, range error and random noise of the sensor, additional disturbances are superimposed in the original data, making it difficult to distinguish between real process anomalies and instrument errors, thereby reducing the accuracy of anomaly classification; therefore, PCA (Principal Component Analysis) decomposition is introduced, which decomposes the data energy into the principal component space according to the degree of cooperation across sensors through eigen decomposition of the covariance matrix, and reconstructs the data by retaining high-energy principal components and discarding low-energy components, thereby eliminating noise caused by sensor errors while retaining the inherent statistical correlation of each sensor due to process coupling, ensuring that subsequent anomaly recognition is only for real process changes, and improving system robustness and diagnostic accuracy.

[0043] Since the essence of PCA reconstruction is to suppress sensor noise while retaining interpretable information, the number of retained principal components directly determines the trade-off between denoising and feature fidelity: if is too small, the reconstruction error increases significantly, causing the process anomaly details to be excessively smoothed, and the anomaly detection sensitivity to decrease; on the contrary, if is too large, the instrument noise carried by the low-order principal components cannot be fully filtered out, but rather masks the high correlation of each sensor due to process coupling. Therefore, by constructing an objective function of reconstruction error and correlation distortion, and taking the minimization of the objective function as the criterion, the optimal value of is adaptively solved, thereby ensuring effective noise removal while retaining the inherent statistical correlation across sensors, providing a clean and discriminative feature space for subsequent anomaly recognition, and significantly improving system robustness and diagnostic accuracy. The specific implementation steps are as follows:

[0044] The reconstruction data matrix includes the following steps:

[0045] Set the number of principal components to be retained as , select the eigenvectors corresponding to the first eigenvalues to construct a dimension reduction projection matrix; multiply the standardized original data matrix by the dimension reduction projection matrix to obtain the principal component score matrix projected into the dimensional principal component space; multiply the principal component score matrix by the transpose of the dimension reduction projection matrix to obtain the reconstruction data matrix back to the original space.

[0046] S3: Calculate the difference between all corresponding elements in the original data matrix and the reconstructed data matrix, quantitatively analyze the reconstruction error before and after reconstruction, analyze the correlation distortion index based on the similarity of corresponding elements between row vectors in the original data matrix and the reconstructed data matrix, construct an objective function taking the reconstruction error and the correlation distortion index as variables, and determine the optimal solution of the number of retained principal components.

[0047] The square error of all corresponding elements in the original data matrix and the reconstructed data matrix is summed and then square rooted to obtain the reconstruction error of the data before and after reconstruction.

[0048] Specifically, the reconstruction error satisfies the following relationship:

[0049] ;

[0050] In the formula, represents the reconstruction error when retaining principal components, represents the standardized matrix of the original data matrix, represents the standardized matrix of the data matrix reconstructed by retaining principal components, is the Frobenius norm, represents the element in the th row and the th column of the original data matrix, represents the element in the th row and the th column of the reconstructed data matrix.

[0051] In online analysis of inkjet printing wastewater treatment data, the reconstruction error is quantified by calculating the Frobenius norm between the standardized original data matrix and the data matrix reconstructed by retaining principal components . The larger the reconstruction error, the fewer the number of retained principal components , the greater the deviation between the reconstructed data and the original data, which may lead to excessive smoothing of the data and thus reduce the sensitivity of anomaly detection.

[0052] The calculation method of the correlation distortion index includes:

[0053] Calculate the correlation coefficient between every two column vectors in the original data matrix and the reconstructed data matrix to obtain the correlation coefficient matrix of the original data matrix and the correlation coefficient matrix of the reconstructed data matrix. Calculate the correlation coefficient between each pair of rows in the standardized matrix and construct a correlation coefficient matrix from the correlation coefficients. Calculate the reconstructed correlation coefficient matrix of the reconstructed data matrix. Sum the squared errors between each pair of rows in the correlation coefficient matrix and the reconstructed correlation coefficient matrix as the total error. Take the square root of the total error to obtain the correlation distortion index.

[0054] Specifically, the correlation distortion index satisfies the following relationship:

[0055] ;

[0056] In the formula, Indicates reservation The correlation distortion index of data matrix reconstruction with principal components. Represents the first element in the original data matrix. row and number Correlation coefficient between columns Indicates reservation The first principal component is the data matrix after reconstruction of the data matrix. row and number The correlation coefficient between columns.

[0057] It should be noted that in the online analysis of inkjet printing wastewater treatment data, the correlation distortion index... The calculation is based on the correlation coefficient matrix of the original data matrix. The correlation coefficient matrix of the reconstructed data matrix Since the correlation coefficient matrix is ​​a symmetric matrix, that is... When calculating the correlation distortion before and after reconstruction, only the elements of the upper triangular matrix need to be considered, avoiding redundant calculations. In the original data matrix, noise masks the correlation between data points. As the number of retained principal components increases... With the reduction of noise, the reconstructed data matrix removes more noise, and the correlation between data gradually increases, leading to a distortion of the correlation index. Value follows Decrease and increase.

[0058] The methods for calculating the optimal solution that retains the number of principal components include:

[0059] Set the minimum number of principal components to retain in the objective function, and calculate the retention rate. Reconstruction error for each principal component, calculation of retention For each principal component, the correlation distortion index of the data matrix reconstruction is obtained. A negative exponential function is used to perform an exponential mapping on the correlation distortion index. The mapping result is multiplied by the reconstruction error to obtain the retained... The objective function for a given number of principal components is used to select the optimal number of principal components based on minimizing the objective function.

[0060] Specifically, the optimal solution satisfies the following relationship:

[0061] ;

[0062] In the formula, This represents the optimal solution for retaining the number of principal components. Describe the objective function. This indicates setting a lower limit for the number of principal components. This represents the upper rank of the data matrix. Indicates reservation Reconstruction error when there are principal components Indicates reservation The correlation distortion index of data matrix reconstruction with principal components. Represented by natural numbers An exponential function with base 0.

[0063] It should be noted that in the online analysis method for inkjet printing wastewater treatment data, the number of principal components retained is optimized. To balance reconstruction errors with data correlation preservation. Specifically, By minimizing the objective function exist The range of values The optimal solution is determined internally. Among them, This is a set hyperparameter, which is set to 3 in this embodiment. It can be adjusted according to specific circumstances, representing the minimum number of principal components to retain; while It is the dimension of the data matrix. and The smaller value in the range represents the maximum possible number of principal components to be retained.

[0064] This represents the quantized error value of the data matrix before and after reconstruction. The larger the value, the more principal components are retained. If the value is too small, the data will be over-smoothed, increasing the reconstruction error and thus reducing the sensitivity of anomaly detection. It is an index of correlation distortion before and after data matrix reconstruction; the larger the value, the more principal components are retained. While a smaller value results in better noise reduction, it also enhances the correlation between reconstructed data, potentially masking the high correlation between data from different sensors. Therefore, minimizing... This can reduce reconstruction errors while preserving the correlation between data, thereby enabling accurate identification of abnormal data.

[0065] S4: Clustering is performed based on the error and distance between the original data matrix and the reconstructed data matrix of the optimal solution with the number of retained principal components. Abnormal data points are screened, and the process parameters for inkjet printing wastewater treatment are adjusted according to the abnormal data points.

[0066] The data matrix is ​​reconstructed based on the optimal solution that retains the number of principal components, and the error of the corresponding elements of the normalized vector of the original data matrix and the normalized vector of the reconstructed data matrix is ​​calculated.

[0067] Specifically, the error satisfies the following relationship:

[0068] ;

[0069] In the formula, Indicates the first The reconstruction error at the corresponding time point. Represents the first element in the original data matrix. row and number Column elements, Represents the first element in the reconstructed data matrix. row and number The elements of the column.

[0070] The error is used as the influence weight of the distance between each point in the original data matrix and the reconstructed data matrix and the distance between other points in the clustering process to obtain the corrected distance;

[0071] Specifically, the corrected distance satisfies the following relationship:

[0072] ;

[0073] In the formula, Indicates the first Line corresponding time and number The corrected distance between data points at the corresponding time in the row. Represents the maximum value function. Indicates the first The reconstruction error at the corresponding time point. Indicates the first The reconstruction error at the corresponding time point. This represents the mean of the reconstruction error between the standardized vectors before and after reconstruction at all time points. Indicates the first Line corresponding time and number The Euclidean distance between data points at the corresponding time points.

[0074] It should be noted that, Take the first Line corresponding time and number The maximum value of the reconstruction error at the corresponding time point amplifies the impact of outliers, while As the mean of reconstruction error at all times, for standardization processing, the error of different data points has comparability; That is, the Euclidean distance between two data points, reflecting their geometric distance in the original space, which can more effectively identify abnormal data points.

[0075] Based on the corrected distance, the corresponding data points at all times are clustered to obtain a plurality of clustering clusters, and the isolated points not belonging to any clustering cluster are marked as abnormal data points;

[0076] For example, the clustering radius is set to 0.2, and the minimum number of points is set to 5, which can be adjusted according to the actual data characteristics and analysis requirements. After clustering, the isolated points not belonging to any clustering cluster are marked as abnormal data points.

[0077] The abnormal data point corresponding to the time point is extracted, and the abnormal reason is analyzed and the process parameters are adjusted according to the performance of the sensor data.

[0078] The application also provides an inkjet printing wastewater treatment data online analysis system. As shown in the figure, the system comprises a processor and a memory, and the memory stores computer program instructions. When the computer program instructions are executed by the processor, the method for analyzing the inkjet printing wastewater treatment data online according to the first aspect of the application is realized. The system also includes communication bus and communication interface and other components familiar to those skilled in the art. Its setting and function are known in the art, so it will not be repeated here. Figure 2

[0079] It should be pointed out that for those skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are within the scope of the present application. Therefore, the protection scope of the present application patent should be subject to the appended claims.​

Claims

1. An online analysis method of inkjet printing wastewater treatment data, characterized in that, The method comprises the steps of: Based on the parameters of each sensor, the parameter data of inkjet printing wastewater is obtained, preprocessed, and an original data matrix is constructed, wherein the parameter data includes water quality parameters, process parameters and equipment state parameters; The preprocessed original data matrix is decomposed to obtain an orthogonal characteristic vector matrix, and the parameter data of a preset reserved principal component number is reconstructed to generate a reconstructed data matrix; The difference between all corresponding elements in the original data matrix and the reconstructed data matrix is calculated, the reconstruction error before and after reconstruction is quantitatively analyzed, the similarity between the corresponding elements in the row vectors of the original data matrix and the reconstructed data matrix is analyzed, the correlation distortion index is analyzed, the reconstruction error and the correlation distortion index are taken as variables to construct an objective function, and the optimal solution of the reserved principal component number is determined; Based on the error and distance between the original data matrix and the reconstructed data matrix of the optimal solution of the reserved principal component number, the abnormal data points are screened, and the process parameters during the treatment of inkjet printing wastewater are adjusted according to the abnormal data points.

2. The method according to claim 1, wherein, The step of constructing the original data matrix comprises: The missing values of each parameter data are filled by linear interpolation, and the supplemented parameter data is time-sequenced according to the same time sequence as the reference to complete time sequence alignment, and each parameter sequence after alignment is spliced into an original data matrix in time sequence, wherein each row of the original data matrix corresponds to the same sampling time, and each column corresponds to the same sensing parameter.

3. The method according to claim 1, characterized in that, The orthogonal characteristic vector matrix comprises the steps of: The mean and standard deviation of the elements in each column vector of the original data matrix are calculated, the corresponding column vector is standardized based on the mean and standard deviation to obtain a standardized matrix, the covariance matrix of the standardized matrix is calculated and feature decomposition is performed to obtain an orthogonal characteristic vector matrix.

4. The method according to claim 1, characterized in that, The reconstructed data matrix comprises the steps of: The number of principal components to be reserved is set as , and eigenvectors corresponding to the first eigenvalues are selected to construct a dimension reduction projection matrix; the normalized original data matrix is multiplied by the dimension reduction projection matrix to obtain a principal component score matrix projected to a dimensional principal component space; and the principal component score matrix is multiplied by the transpose of the dimension reduction projection matrix to obtain a reconstructed data matrix projected back to the original space.

5. The method according to claim 1, wherein, The calculation method of the reconstruction error comprises: The square errors of all corresponding elements in the original data matrix and the reconstructed data matrix are summed and then square rooted to obtain the reconstruction error of the data before and after reconstruction.

6. The method according to claim 1, wherein, The calculation method of the correlation distortion index comprises: The correlation coefficients between each two column vectors in the original data matrix and the reconstructed data matrix are calculated respectively to obtain the correlation coefficient matrix of the original data matrix and the correlation coefficient matrix of the reconstructed data matrix, the correlation coefficients between each pair of rows in the standardized matrix are calculated, and the correlation coefficient matrix is constructed; the reconstruction correlation coefficient matrix of the reconstructed data matrix is calculated, the square error between each pair of rows in the correlation coefficient matrix and the reconstruction correlation coefficient matrix is summed as the total error, and the total error is square rooted to obtain the correlation distortion index.

7. The method according to claim 1, wherein the method is characterized by, The calculation method of the optimal solution of the reserved principal component number comprises: The minimum value of the number of principal components reserved in the objective function is set, the reconstruction error when the number of principal components reserved is calculated The correlation distortion index of the data matrix reconstruction when the number of principal components reserved is calculated The correlation distortion index of the data matrix reconstruction when the number of principal components reserved is calculated The objective function when the number of principal components reserved is multiplied by the mapping result of the negative exponential function to the correlation distortion index, and the mapping result and the reconstruction error are multiplied to obtain the objective function when the number of principal components reserved is selected based on the minimum objective function. 8.The online analysis method of inkjet printing wastewater treatment data according to claim 1, characterized in that, The step of obtaining the abnormal data points comprises: Based on the optimal solution of the reserved principal component number, the data matrix is reconstructed, and the error between the corresponding elements of the standardized vector of the original data matrix and the standardized vector of the reconstructed data matrix is calculated; The error is taken as the influence weight of the distance between each point data and other point data in the clustering process between the original data matrix and the reconstructed data matrix to obtain a modified distance; The data points corresponding to all time points are clustered based on the corrected distance to obtain multiple clustering clusters, and an isolated point not belonging to any clustering cluster is marked as an abnormal data point; A data sequence collected by each sensor at the time corresponding to the abnormal data point is extracted, and the abnormal reason is analyzed and the process parameters are adjusted according to the performance of the data collected by each sensor.

9. An inkjet printing wastewater treatment data online analysis system, characterized in that, The method comprises the following steps: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the method for online analysis of inkjet printing wastewater treatment data according to any one of claims 1-8 is realized.

Images