A method for monitoring alum flower morphology based on optical detection and a water treatment dosing method and system

Abstract

The present invention discloses a method for monitoring alum flower morphology based on optical detection, and a method and system for dosing water treatment drugs, which relate to the technical field of water treatment. In the alum flower morphology monitoring method, an alum flower absorbance detection model of a monochromatic light source is constructed to obtain absorbance scanning data in three-dimensional space. Edge detection is performed on the absorbance scanning data to obtain the inflection point of the alum flower shape, and the position information of the alum flower entity boundary is recorded; the wavelength of the light source for multi-band measurement is determined based on the alum flower structure, and an alum flower equipotential surface data set corresponding to different light source wavelengths is obtained; based on the alum flower equipotential surface data set corresponding to different light source wavelengths, the metal ion hydrate polymerized in the alum flower core layer is used as the center, and the epitaxial extension layer is used as a segmentation carrier to screen and determine the alum flower data set, thereby determining the equivalent parameters of the alum flower morphology. The present invention monitors the alum flower morphology from the perspective of optical absorbance detection, and the analysis accuracy is faster and more reliable, with good practicality.

Description

Technical Field

[0001] The invention belongs to the technical field of water treatment, and in particular relates to an alum floc morphology monitoring method based on optical detection and a water treatment dosing method and system. Background Art

[0002] During water purification, flocculants are typically added to the water for coagulation and sedimentation. The real-time flocculant dosage requires precise control based on the coagulation effect. The coagulation effect has a decisive impact on subsequent process control, ensuring optimal dosage while minimizing chemical consumption and improving effluent quality. Flocculant dosing is a core step in water treatment. However, due to its inherent hysteresis and nonlinear characteristics, automated dosing control has always been a major challenge in water treatment.

[0003] Traditional dosing control methods, for example, most water plants still use a control strategy based on the inlet water flow rate, combined with manual observation of alum floc morphology and sedimentation tank outlet turbidity to evaluate the dosing effect, and mainly rely on the flow ratio to control the coagulant dosage. However, the manually observed alum floc morphology is uncertain and more dependent on work experience and work ability; secondly, the fixed programmed flow ratio automatic control system, its test results only reflect the water quality status at the moment of sampling, and the determined coagulant dosage is discontinuous and hysteretic, making it difficult to achieve optimization and precise control.

[0004] With the rapid development of artificial intelligence technology, especially the widespread application of deep learning, underwater particulate matter imaging technology and computer artificial intelligence control have provided new ideas for the intelligent control of coagulation dosing. For example, existing Chinese patent CN112101352B discloses a method for identifying underwater flocculation status, a monitoring device, a computer device, and a storage medium that can automatically identify underwater flocculation and guide the coagulant dosage based on the flocculation status. This method performs image preprocessing on underwater flocculation images, performs multiple multi-scale sampling, extracts features from the sampling images based on different feature description factors, obtains feature vectors, selects highly correlated feature vectors and inputs them into a fuzzy layer to obtain a flocculation value. The flocculation value is then compensated based on the difference between turbidity meter data and the flocculation value and the process data of the water plant, and iterative training is performed until accurate flocculation value data is obtained. This technical solution, through the combination of machine vision and artificial intelligence, objectively analyzes the state of underwater flocculation in real time and provides guidance on the dosage of coagulant. However, this method is highly dependent on high-quality images and is easily affected by external factors such as lighting and noise. In addition, feature extraction may not be comprehensive, the generalization ability of the machine learning model is also limited, and the response speed of real-time monitoring is relatively slow. Summary of the Invention

[0005] The present invention aims to provide a method for monitoring alum floc morphology based on optical detection, aiming to analyze the edges of alum flocs based on optical absorbance detection to determine their morphology. The present invention also aims to provide a method and system for water treatment dosing based on optical detection, aiming to control the dosage of flocculant based on the alum floc morphology detected by optical absorbance analysis.

[0006] The present invention is mainly achieved through the following technical solutions:

[0007] A method for monitoring alum flower morphology based on optical detection comprises the following steps:

[0008] Step T1: Constructing a monochromatic light source alum flower absorbance detection model;

[0009] Step T2: Scan the three-dimensional space to be measured and obtain absorbance scanning data of the xoz plane and the yoz plane;

[0010] Step T3: performing edge detection on the absorbance scanning data of the xoz plane and the yoz plane respectively to obtain the inflection point of the alum flower shape and record the position information of the alum flower entity boundary;

[0011] Step T4: Determine the wavelength of the light source for multi-band measurement based on the alum flower structure, and obtain alum flower equipotential surface dataset corresponding to different light source wavelengths;

[0012] Step T5: Based on the equipotential surface datasets of the alum flower corresponding to different light source wavelengths, the metal ion hydrate polymerized in the alum flower core layer is taken as the center, and the epitaxial extension layer is used as the segmentation carrier. The alum flower dataset is screened and determined as follows:

[0013] S floc (x, y, z) = {p1(Am s ), p2(Am s ),…,p q (Am s )},

[0014] Where: p q (Am s ) is the spatial position information of the qth alum flower geometric body screened;

[0015] q is the number of alum flowers screened;

[0016] Step T6: Determine the equivalent parameters of the alum flower morphology based on the alum flower data set.

[0017] In order to better implement the present invention, further, step T3 includes the following steps:

[0018] Step T31: Derivative analysis and dynamic threshold filtering are used to distinguish the true alum flower signal from the ambient noise on the absorbance scan data of the xoz plane and the yoz plane respectively; the projections of overlapping structures are separated by analyzing the spatial continuity of the edge points of adjacent z layers;

[0019] Step T32: Based on the alum flower position data set of the entity in the xoz plane and the yoz plane, edge data adaptation is performed to achieve three-dimensional positioning and quantity statistics.

[0020] In order to better implement the present invention, step T31 further includes the following steps:

[0021] Step T311: Calculate the average single-dimensional absorbance of the xoz plane and the yoz plane when the number of subdivision steps of the z-axis is n;

[0022] Step T312: Calculate the single-dimensional fast overlay screening values ​​of the xoz plane and the yoz plane respectively:

[0023]

[0024]

[0025] in: is the average absorbance value in one dimension of the xoz plane when the number of subdivision steps on the z-axis is n;

[0026] The average absorbance value in one dimension of the yoz plane when the number of subdivision steps of the z-axis is n;

[0027] olpA n-zy It is the single-dimensional fast overlay screening value of the xoz plane when the number of subdivision steps of the z axis is n;

[0028] olpA n-zy When the number of subdivision steps is n, the single-dimensional fast overlay screening value of the yoz plane;

[0029] k olp is the precision coefficient of overlapping screening;

[0030] Step T313: Traverse the single-dimensional data of the xoz plane and the yoz plane to screen the absorbance extreme value points of the alum flowers:

[0031] (|Pd1(r)|<ε0∨Pd1(t)=0)∧A(r)>A uv ,

[0032] Where: A(r) is the absorbance corresponding to the spatial position r;

[0033] ε0 is the zero-point noise threshold of discrete derivative;

[0034] A uv is the absorbance calibration value;

[0035] Pd1(r) is the first-order partial derivative value of the absorbance at the spatial position r;

[0036] ∧ is the logical operator of AND;

[0037] ∨ is the logical operator of OR;

[0038] < is the less than operator;

[0039] > is the greater than operator;

[0040] Step T314: If the second-order partial derivative value of the absorbance of the extreme point r, Pd2(r)>0, the current extreme point is located at the edge, and the extreme point located at the edge is screened and the process proceeds to step T315;

[0041] Step T315: If the absorbance of the current extreme point is greater than the ghost screening value, determine whether there are ghosts on both sides of it; if the third-order partial derivative values ​​of the absorbance of the extreme points on both sides of the current extreme point exceed the third-order threshold range, determine that ghosts exist and proceed to step T316;

[0042] Step T316: Perform overlapping image segmentation: If the absolute value of the difference between the overlapping absorbance gradient values ​​of point rn and point r is less than or equal to the overlapping image judgment accuracy τ d , then the marked points rn and r are two boundaries of the same entity;

[0043] Step T317: Repeat steps T311 to T316 to traverse each z-axis subdivided single-dimensional data set; record edge data and overlap marks;

[0044] Step T318: vectorize the edge data and filter the data based on the convergence condition and the closure condition; then, based on the left and right boundary definitions, record the alum flower projection clusters of the xoz plane and the yoz plane.

[0045] In order to better implement the present invention, further, step T32 includes the following steps:

[0046] Step T321: traverse the entity's flower position data set on the xoz plane from the z-axis subdivision step number, and adapt it to the data on the yoz plane;

[0047] Step T322: extract the z-axis range of the entity; match the z-axis range entity corresponding to the yoz plane; synthesize the alum flower entity cluster and mark the usage data of the yoz plane;

[0048] The data sets filtered out by the xoz plane and the yoz plane are:

[0049] Sx ″(x 1 , x 2 )={s x1 ″(x 1 , x 2 ), s x2 ″(x 1 , x 2 ),…,s xn ″(x 1 , x 2 )},

[0050] S y ″(x1, x2)={s y1 ″(x1, x2), s y2 ″(x 1 , x 2 ),…,s yn ″(x 1 , x 2 )},

[0051] If S x ″(x 1 , x 2 ) and S y ″(x 1 , x 2 ) elements, respectively, perform subdivision steps on the z-axis to match, and the number of elements can also be matched, then the coordinates of the two elements are merged;

[0052] Step T323: If there is an unused data set in the yoz plane, traverse the unused flower position data set in the yoz plane;

[0053] Step T324: Extract the z-axis range of the entity, match the z-axis range entity corresponding to the xoz plane, and synthesize the alum flower entity cluster.

[0054] In order to better realize the present invention, further, in the step T4, the alum flower structure includes a core layer, a hydrophilic colloid coating layer, a microorganism aggregation layer, a suspended matter adsorption layer and an epitaxial extension layer arranged in sequence from the inside to the outside; the detection light source wavelength of the core layer is 380-450nm, the detection light source wavelength of the hydrophilic colloid coating layer is 100-280nm and 280-315nm, the detection light source wavelength of the microorganism aggregation layer is 100-280nm and 280-315nm, the detection light source wavelength of the suspended matter adsorption layer is 450-495nm, and the detection light source wavelength of the epitaxial extension layer is 450-495nm and the near-infrared light band.

[0055] In order to better realize the present invention, further, the wavelength of the light source for multi-band measurement includes 400nm, and the alum flower equipotential surface data set includes the data set of the core layer Specifically, the wavelengths of the light source for the multi-band measurement also include 254nm, 290nm, 485nm and 890nm; the alum flower equipotential surface dataset also includes:

[0056] The dataset of the middle layer:

[0057] and

[0058]

[0059] Datasets of the extension layer:

[0060] and

[0061]

[0062] in, The spatial position information of the nth alum flower geometric body screened under the wavelength of 400nm light source;

[0063] The spatial position information of the nth alum flower geometric body screened under the wavelength of 254nm light source;

[0064] The spatial position information of the nth alum flower geometric body screened under the wavelength of 290nm light source;

[0065] The spatial position information of the nth alum flower geometric body screened under the wavelength of 485nm light source;

[0066] It is the spatial position information of the nth alum flower geometric body screened under the wavelength of 890nm light source.

[0067] In order to better implement the present invention, further, step T5 includes the following steps:

[0068] Step T51: traverse the core layer dataset S 400 (x, y, z), if and If there are two independent extension layer entities, then go to step T52; otherwise, go to step T54;

[0069] Step T52: If and If there are two independent extension layer entities, then go to step T55; otherwise, go to step T53;

[0070] Step T53: If and If there are two independent suspended matter adsorption layers and epitaxial extension layers, then go to step T55; otherwise, discard the data;

[0071] Step T54: If and There are two independent suspended matter adsorption layer and epitaxial expansion layer entities, then if and If there are two independent extension layer entities, proceed to step T55; otherwise, discard the data;

[0072] Step T55: Record all data connected to the spatial extent of the two extension layers as equivalent data of a single flower, forming a data set of the flower;

[0073] in: is the spatial position information of the i-th alum flower geometric body screened under the wavelength of 400nm light source;

[0074] The spatial position information of the i+1th alum flower geometric body screened under the wavelength of 400nm light source;

[0075] It is the spatial position information of the i-1th alum flower geometric body screened under the wavelength of 400nm light source.

[0076] In order to better implement the present invention, further, in step T6, if the alum flower is a spherical model, the equivalent diameter of the alum flower is:

[0077]

[0078] If the alum flower is an ellipsoid model, the equivalent diameter of the alum flower is:

[0079]

[0080] Where: x L is the length of the alum flower geometry in the x direction;

[0081] y L is the length of the alum flower geometry in the y direction;

[0082] Z L is the length of the flower geometry in the z direction.

[0083] The volume density of alum flowers is:

[0084]

[0085] Where: V i is the volume of the i-th alum flower;

[0086] Dt x is the total x-axis length of the scan measurement;

[0087] Dt y is the total y-axis length of the scan measurement;

[0088] Dt z The total z-axis length of the scan.

[0089] The present invention is mainly achieved through the following technical solutions:

[0090] A water treatment dosing method based on optical detection is implemented based on the above-mentioned alum flower morphology monitoring method, comprising the following steps:

[0091] Step S1: real-time monitoring of the morphology of alum flowers to obtain equivalent parameters of the alum flower morphology;

[0092] Step S2: Derivation of the Zeta potential based on the equivalent parameters of the alum flower morphology, and dynamic control of the dosage based on the metal ion mass balance and potential feedback.

[0093] In order to better implement the present invention, further, step S2 includes the following steps:

[0094] Step S21: setting constraints;

[0095] Power constraints:

[0096]

[0097] Concentration boundary constraints:

[0098]

[0099] Constraints on dosage Q(t):

[0100] 0≤Q(t)≤Q max ,

[0101] in: is the average concentration of metal ions;

[0102] C Al，max It is the limit value of metal ions in sewage treatment environment;

[0103] Q max is the maximum dosage;

[0104] k in Enter the rate constant for drug addition;

[0105] k loss is the nonspecific loss rate (such as precipitation, adsorption)

[0106] k outis the consumption rate constant, k out =k0ρ(t)(d(q,t)) 2 ;

[0107] k0 is the reaction activity coefficient;

[0108] ρ(t) is the number density of alum flowers at time t;

[0109] d(q, t) is the average equivalent diameter corresponding to time t;

[0110] Step S22: Set the objective function as: p Within, the Zeta potential tracking error and the dosage cost model value J are minimized;

[0111]

[0112] Where: k is the number of steps;

[0113] N is the number of prediction steps;

[0114] is the average Zeta potential calculated based on the equivalent parameters of alum flower morphology in step k;

[0115] is the target zeta potential set;

[0116] w is the dosage weight;

[0117] Q k is the dosage of the kth step;

[0118] Δt is the time step;

[0119] Step S23: Solve to obtain the matrix u of the optimal dosage:

[0120] u=[Q0 Q1…Q N-1 ] T ,

[0121] Where: Q N-1 is the optimal dosage for step N;

[0122] The dosage is dynamically controlled based on the optimal dosage matrix u.

[0123] A water treatment dosing system based on optical detection is used to implement the above-mentioned water treatment dosing method based on optical detection, comprising an optical monitoring module, a data processing module and a dosing control module. The optical monitoring module is used to monitor the morphology of alum flowers in real time based on the above-mentioned alum flower morphology monitoring method to obtain accurate equivalent parameters of the alum flower morphology;

[0124] The data processing module is used to derive the Zeta potential based on the equivalent parameters of the alum flower morphology, and to solve the matrix u of the optimal dosage through the set objective function and constraint conditions;

[0125] The dosing control module is used to control the dosing amount of each step according to the matrix u of the optimal dosing amount.

[0126] The beneficial effects of the present invention are as follows:

[0127] (1) The present invention performs edge detection of alum flowers in water based on optical absorbance. Secondly, the present invention further integrates multi-light source detection to obtain alum flower equipotential surface datasets corresponding to different light source wavelengths. Based on the structural characteristics of alum flowers, the alum flower dataset is finally segmented to record the spatial position information of the alum flower geometry, and then the equivalent parameters of the alum flower morphology are determined. The present invention monitors the alum flower morphology from the perspective of optical absorbance detection, and no longer relies on high-precision underwater image capture. The morphological analysis accuracy of the present invention is faster and more reliable, and has better practicality.

[0128] (2) Based on the monitored alum flower morphology, the present invention integrates the objective function and the constraint conditions to solve the optimal dosing data set and realize dynamic control of dosing. Specifically, the present invention uses the explicit sensitivity analysis mechanism and the chain rule to establish a dynamic quantitative model of dosing amount and potential change, thereby significantly improving the control accuracy; secondly, the present invention develops a modular matrix architecture to enable the Hessian matrix and the gradient matrix to have online adaptive update capabilities, which can dynamically respond to process fluctuations of time-varying parameters; finally, a hard constraint safety boundary is implanted, and by constructing a multi-dimensional constraint matrix, the risk of over-dosing and concentration exceeding the limit are effectively prevented to ensure the safe operation of the process. The present invention realizes the closed-loop feedback mapping of the objective function to the control instruction through matrix operation, providing a quantifiable and verifiable intelligent optimization engine for the water treatment system, which significantly improves the adaptive control capability of the system while ensuring process stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0129] Figure 1 This is the principle block diagram of the absorbance measurement of alum flowers;

[0130] Figure 2 Schematic diagram of the structure of alum flower;

[0131] Figure 3 This is a flow chart of the water treatment dosing method based on optical detection of the present invention;

[0132] Figure 4 This is a flow chart of the alum flower morphology monitoring method based on optical detection of the present invention;

[0133] Figure 5 Flowchart for distinguishing real alum bloom signals;

[0134] Figure 6 Flowchart for adapting edge data into alum entity clusters;

[0135] Figure 7 Flowchart for screening datasets to identify alum flowers;

[0136] Figure 8 This is an architectural diagram of the water treatment dosing system based on optical detection according to the present invention. DETAILED DESCRIPTION

[0137] Example 1:

[0138] A method for monitoring alum flower morphology based on optical detection, such as Figure 4 As shown, the following steps are included:

[0139] Step T1, constructing a monochromatic light source alum flower absorbance detection model;

[0140] Step T2, scanning the three-dimensional space to be measured to obtain absorbance scanning data of the three-dimensional space; specifically, performing step-by-step planar scanning of the xoz plane and the yoz plane to obtain absorbance scanning data of the xoz plane and the yoz plane;

[0141] Step T3, using derivative analysis to obtain the inflection point of the alum flower shape and record the location information of the alum flower entity boundary; performing edge detection on the absorbance scanning data of the xoz plane and the yoz plane respectively, and calculating the first-order partial derivative, second-order partial derivative and third-order partial derivative of the absorbance scanning data;

[0142] Step T4, constructing an alum flower structure model, determining the wavelength of the light source for multi-band measurement based on the alum flower structure; obtaining alum flower equipotential surface dataset corresponding to different light source wavelengths;

[0143] Step T5, based on the alum flower equipotential surface dataset corresponding to different light source wavelengths, the metal ion hydrate polymerized in the alum flower core layer is taken as the center, and the epitaxial extension layer is used as the segmentation carrier to screen and determine the alum flower dataset; Step T6, based on the alum flower dataset, determine the equivalent parameters of the alum flower morphology.

[0144] Preferably, in the water treatment process, the morphology and distribution of alum flocs are key indicators for evaluating the flocculation effect. When alum flocs have overlapping or multi-layer structures, it is difficult for traditional optical measurement methods to accurately identify the boundaries and parameters of a single alum floc. Based on the transmitted light detection technology of Mie scattering, the present invention realizes three-dimensional tomography by co-scanning the light source and the detector in orthogonal planes, combined with the z-axis step displacement, and can obtain high-resolution absorbance data. In order to solve the interference of overlapping images, the present invention adopts derivative analysis (such as the first-order derivative to locate the absorbance extreme point, and the second-order derivative to identify the edge inflection point) combined with dynamic threshold filtering to distinguish the real alum floc signal from the environmental noise. By analyzing the spatial continuity of the edge points of adjacent z layers, the projection of the overlapping structure can be separated to achieve three-dimensional positioning and quantitative statistics. Further based on the scanning step size and spatial coordinates, the isotropic size, equivalent diameter and distribution density of the alum floc can be quantified. The present invention improves the accuracy of complex floc structure characterization by integrating multi-axis motion control, scattered signal analysis and mathematical morphology processing, and provides a highly reliable online monitoring solution for water treatment process optimization.

[0145] Preferably, the specific contents of the alum flower absorbance detection model for constructing a monochromatic light source are as follows:

[0146] The alum flower absorbance detection model is based on the Mie scattering attenuation theory, that is, alum flowers are regarded as a multi-scale particle system formed in the coagulation process, and its scattering characteristics play a dominant role in the attenuation of light signals. The Mie scattering attenuation theory can accurately describe the scattering and absorption behavior of spherical particles (such as alum flowers) to light, and is particularly suitable for inhomogeneous medium systems when the particle size is comparable to the wavelength of light. Compared with Rayleigh scattering (applicable to the case where the particle size is much smaller than the wavelength), Mie scattering takes into account the combined effects of particle size, complex refractive index (including absorption characteristics) and incident wavelength, and can quantify the total light intensity attenuation caused by alum flower particle groups in turbid water bodies. Figure 1 As shown, when measuring absorbance, light can be emitted through the light source array, and then passes through the focusing lens, the flocculation tank and the focusing lens in sequence before reaching the detector array.

[0147] Preferably, the specific contents of scanning the three-dimensional space to be measured and analyzing the position information of the boundary of the alum flower entity are as follows:

[0148] 1. Scan the three-dimensional space of the Cartesian coordinate system and obtain absorbance scan data.

[0149] 1.1. Scan to obtain the absorbance of the monochromatic light source:

[0150] The plane scanning process is performed step by step, and the x-axis motion scanning is performed first, then the y-axis motion scanning is performed, and then the z-axis is moved, and then the plane scanning is repeated. Preferably, a surface light source can be used for one transmission to obtain all three-dimensional transmission absorbance at the same time.

[0151] The absorbance scanning data of the xoz plane is obtained by spatial scanning:

[0152]

[0153] The absorbance scanning data of the yoz plane is obtained by spatial scanning:

[0154]

[0155] Where: l is the number of subdivision steps on the x-axis;

[0156] n is the number of subdivision steps on the z axis;

[0157] m is the number of subdivision steps on the y-axis.

[0158] 1.2. Orthogonal plane projection constraints:

[0159] When scanning in the xoz and yoz planes, the absorbance distribution can be decomposed into:

[0160]

[0161] 2. Calculate the derivative (gradient) of the absorbance scan data of the xoz plane and yoz plane respectively.

[0162] The necessary and sufficient conditions for a closed geometric body include: ① closed set characteristics: all limit points belong to themselves; ② spatial convergence: the limit points of any convergent sequence are still inside the set.

[0163] 2.1 Derivative the absorbance scan data of the xoz plane:

[0164] 1) The first-order partial derivative on the x-axis is:

[0165]

[0166] Where: Δi is the subdivision step size of the x-axis in the xoz plane dimension;

[0167] 2) The second-order partial derivative on the x-axis is:

[0168]

[0169] 3) The third-order partial derivative on the x-axis is:

[0170]

[0171] 4) The first-order partial derivative on the z-axis is:

[0172]

[0173] Where: Δj is the subdivision step size of the z-axis in the xoz plane dimension;

[0174] 5) The second-order partial derivative on the z-axis is:

[0175]

[0176] 6) The third-order partial derivative on the z-axis is:

[0177]

[0178] In summary, let the xoz plane be:

[0179] The first-order partial derivative in the x direction is:

[0180] The second-order partial derivative in the x direction is:

[0181] The third-order partial derivative in the x-direction is:

[0182] The first-order partial derivative in the z direction is:

[0183] The second-order partial derivative in the z direction is:

[0184] The third-order partial derivative in the z direction is:

[0185] 2.2 Derivative the absorbance scan data of the yoz plane:

[0186] Using the method in the xoz direction, we can get the derivative in the yoz plane:

[0187] The first-order partial derivative in the y direction is:

[0188] The second-order partial derivative in the y direction is:

[0189] The third-order partial derivative in the y direction is:

[0190] The first-order partial derivative in the z direction is:

[0191] The second-order partial derivative in the z direction is:

[0192] The third-order partial derivative in the z direction is:

[0193] 3. Perform edge detection to obtain the inflection point of the alum flower shape and record the boundary position information of the alum flower entity.

[0194] In the light scattering analysis of alum flower structures, the zero values ​​of partial derivatives of each order have hierarchical physical meanings: the theoretical zero value of the first-order partial derivative corresponds to the extreme position of the structural gap and the inflection point of the homogenization of the overlapping shadows when the ideal alum flower (Mie scattering sphere model) is scanned along the x-axis in the z-axis plane. For actual non-uniform convergence geometries, dynamic threshold filtering of environmental noise is used to achieve quantitative characterization of structural heterogeneity by counting the number and spatial coordinates of the first and last equivalent zero-value points of the first-order derivative; the zero value of the second-order partial derivative strengthens the edge feature extraction based on the first-order detection, revealing the sudden change boundary of the alum flower contour; and the zero value of the third-order partial derivative is used as a high-order feature criterion specifically for identifying the composite structure formed by overlapping interference, and its sign change can distinguish between real alum flowers and optical overlapping shadow artifacts. This multi-order derivative system constructs a complete analytical framework from geometric morphology to optical response through the multi-scale coupling of mathematical analysis and physical mechanisms. Preferably, if Figure 5 and Figure 6 The specific steps are as follows:

[0195] Step a1: Scan the absorbance data of the xoz plane and yoz plane, such as Figure 5 As shown, perform the following steps respectively:

[0196] (1) Calculate the average absorbance of the z-axis subdivision;

[0197]

[0198] Where: l: total number of subdivision steps on the x-axis;

[0199] m: total number of subdivision steps on the y-axis;

[0200] When the number of subdivision steps of the z-axis is n, the average absorbance value of a single dimension of the xoz plane;

[0201] The average absorbance value in one dimension of the yoz plane when the z-axis subdivision step number is n;

[0202] A ni : Absorbance in the z-axis of the xoz or yoz plane, A ni ∈A zx ∪A zy .

[0203] (2) Calculate the single-dimensional fast overlap screening value;

[0204]

[0205] Among them: olpA n-zx : When the number of subdivision steps on the z-axis is n, the single-dimensional fast overlapping screening value of the xoz plane;

[0206] olpAn-zy : The single-dimensional fast overlay screening value of the yoz plane when the z-axis subdivision step number is n;

[0207] k olp : Overlapping screening accuracy coefficient, set by the engineer, the default is 0.2.

[0208] (3) Traverse the single-dimensional data and select the extreme absorbance points of alum flowers;

[0209] The screening condition is: (|Pd1(r)|<ε0∨Pd1(r)=0)∧A(r)>A uv ;

[0210] Where: A(r): absorbance corresponding to the spatial position point r (i.e., a point on the xoz plane and the yoz plane);

[0211] ε0 is the zero-point noise threshold of discrete derivative;

[0212] A uv : Absorbance calibration value; it can be the absorbance value detected by the online UVCOD sensor, or the value calibrated in the laboratory after filtering water, or an empirical value;

[0213] Pd1(r): the first-order partial derivative of the absorbance at the spatial position r;

[0214] And Pd1(r)∈Pd1 zx (x)∪Pd1 zx (z)∪Pd1 zy (y)∪Pd1 zy (z).

[0215] ∧: logical operator of AND;

[0216] ∨: logical operator of OR;

[0217] < is the less than operator;

[0218] > is the greater than operator.

[0219] (4) Extract the absorbance of the extreme point;

[0220] (5) If the second-order partial derivative value Pd2(r)>0, the current extreme point is an edge, and the absorbance of the current extreme point is determined to be greater than the fast overlay screening value. If so, it is determined whether there is overlay. If so, the process proceeds to step (6), otherwise, the process proceeds to step (7);

[0221] Among them, the judgment condition for the existence of overlapping images is: if the absorbance curve of the overlapping area presents a "double peak" or "platform" shape, the third-order derivative may have extreme points with opposite signs on both sides, indicating the existence of multiple objects.

[0222]

[0223] Where: Pd2(r)∈Pd2 zx (x)∪Pd2 zx (z)∪Pd2 zy (y)∪Pd2 zy (z);

[0224] Pd3(r): the third-order partial derivative of the absorbance at the spatial position r;

[0225] And Pd3(r)∈Pd3 zx (x)∪Pd3 zx (z)∪Pd3 zy (y)∪Pd3 zy (z);

[0226] x H : overlap bandwidth;

[0227] T and -T are the third-order thresholds respectively.

[0228] (6) Perform overlapping image segmentation;

[0229] The absorbance gradient value of the overlapping images with overlapping images is:

[0230]

[0231] If there is another point rn whose gradient difference with point r satisfies the accuracy range, then they are marked as two boundaries of the same entity; the judgment formula is as follows:

[0232]

[0233] Where: τ d : Overlap judgment accuracy;

[0234] The bandwidth of point r is x H The absorbance gradient value of the overlapping shadow;

[0235] The bandwidth of point rn is x H The absorbance gradient value of the overlap.

[0236] (7) Record edge data and overlap marks;

[0237] (8) Follow steps (1) to (7) to traverse each z-axis to subdivide the single-dimensional data set.

[0238] (9) After plane fusion of edge data, data screening is performed based on the necessary and sufficient conditions (convergence conditions and closure conditions) of the closed geometry of the alum flower;

[0239] (10) If there is continuous data that meets the convergence condition and the closure condition, then go to step (11);

[0240] The data set D″(x1, x2)={D1, D2, ..., D a} Perform vectorization processing to form a vector set

[0241] Where: D a : Plane data individual, a is the number of data;

[0242] The filtered vectorized plane data; The format is (x 1 g1 x 2 g2), where x 1 , x 2 are the two-dimensional basis vectors of the plane, and g1 and g2 are the plane coordinate values.

[0243] If there is a data vector in a region:

[0244] If the convergence condition is satisfied (the modulus of the new vector formed by subtracting adjacent coordinate vectors is less than a constant value), it will always exist:

[0245]

[0246] And it satisfies the closed condition (discrete points form a polygon, and the sum of the side vectors is zero), and it always exists:

[0247]

[0248] Among them: v, i, j: intermediate process data, v∈[i, j];

[0249] Δx 1 , Δx 2 : subdivision step size in plane dimension;

[0250] k D : Convergence precision, the default value is 2.

[0251] The vector that satisfies the condition The restored data set is: i ′, r′ i+1 ,…,r j ′}, i≤j≤a∧i>0∧j>0, the partial derivatives in the z-axis direction all satisfy: (|Pd1(r)|<ε0∨Pd1(r)=0)∧A(r)>A uv ;

[0252] Where: ε0 is the zero-point noise threshold of discrete derivative;

[0253] A uv : Absorbance calibration value; it can be the absorbance value detected by the online UVCOD sensor or the value calibrated by filtered water in the laboratory, or an empirical value;

[0254] Pd1(r): the first-order partial derivative value of the absorbance at the spatial position r,

[0255] and

[0256] Pd1(r)∈Pd1 zx (x)∪Pd1 zx (z)∪Pd1 zy (y)∪Pd1 zy (z).

[0257] Mark the data set that satisfies

[0258] s″(x 1 , x 2 )={r′i, r′i+1,…, r′j}.

[0259] (11) Record the projection cluster of alum flowers.

[0260] 1) Method for determining left and right boundaries

[0261] When the preconditions are met: the left boundary method:

[0262] A(v-1)<A(v)<A(v+1)

[0263] When the preconditions are met: the right boundary method:

[0264] A(v-1)>A(v)>A(v+1)

[0265] Where: A(v): absorbance measured at a certain spatial position, A(v)∈A zx ∪A zy ;

[0266] v: Bandwidth variable for range positions.

[0267] 2) Process the xoz plane using the above method to obtain the alum flower position data of a single dimension:

[0268] [Lotz x (x)1, Lot zx (x)2,…,Lot zx (x) i ];

[0269] [Lotzx (z)1, Lot zx (z)2,…,Lot zx (z) j ];

[0270] The above method is used to process the yoz plane to obtain the alum flower position data of a single dimension:

[0271] [Lot zy (x)1, Lot zy (x)2,…,Lot zy (x) a ;

[0272] [Lot yx (z)1, Lot yx (z)2,…,Lot yx (z) a ];

[0273] Where: i, j, a, b: the number of alum flowers with complete boundaries counted in different dimensions;

[0274] Lot zx (x) i , Lot yx (x) j , Lot zy (x) a , Lot yx (x) b : Single-dimensional position information of alum flowers, whose elements contain the position information of a certain dimension of alum flowers, for example:

[0275] Lot zx (x) i ={x left , x middle , x right}x left , x middle , x right ∈l.

[0276] Where: x right is the right critical value of the alum flower in the x direction;

[0277] x left is the left critical value of the alum flower in the x direction;

[0278] x middle is the measurement value of the middle part of the alum flower in the x direction;

[0279] l is the number of subdivision steps on the x-axis.

[0280] Step a2: Dual-plane entity records data and performs edge data adaptation;

[0281] Step a3: traverse the xoz plane entity data from the z-axis subdivision step number and adapt it to the yoz plane data;

[0282] Step a4: extract the z-axis range of the entity;

[0283] Step a5: Match the z-axis range entity corresponding to the yoz plane;

[0284] The data sets filtered out on the xoz and yoz planes are:

[0285] S x ″(x 1 , x 2 )={s x1 ″(x 1 , x 2 ), s x2 ″(x 1 , x 2 ),…,s xn ″(x 1 , x 2 )};

[0286] S y ″(x 1 ,x 2 )={s y1 ″(x 1 ,x 2 ),s y2 ″(x 1 ,x 2 ),…,s yn ″(x 1 ,x 2 )};

[0287] If s x ″(x 1 ,x 2 ) and any element in s y ″(x 1 ,x 2 ) elements, the subdivision steps on the z-axis can match, and the number of elements can also be matched, then the coordinates of the two elements are merged. That is:

[0288] Let function f z (s″(x 1 , x 2 )) is to get the data set

[0289] s″(x 1 , x 2 )={r i ′, r′ i+1 ,…,r j ′} in the z-axis step size;

[0290] f z (s xn ″(x 1 , x 2 ))-f z (s yn ″(x 1 , x 2 ))≤τ z ;

[0291] Where: τ Z : z-axis step matching accuracy, the default is 0;

[0292] Let function f Num (s″(x 1 , x 2 )) is to get the data set

[0293] s″(x 1 , x 2 )={r i ′, r′ i+1 ,…,r j '} in the elements;

[0294] f Num (s xn ″(x 1 ,x 2 ))-f Num (s yn ″(x 1 ,x 2 ))≤τ Num .

[0295] Where: τ Num : The matching accuracy of alum flower entities, the default is 0.

[0296] Element Fusion:

[0297] If the set S x ″(x 1 ,x 2 ) and the set S y ″(x 1 , x 2 ) matches successfully, then the sets are combined in ascending or descending order; that is, if S x ″(x 1 , x 2 ) in the element r x ″(x1, z1) and S y ″(x 1 ,x 2 ) in the element r yIf ″(y2,z2) exists and z1=z2, then the synthesized spatial element r(x,y,z) has corresponding values ​​of x1,y2,z1. Of course, if there is a difference in matching accuracy, it can be filled by moving in the order of front and back.

[0298] Step a6: synthesize the alum flower entity cluster and mark the yoz usage data;

[0299] Step a7: If there is unused data in the yoz plane, traverse all unused entity data in the yoz plane and extract the z-axis range; match the z-axis range entity corresponding to the xoz plane as in step a5; and synthesize the alum flower entity cluster.

[0300] The spatial position of the cluster of alum flowers that has been screened by the process is expressed by the function p(Am s ) annotation, function p(Am s ) is the spatial position information of each geometric body, and the new statistical data set of alum flowers is:

[0301] S(x,y,z)={p1(Am s ),p2(Am s ),…,p q (Am s )};

[0302] Where: p q (Am s ): The spatial position information of the qth alum flower geometry screened;

[0303] q: the number of alum flowers counted;

[0304] S(x, y, z): A dataset containing the spatial locations of alum flowers.

[0305] The rest of this embodiment is the same as that of Embodiment 1 or 2, and therefore will not be described in detail.

[0306] Example 2:

[0307] This embodiment is based on the optimization of embodiment 1, and the alum flower structure is constructed. The specific content of adapting the wavelength of the light source according to the alum flower structure is as follows:

[0308] 1. Construct alum flower structure model.

[0309] 1.1 Alum flower formation mechanism:

[0310] A system formed by dispersing one substance in another is called a dispersion. The former is called the dispersant, while the latter is called the dispersant. Different types of dispersions are simply different forms of matter, and they can be converted into each other. When the particle size of the particles in the system is between 1 and 100 nm, it is a colloidal dispersion. There are two types of particles within this size range in natural water, industrial wastewater, and domestic sewage: soluble macromolecules, such as proteins, and insoluble particles, such as clay minerals. Both form colloidal dispersions with water. The former, without an interface between them and the aqueous dispersion medium, are thermodynamically stable, reversible, and can be redispersed after precipitation. They are called lyophilic colloids. The latter, with an interface between them and the aqueous dispersion medium, are thermodynamically unstable due to the presence of interfacial free energy, irreversible, and cannot be redispersed after precipitation. They are called lyophobic colloids. These two types of colloidal dispersions share both similarities and differences. Because the particle sizes are similar, some properties related to particle size are also similar, such as dynamic properties, optical properties, and rheological properties. However, properties related to interfaces, such as electrical properties and surface chemistry, differ. The particle size of particles in a suspension system is larger than that of particles in a colloidal dispersion system. Because the particle size of particles in a suspension system is larger and there is a distinct interface between the dispersant and the dispersant, many properties remain similar to those of a colloidal dispersion system. The substances that form suspension systems in natural water, domestic sewage, and industrial wastewater are primarily silt and oil. In addition to forming true solutions due to the presence of dissolved salts, natural water, domestic sewage, and industrial wastewater often contain colloids and suspended matter. Therefore, they are often all at once true solutions, colloidal dispersions, and suspension systems—complex, integrated systems.

[0311] Coagulation is the aggregation of dispersed particles in water. It's essentially the process by which destabilized particles collide with each other and are brought together by van der Waals forces. These aggregates are approximately 1 nm or larger in size. Flocculation, on the other hand, occurs when destabilized or partially destabilized particles produced during coagulation collide with each other, further aggregate, or form larger flocs (or flocs) through chemical interactions. In practice, coagulation and flocculation are two distinct processes and effects that are difficult to distinguish.

[0312] Alum flocs exhibit a multilayered composite structure centered around a flocculant hydrolysis product (e.g., Al(OH)3): an inner core consisting of a positively charged metal hydroxide skeleton, followed outward by a hydrophilic colloid coating (which adsorbs soluble macromolecules such as proteins), a microbial aggregation layer (cross-linking extracellular polymers through electrostatic interactions), a suspended matter adsorption layer (clay minerals and oil droplets embedded in the pores), and a loose, epitaxial outer layer (formed by the collision and expansion of partially destabilized particles). These layers are combined through electrostatic interactions and chemical bonding to form a porous, network-like aggregate with fractal characteristics, exhibiting charge heterogeneity, porosity gradients, and particle size polydispersity.

[0313] 1.2 Alum flower structure:

[0314] like Figure 2 As shown in the figure, based on the theory of dispersed systems and the characteristics of the water treatment process, it can be deduced that the alum flowers present a multi-layer composite structure, from the inner layer to the outer layer:

[0315] The core layer of alum flocs: composed of the hydrolysis products of flocculants, such as a network skeleton formed by metal hydroxides such as Al(OH)3, which has a large number of positively charged active sites.

[0316] Hydrophilic colloid coating layer: Soluble macromolecules such as proteins and polysaccharides are bound to the flocculant skeleton through electrostatic adsorption, hydrogen bonds, etc., forming a hydrophilic colloid interface layer. This layer is tightly bound to water molecules to maintain colloid stability.

[0317] Microbial aggregation layer: Microbial cells in the water body are wrapped into the colloidal network through surface charge neutralization and bridging, and their extracellular polymers are cross-linked with flocculants.

[0318] Suspended matter adsorption layer: Clay minerals, metal oxides and other hydrophobic colloidal particles are embedded in the pores of the colloidal network through van der Waals forces and hydrophobic effects, while adsorbing larger particles such as oil droplets and sediment in the water to form a rough surface.

[0319] Epitaxial expansion layer: The particles that are not completely destabilized continue to collide under the action of fluid shear, and expand outward through edge effects, polymer bridging, etc., forming a loose flocculent branch structure.

[0320] 2. Adapt the light source wavelength for multi-band measurement based on the alum flower structure.

[0321] The core layer of alum flocs is composed of the hydrolysis products of flocculants, such as the network skeleton formed by metal hydroxides such as Al(OH)3. The structure can be detected by ultraviolet light (380-450nm); the charge transfer absorption band (Fe 3 +→O 2 -), reflecting a positive correlation between absorbance and the metal hydroxide content in the alum flocs, and the shift in the absorbance peak position can reflect the particle crystallinity or degree of hydroxylation. Preferably, this embodiment uses a 400nm light source wavelength for the core layer, though engineers can adapt it to different water bodies.

[0322] Hydrophilic colloid coating: Soluble macromolecules such as proteins and polysaccharides are bound to the flocculant skeleton through electrostatic adsorption and hydrogen bonding, forming a lyophilic colloid interface layer. This structure can use ultraviolet light UVC (100-280nm) and ultraviolet light UVB (280-315nm). This layer primarily exhibits UVC (100-280nm) absorption peaks associated with conjugated double bonds (π-π* transitions) of organic matter (such as humic acid and protein) and pyrimidine base absorption of microbial DNA / RNA; UVB (280-315nm) absorption peaks are associated with n-π* transitions of aromatic compounds (such as phenols) and some microbial metabolites. Preferably, this embodiment uses light source wavelengths of 254nm and 290nm, although engineers can adapt according to different water bodies.

[0323] Microbial aggregation layer: The microbial cells in the water body are wrapped into the colloidal network through surface charge neutralization and bridging, and their extracellular polymers are cross-linked with the flocculant. This structure can use ultraviolet light UVC (100-280nm) and ultraviolet light UVB (280-315nm); this layer mainly shows that the UVC (100-280nm) ultraviolet absorption peak is related to the conjugated double bonds (π-π* transitions) of organic matter (such as humic acid and protein) and the pyrimidine base absorption of microbial DNA / RNA. Preferably, this embodiment corresponds to the use of a 254nm light source wavelength; of course, engineers can adapt it according to different water bodies.

[0324] Suspended matter adsorption layer: Clay minerals, metal oxides and other hydrophobic colloidal particles are embedded in the pores of the colloidal network through van der Waals forces and hydrophobic interactions, while adsorbing larger particles such as oil droplets and mud in the water to form a rough surface. This structure can use blue light (450-495); the Mie scattering of colloidal particles (scattering is strongest when the particle size is close to the wavelength), the absorbance is related to the average particle size of the alum flower (particle size increases → scattering increases → apparent absorbance increases), and the dynamic light attenuation curve can be used to deduce the particle aggregation rate. Preferably, this embodiment corresponds to the use of a 485nm light source wavelength; of course, engineers can adapt it according to different water bodies.

[0325] Epitaxial expansion layer: Particles that are not completely destabilized continue to collide under the action of fluid shear, and expand outward through edge effects, polymer bridging, etc., forming a loose flocculent branching structure. This layer mainly shows the scattering absorbance of colloidal particles from infrared light to near-infrared light, which is used to reflect the porosity of alum flocs. Its scattering signal is related to the particle size distribution (wide distribution leads to multi-wavelength scattering differences). This structure can use blue light (450-495) and near-infrared light; blue light and near-infrared light mainly show: Mie scattering of colloidal particles (scattering is strongest when the particle size is close to the wavelength), and the absorbance is related to the average particle size of alum flocs (particle size increases → scattering increases → apparent absorbance increases), and the dynamic light attenuation curve can be used to deduce the particle aggregation rate. Preferably, this embodiment corresponds to the use of 485nm and 890nm light source wavelengths; of course, engineers can adapt according to different water bodies.

[0326] Preferably, the specific content of obtaining the alum flower equipotential surface data set corresponding to different light source wavelengths is as follows:

[0327] (1) Statistics of alum flowers:

[0328] Get the alum flower boundary set of different bands through different bands:

[0329] Core layer: Alum flower equipotential surface dataset with a wavelength of 400nm:

[0330]

[0331] Middle layer: Alum flower equipotential surface dataset with a wavelength of 254nm:

[0332]

[0333] Alum flower equipotential surface dataset for 290nm light source wavelength:

[0334]

[0335] Core range of the extension layer:

[0336] Alum flower equipotential surface dataset for 485nm light source wavelength:

[0337]

[0338] Alum flower equipotential surface dataset for 890nm light source wavelength:

[0339]

[0340] in, The spatial position information of the nth alum flower geometric body screened under the wavelength of 400nm light source;

[0341] The spatial position information of the nth alum flower geometric body screened under the wavelength of 254nm light source;

[0342] The spatial position information of the nth alum flower geometric body screened under the wavelength of 290nm light source;

[0343] The spatial position information of the nth alum flower geometric body screened under the wavelength of 485nm light source;

[0344] It is the spatial position information of the nth alum flower geometric body screened under the wavelength of 890nm light source.

[0345] (2) Neutral polymerization of alum flower morphology, with the hydrate of the core polymerized metal ion as the center and the epitaxial extension layer as the dividing carrier, the ultimate alum flower morphology combination is carried out, such as Figure 7 The specific screening process is as follows:

[0346] Step T51: traverse the core layer dataset S 400 (x, y, z), if and If there are two independent extension layer entities, then go to step T52; otherwise, go to step T54;

[0347] Step T52: If and If there are two independent extension layer entities, then go to step T55; otherwise, go to step T53;

[0348] Step T53: If and If there are two independent suspended matter adsorption layers and epitaxial extension layers, then go to step T55; otherwise, discard the data;

[0349] Step T54: If and There are two independent suspended matter adsorption layer and epitaxial expansion layer entities, then if and If there are two independent extension layer entities, proceed to step T55; otherwise, discard the data;

[0350] Step T55: Record all the data connected to the two extension layers as the equivalent data of a single flower, forming a data set of the flower:

[0351] S floc (x, y, z) = {p1(Am s ), p2(Am s ),…,p q (Ams )};

[0352] Where: p q (Am s ) is the spatial position information of the qth alum flower geometric body screened;

[0353] q is the number of alum flowers screened;

[0354] Preferably, if q blooms are detected at a certain moment, the equivalent diameter of the bloom is:

[0355] 1) If the alum flower is a spherical model, the equivalent diameter is:

[0356]

[0357] 2) If the alum flower is an ellipsoidal model, the equivalent diameter is:

[0358]

[0359] Where: x L is the length of the alum flower geometry in the x direction;

[0360] y L is the length of the alum flower geometry in the y direction;

[0361] Z L is the length of the flower geometry in the z direction.

[0362] The volume density of alum flowers is:

[0363]

[0364] Where: V i is the volume of the i-th alum flower;

[0365] Dt x : Total x-axis length of the scan measurement;

[0366] Dt y : The total length of the y-axis of the scan measurement;

[0367] Dt z : The total z-axis length of the scan measurement.

[0368] The rest of this embodiment is the same as that of embodiment 1, so it will not be described again.

[0369] Example 3:

[0370] A water treatment dosing method based on optical detection, such as Figure 3As shown in the figure, the aim is to monitor the equivalent parameters of alum floc morphology in real time through the above-mentioned alum floc morphology monitoring method based on optical detection. The van der Waals forces between alum floc particles are derived based on the equivalent parameters of alum floc morphology; the zeta potential is derived based on the van der Waals forces; based on the metal ion mass balance and potential feedback, constraints and objective functions are set to solve the optimal dosing matrix u, thereby achieving dynamic control of the dosing amount Q(t).

[0371] Specifically, the present invention aims to adjust the dosage Q(t) based on a dynamic model, so that the zeta potential tracks the set value, while minimizing the dosing cost. The dynamic model is based on metal ion mass balance and potential feedback.

[0372] The present invention comprises the following steps:

[0373] 1. Construct the dynamic equation of metal ion concentration:

[0374]

[0375] in: is the average concentration of metal ions;

[0376] k in Enter the rate constant for drug addition;

[0377] k out : Consumption rate constant (s -1 ), which is related to the rate of floc formation:

[0378] k out =k0ρ(t)(d(q,t)) 2 ;

[0379] k0: Reaction activity coefficient (m -2 ·s -1 );

[0380] ρ(t): Number density of alum flowers (pieces / m 3 );

[0381]

[0382] q(t) is the number of blooms in the periodic scanning result at time t;

[0383] d(q, t) is the average equivalent diameter corresponding to time t;

[0384] k loss : Nonspecific loss rate (such as precipitation, adsorption).

[0385] 2. Constructing Zeta potential tracking model:

[0386] Based on the simplification of the Stern layer model, we can get:

[0387]

[0388] in: The average concentration of metal ions at time t;

[0389] k ads :Metal ion adsorption equilibrium constant (m 3 / mol), obtained from experimental data, empirical data or table lookup;

[0390] The average zeta potential at time t;

[0391] is the set maximum zeta potential.

[0392] 3. Control objectives and discretization:

[0393] Constructing the objective function: In the prediction domain T p Zeta potential tracking error and dosage cost are minimized:

[0394]

[0395] Where: k is the number of steps;

[0396] N is the number of prediction steps;

[0397] is the average Zeta potential calculated based on the equivalent parameters of alum flower morphology in step k;

[0398] is the target zeta potential set;

[0399] w is the dosage weight;

[0400] Q k is the dosage of the kth step;

[0401] Δt is the time step;

[0402] Then, the state equation is obtained by the discrete Euler method:

[0403]

[0404] in, are the average concentrations of metal ions monitored in the kth step and k+1th step, respectively.

[0405] Preferably, the optimal dosage is solved based on the Quadratic Programming (QP) form. Quadratic Programming (QP) is an optimization problem whose objective function is a quadratic function and whose constraints are linear functions. Specifically:

[0406] (1) Constructing the objective function:

[0407] Let the matrix u of the optimal dosage be:

[0408] u=[Q0 Q1 … Q N-1 ] T ;

[0409] Where: Q N-1 : The dosage of step N;

[0410] u: matrix of optimal dosage, in matrix form;

[0411] T: matrix transpose;

[0412] Substituting into the objective function, we get:

[0413]

[0414] Where: f: gradient vector;

[0415] H: Hessian matrix, that is, the diagonal matrix composed of the dosage cost items;

[0416] H=2wΔt·I N ;

[0417] I N : is the N×N identity matrix, where N is determined by the engineer; for example, when N=3, the identity matrix is ​​as follows:

[0418]

[0419] The vector gradient f is the linear contribution of the error term and needs to be calculated for each Q k right The sensitivity is as follows:

[0420]

[0421] Where: f k :Each Q k right sensitivity;

[0422] Q k : dosage of the kth step;

[0423] is the average Zeta potential calculated based on the equivalent parameters of the alum flower morphology in step k.

[0424] a1. Sensitivity chain rule:

[0425]

[0426] a2. Electrochemical sensitivity (Stern layer model):

[0427]

[0428] Where: k ads :Metal ion adsorption equilibrium constant (m 3 / mol), obtained from experimental data, empirical data or table lookup;

[0429] a3. Concentration sensitivity (equation of state):

[0430]

[0431] The expression example of the gradient vector is:

[0432] Assume N = 3,

[0433]

[0434] in: is the set maximum zeta potential.

[0435] (2) Constructing constraints

[0436] b1. Dynamic constraints:

[0437]

[0438] b2. Concentration boundary:

[0439]

[0440] b3. Dosage limit:

[0441] 0≤Q(t)≤Q max ;

[0442] Where: C Al，max :The metal ion limit value in the sewage treatment environment is set by engineers through experience, which can be theoretically

[0443] Derivation;

[0444] Q max : Maximum dosage, set by the engineer.

[0445] (3) Solving the goal: using the interior point method or the effective set method, find the optimal dosing matrix u with the minimum value of J, and update the optimal dosing matrix u in real time.

[0446] Specifically, the operation control effect of a sewage treatment plant is verified as follows:

[0447] Setting the average target zeta potential Initial Collection w=0.01s 2 / mol 2 .

[0448] The operation results are: adjustment time ts = 120s, overshoot <5%, steady-state error <1mV, and the dosage Q(t) smoothly transitions to the steady-state value of 0.08L / s.

[0449] Preferably, the specific steps for deriving the Zeta potential based on the equivalent parameters of the alum flower morphology are as follows:

[0450] (1) Derivation of the van der Waals forces between alum floc particles based on the equivalent parameters of alum floc morphology;

[0451] (2) Derivation of Zeta potential based on van der Waals force.

[0452] Basic theory: The lyophobic colloid stability theory is commonly known as the DLVO theory. Colloidal particles attract each other due to the van der Waals force. When they approach, the double layers overlap and repel each other. The stability of the colloid depends on the relative size of the two. This theory calculates the change of attractive energy and repulsive energy between particles of different shapes with distance. If the repulsive energy is greater than the attractive energy at a certain distance, the colloid is stable; if the attraction is always greater than the repulsion, the colloid will precipitate. As the ion concentration or the valence of the counterion in the solution increases, the van der Waals force remains basically unchanged, the double layer repulsive energy decreases, and the colloid stability decreases until it precipitates. According to the dynamic DLVO model theory, the colloidal force field is derived from the dynamic morphological data. Preferably, the derivation of the van der Waals force specifically includes the following steps:

[0453] set up:

[0454] 1. Equivalent the shape of alum flowers to spherical particles;

[0455] 2. The effects of Brownian motion and fluid shear on the mechanics of alum flowers are ignored in the derived model;

[0456] 3. The Hamaker constant is affected by the ionic strength of the solution;

[0457] The van der Waals force is deduced based on the equivalent parameters of the alum flower morphology:

[0458]

[0459] Among them: F vdW : Van der Waals force between alum flake particles;

[0460] r i and r j are the equivalent radii of alum floc particles i and j respectively;

[0461] h: the equivalent distance between the surfaces of adjacent alum floc particles, obtained by the spatial positions of adjacent alum flocs;

[0462] A: Hamaker constant:

[0463]

[0464] A0: Hamaker constant in pure solvent, obtained experimentally;

[0465] α: Ion shielding coefficient, which can be obtained by empirical analysis, table lookup, or experimental calibration;

[0466] I(t): ionic strength (unit: mol / m 3 ), which is obtained by online ion selective electrode acquisition and conversion; it can also be obtained by accumulation.

[0467] Specifically, the online ion selective electrode acquisition conversion method is:

[0468] I(t)=I0+β*Q(t);

[0469] Where: β: drug addition-ion strength conversion coefficient, look up the table or calibrate experimentally;

[0470] Q(t): real-time dosage;

[0471] I0: The ionic strength at the previous moment, or the initial ionic strength.

[0472] Preferably, deriving the zeta potential comprises the following steps:

[0473] 1) Assumption: The double layer effect can be described by the Debye-Hückel approximation (low surface potential condition). Based on the DLVO theory, the double layer force formula is:

[0474]

[0475] Among them: F electric : Double layer potential force;

[0476] ε: vacuum dielectric constant;

[0477] k D : Debye reciprocal length (unit: m -1 );

[0478] Zeta potential;

[0479] k1: Equivalent conversion empirical constant;

[0480] h: the equivalent distance between the surfaces of adjacent alum floc particles, obtained by the spatial positions of adjacent alum flocs;

[0481] e v : elementary charge;

[0482] N A : Avogadro's constant;

[0483] k B : Boltzmann constant;

[0484] T: absolute temperature (unit: K).

[0485] Specifically, the conversion method for ionic strength is:

[0486] I(t)=∑ i C i Z i ;

[0487] Where: C i : Electrode detection of cation i molar concentration;

[0488] Z i : The charge number corresponding to cation i.

[0489] 2) Derivation of surface potential through force balance relationship:

[0490] F vdW +F electric =0;

[0491] 3) Substituting into the double layer potential formula, the Zeta potential can be derived.

[0492] The innovative core of this invention is reflected in three dimensions:

[0493] First, through the explicit sensitivity analysis mechanism, a dynamic quantitative model of the dosage and potential change is established using the chain rule, achieving a significant improvement in control accuracy.

[0494] Secondly, a modular matrix architecture is developed to enable online adaptive updating of the Hessian matrix and the gradient matrix, which can dynamically respond to process fluctuations of time-varying parameters;

[0495] Finally, a hard-constrained safety boundary is implanted, and by constructing a multi-dimensional constraint matrix, the risk of over-dosing and concentration exceeding the limit are effectively prevented to ensure the safe operation of the process.

[0496] The present invention realizes closed-loop feedback mapping from objective function to control instructions through matrix operations, providing a quantifiable and verifiable intelligent optimization engine for the water treatment system, significantly improving the system's adaptive control capabilities while ensuring process stability.

[0497] Example 4:

[0498] A water treatment dosing system based on optical detection, such as Figure 8 As shown, it includes a main control unit, a light detector, a light source, a signal acquisition unit, a signal processing unit, and a light source driving unit. The light detector is connected to the signal acquisition unit, and the signal acquisition unit is connected to the isolation unit and the active unit through the signal processing unit. The main control unit is connected to the light source through the light source driving unit.

[0499] The main control unit includes an optical monitoring module, a data processing module and a dosing control module;

[0500] The optical monitoring module is used to monitor the morphology of alum flowers in real time based on the above-mentioned alum flower morphology monitoring method, and obtain accurate equivalent parameters of the alum flower morphology;

[0501] The data processing module is used to derive the Zeta potential based on the equivalent parameters of the alum flower morphology, and to solve the matrix u of the optimal dosage through the set objective function and constraint conditions;

[0502] The dosing control module is used to control the dosing amount of each step according to the matrix u of the optimal dosing amount.

[0503] Preferably, the system further comprises a communication unit, a data storage unit, a motor drive unit and a motor respectively connected to the main control unit, wherein the motor drive unit is used to drive the motor to drive the light detector to scan and measure in three-dimensional space. The main control unit is connected to the communication device via the communication unit.

[0504] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any simple modification or equivalent change made to the above embodiment based on the technical essence of the present invention shall fall within the scope of protection of the present invention.

Claims

1. A method for monitoring alum flower morphology based on optical detection, characterized in that: The following steps are involved: Step T1: Constructing a monochromatic light source alum flower absorbance detection model; Step T2: Scan the three-dimensional space to be measured and obtain absorbance scanning data of the xoz plane and the yoz plane; Step T3: performing edge detection on the absorbance scanning data of the xoz plane and the yoz plane respectively to obtain the inflection point of the alum flower shape and record the position information of the alum flower entity boundary; Step T31: Derivative analysis and dynamic threshold filtering are used to distinguish the true alum flower signal from the ambient noise on the absorbance scan data of the xoz plane and the yoz plane respectively; the projections of overlapping structures are separated by analyzing the spatial continuity of the edge points of adjacent z layers; Step T32: Based on the alum flower position data set of the entity in the xoz plane and the yoz plane, edge data adaptation is performed to achieve three-dimensional positioning and quantity statistics; Step T4: Determine the wavelength of the light source for multi-band measurement based on the alum flower structure, and obtain alum flower equipotential surface dataset corresponding to different light source wavelengths; Step T5: Based on the equipotential surface datasets of the alum flower corresponding to different light source wavelengths, the metal ion hydrate polymerized in the alum flower core layer is taken as the center, and the epitaxial extension layer is used as the segmentation carrier. The alum flower dataset is screened and determined as follows: S floc (x,y,z)={p1(Am s ),p2(Am s ),…,p q (Am s )}, Where: p q (Am s ) is the spatial position information of the qth alum flower geometric body screened; q is the number of alum flowers screened; Step T6: determining equivalent parameters of alum flower morphology based on the alum flower data set; If the alum flower is a spherical model, the equivalent diameter of the alum flower is: If the alum flower is an ellipsoid model, the equivalent diameter of the alum flower is: Where: x L is the length of the alum flower geometry in the x direction; y L is the length of the alum flower geometry in the y direction; z L is the length of the alum flower geometry in the z direction; The volume density of alum flowers is: Where: V i is the volume of the i-th alum flower; Dt x is the total x-axis length of the scan measurement; Dt y is the total y-axis length of the scan measurement; Dt z The total z-axis length of the scan.

2. A method for monitoring alum flower morphology based on optical detection according to claim 1, characterized in that, Step T31 includes the following steps: Step T311: Calculate the average single-dimensional absorbance of the xoz plane and the yoz plane when the number of subdivision steps of the z-axis is n; Step T312: Calculate the single-dimensional fast overlay screening values ​​of the xoz plane and the yoz plane respectively: in: is the average absorbance value in one dimension of the xoz plane when the number of subdivision steps on the z-axis is n; The average absorbance value in one dimension of the yoz plane when the number of subdivision steps of the z-axis is n; olpA n-zx It is the single-dimensional fast overlay screening value of the xoz plane when the number of subdivision steps of the z axis is n; olpA n-zy The single-dimensional fast overlay screening value of the yoz plane when the number of subdivision steps of the z axis is n; k olp is the precision coefficient of overlapping screening; Step T313: Traverse the single-dimensional data of the xoz plane and the yoz plane to screen the absorbance extreme value points of the alum flowers: (|Pd1(r)|<ε0∨Pd1(r)=0)∧A(r)>A uv , Where: A(r) is the absorbance corresponding to the spatial position r; ε0 is the zero-point noise threshold of discrete derivative; A uv is the absorbance calibration value; Pd1(r) is the first-order partial derivative value of the absorbance at the spatial position r; ∧ is the logical operator of AND; ∨ is the logical operator of OR; < is the less than operator; > is the greater than operator; Step T314: If the second-order partial derivative value of the absorbance of the extreme point r, Pd2(r)>0, the current extreme point is located at the edge, and the extreme point located at the edge is screened and the process proceeds to step T315; Step T315: If the absorbance of the current extreme point is greater than the ghost screening value, determine whether there are ghosts on both sides of it; if the third-order partial derivative values ​​of the absorbance of the extreme points on both sides of the current extreme point exceed the third-order threshold range, determine that ghosts exist and proceed to step T316; Step T316: Perform overlapping image segmentation: If the absolute value of the difference between the overlapping absorbance gradient values ​​of point rn and point r is less than or equal to the overlapping image judgment accuracy τ d , then the marked points rn and r are two boundaries of the same entity; Step T317: Repeat steps T311 to T316 to traverse each z-axis subdivided single-dimensional data set; record edge data and overlap marks; Step T318: vectorize the edge data and filter the data based on the convergence condition and the closure condition; then, based on the left and right boundary limitations, record the alum flower projection clusters of the xoZ plane and the yoz plane.

3. A method for monitoring alum flower morphology based on optical detection according to claim 1 or 2, characterized in that, The step T32 includes the following steps: Step T321: traverse the entity's flower position data set on the xoZ plane from the z-axis subdivision step number, and adapt it to the data on the yoz plane; Step T322: extract the z-axis range of the entity; match the z-axis range entity corresponding to the yoz plane; synthesize the alum flower entity cluster and mark the usage data of the yoz plane; The data sets filtered out by the xoz plane and the yoz plane are: S x ″(x 1 ,x 2 )={s x1 ″(x 1 ,x 2 ),s x2 ″(x 1 ,x 2 ),…,s xn ″(x 1 ,x 2 )}, S y ″(x 1 ,x 2 )={s y1 ″(x 1 ,x 2 ),s y2 ″(x 1 ,x 2 ),…,s yn ″(x 1 ,x 2 )}, If S x ″(x 1 , x 2 ) and any element in S y ″(x 1 , x 2 ) elements, respectively, perform subdivision steps on the z-axis to match, and the number of elements can also be matched, then the coordinates of the two elements are merged; Step T323: If there is an unused data set in the yoz plane, traverse the unused flower position data set in the yoz plane; Step T324: Extract the z-axis range of the entity, match the z-axis range entity corresponding to the xoz plane, and synthesize the alum flower entity cluster.

4. A method for monitoring alum flower morphology based on optical detection according to claim 1, characterized in that, In step T4, the alum flower structure includes a core layer, a hydrophilic colloid coating layer, a microorganism aggregation layer, a suspended matter adsorption layer and an epitaxial extension layer, which are arranged in sequence from the inside to the outside; the detection light source wavelength of the core layer is 380-450nm, the detection light source wavelength of the hydrophilic colloid coating layer is 100-280nm and 280-315nm, the detection light source wavelength of the microorganism aggregation layer is 100-280nm and 280-315nm, the detection light source wavelength of the suspended matter adsorption layer is 450-495nm, and the detection light source wavelength of the epitaxial extension layer is 450-495nm and the near-infrared light band.

5. A method for monitoring alum flower morphology based on optical detection according to claim 4, characterized in that, The wavelength of the light source for multi-band measurement includes 400nm, and the alum flower equipotential surface dataset includes the core layer dataset. in, It is the spatial position information of the nth alum flower geometric body screened under the wavelength of 400nm light source.

6. A method for monitoring alum flower morphology based on optical detection according to claim 5, characterized in that, The step T5 comprises the following steps: Step T51: traverse the core layer dataset S 400 (x, y, z), if and If there are two independent extension layer entities, then go to step T52; otherwise, go to step T54; Step T52: If and If there are two independent extension layer entities, then go to step T55; otherwise, go to step T53; Step T53: If and If there are two independent suspended matter adsorption layers and epitaxial extension layers, then go to step T55; otherwise, discard the data; Step T54: If and There are two independent suspended matter adsorption layer and epitaxial expansion layer entities, then if and If there are two independent extension layer entities, proceed to step T55; otherwise, discard the data; Step T55: Record all data connected to the spatial extent of the two extension layers as equivalent data of a single flower, forming a data set of the flower; in: is the spatial position information of the i-th alum flower geometric body screened under the wavelength of 400nm light source; The spatial position information of the i+1th alum flower geometric body screened under the wavelength of 400nm light source; It is the spatial position information of the i-1th alum flower geometric body screened under the wavelength of 400nm light source.

7. A water treatment dosing method based on optical detection, implemented based on the alum flower morphology monitoring method according to any one of claims 1 to 6, characterized in that: The following steps are involved: Step S1: real-time monitoring of the morphology of alum flowers to obtain equivalent parameters of the alum flower morphology; Step S2: Derivation of the Zeta potential based on the equivalent parameters of the alum flower morphology, and dynamic control of the dosage based on the metal ion mass balance and potential feedback.

8. The water treatment dosing method based on optical detection according to claim 7, characterized in that: The step S2 comprises the following steps: Step S21: setting constraints; Power constraints: Concentration boundary constraints: Constraints on dosage Q(t): 0≤Q(t)≤Q max , in: is the average concentration of metal ions; C Al，max It is the limit value of metal ions in sewage treatment environment; Q max is the maximum dosage; k in Enter the rate constant for drug addition; k loss is the nonspecific loss rate; k out is the consumption rate constant, k out =k0ρ(t)(d(q,t)) 2 ; k0 is the reaction activity coefficient; ρ(t) is the number density of alum flowers at time t; d(q, t) is the average equivalent diameter corresponding to time t; Step S22: Set the objective function as: p Within, the Zeta potential tracking error and the dosage cost model value J are minimized; Where: k is the number of steps; N is the number of prediction steps; is the average Zeta potential calculated based on the equivalent parameters of alum flower morphology in step k; is the target zeta potential set; w is the dosage weight; Q k is the dosage of the kth step; △t is the time step; Step S23: Solve to obtain the matrix u of the optimal dosage: u=[Q0 Q1 … Q N-1 ] T , Where: Q N-1 is the optimal dosage for step N; The dosage is dynamically controlled based on the optimal dosage matrix u.

9. A water treatment dosing system based on optical detection, used to implement the water treatment dosing method based on optical detection according to claim 7 or 8, characterized in that: It includes optical monitoring module, data processing module and dosing control module; The optical monitoring module is used to monitor the morphology of alum flowers in real time based on the alum flower morphology monitoring method according to any one of claims 1 to 6, and obtain accurate equivalent parameters of the alum flower morphology; The data processing module is used to derive the Zeta potential based on the equivalent parameters of the alum flower morphology, and to solve the matrix u of the optimal dosage through the set objective function and constraint conditions; The dosing control module is used to control the dosing amount of each step according to the matrix u of the optimal dosing amount.

Images