Intelligent grading fineness regulation and control method and system based on mathematical model

Abstract

The invention relates to the technical field of hierarchical control, and discloses a mathematical model-based hierarchical fineness intelligent regulation and control method and system, and the method comprises the steps: obtaining an initial hierarchical scheme, analyzing a first structure and a working condition parameter set, and calibrating the first structure parameter set to obtain a second structure parameter set; constructing a distributed particle size analysis network to obtain a second particle size feature set, measuring multi-source data to obtain a second working condition parameter set, evaluating grading fluctuation, and optimizing parameters to obtain an optimized parameter set, so as to realize accurate control on the grading process of the cyclone; the method can effectively cope with fluctuation in the cyclone classification process, the classification precision is improved, the stability and efficiency of a classification system are improved, resource waste is reduced, and the production cost is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of hierarchical control, and more specifically, to an intelligent regulation method and system for hierarchical fineness based on a mathematical model. Background Art

[0002] In the field of industrial production, especially in the process of material classification, precise control of the classification fineness is crucial for improving product quality, enhancing production efficiency, and reducing production costs. As a commonly used classification device, the classification effect of a hydrocyclone is affected by various factors. How to achieve precise control of the classification fineness of a hydrocyclone has always been the focus and difficulty in the industry.

[0003] The Chinese patent application with the publication number CN101320257A discloses an intelligent coordinated control system for ball milling classification in a concentrator, proposing to integrate multiple decentralized control loops of ball milling classification into an intelligent coordinated control system based on PLC programming and upper computer configuration control, and realizing the intelligent coordinated control of the ball milling classification system by recording process variables and selecting the optimal process function. However, this system mainly focuses on converting expert experience into computer control, and has limitations in dealing with the complex and changeable working conditions in the classification process of the hydrocyclone. There is a lack of effective means for real-time dynamic optimization of the structural parameters and working condition parameters of the hydrocyclone. In actual production, the structural parameters of the hydrocyclone (such as the cone angle ratio, overflow pipe diameter ratio) will affect the classification effect due to factors such as equipment wear and changes in material properties, but this system cannot adjust the parameters in a timely manner according to these changes, making it difficult to ensure the stability and accuracy of the classification fineness.

[0004] The Chinese patent with the authorization announcement number CN101950171B discloses a method and control equipment for intelligent classification control of grinding in a concentrator, giving a set of control processes covering multiple links such as feed amount, feed water, discharge water, and pulp concentration at the hydrocyclone inlet, with advantages such as high control accuracy and fast adjustment speed. However, this technical solution has deficiencies in the evaluation of classification fluctuations and in dealing with complex fluctuation scenarios. It does not fully consider the correlation of particle size characteristics between different hydrocyclones and various fluctuations in the classification process. When complex fluctuations such as cross-section particle size mutual feedback and coupling instability of feed pressure and flow rate occur, it is difficult to quickly make accurate judgments and carry out targeted optimizations, and it cannot effectively suppress the impact of fluctuations on the classification fineness, thereby affecting product quality and production efficiency.

[0005] In summary, the existing hierarchical control technologies have obvious deficiencies in the control of the classification fineness of the hydrocyclone, and cannot meet the strict requirements for classification accuracy, system stability, and dealing with complex working conditions in industrial production. Summary of the Invention

[0006] To overcome the above defects of the prior art, the present invention provides a method and system for intelligent regulation of classification fineness based on a mathematical model. By obtaining the parameter sets related to the analysis of the initial classification scheme, adaptively calibrating the first set of structural parameters, constructing a network to obtain the particle size feature set, measuring multi-source data to obtain the operating condition parameter set, evaluating the fluctuations and optimizing the parameters, precise control of the cyclone classification process is achieved, effectively suppressing the classification fluctuations, greatly improving the classification accuracy, ensuring that the product particle size meets the production requirements, and improving the overall efficiency and stability of the classification system.

[0007] The present invention is widely applicable to various industrial scenarios involving cyclone classification, such as the mining and beneficiation industry, building materials processing industry, chemical raw material processing industry, etc. In the process of mining and beneficiation, it is necessary to finely classify the ore to separate useful minerals of different particle sizes; during the processing of building materials, the classification accuracy of raw materials such as sand and gravel affects the quality and performance of the products; in the chemical field, the classification treatment of raw materials is also related to the effects of subsequent chemical reactions and product quality. In these scenarios, the present invention can play a key role in ensuring the efficient and stable operation of the classification process.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A method for intelligent regulation of classification fineness based on a mathematical model, including:

[0010] Obtain an initial classification scheme, and parse the first set of structural parameters and the first set of operating condition parameters from the initial classification scheme; based on the first set of structural parameters and the first set of operating condition parameters, adaptively calibrate the first set of structural parameters to generate a second set of structural parameters; construct a distributed particle size analysis network, and obtain a second particle size feature set through the distributed particle size analysis network; measure multi-source data during the operation of the n-stage cyclone in real time, and obtain a second set of operating condition parameters according to the multi-source data; evaluate the classification fluctuations based on the second particle size feature set and the second set of operating condition parameters to obtain a classification fluctuation evaluation result;

[0011] According to the classification fluctuation evaluation result, optimize the first set of operating condition parameters and the second set of structural parameters to obtain an optimized parameter set, where the optimized parameter set includes a first optimized parameter set, a second optimized parameter set, and a third optimized parameter set; precisely control the cyclone classification process according to the optimized parameter set.

[0012] Further, the first set of structural parameters includes the cone angle ratio sequence [α'1, α'2,..., α' n and the overflow pipe diameter ratio sequence [D'1, D'2,..., D' n ; where α' i is the cone angle ratio of the i-th stage cyclone, specifically the ratio of the cone angle α i of the i-th stage cyclone to the reference cone angle α0, and D' iis the overflow pipe diameter ratio of the i-th stage cyclone, specifically the diameter D of the overflow pipe of the i-th stage cyclone i The ratio to the feed inlet diameter D0, where 1 ≤ i ≤ n; the first set of operating condition parameters includes the reference value C0 of the feed concentration, the reference value ΔP0 of the pressure gradient, and the reference value Q' of the flow rate ratio.

[0013] Furthermore, the adaptive calibration of the first set of structural parameters to generate the second set of structural parameters includes:

[0014] Based on the first set of structural parameters and the first set of operating condition parameters, obtain the structural error compensation coefficient matrix k; based on the structural error compensation coefficient matrix k, perform adaptive calibration on the first set of structural parameters to generate the second set of structural parameters, and perform iterative optimization on the second set of structural parameters.

[0015] Furthermore, obtaining the structural error compensation coefficient matrix k includes:

[0016] Under standard material conditions, with the cone angle ratio sequence [α'1, α'2,..., α' n and the overflow pipe diameter ratio sequence [D'1, D'2,..., D' n as the reference, under the operating conditions of the first set of operating condition parameters, collect the overflow particle size distribution curve and the underflow concentration curve of the n-stage cyclone, and extract the actual overflow particle size median sequence and the actual underflow concentration sequence ; where is the actual overflow particle size median of the -th stage cyclone, is the actual underflow concentration of the i-th stage cyclone;

[0017] Obtain the theoretical overflow particle size median sequence of the n-stage cyclone, where is the theoretical overflow particle size median of the -th stage cyclone; according to the theoretical overflow particle size median sequence and the actual overflow particle size median sequence , calculate the overflow particle size median deviation sequence of the n-stage cyclone, where is the overflow particle size median deviation of the i-th stage cyclone;

[0018] Calculate the standard deviation sequence of the actual underflow concentration of the n-stage cyclone, and calculate the underflow concentration fluctuation coefficient sequence of the n-stage cyclone according to the standard deviation sequence and the actual underflow concentration sequence , where is the underflow concentration fluctuation coefficient of the i-th stage cyclone;

[0019] Construct a structural error compensation coefficient matrix based on the overflow particle size median deviation sequence and the underflow concentration fluctuation coefficient sequence , where is the structural error compensation coefficient of the i-th cyclone stage

[0020] Furthermore, the adaptive calibration of the first set of structural parameters based on the structural error compensation coefficient matrix k to generate a second set of structural parameters and the iterative optimization of the second set of structural parameters include

[0021] Step S1131, calibrate the cone angle ratio sequence [α'1, α'2,..., α' n in the first set of structural parameters according to the structural error compensation coefficient matrix k to obtain the calibrated cone angle ratio sequence [α''1, α''2,..., α'' n ;

[0022] Step S1132, calibrate the overflow pipe diameter ratio sequence [D'1, D'2,..., D' n in the first set of structural parameters according to the structural error compensation coefficient matrix k to obtain the calibrated overflow pipe diameter ratio sequence [D''1, D''2,..., D'' n ;

[0023] Step S1133, generate a second set of structural parameters based on the calibrated cone angle ratio sequence [α''1, α''2,..., α'' n and the calibrated overflow pipe diameter ratio sequence [D''1, D''2,..., D'' n ;

[0024] Step S1134, recalculate the overflow particle size median deviation sequence and the underflow concentration fluctuation coefficient sequence under the second set of structural parameters according to the overflow particle size distribution curve and the underflow concentration curve, where is the overflow particle size median deviation of the i-th cyclone stage under the second set of structural parameters, is the underflow concentration fluctuation coefficient of the i-th cyclone stage under the second set of structural parameters;

[0025] Step S1135, preset the deviation convergence accuracy and the fluctuation convergence accuracy , traverse the overflow particle size median deviation sequence and the underflow concentration fluctuation coefficient sequence under the second set of structural parameters, and judge whether it satisfies and . If not, return to Step S1131 for repeated iteration until it is satisfied

[0026] Further, the construction of the distributed particle size analysis network and the acquisition of the second particle size feature set through the distributed particle size analysis network include:

[0027] Deploy on-line particle size analyzers at the overflow and underflow outlets of the n-stage cyclone to form a distributed particle size analysis network;

[0028] Collect the particle size distribution data of the overflow and underflow of the n-stage cyclone in real time through the distributed particle size analysis network to obtain the second particle size feature set; the second particle size feature set includes the overflow particle size skewness index sequence S = [S1, S2,..., S n , the underflow particle size kurtosis index sequence K = [K1, K2,..., K n , and the cross-stage particle size covariance matrix ; where S i is the overflow particle size skewness index of the i-th stage cyclone, and K i is the underflow particle size kurtosis index of the i-th stage cyclone.

[0029] Further, the real-time measurement of multi-source data during the operation of the n-stage cyclone and the obtaining of the second operating condition parameter set according to the multi-source data include:

[0030] Deploy pressure sensors, flow meters and concentration meters at the corresponding positions of the n-stage cyclone to measure the multi-source data during the operation of the n-stage cyclone in real time. The multi-source data includes the feed pressure, ore volume flow and feed concentration;

[0031] Calculate the dynamic feed concentration gradient according to the feed concentration ;

[0032] Calculate the pressure pulsation amplitude according to the feed pressure ;

[0033] Calculate the flow variation coefficient according to the ore volume flow .

[0034] Further, the classification fluctuation evaluation results include primary fluctuation, secondary fluctuation and tertiary fluctuation;

[0035] The obtaining of the classification fluctuation evaluation results includes:

[0036] Based on the cross-stage particle size covariance matrix in the second particle size feature set judge whether it is a primary fluctuation;

[0037] Based on the second operating condition parameter set, judge whether it is a secondary fluctuation;

[0038] Based on the overflow particle size skewness index sequence S in the second particle size feature set, judge whether it is a tertiary fluctuation.

[0039] Further, the cross-segment granularity covariance matrix based on the second granularity feature set Determining whether it is a first-level fluctuation includes: calculating the cross-segment granularity covariance matrix in the second granularity feature set The maximum eigenvalue λ max of it. If λ max > λ α , it is a first-level fluctuation, where λ α is the preset first-level fluctuation threshold.

[0040] Further, determining whether it is a second-level fluctuation based on the second operating parameter set includes:

[0041] Presetting a second-level fluctuation threshold λ β . If the product of the pressure pulsation amplitude and the flow coefficient of variation exceeds λ β , it is a second-level fluctuation.

[0042] Further, determining whether it is a third-level fluctuation based on the overflow granularity skewness index sequence S in the second granularity feature set includes:

[0043] Setting the normal range [L0, H0] of the overflow granularity skewness index. If S exceeds the normal range [L0, H0] continuously for λ N times, it is a third-level fluctuation, where L0 is the lower limit of the normal range of the overflow granularity skewness index, H0 is the upper limit of the normal range of the overflow granularity skewness index, and λ N is the preset third-level fluctuation threshold.

[0044] Further, optimizing the first operating parameter set and the second structural parameter set to obtain an optimized parameter set includes:

[0045] If the hierarchical fluctuation evaluation result is a first-level fluctuation, perform transfer learning optimization on the first operating parameter set and the second structural parameter set to obtain a first optimized parameter set;

[0046] If the hierarchical fluctuation evaluation result is a second-level fluctuation, optimize the first operating parameter set to obtain a second optimized parameter set;

[0047] If the hierarchical fluctuation evaluation result is a third-level fluctuation, optimize the first operating parameter set to obtain a third optimized parameter set.

[0048] Further, obtaining the third optimized parameter set includes:

[0049] Calculating the cross-segment granularity equilibrium factor ξ according to the second granularity feature set;

[0050] Based on the cross-segment granularity equilibrium factor ξ, calculating the correction weight of the cone angle ratio of the i-th cyclone section;

[0051] According to the correction weight calculate the correction cone angle ratio of the n-th stage cyclone ;

[0052] From the correction cone angle ratio of the n-th stage cyclone obtain the correction cone angle ratio sequence ;

[0053] According to the calibrated overflow pipe diameter ratio sequence [D''1, D''2,..., D'' n , the correction cone angle ratio sequence , the pressure gradient reference value ΔP0, the ore concentration reference value C0, and the flow ratio reference value Q', construct the third optimization parameter set.

[0054] The intelligent regulation system for classification fineness based on a mathematical model, which is used to implement the above-mentioned intelligent regulation method for classification fineness based on a mathematical model, the system includes:

[0055] Structure parameter correction module: Obtain the initial classification scheme, and parse the first structure parameter set and the first operating condition parameter set from the initial classification scheme; Based on the first structure parameter set and the first operating condition parameter set, perform adaptive calibration on the first structure parameter set to generate the second structure parameter set;

[0056] Fluctuation evaluation module: Construct a distributed particle size analysis network, and obtain the second particle size feature set through the distributed particle size analysis network; Real-time measure the multi-source data during the operation of the n-stage cyclone, and obtain the second operating condition parameter set according to the multi-source data; Based on the second particle size feature set and the second operating condition parameter set, perform classification fluctuation evaluation to obtain the classification fluctuation evaluation result;

[0057] Classification control module: According to the classification fluctuation evaluation result, optimize the first operating condition parameter set and the second structure parameter set to obtain the optimization parameter set, the optimization parameter set includes the first optimization parameter set, the second optimization parameter set, and the third optimization parameter set; Accurately control the cyclone classification process according to the optimization parameter set.

[0058] Compared with the prior art, the beneficial effects of the present invention are:

[0059] The present invention realizes precise control of the cyclone classification process by obtaining the relevant parameter set for parsing the initial classification scheme, adaptively calibrating the first set of structural parameters, constructing a network to obtain the particle size feature set, measuring multi-source data to obtain the operating condition parameter set, evaluating the fluctuations and optimizing the parameters. This enables the cyclone to operate stably under different operating conditions, effectively suppresses classification fluctuations, greatly improves the classification accuracy, ensures that the particle size of the product meets the production requirements, reduces the output of unqualified products; at the same time, it improves the overall efficiency of the classification system, reduces energy consumption and equipment losses, reduces resource waste and production costs, enhances the reliability and controllability of the production process, and improves the economic benefits and competitiveness of the enterprise. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0061] Figure 1 It is the principle flow chart of the intelligent regulation method for classification fineness based on a mathematical model in the present invention;

[0062] Figure 2 It is the method flow chart for parsing the first set of structural parameters and the first set of operating condition parameters from the initial classification scheme in the intelligent regulation method for classification fineness based on a mathematical model of the present invention;

[0063] Figure 3 It is the method flow chart for constructing a distributed particle size analysis network and obtaining a second particle size feature set through the distributed particle size analysis network in the intelligent regulation method for classification fineness based on a mathematical model of the present invention;

[0064] Figure 4 It is the method flow chart for obtaining a second set of operating condition parameters based on multi-source data in the intelligent regulation method for classification fineness based on a mathematical model of the present invention;

[0065] Figure 5 It is the method flow chart for evaluating classification fluctuations based on the second particle size feature set and the second set of operating condition parameters to obtain the classification fluctuation evaluation result in the intelligent regulation method for classification fineness based on a mathematical model of the present invention;

[0066] Figure 6 It is the method flow chart for obtaining the first set of optimized parameters in the intelligent regulation method for classification fineness based on a mathematical model of the present invention;

[0067] Figure 7 It is the method flow chart for obtaining the second set of optimized parameters in the intelligent regulation method for classification fineness based on a mathematical model of the present invention;

[0068] Figure 8 This is a functional module diagram of the intelligent regulation system for classification fineness based on a mathematical model in the present invention. Specific implementation manners

[0069] Next, with reference to the accompanying drawings in the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0070] Embodiment 1

[0071] Please refer to Figure 1 As shown, this embodiment provides an intelligent regulation method for classification fineness based on a mathematical model, including:

[0072] Step S1000, obtaining an initial classification scheme, parsing a first set of structural parameters and a first set of operating parameters from the initial classification scheme; based on the first set of structural parameters and the first set of operating parameters, performing adaptive calibration on the first set of structural parameters to generate a second set of structural parameters; constructing a distributed particle size analysis network, and obtaining a second set of particle size characteristics through the distributed particle size analysis network; measuring multi-source data during the operation of the n-stage cyclone in real time, and obtaining a second set of operating parameters according to the multi-source data; performing classification fluctuation evaluation based on the second set of particle size characteristics and the second set of operating parameters to obtain a classification fluctuation evaluation result;

[0073] Further, step S1000 includes:

[0074] Step S1100, obtaining an initial classification scheme, parsing a first set of structural parameters and a first set of operating parameters from the initial classification scheme; based on the first set of structural parameters and the first set of operating parameters, performing adaptive calibration on the first set of structural parameters to generate a second set of structural parameters;

[0075] Further, step S1100 includes:

[0076] Step S1110, obtaining an initial classification scheme, parsing a first set of structural parameters and a first set of operating parameters from the initial classification scheme;

[0077] Further, as Figure 2 shown, step S1110 includes:

[0078] Step S1111, the first set of structural parameters includes the cone angle ratio sequence [α'1, α'2,..., α' n and the overflow pipe diameter ratio sequence [D'1, D'2,..., D'n ]; where α' i is the cone angle ratio of the i-th cyclone, specifically the cone angle α of the i-th cyclone i Ratio to the reference cone angle α0, D' i is the overflow pipe diameter ratio of the i-th cyclone, specifically the overflow pipe diameter D of the i-th cyclone i The ratio of D0 to the feed opening diameter is 1≤i≤n;

[0079] Step S1112: the first operating condition parameter set includes a feed concentration reference value C0, a pressure gradient reference value ΔP0, and a flow ratio reference value Q'.

[0080] Further, step S1112 includes:

[0081] Step S11121, the feed concentration reference value C0 is dynamically calibrated according to the material density distribution histogram;

[0082] Step S11122, the pressure gradient reference value ΔP0 is calculated by a Reynolds number correction formula;

[0083] Step S11123, the flow ratio reference value Q'=Q0 / Q1, wherein Q0 is the design flow, and Q1 is the actual flow; the flow ratio reference value Q' is dynamically adjusted through the Bernoulli equation.

[0084] Specifically, the initial classification scheme is the starting setting for the operation of the entire classification system, which takes into account a variety of factors to determine the parameters. For example, when processing ore from a specific mine, the particle size distribution of the ore in the mine presents a bimodal characteristic, that is, there are more coarse particles and fine particles, and fewer particles of intermediate size. In view of this situation, when designing the initial classification scheme, it is necessary to consider the effective separation of coarse particles and fine particles. For coarse particles, a cyclone with a larger cone angle may be required to throw them to the bottom flow more quickly under the action of centrifugal force; for fine particles, a suitable overflow pipe diameter ratio is required to ensure that they can be discharged smoothly through the overflow. This reflects the impact of material characteristic factors on the initial classification scheme.

[0085] If the expected classification fineness requirement is to obtain a very fine concentrate product for the preparation of high-end materials. Then in the initial classification scheme, a smaller overflow pipe diameter ratio tends to be set. Because a smaller overflow pipe diameter ratio allows fine particles to pass through the overflow more easily, thus improving the fineness of the concentrate. At the same time, in order to ensure that the particles have sufficient dispersion for classification, the feed concentration reference value will also be appropriately reduced. For example, if the original feed concentration is 30%, in order to achieve a finer classification goal, the feed concentration may be adjusted to about 25%. Equipment characteristic factors also affect the initial classification scheme. Cyclones of different models and specifications have different structural parameters and different classification performances. For example, the cone angle of a certain model of cyclone is designed at a specific angle, and at this cone angle, its classification efficiency is optimal when processing a certain ore. Then when setting the cone angle ratio sequence in the initial classification scheme, this characteristic will be referred to in order to fully utilize the performance advantages of the cyclone.

[0086] In sub-step S1111, the first set of structural parameters includes the cone angle ratio sequence [α'1, α'2,..., α' n and the overflow pipe diameter ratio sequence [D'1, D'2,..., D' n of the n-stage cyclone. The cone angle ratio α' i is the ratio of the cone angle α i of the i-th stage cyclone to the reference cone angle α0, and the reference cone angle α0 is dynamically set according to the material hardness characteristics. This is because materials of different hardnesses require different centrifugal forces to achieve effective separation during classification. For example, for ores with greater hardness, a greater centrifugal force is required during classification, which means a relatively larger cone angle for the cyclone, so the reference cone angle α0 will increase accordingly; while for ores with smaller hardness, a smaller centrifugal force can meet the classification requirements, and the reference cone angle α0 will decrease. By dynamically setting the reference cone angle α0, the cone angle ratio sequence can better adapt to the classification requirements of different materials, improving the classification efficiency and effect. The overflow pipe diameter ratio D' i is the ratio of the overflow pipe diameter D i of the i-th stage cyclone to the feed inlet diameter D0, and D0 is the design reference value. This ratio has an important impact on the classification fineness. A larger overflow pipe diameter ratio means that more particles have the opportunity to pass through the overflow, which is suitable for separating out more fine particles; a smaller overflow pipe diameter ratio is more conducive to obtaining a coarser overflow product. In practical applications, the overflow pipe diameter ratio can be flexibly adjusted according to different classification goals.

[0087] In sub-step S11121, the feed concentration reference value C0 is dynamically calibrated according to the material density distribution histogram. The material density distribution histogram reflects the distribution of particles with different densities in the material. For example, when there are more high-density particles in the material, if the feed concentration is too high, these high-density particles may aggregate with each other, affecting the classification effect. By analyzing the material density distribution histogram, a suitable feed concentration reference value C0 can be determined, so that particles with different densities can be better dispersed during the classification process, improving the classification accuracy. Suppose the density distribution histogram of a certain material shows that high-density particles are mainly concentrated in a certain interval. After analysis, the feed concentration reference value C0 is set to 28%. Compared with the concentration set without considering the density distribution histogram before, in subsequent classification experiments, the purity of the concentrate has increased by 5%, which fully demonstrates the importance and effectiveness of dynamically calibrating the feed concentration reference value according to the material density distribution histogram.

[0088] In sub-step S11122, the pressure gradient reference value ΔP0 is calculated by the Reynolds number correction formula. The Reynolds number is a dimensionless number that reflects the relative magnitude of inertial force and viscous force in fluid flow. During the hydrocyclone classification process, the Reynolds number can help us judge the flow state of the pulp, which in turn affects the calculation of the pressure gradient reference value. For example, when the Reynolds number of the pulp is large, it means that the inertial force dominates and the flow of the pulp is closer to the turbulent state. At this time, a larger pressure gradient is required to ensure the effective classification of the pulp in the hydrocyclone; on the contrary, when the Reynolds number is small, the viscous force dominates and the required pressure gradient is relatively small. Calculating the pressure gradient reference value ΔP0 by the Reynolds number correction formula can make the classification process more in line with the actual flow characteristics of the pulp, improving the stability and reliability of classification.

[0089] In sub-step S11123, the flow rate ratio reference value Q' = Q0 / Q1, where Q0 is the designed flow rate and Q1 is the actual flow rate, and the flow rate ratio reference value Q' is dynamically adjusted by the Bernoulli equation. The Bernoulli equation describes the relationship between pressure, flow velocity and height of an ideal fluid in steady flow. In hydrocyclone classification, the Bernoulli equation can be used to consider the influence of pressure and flow velocity changes of the pulp at different positions on classification. For example, when the actual flow rate Q1 changes, according to the Bernoulli equation, the corresponding pressure change can be calculated, and then the flow rate ratio reference value Q' can be dynamically adjusted. This can ensure that the hydrocyclone can maintain good classification performance under different actual flow rate conditions, improving the adaptability of the classification system to flow rate fluctuations.

[0090] Through the operations of step S1110 and its sub-steps S1111 - S1112, comprehensive and accurate initial classification parameters can be obtained. The reasonable setting and dynamic calibration of these parameters lay a solid foundation for the subsequent adaptive calibration of the first set of structural parameters and the stable and efficient operation of the entire classification system. From the perspective of the operation effect of the overall classification system, accurate initial parameter setting can reduce the frequency and amplitude of subsequent parameter adjustments, and reduce the energy consumption and equipment wear of the system. For example, by accurately setting the feed concentration reference value C0 and the pressure gradient reference value ΔP0, in the actual application of a certain concentrator, the energy consumption of the equipment is reduced by 10%, and the equipment maintenance cycle is extended by 20%, greatly improving the production efficiency and economic benefits. At the same time, this method of dynamically setting and calibrating parameters based on multiple factors also improves the adaptability of the classification system to different materials and production conditions, and enhances the versatility and flexibility of the classification system.

[0091] Step S1120: Obtain the structural error compensation coefficient matrix k according to the first set of structural parameters and the first set of operating condition parameters.

[0092] Furthermore, step S1120 includes:

[0093] Step S1121: Under standard material conditions, use the cone angle ratio sequence [α'1, α'2,..., α' n and the overflow pipe diameter ratio sequence [D'1, D'2,..., D' n as the benchmark to conduct baseline tests on the n-stage cyclone; the standard material is a material with a particle size distribution conforming to a Gaussian distribution.

[0094] Step S1122: Under the operating conditions of the first set of operating condition parameters, collect the overflow particle size distribution curve and the underflow concentration curve of the n-stage cyclone, and extract the actual overflow particle size median sequence and the actual underflow concentration sequence ; where, is the actual overflow particle size median of the -th stage cyclone, is the actual underflow concentration of the i-th stage cyclone.

[0095] Step S1123: Obtain the theoretical overflow particle size median sequence of the n-stage cyclone, where, is the theoretical overflow particle size median of the -th stage cyclone; according to the theoretical overflow particle size median sequence and the actual overflow particle size median sequence , calculate the overflow particle size median deviation sequence of the n-stage cyclone, where, is the overflow particle size median deviation of the i-th stage cyclone.

[0096] Calculation The method for

[0097]

[0098] Step S1124: Calculate the standard deviation sequence of the actual underflow concentration of the n-stage cyclone. Based on the standard deviation sequence and the actual underflow concentration sequence calculate the underflow concentration fluctuation coefficient sequence of the n-stage cyclone , where is the underflow concentration fluctuation coefficient of the i-th stage cyclone;

[0099] Calculation The method for

[0100]

[0101] where is the standard deviation of the actual underflow concentration of the i-th stage cyclone;

[0102] Step S1125: Based on the overflow particle size median deviation sequence and the underflow concentration fluctuation coefficient sequence, construct a structural error compensation coefficient matrix , where is the structural error compensation coefficient of the i-th stage cyclone.

[0103] Calculation The method for

[0104]

[0105] where is a non-linear weight coefficient used to capture the non-linear relationship in error compensation; is a high-order non-linear coefficient used to further capture the complex relationship in error compensation, and are respectively obtained by performing curve fitting on the non-linear relationship between the overflow particle size median deviation measured under standard material conditions and the actual calibration parameters using the least squares method; the fitting data selects error samples under different combinations of structural parameters to ensure that the fitting accuracy meets the mean square error less than 0.01. For example = 0.12, = 0.02. and are non-linear terms used to capture more complex error compensation relationships.

[0106] reflects the structural error degree of the i-th stage cyclone under the reference parameters, The smaller it is, the greater the structural error of the i-th stage cyclone, and greater compensation is required, providing a quantitative basis for subsequent adjustment of structural parameters.

[0107] Specifically, the standard material is the basic material for the experiment. Its particle size distribution conforms to the Gaussian distribution, and its properties such as density and hardness are close to those of the actual application material. Selecting the standard material can ensure that the experimental results are closer to the actual working conditions. Taking the cone angle ratio sequence [α'1, α'2,..., α' n and the overflow pipe diameter ratio sequence [D'1, D'2,..., D' n as the basis for the baseline test of the n-stage cyclone means operating the cyclone according to these baseline parameters under the condition of the standard material, which serves as the basis for subsequent comparative analysis. Through the baseline test, the basic performance data of the cyclone under standard conditions can be obtained, providing a reference for subsequent analysis of the differences under actual working conditions.

[0108] In step S1122, the overflow particle size distribution curve and the underflow concentration curve of the n-stage cyclone are collected under the working conditions of the first set of working parameters. These curves can intuitively reflect the classification effect of the cyclone under the current working conditions. The actual overflow particle size median sequence and the actual underflow concentration sequence are the key data extracted from the collected curves. The actual overflow particle size median represents the median value of the particle size distribution in the overflow of the i-th stage cyclone, which reflects the central tendency of the particle size in the overflow; the actual underflow concentration sequence reflects the concentration of the material in the underflow. The accurate acquisition of these data provides a direct basis for subsequent calculation of the deviation and the fluctuation coefficient.

[0109] The median sequence of the overflow particle size of the n-stage hydrocyclone is calculated based on the theoretical model or empirical formula of the hydrocyclone. These theoretical models or empirical formulas comprehensively consider various factors such as the structural parameters, operating parameters, and material properties of the hydrocyclone. Structural parameters such as the cone angle, overflow pipe diameter, and feed inlet diameter of the hydrocyclone have an important impact on the median of the theoretical overflow particle size. Taking the cone angle as an example, different cone angles will change the flow path and centrifugal force distribution of the pulp in the hydrocyclone. A larger cone angle may increase the axial velocity of the pulp in the hydrocyclone, causing coarser particles to more easily enter the underflow, while a smaller cone angle is beneficial for more fine particles to be discharged from the overflow, thus affecting the median of the theoretical overflow particle size. When calculating the median of the theoretical overflow particle size, these structural parameters will be taken into account, and their effects on particle size classification will be reflected through corresponding mathematical relationships. Operating parameters such as the reference value of the feed concentration, the reference value of the pressure gradient, and the reference value of the flow ratio are also closely related to the median of the theoretical overflow particle size. Characteristics of the material such as the particle size distribution, density, and hardness are factors that cannot be ignored in calculating the median of the theoretical overflow particle size. For materials with a wide particle size distribution, the separation difficulty of particles of different sizes needs to be considered during classification, which will affect the calculation of the median of the theoretical overflow particle size; different material densities result in different centrifugal and gravitational forces acting on them in the hydrocyclone, thus affecting their distribution in the overflow and underflow and having an impact on the median of the theoretical overflow particle size; harder materials may require a higher pressure gradient to ensure the crushing and classification effects, which also indirectly affects the calculation result of the median of the theoretical overflow particle size. During the calculation process, corresponding parameter adjustments and calculations will be carried out for these material characteristics to obtain the median of the theoretical overflow particle size that conforms to the actual situation.

[0110] The deviation sequence of the median of the overflow particle size of the n-stage hydrocyclone can clearly show the degree of difference between the actual classification result and the theoretical expectation. The larger the deviation, the greater the gap between the actual classification effect and the ideal situation, and relevant parameters need to be adjusted and optimized.

[0111] In step S1124, the standard deviation sequence of the actual underflow concentration of the n-stage hydrocyclone is calculated. The standard deviation can measure the degree of dispersion of data. In the underflow concentration data, the larger the standard deviation, the greater the fluctuation of the underflow concentration. The underflow concentration fluctuation coefficient sequence is calculated based on the standard deviation sequence and the actual underflow concentration sequence. The underflow concentration fluctuation coefficient can more intuitively reflect the fluctuation of the underflow concentration. It is a relative index and is not affected by the absolute value of the concentration, which is convenient for comparing the fluctuation degrees of the underflow concentration under different operating conditions.

[0112] Step S1120 can accurately evaluate the difference between the operating state of each section of the hydrocyclone and the ideal state by precisely calculating the structural error compensation coefficient matrix k. Based on this, the structural parameters of the hydrocyclone can be adjusted accordingly to improve the classification accuracy. For example, when the structural error compensation coefficient of a certain section of the hydrocyclone is small, it indicates that the structural error of this section is large. Through subsequent parameter calibration, the cone angle ratio and the overflow pipe diameter ratio can be optimized, making the classification effect of the hydrocyclone closer to the theoretical expectation. At the same time, this precise evaluation and adjustment helps to improve the overall operating efficiency of the hydrocyclone, reduce energy waste, and lower production costs. In the long-term operation, it can extend the service life of the hydrocyclone, enhance the stability and reliability of the entire classification system, and provide a more reliable classification guarantee for industrial production.

[0113] Step S1130: Based on the structural error compensation coefficient matrix k, adaptively calibrate the first set of structural parameters to generate a second set of structural parameters, and iteratively optimize the second set of structural parameters.

[0114] Furthermore, step S1130 includes:

[0115] Step S1131: According to the structural error compensation coefficient matrix , calibrate the cone angle ratio sequence [α'1, α'2,..., α' n in the first set of structural parameters to obtain the calibrated cone angle ratio sequence [α''1, α''2,..., α'' n ;

[0116] The calibration of the cone angle ratio sequence [α'1, α'2,..., α' n in the first set of structural parameters includes:

[0117]

[0118] where α'' i is the cone angle ratio of the i-th section of the calibrated hydrocyclone, is the hyperbolic tangent function, which is used to smooth the correction amplitude of the cone angle ratio and avoid excessive parameter jumps;

[0119] Step S1132: According to the structural error compensation coefficient matrix , calibrate the overflow pipe diameter ratio sequence [D'1, D'2,..., D' n in the first set of structural parameters to obtain the calibrated overflow pipe diameter ratio sequence [D''1, D''2,..., D'' n ;

[0120] The calibration of the overflow pipe diameter ratio sequence [D'1, D'2,..., D' n in the first set of structural parameters includes:

[0121]

[0122] wherein is the overflow pipe diameter ratio of the i-th cyclone after calibration, Item reflects the non-linear correction characteristic of the overflow pipe diameter ratio, that is, when the overflow pipe diameter ratio is larger, the correction amount of the overflow pipe diameter ratio caused by the same overflow particle size median deviation is larger;

[0123] Step S1133, based on the calibrated cone angle ratio sequence [α''1, α''2,..., α'' n and the calibrated overflow pipe diameter ratio sequence [D''1, D''2,..., D'' n , generate the second set of structural parameters;

[0124] Step S1134, according to the overflow particle size distribution curve and the underflow concentration curve, recalculate the overflow particle size median deviation sequence and the underflow concentration fluctuation coefficient sequence under the second set of structural parameters, wherein, is the overflow particle size median deviation of the i-th cyclone under the second set of structural parameters, is the underflow concentration fluctuation coefficient of the i-th cyclone under the second set of structural parameters;

[0125] Step S1135, preset the deviation convergence accuracy and the fluctuation convergence accuracy , traverse the overflow particle size median deviation sequence and the underflow concentration fluctuation coefficient sequence under the second set of structural parameters, and judge whether it satisfies and . If not, return to step S1131 to repeat the iteration until it is satisfied.

[0126] Specifically, in step S1130, the adaptive calibration of the first set of structural parameters based on the structural error compensation coefficient matrix k is the key link of the entire optimization process. The structural error compensation coefficient matrix k is obtained through the previous baseline test, data collection and analysis, and it reflects the structural error degree of each section of the cyclone under the reference parameters. By calibrating the first set of structural parameters, the structural parameters of the cyclone can be adjusted to make it more adaptable to the actual classification working conditions, reduce the classification error caused by unreasonable structural parameters, and thus improve the overall performance of the classification system.

[0127] In sub-step S1131, the cone angle ratio sequence [α'1, α'2,..., α' n in the first set of structural parameters is calibrated, and the formula used is , wherein, is a hyperbolic tangent function; its main function is to smooth the correction amplitude of the cone angle ratio to avoid excessive parameter jumps. In the actual classification process, if the cone angle ratio is adjusted significantly directly according to the error, the classification effect may be unstable or even deteriorated. For example, suppose that when a certain ore concentrator is processing a certain ore, the initial cone angle ratio is set to a certain value, but through the previous test, it is found that there is a large deviation in the median overflow particle size. If the hyperbolic tangent function is not used for smoothing, and the cone angle ratio is adjusted directly according to the error ratio, the cone angle ratio may change too much, resulting in the subsequent classification process. The separation effect of fine particles becomes worse and the concentrate quality decreases. Through the hyperbolic tangent function, the correction amount of the cone angle ratio can be reasonably adjusted according to the size of the median overflow particle size deviation. When the deviation is small, the correction amount is also small; when the deviation is large, the correction amount will increase, but it will not be too large, thereby ensuring the stability of the classification process. This smoothing method makes the adjustment of the cone angle ratio more reasonable and stable, which helps to improve the reliability and stability of the classification system, and thus improve the quality and output of the concentrate.

[0128] Sub-step S1132 calculates the overflow pipe diameter ratio sequence [D'1, D'2, ..., D' n ] for calibration, the calibration formula is , in this formula The item reflects the nonlinear correction characteristics of the overflow diameter ratio. This means that when the overflow diameter ratio is larger, the overflow diameter ratio correction caused by the same overflow particle size median deviation is larger. For example, in a multi-stage cyclone classification system, the overflow diameter of a certain cyclone is relatively large. During the calibration process, if the same overflow particle size median deviation occurs, the overflow diameter ratio correction of this cyclone will be larger than that of the cyclone with a smaller overflow diameter. This nonlinear correction method is designed based on the complex relationship between the overflow diameter ratio and the overflow particle size in the actual classification process. A larger overflow diameter ratio means that more particles have the opportunity to be discharged through the overflow. When the overflow particle size median deviation occurs, a larger correction is made to the cyclone with a larger diameter ratio, which can more effectively adjust the classification effect and improve the classification accuracy. Through this nonlinear correction method, the overflow diameter ratio can be adjusted more accurately, the classification process can be optimized, the adaptability of the classification system to different working conditions can be improved, and the classification effect can be ensured to be more stable and reliable.

[0129] Sub-step S1133 integrates the two calibrated key structural parameter sequences before to form a new set of structural parameters. This second set of structural parameters comprehensively considers the optimization results of the cone angle ratio and the overflow pipe diameter ratio of each section of the hydrocyclone. Compared with the first set of structural parameters, it better meets the actual classification requirements. For example, in a classification system for processing complex ores, through the calibration of the cone angle ratio and the overflow pipe diameter ratio, the newly generated second set of structural parameters can better adapt to the particle size distribution and properties of the ores, enabling more effective separation of particles of different sizes, and improving the classification efficiency and concentrate quality.

[0130] Sub-step S1134 recalculates the overflow particle size median deviation sequence and the underflow concentration fluctuation coefficient sequence under the second set of structural parameters according to the overflow particle size distribution curve and the underflow concentration curve. This is to evaluate the optimization effect of the second set of structural parameters. By comparing with the theoretical values, the new overflow particle size median deviation is calculated, and the fluctuation of the underflow concentration is analyzed to obtain the underflow concentration fluctuation coefficient. These data can intuitively reflect the operation of the classification system under the second set of structural parameters. For example, if the recalculated overflow particle size median deviation is significantly smaller than that under the first set of structural parameters, it indicates that the calibrated structural parameters make the classification result closer to the theoretical expectation, and the classification accuracy has been improved; similarly, if the underflow concentration fluctuation coefficient decreases, it indicates that the classification process is more stable.

[0131] The deviation convergence accuracy in sub-step S1135 is set as the limit value when the particle size deviation does not exceed 5%, and the fluctuation convergence accuracy is set as the underflow concentration fluctuation coefficient not higher than 0.02, which is used as the judgment standard for the stable operation of the system. The iterative optimization method can continuously adjust the structural parameters to gradually improve the performance of the classification system. Taking an actual ore dressing classification project as an example, under the initially set structural parameters, the overflow particle size median deviation of the classification system is large, and the underflow concentration fluctuation is also unstable. After the first iteration of calibrating the structural parameters, the calculated deviation and fluctuation coefficient are improved, but still do not reach the preset accuracy. After multiple iterations, continuously adjusting the cone angle ratio and the overflow pipe diameter ratio, finally the deviation and the fluctuation coefficient meet the preset accuracy requirements. This iterative optimization mechanism can ensure that the classification system can achieve better classification effects under different working conditions, improve the adaptability and reliability of the classification system, ensure that the classification process is more stable and accurate, and thus improve the efficiency and economic benefits of the entire classification system. Through continuous iterative optimization, the classification system can better adapt to different ore properties and production requirements, reduce resource waste, and improve production efficiency.

[0132] Step S1200: Construct a distributed particle size analysis network, and obtain a second particle size feature set through the distributed particle size analysis network; Measure the multi-source data of the n-stage cyclone during operation in real time, and obtain a second set of operating condition parameters based on the multi-source data; Perform hierarchical fluctuation evaluation based on the second particle size feature set and the second set of operating condition parameters to obtain a hierarchical fluctuation evaluation result.

[0133] Further, step S1200 includes:

[0134] Step S1210: Construct a distributed particle size analysis network, and obtain a second particle size feature set through the distributed particle size analysis network;

[0135] Further, as Figure 3 shown, step S1210 includes:

[0136] Step S1211: Deploy on-line particle size analyzers at the overflow and underflow outlets of the n-stage cyclone to form a distributed particle size analysis network;

[0137] Step S1212: Real-time collect the particle size distribution data of the overflow and underflow of the n-stage cyclone through the distributed particle size analysis network to obtain a second particle size feature set; The second particle size feature set includes an overflow particle size skewness index sequence S = [S1, S2,..., S n , an underflow particle size kurtosis index sequence K = [K1, K2,..., K n and a cross-stage particle size covariance matrix ; where S i is the overflow particle size skewness index of the i-th stage cyclone, and K i is the underflow particle size kurtosis index of the i-th stage cyclone.

[0138] Specifically, an on-line particle size analyzer is an instrument that can measure the particle size distribution in real time. It works based on various principles, such as the laser diffraction principle, the image analysis principle, etc. By deploying these instruments at the overflow and underflow outlets, the particle size distribution of the overflow and underflow of the cyclone during operation can be monitored in real time. Connecting multiple on-line particle size analyzers forms a distributed particle size analysis network, which can comprehensively and real-time collect the particle size data of each stage of the cyclone, providing rich and accurate data support for subsequent analysis. For example, in a classification system containing 5-stage cyclones, on-line particle size analyzers are installed at the overflow and underflow outlets of each stage of the cyclone respectively. These instruments can work simultaneously, continuously collect particle size data, and transmit the data to the central processing unit for unified processing and analysis.

[0139] The overflow particle size skewness index , which is used to characterize the asymmetry of the overflow particle size distribution of the i-th stage cyclone. Among them, is the third-order central moment of the overflow particle size distribution of the i-th stage cyclone, which reflects the asymmetry degree of the particle size distribution on both sides of the mean value. If is positive, it indicates that the right side (in the direction of larger particle sizes) of the particle size distribution has a longer tail; if it is negative, the left side (in the direction of smaller particle sizes) has a longer tail; if it is close to 0, the particle size distribution is relatively symmetric. And is the standard deviation of the particle size distribution of the -th stage cyclone; it is used to characterize the degree of dispersion of the data. The larger the standard deviation, the more dispersed the particle size distribution. The kurtosis index of the underflow particle size is used to characterize the sharpness of the underflow particle size distribution of the i-th stage cyclone. is the fourth-order central moment of the underflow particle size distribution of the i-th stage cyclone, which is used to measure the sharpness or flatness of the distribution. When is relatively large, it indicates that the particle size distribution is relatively sharp, that is, most particles are concentrated near a certain specific particle size; when is relatively small, the particle size distribution is relatively flat and the particle distribution is relatively uniform.

[0140] The cross-stage particle size covariance matrix M = cov(S, K) reflects the correlation between the overflow particle size skewness and the underflow particle size kurtosis of different cyclones. Its matrix element , where i and j are the indices of the cyclones; taking i = 1, j = 2 as an example, represents the covariance between the overflow particle size skewness index S1 of the first-stage cyclone and the underflow particle size kurtosis index K2 of the second-stage cyclone. If is positive, it indicates that the change trends of S1 and K2 have a certain co-directionality; if it is negative, the change trends are opposite; if it is close to 0, the correlation between the two is weak. By calculating the cross-stage particle size covariance matrix, the co-variation relationship of the particle size characteristics between different-stage cyclones can be deeply understood, providing an important basis for analyzing the overall performance of the classification system.

[0141] Step S1210 can comprehensively and in real time grasp the particle size distribution and its variation trend in each stage during the cyclone classification process by constructing a distributed particle size analysis network and obtaining the second particle size feature set. Taking the overflow particle size skewness index and the underflow particle size kurtosis index as examples, they can help operators promptly detect abnormal particle size distributions, such as excessive asymmetry or sharpness in the particle size distribution, which may indicate problems with the working state of the cyclone and require adjustment. The cross-stage particle size covariance matrix can reflect the mutual influence between different stages of the cyclone. When abnormal particle size correlation is found between two stages, measures can be taken in advance to prevent problems from escalating and ensure the stable operation of the classification system. At the same time, these particle size feature data provide an accurate data basis for subsequent classification fluctuation evaluation, making the evaluation results more reliable. Based on accurate evaluation results, the operating parameters of the cyclone can be adjusted more precisely, improving the classification efficiency, reducing the output of unqualified products, and thus enhancing the economic benefits and production quality of the entire classification system.

[0142] Step S1220: Real-time measure the multi-source data during the operation of the n-stage cyclone, and obtain the second set of operating condition parameters according to the multi-source data. The second set of operating condition parameters includes the dynamic feed concentration gradient , the pressure pulsation amplitude and the flow rate coefficient of variation .

[0143] Furthermore, as Figure 4 shown, Step S1220 includes:

[0144] Step S1221: Deploy pressure sensors, flow meters and concentration meters at corresponding positions of the n-stage cyclone to real-time measure the multi-source data during the operation of the n-stage cyclone. The multi-source data includes the feed pressure, the ore volume flow rate and the feed concentration;

[0145] Step S1222: Calculate the dynamic feed concentration gradient according to the feed concentration .

[0146]

[0147] where C(t) is the feed concentration at the current sampling time t, C(t - Δt) is the feed concentration at the previous sampling time t - Δt, and Δt is the sampling period;

[0148] Step S1223: Calculate the pressure pulsation amplitude according to the feed pressure .

[0149] Step S1224: Calculate the flow rate coefficient of variation according to the ore volume flow rate .

[0150] Specifically, in step S1221, pressure sensors, flow meters, and concentration meters need to be deployed at corresponding positions of the n-stage cyclone to measure multi-source data in real time, which includes feed pressure, ore volume flow rate, and feed concentration. Selecting appropriate sensors is crucial. A high-frequency response pressure sensor is selected, and the sampling frequency is not less than 10 kHz because the feed pressure has pulsation characteristics, and a higher sampling frequency can capture the instantaneous changes of the pressure more accurately. For example, in some ore dressing plants, the working process of the feed pump causes periodic fluctuations in the feed pressure. If the sampling frequency is too low, these fluctuation information may be missed, and key parameters such as the pressure pulsation amplitude cannot be accurately obtained, affecting the accurate evaluation of the classification process. An electromagnetic flow meter with a measurement lower limit of 0.3 m / s is selected to avoid measurement distortion at low flow rates. In the actual cyclone classification process, when the pulp flow rate is low, the measurement accuracy of some ordinary flow meters will drop significantly, and the flow rate data cannot be accurately measured. However, this electromagnetic flow meter can still maintain a high measurement accuracy at low flow rates, ensuring the reliability of the flow rate data. An online concentration meter based on the γ-ray principle or ultrasonic principle is selected because concentration meters based on these two principles can balance the measurement range and accuracy requirements. In ore dressing production, the change range of the feed concentration is relatively wide, from a relatively low concentration to a relatively high concentration may occur. The online concentration meter based on the γ-ray principle or ultrasonic principle can accurately measure within a wide concentration range, providing an accurate data basis for subsequent concentration-related calculations and controls.

[0151] In step S1222, the feed concentration data is collected in real time through the concentration meter, and the dynamic feed concentration gradient is calculated. This parameter reflects the change rate of the feed concentration over time and is crucial for understanding the dynamic changes of the feed concentration during the classification process. For example, in a continuously operating cyclone classification system, if the dynamic feed concentration gradient suddenly increases, it may mean that the feed concentration rises rapidly in a short period, which may lead to problems such as coarser overflow particle size and increased underflow concentration fluctuation in the classification effect. By monitoring the dynamic feed concentration gradient, operators can timely detect this change trend and take corresponding measures to adjust the feed concentration to ensure the stability and efficiency of the classification process. From the beneficial effects, accurately calculating the dynamic feed concentration gradient can provide a key basis for subsequent control strategies. By timely adjusting the feed concentration, the classification error caused by concentration fluctuations can be effectively reduced, and the quality and output of the concentrate can be improved.

[0152] Within the time window T, the pressure sensor collects a series of pressure data P(t). By finding the maximum pressure value max(P(t)) and the minimum pressure value min(P(t)) within this time window and subtracting the two, the pressure pulsation amplitude is obtained. The pressure pulsation amplitude is used to measure the magnitude of pressure fluctuations, which is an important indicator during the hydrocyclone classification process. For example, a larger pressure pulsation amplitude may indicate unstable operation of the feed pump or problems such as blockage or leakage in the pipeline system. When the pressure pulsation amplitude is too large, it will have an adverse impact on the internal flow field of the hydrocyclone, leading to deterioration of the classification effect, such as inaccurate particle size classification and disordered movement trajectories of mineral particles within the hydrocyclone. By monitoring the pressure pulsation amplitude, these potential problems can be detected in a timely manner, enabling inspection and maintenance of the equipment to ensure the normal operation of the classification process. Reasoning from the beneficial effects, accurately obtaining the pressure pulsation amplitude helps to detect potential equipment failures in advance, reduce downtime caused by equipment failures, lower maintenance costs, and improve production efficiency. Taking a certain ore dressing plant as an example, through real-time monitoring of the pressure pulsation amplitude, the wear problem of the feed pump impeller was detected in advance and replaced in a timely manner, avoiding a large fluctuation in the feed pressure caused by severe wear of the impeller, thus ensuring the stable operation of the classification system, reducing production losses caused by equipment failures, and saving the enterprise about 500,000 yuan in maintenance costs and production loss expenses annually.

[0153] Based on sliding window statistics, using the flow rate data collected by the flow meter, the standard deviation σ of the flow rate is calculated within the sliding window Q and the mean value μ Q , and the two are divided to obtain the flow rate coefficient of variation. The window length is consistent with the pressure pulsation period T. This setting is to analyze the variation characteristics of pressure and flow rate on the same time scale and better reflect the relationship between the two. The flow rate coefficient of variation reflects the relative dispersion degree of the flow rate, that is, the stability of the flow rate. During the hydrocyclone classification process, a stable flow rate is crucial for ensuring the classification effect. For example, if the flow rate coefficient of variation is large, it indicates that the flow rate fluctuates greatly, which will make the amount of pulp entering the hydrocyclone unstable, thereby affecting the classification accuracy and resulting in uneven particle size distributions of the overflow and underflow products. By monitoring the flow rate coefficient of variation, the unstable situation of the flow rate can be detected in a timely manner and corresponding measures can be taken for adjustment, such as adjusting the rotational speed of the feed pump or the opening degree of the valve. From the beneficial effects, by monitoring the flow rate coefficient of variation and stabilizing the flow rate, the quality stability of the classification products can be improved, and the economic losses caused by product quality fluctuations can be reduced. In the actual production of a certain ore dressing plant, by monitoring and controlling the flow rate coefficient of variation, the particle size qualification rate of the concentrate product was increased by 15%, reducing the rework and scrap losses caused by unqualified product quality and improving the economic benefits of the enterprise.

[0154] Step S1230: Perform hierarchical fluctuation evaluation based on the second particle size feature set and the second operating condition parameter set to obtain a hierarchical fluctuation evaluation result, where the hierarchical fluctuation evaluation result includes first-level fluctuation, second-level fluctuation, and third-level fluctuation;

[0155] Further, as Figure 5 shown, Step S1230 includes:

[0156] Step S1231: Judge whether it is a first-level fluctuation based on the cross-section particle size covariance matrix in the second particle size feature set;

[0157] The judgment of whether it is a first-level fluctuation based on the cross-section particle size covariance matrix in the second particle size feature set includes:

[0158] Calculate the maximum eigenvalue λ of the cross-section particle size covariance matrix max in the second particle size feature set. If λ max > λ α , then it is a first-level fluctuation, where λ α is a preset first-level fluctuation threshold.

[0159] Step S1232: Judge whether it is a second-level fluctuation based on the second operating condition parameter set;

[0160] The judgment of whether it is a second-level fluctuation based on the second operating condition parameter set includes:

[0161] Preset a second-level fluctuation threshold λ β . If the product of the pressure pulsation amplitude and the flow variation coefficient exceeds λ β , then it is a second-level fluctuation.

[0162] Step S1233: Judge whether it is a third-level fluctuation based on the overflow particle size skewness index sequence S in the second particle size feature set.

[0163] Set the normal range [L0, H0] of the overflow particle size skewness index. If S exceeds the normal range [L0, H0] continuously for λ N times, then it is a third-level fluctuation, where L0 is the lower limit of the normal range of the overflow particle size skewness index, H0 is the upper limit of the normal range of the overflow particle size skewness index, and λ N is a preset third-level fluctuation threshold.

[0164] Specifically, in Step S1231, judge whether it is a first-level fluctuation based on the cross-section particle size covariance matrix M in the second particle size feature set. The cross-section particle size covariance matrix M reflects the correlation between the overflow particle size skewness and the underflow particle size kurtosis among different cyclones, and its maximum eigenvalue λ maxis the key indicator for measuring the strength of this correlation. Calculate the maximum eigenvalue λ of the cross-section granularity covariance matrix M max , which is achieved through a specific mathematical algorithm, such as the commonly used eigenvalue decomposition algorithm. The first-level fluctuation threshold λ α is set by statistically adding 1 times the standard deviation to the mean of the maximum eigenvalue of the cross-section granularity covariance matrix under stable operating conditions. Compare the calculated λ max with the preset first-level fluctuation threshold λ α . If λ max >λ α , it is determined as the first-level fluctuation. The first-level fluctuation means that there is a strong granularity mutual feedback between different cyclones, which may lead to the deterioration of the classification effect. For example, in a classification system composed of 5 sections of cyclones, when the skewness of the overflow particle size or the kurtosis of the underflow particle size of a certain section of cyclone changes abnormally, if the maximum eigenvalue λ of the cross-section granularity covariance matrix M max exceeds the preset first-level fluctuation threshold λ α , it indicates that the granularity correlation between this section of cyclone and other sections is abnormal, which may be caused by reasons such as blockage of the overflow and underflow pipelines. Timely discovery of the first-level fluctuation and taking corresponding measures, such as checking whether the pipeline is blocked, can avoid further affecting the classification quality due to the granularity mutual feedback problem, ensure the normal operation of the classification system. This step judges the operating state of the system through quantitative indicators, provides a basis for subsequent targeted adjustments, helps to maintain the stability and reliability of the classification system, improves production efficiency, and reduces the downtime and production cost caused by equipment failures.

[0165] Step S1232 is to judge whether it is a second-level fluctuation based on the second set of operating parameters. The second set of operating parameters includes the dynamic feed concentration gradient , the pressure pulsation amplitude and the flow variation coefficient and other parameters. Preset the second-level fluctuation threshold λ β at 90% of the peak value of the product of the pressure pulsation and the flow variation during historical operation. Determine whether it is a second-level fluctuation by judging whether the product of the pressure pulsation amplitude and the flow variation coefficient exceeds λ β . When the product of the pressure pulsation amplitude and the flow variation coefficient exceeds λ β , it indicates that the feed pump has unstable pumping. For example, in the cyclone classification operation of a certain mine, if the impeller of the feed pump is worn, it will cause an increase in the pressure pulsation amplitude, and at the same time, the stability of the flow will also be affected, resulting in a change in the flow variation coefficient. When the product of the two exceeds the second-level fluctuation threshold λ βWhen it occurs, a secondary fluctuation warning is triggered. At this time, it is necessary to check for abnormal wear or cavitation of components such as the pump chamber and impeller, because these problems will affect the stability of the ore feeding, and thus affect the classification effect. By accurately judging and timely handling the secondary fluctuation, the normal operation of the ore feeding pump can be ensured, the working conditions of the classification system can be stabilized, the quality of the classified products can be improved, the equipment loss can be reduced, the service life of the equipment can be extended, and the continuity and stability of production can be guaranteed.

[0166] Step S1233 determines whether it is a tertiary fluctuation based on the overflow particle size skewness index sequence S in the second particle size feature set. First, set the normal range [L0, H0] of the overflow particle size skewness index. The overflow particle size skewness index S i is used to characterize the asymmetry of the overflow particle size distribution of the i-th stage cyclone. If S exceeds the normal range [L0, H0] continuously for λ N (usually the empirical value is set to 3) times, it is determined as a tertiary fluctuation. The tertiary fluctuation indicates that the overflow particle size distribution shows a continuous deviation, which means a long-term decline in classification performance. For example, during a certain period, if the overflow particle size skewness index of a certain stage of the cyclone continuously exceeds the normal range, it may be due to abnormal structural parameters such as the cone angle and overflow pipe diameter of this stage of the cyclone, resulting in a change in the movement trajectory of particles during the classification process, thereby affecting the classification effect. At this time, checking whether the structural parameters such as the cone angle and overflow pipe diameter of the single cyclone are abnormal can timely discover and solve potential problems existing in the classification system, optimize the classification process, improve the classification accuracy, ensure that the product particle size meets the production requirements, and avoid economic losses caused by unqualified product quality.

[0167] Step S1230 realizes an effective evaluation of the classification fluctuation through the comprehensive analysis of the second particle size feature set and the second operating parameter set. By setting reasonable fluctuation thresholds, different levels of fluctuations can be judged timely and accurately, and then targeted measures can be taken. This not only helps to ensure the stable operation of the cyclone classification system, improve the classification quality and efficiency, but also reduces equipment failures and maintenance costs, enhances the reliability and controllability of the production process, provides strong support for the optimization of the entire production process, and improves the economic benefits and competitiveness of the enterprise.

[0168] Step S2000, according to the classification fluctuation evaluation result, optimizes the first operating parameter set and the second structural parameter set to obtain an optimized parameter set, and the optimized parameter set includes a first optimized parameter set, a second optimized parameter set, and a third optimized parameter set; accurately controls the cyclone classification process according to the optimized parameter set.

[0169] Furthermore, step S2000 includes:

[0170] Step S2100, if the hierarchical fluctuation evaluation result is a first-level fluctuation, perform transfer learning optimization on the first working condition parameter set and the second structural parameter set to obtain a first optimized parameter set;

[0171] Further, as Figure 6 shown, step S2100 includes:

[0172] Step S2110, retrieve historical optimization cases similar to the current working condition from the historical working condition library according to the preset matching conditions;

[0173] The preset matching conditions are:

[0174] |λ max -λ history |<δ1 and the correlation coefficient ρ>0.8, where λ max is the maximum eigenvalue of the current cross-section granularity covariance matrix M, λ history is the maximum eigenvalue of the covariance matrix corresponding to the historical optimization case, δ1 is the preset eigenvalue difference threshold, and ρ is the correlation coefficient between the current working condition and the historical optimization case;

[0175] Step S2120, extract the historical cone angle ratio sequence corresponding to the historical optimization case, and optimize the calibrated cone angle ratio sequence [α''1, α''2,..., α'' n in the second structural parameter set to obtain an optimized cone angle ratio sequence [α'''1, α'''2,..., α''' n ; where α''' i is the optimized cone angle ratio of the i-th stage cyclone;

[0176]

[0177] where α''' i is the optimized cone angle ratio of the i-th stage cyclone, is the historical cone angle ratio of the i-th stage cyclone in the historical optimization case, α'' i is the cone angle ratio of the i-th stage cyclone calibrated in the second structural parameter set, is the learning rate, which is used to control the influence degree of historical parameters on the current optimization, and its appropriate value is usually determined through multiple experiments and data analysis to balance the optimization speed and accuracy; is the sigmoid function, which is used to smooth the correction amplitude of the cone angle ratio.

[0178] Step S2130, according to λ max and the first-level fluctuation threshold λ αCalculate the overrun amplitude of the eigenvalue, and optimize the reference value ΔP0 of the pressure gradient in the first operating condition parameter set according to the overrun amplitude of the eigenvalue and the structural error compensation coefficient matrix k to obtain the first compensated pressure gradient ΔP c ;

[0179]

[0180] Among them, is the overrun amplitude of the eigenvalue.

[0181] Step S2140: Construct the first optimized parameter set according to the calibrated overflow pipe diameter ratio sequence [D''1, D''2,..., D'' n , the optimized cone angle ratio sequence [α'''1, α'''2,..., α''' n , the first compensated pressure gradient ΔP c , the reference value C0 of the ore concentration, and the reference value Q' of the flow ratio.

[0182] Specifically, in step S2100, when the classification fluctuation evaluation result is judged to be a first-level fluctuation, it means that there is a strong particle size mutual feedback between different cyclones, which will seriously affect the classification effect. At this time, the strategy of optimizing by transfer learning is adopted. The purpose is to draw on the successful experience of historical optimization cases, combine the current working conditions for targeted adjustment, improve the ability of the classification system to cope with fluctuations, and then improve the overall classification performance.

[0183] Sub-step S2110 needs to retrieve historical optimization cases similar to the current working condition from the historical working condition database according to the preset matching conditions. The preset matching conditions are |λ max -λ history |<δ1 and the correlation coefficient ρ>0.8, where λ max is the maximum eigenvalue of the current cross-section particle size covariance matrix M, which reflects the degree of association of the particle size distribution between different cyclones at present; λ historyis the maximum eigenvalue of the covariance matrix corresponding to the historical optimization case, which is used to compare historical situations; δ1 is the preset eigenvalue difference threshold, which is used to measure the acceptable range of the difference between the current and historical eigenvalues, ensuring that the retrieved historical cases are similar to the current operating conditions in terms of particle size correlation characteristics; the setting method of δ1 is: statistically calculate the average value of the maximum eigenvalue difference between historical cases, and add 0.1 times the standard deviation as δ1 on this basis; ρ is the correlation coefficient between the current operating condition and the historical optimization case, which is obtained by calculating the correlation between multiple key parameters (such as the particle size characteristics of each section of the cyclone, operating parameters, etc.). If it is greater than 0.8, it means that the current operating condition and the historical case have a high similarity in overall characteristics. Through such strict matching condition screening, cases similar to the current operating condition can be accurately found from the historical operating condition library, providing a reliable reference for subsequent optimization. For example, in the grading process of a certain mineral processing plant, the maximum eigenvalue λ of the current cross-segment particle size covariance matrix M max is 5.5. There are many cases in the historical operating condition library. By setting δ1 to 0.5, a historical optimization case is found after screening, and its corresponding covariance matrix maximum eigenvalue λ history is 5.2, and the correlation coefficient ρ between the current working condition and the historical case is 0.85, which meets the preset matching conditions, so that the historical case can be used as a reference for the current optimization. The beneficial effect of this retrieval method is that it can make full use of the historical accumulated experience data to avoid starting from scratch to explore the optimization plan when encountering new fluctuating working conditions, which greatly saves optimization time and cost. By comparing historical cases, possible effective adjustment directions can be quickly found to improve the response speed of the grading system to fluctuations, thereby reducing the impact of fluctuations on the grading effect and improving the quality and output of the concentrate. From the reasoning process, since the historical optimization case is an effective solution that has been verified in actual production, when a case similar to the current working condition is found, its optimization ideas and parameter adjustment methods can be used to improve the current grading situation with a high probability. In the above example, after learning from the optimization plan of the historical case, after actual operation monitoring, the quality of the concentrate has been significantly improved and the output has also increased, which proves the effectiveness of this retrieval method.

[0184] Sub-step S2120 extracts the historical cone angle ratio sequence corresponding to the historical optimization case, and calibrates the calibrated cone angle ratio sequence [α''1, α''2, ..., α'' in the second structural parameter set according to the historical cone angle ratio sequence. n ] to optimize and obtain the optimized cone angle ratio sequence [α'''1, α'''2, ..., α''' nSince there are differences between the current working conditions and historical working conditions, historical parameters cannot be directly applied. Instead, a feature transfer algorithm is used for optimization. This algorithm is corrected by weighted learning rate and dynamically adjusts the correction amplitude by introducing the sigmoid function of the relative eigenvalue. The optimization method in step S2120 not only fully draws on historical successful experiences but also takes into account the particularity of the current working conditions. By dynamically adjusting the correction amplitude, the optimized cone angle ratio is more in line with the current actual situation, effectively improving the adaptability of the classification system to primary fluctuations. By reasonably adjusting the cone angle ratio, the flow field distribution inside the cyclone can be optimized, the separation efficiency of particles can be improved, and thus the size qualification rate and output of the concentrate can be increased. From the reasoning process, the historical cone angle ratio sequence represents the successful parameter settings under similar fluctuations. By adjusting through the feature transfer algorithm, the historical parameters can be reasonably corrected according to the differences in the current working conditions.

[0185] In sub-step S2130, It reflects the degree to which the maximum eigenvalue of the current cross-section particle size covariance matrix exceeds the normal range. The structural error compensation coefficient matrix k reflects the structural error degree of each section of the cyclone under the reference parameters, and its value is related to the actual structure and operating conditions of the cyclone. Through such a calculation method, the pressure gradient reference value can be adjusted targeted according to the current fluctuation situation and the structural characteristics of the cyclone. The beneficial effect of the adopted pressure gradient compensation mechanism is that it can suppress the diffusion of local fluctuations to the whole, ensuring the stable operation of the classification process. When primary fluctuations occur, appropriately adjusting the pressure gradient can change the flow state of the pulp in the cyclone, making the movement trajectory of the particles more stable and reducing the classification chaos caused by particle size mutual feedback. From the reasoning process, when the over-limit amplitude of the eigenvalue is relatively large, it indicates that the current fluctuation is relatively serious. At this time, by magnifying and adjusting the pressure gradient through the structural error compensation coefficient matrix k, the fluctuation can be more effectively dealt with and the classification process can be stabilized. In actual production, through the monitoring of multiple classification systems, it is found that after adopting this pressure gradient compensation mechanism, in the case of primary fluctuations, the stability of the classification system is significantly improved, the particle size distributions of the overflow and underflow products are more uniform, the product quality fluctuations caused by fluctuations are reduced, and the production efficiency and product quality are improved.

[0186] Step S2200, if the classification fluctuation evaluation result is secondary fluctuation, optimize the first set of working condition parameters to obtain the second optimized parameter set;

[0187] Furthermore, as Figure 7 shown, step S2200 includes:

[0188] Step S2210, construct a pressure-flow coupling equation according to the second set of working condition parameters;

[0189] The construction of the pressure-flow coupling equation includes:

[0190] Establish the initial pressure-flow coupling equation:

[0191]

[0192] wherein, is the first model coefficient to be identified, is the second model coefficient to be identified, controls the non-linear amplification effect of the flow rate variation on the pressure pulsation, controls the linear contribution of the feed concentration change to the pressure pulsation.

[0193] Through the least squares method, calculate and obtain and ;

[0194] According to , and the initial pressure-flow coupling equation, obtain the final pressure-flow coupling equation.

[0195] Specifically, the second set of operating parameters includes the dynamic feed concentration gradient , the pressure pulsation amplitude A p and the flow rate variation coefficient CV Q . These parameters reflect the dynamic operating characteristics of the hydrocyclone during operation. The purpose of constructing the pressure-flow coupling equation is to reveal the internal relationship between the pressure pulsation amplitude, the flow rate variation coefficient, and the dynamic feed concentration gradient, so as to provide a theoretical basis for the subsequent design of control strategies.

[0196] First, establish the initial pressure-flow coupling equation: , wherein, is the first model coefficient to be identified, which controls the non-linear amplification effect of the flow rate variation on the pressure pulsation. The flow rate variation coefficient CV Q reflects the relative dispersion degree of the flow rate, and the product of its square and embodies the complex influence of the flow rate fluctuation on the pressure pulsation. For example, when the flow rate variation coefficient is relatively large, if is also large, then the fluctuation of the flow rate will cause the pressure pulsation amplitude to increase significantly. is the second model coefficient to be identified, which is used to control the linear contribution of the feed concentration change to the pressure pulsation. The dynamic feed concentration gradient reflects the change rate of the feed concentration with time, and its product with represents the direct influence of the feed concentration change on the pressure pulsation.

[0197] Calculate and obtain and , The least squares method is a commonly used mathematical method in data processing and model parameter estimation. Its principle is to determine the optimal estimated values of the unknown parameters in the model by minimizing the sum of the squares of the errors between the observed values and the model predicted values. In this step, a large number of historical data of the pressure pulsation amplitude A p , the flow coefficient of variation CV Q and the dynamic ore feeding concentration gradient are collected, and these data are fitted using the least squares method to obtain the coefficients and that can best describe the pressure-flow coupling relationship.

[0198] According to , and the initial pressure-flow coupling equation, the final pressure-flow coupling equation is obtained. This final equation accurately describes the quantitative relationship between the pressure pulsation amplitude, the flow coefficient of variation, and the dynamic ore feeding concentration gradient under the current working conditions. Its beneficial effect is to provide a mathematical model basis for deeply understanding the interaction between pressure and flow during the operation of the hydrocyclone. Through this model, the change trend of pressure pulsation under different working conditions can be analyzed and predicted more accurately, providing strong theoretical support for formulating effective control strategies subsequently, helping to optimize the operating parameters of the hydrocyclone, and improving the stability and classification efficiency of the classification system. For example, in actual production, by monitoring the real-time flow coefficient of variation and the dynamic ore feeding concentration gradient, the change of the pressure pulsation amplitude can be predicted using this coupling equation, and measures can be taken in advance to avoid excessive pressure fluctuations from damaging the equipment or affecting the classification effect.

[0199] Step S2220, design a pressure-flow decoupling control strategy based on the pressure-flow coupling equation; according to the pressure-flow decoupling control strategy, obtain the adjusted ore feeding concentration reference value C'0;

[0200] Further, step S2220 includes:

[0201] Step S2221, establish a pressure-flow trade-off objective function:

[0202]

[0203] Wherein, is the pressure-flow trade-off coefficient, is 's lower limit of value, is 's upper limit of value;

[0204] Step S2222, solve the pressure-flow trade-off objective function to obtain the optimal pressure-flow trade-off coefficient ;

[0205] Step S2223, according to the optimal pressure-flow trade-off coefficient correct the feed concentration reference value C0 to obtain the adjusted feed concentration reference value C'0;

[0206]

[0207] wherein, C'0 is the adjusted feed concentration reference value, is the sign function.

[0208] Specifically, in step S2220, a pressure-flow decoupling control strategy is designed based on the pressure-flow coupling equation, and the adjusted feed concentration reference value C'0 is obtained according to this strategy. This series of operations aims to solve the strong coupling problem between pressure and flow during the operation of the hydrocyclone, achieve precise control of the feed concentration, and thus improve the overall performance of the classification system.

[0209] First, establish a pressure-flow trade-off objective function. In the pressure-flow trade-off objective function, is the pressure-flow trade-off coefficient, which is a key design variable used to balance the pressure pulsation amplitude A p and the coefficient of variation of flow rate CV Q the weights in the objective function. is the lower limit of the value of is the upper limit of the value of. These two limits are determined according to the equipment characteristics, process requirements and actual operation experience of the hydrocyclone. The lower limit ensures that the inhibitory effect on the coefficient of variation of flow rate during adjustment will not be too small to effectively control the flow rate fluctuation, and the upper limit prevents excessive relaxation of the pressure pulsation from causing unstable operation of the equipment. For example, in a specific hydrocyclone classification system, through a large number of experiments and data analysis, it is determined that the value range of is [0.1, 0.9] to ensure that the fluctuations of pressure and flow rate can be balanced to a certain extent under different working conditions.

[0210] The physical meaning of this objective function is to find an optimal trade-off coefficient within the value range of such that the sum of the pressure pulsation amplitude and the weighted coefficient of variation of flow rate is minimized, that is, the weighted sum of the pressure pulsation and the flow rate fluctuation is minimized. This means that in the control process, the stability of both pressure and flow rate is considered comprehensively, avoiding the situation of simply pursuing one index while ignoring the other. When the value is larger, it indicates that in the control process, it is more inclined to stabilize the flow rate and relax the pressure pulsation; on the contrary, When the value is smaller, more attention is paid to reducing pressure pulsation while having a relatively higher tolerance for flow fluctuations. By adjusting the value, the operating state of the system can be flexibly optimized according to actual requirements.

[0211] Due to the complexity of this objective function, heuristic algorithms such as genetic algorithms and particle swarm algorithms are usually used for solving. These algorithms simulate natural evolution or swarm intelligence behavior, and gradually approach the optimal solution by continuously searching and iterating in the solution space. Taking the genetic algorithm as an example, it simulates the genetic evolution process of organisms. First, a group of initial solutions (i.e., a group of different values) are randomly generated, and these solutions are regarded as individuals in the population. Each individual has a fitness value, which is calculated by the objective function and used to measure the quality of the individual in the current problem. In each generation of evolution, new populations are generated through operations such as selection, crossover, and mutation. The selection operation selects better individuals to enter the next generation according to the fitness values of the individuals; the crossover operation simulates the gene exchange of organisms, exchanges part of the genes of two individuals to generate new individuals; the mutation operation randomly changes the genes of individuals with a certain probability to increase the diversity of the population and avoid the algorithm falling into a local optimal solution. After multiple generations of evolution, the individuals in the population gradually approach the optimal solution, and the value corresponding to the finally obtained optimal individual is the optimal pressure-flow trade-off coefficient .

[0212] The particle swarm algorithm simulates the foraging behavior of bird flocks. Each solution is regarded as a bird (particle) in the search space, and each particle has its own position and velocity. The particle adjusts its velocity and position according to its own historical optimal position and the global optimal position of the group, and continuously flies in the search space to find the optimal solution. In this problem, the position of the particle is different values. By continuously updating the velocity and position of the particle, the particle gradually approaches the optimal solution, thereby obtaining the optimal pressure-flow trade-off coefficient . The optimal pressure-flow trade-off coefficient obtained by solving through these algorithms can achieve the best balance between pressure pulsation and flow fluctuations under the condition of meeting the constraint conditions, providing an accurate basis for subsequent adjustment of the ore feeding concentration.

[0213] In sub-step S2223, the ore feeding concentration reference value C0 is corrected according to the optimal pressure-flow trade-off coefficient to obtain the adjusted ore feeding concentration reference value C'0; the calculation formula is ; where C'0 is the adjusted ore feeding concentration reference value; is the sign function, which determines the adjustment direction according to the positive and negative of the dynamic ore feeding concentration gradient . When the ore feeding concentration shows an upward trend ( When (), the value of the sign function is 1. At this time, according to the formula, the adjusted feed concentration reference value C'0 will be appropriately reduced, that is , and the reduction amplitude is proportional to . This is because when the feed concentration increases, if no adjustment is made, it may lead to further aggravation of pressure pulsation and flow fluctuation. By reducing the feed concentration reference value, this trend can be alleviated and the stability of the system can be maintained. On the contrary, when the feed concentration shows a downward trend ( ), the value of the sign function is -1, and the adjusted feed concentration reference value C'0 will be appropriately increased, that is , and the increase amplitude is also proportional to .

[0214] Through this adjustment method based on the optimal pressure-flow trade-off coefficient and the trend of feed concentration change, the dynamic optimization of the feed concentration can be achieved, effectively suppressing the secondary fluctuation, and improving the stability and classification effect of the classification system. Its beneficial effect lies in accurately adjusting the feed concentration according to the real-time working conditions of the system, enabling the pressure and flow to operate within a relatively stable range, reducing the interference of pressure-flow coupling instability on the classification process, thereby improving the quality and production efficiency of the product, and reducing equipment loss and production cost.

[0215] Step S2230, construct a second optimization parameter set according to the calibrated overflow pipe diameter ratio sequence [D''1, D''2,..., D'' n , the calibrated cone angle ratio sequence [α''1, α''2,..., α'' n , the adjusted feed concentration reference value C'0, the pressure gradient reference value ΔP0, and the flow ratio reference value Q'.

[0216] Specifically, the calibrated overflow pipe diameter ratio sequence [D''1, D''2,..., D'' n and the cone angle ratio sequence [α''1, α''2,..., α'' n are obtained through previous adjustments. They consider the structural errors and working condition changes during the operation of the cyclone and can better meet the actual operation requirements. The adjusted feed concentration reference value C'0 is obtained based on the pressure-flow decoupling control strategy, aiming to balance the fluctuations of pressure and flow and improve the stability of the classification system. The pressure gradient reference value ΔP0 and the flow ratio reference value Q' are important parameters reflecting the working conditions of the cyclone and have a direct impact on the classification effect.

[0217] The process of constructing the second optimized parameter set is to integrate these optimized and adjusted parameters to form a complete set of parameter combinations. This parameter set can provide a more accurate control basis for the operation of the hydrocyclone under secondary fluctuations. For example, in an actual hydrocyclone classification system, the overflow pipe diameter ratio D'' of a certain section of the hydrocyclone after calibration i is 0.8, and the cone angle ratio α'' i is 1.2. The adjusted feed concentration reference value C'0 is 25%, the pressure gradient reference value ΔP0 is 0.5 MPa, and the flow ratio reference value Q' is 0.9. Combining these parameters together forms the second optimized parameter set for this hydrocyclone under the current working conditions.

[0218] Its beneficial effect is that by comprehensively considering the optimization results of multiple key parameters, the constructed second optimized parameter set can enable the hydrocyclone to operate more efficiently and stably under the working conditions of secondary fluctuations. These parameters cooperate with each other to optimize the flow state of the pulp in the hydrocyclone, improve the classification accuracy of particles, and reduce the output of unqualified products. At the same time, reasonable parameter settings can also reduce the energy consumption of the equipment, extend the service life of the equipment, and enhance the economic benefits and operation reliability of the entire classification system. By using the second optimized parameter set to precisely control the hydrocyclone, the adverse effects brought by secondary fluctuations can be effectively addressed, ensuring the smooth progress of the classification process and providing strong support for industrial production.

[0219] Step S2300, if the classification fluctuation evaluation result is a three-level fluctuation, optimize the first working condition parameter set to obtain the third optimized parameter set;

[0220] Furthermore, step S2300 includes:

[0221] Step S2310, calculate the cross-section particle size equilibrium factor ξ according to the second particle size characteristic set;

[0222]

[0223] Among them, is the determinant of the cross-section particle size covariance matrix , is the modulus of the overflow particle size skewness index sequence , is the modulus of the underflow particle size kurtosis index sequence .

[0224] Step S2320, based on the cross-section particle size equilibrium factor ξ, calculate the correction weight of the cone angle ratio of the i-th section of the hydrocyclone;

[0225] The calculation formula of

[0226]

[0227] Among them, is the actual flow rate of the th stage cyclone, is the proportion of the cone angle ratio of the th stage cyclone in the total cone angle ratio,

[0228] Step S2330, according to the correction weight , calculate the corrected cone angle ratio of the

[0229] th stage cyclone; Step S2340, from the corrected cone angle ratio of the

[0230] th stage cyclone, obtain the corrected cone angle ratio sequence n ; Step S2350, according to the calibrated overflow pipe diameter ratio sequence [D''1, D''2,..., D''

[0231] Specifically, step S2300 and its sub-steps are aimed at the situation where the classification fluctuation evaluation result is a three-level fluctuation, optimizing the first working condition parameter set to obtain the third optimized parameter set, so as to solve problems such as continuous deviation of the overflow particle size distribution and decline in classification performance, and ensure the efficient and stable operation of the cyclone classification process. In step S2300, when the classification fluctuation evaluation result is a three-level fluctuation, it means that the overflow particle size distribution shows a continuous deviation, and relevant parameters need to be optimized and adjusted. At this time, through a series of subsequent sub-steps, the first working condition parameter set is optimized to construct the third optimized parameter set, so that the operating parameters of the cyclone are more adapted to the current classification situation, improve the classification accuracy and stability, reduce production losses caused by abnormal particle size distribution, and ensure the efficient operation of the entire classification system.

[0232] Sub-step S2310 calculates the cross-stage particle size balance factor ξ according to the second particle size feature set; the formula is ; among them, is the cross-stage particle size covariance matrix The determinant comprehensively reflects the overall degree of dispersion among the particle size distributions of each stage of the hydrocyclone. The larger the value of the determinant, the higher the overall degree of dispersion of the particle size distributions of each stage of the hydrocyclone. For example, in a classification system with three stages of hydrocyclones, if the skewness index of the overflow particle size of the first-stage hydrocyclone changes significantly, and at the same time, abnormal fluctuations also occur in the kurtosis indices of the underflow particle sizes of the second and third stages, and these changes are interrelated, resulting in an increase in the determinant value of the cross-stage particle size covariance matrix, indicating a higher overall degree of dispersion. is the sequence of skewness indices of the overflow particle size of the modulus length, which measures the comprehensive degree of asymmetry of the overflow particle size distribution; is the sequence of kurtosis indices of the underflow particle size of the modulus length, reflecting the comprehensive situation of the sharpness of the underflow particle size distribution. The product of the two reflects the independent degree of dispersion of the particle size distributions of each stage of the hydrocyclone. By calculating the ratio of the two and then subtracting the ratio from 1 to obtain the cross-stage particle size balance factor ξ, it can be used to characterize the overall balance degree. When the value of ξ is larger, it indicates that the overall is more balanced, and the influence of the correlation of the particle size distributions between each stage of the hydrocyclone on the overall degree of dispersion is smaller; when the value of ξ is smaller, it indicates that the overall is more unbalanced, and the influence of the correlation of the particle size distributions between each stage of the hydrocyclone on the overall degree of dispersion is larger. By calculating the cross-stage particle size balance factor ξ, the balance state of the particle size distributions of each stage of the hydrocyclone in the classification system can be quantitatively evaluated, providing an important basis for subsequent parameter optimization. This helps to accurately locate the problems existing in the classification system, and by adjusting the relevant parameters, improve the balance of the particle size distribution, and enhance the classification efficiency and product quality. From the reasoning process, through the comprehensive calculation of the cross-stage particle size covariance matrix and the particle size characteristic sequences of each stage, the cross-stage particle size balance factor obtained can reflect the actual operating state of the system. According to this factor for optimization and adjustment, each stage of the hydrocyclone can work better together, reducing resource waste and product quality instability problems caused by uneven particle size distribution.

[0233] In the calculation formula of, the first part is proportional to the cross-stage particle size balance factor ξ. This part considers the overall balance degree of the particle size distributions of each stage. When the value of ξ is larger, it indicates a higher overall balance degree. When correcting the cone angle ratio, more attention will be paid to the proportion of the cone angle ratio of each stage in the total cone angle ratio to maintain the overall balance. For example, if the cone angle ratio of a certain stage of the hydrocyclone accounts for a relatively large proportion in the total cone angle ratio under the overall balance state, when the overall balance degree is higher (i.e., the value of ξ is larger), the correction weight of the cone angle ratio of this stage of the hydrocyclone will increase accordingly, making its contribution to the overall balance greater during the adjustment process. The second part of the formula is related to It is directly proportional, and this part takes into account the actual loads of each section of the cyclone, that is, it is reflected by the proportion of the actual flow rate in the total flow rate. When the value of ξ is small, it indicates a relatively high overall imbalance degree. At this time, more attention is paid to the actual flow rate of each section of the cyclone, and the correction weight is allocated according to the actual flow rate to balance the working loads of each section of the cyclone. For example, if the actual flow rate of a certain section of the cyclone is large, but the overall balance is poor (i.e., the value of ξ is small), through this part of the calculation, the correction weight of the cone angle ratio of this section of the cyclone will be appropriately increased, so that it can better adapt to the actual flow rate and improve the classification effect. By calculating the correction weight in this way, the balance of the classification system and the actual operating conditions of each section of the cyclone can be comprehensively considered, providing a reasonable weight basis for accurately adjusting the cone angle ratio in the follow-up, ensuring that both the overall particle size distribution balance can be improved and the actual working loads of each section of the cyclone can be taken into account during the optimization process, and improving the overall performance of the classification system.

[0234] In sub-step S2330, according to the correction weight , calculate the corrected cone angle ratio of the i-th section of the cyclone ; by combining the correction weight with the original cone angle ratio, the accurate adjustment of the cone angle ratio is realized. The specific calculation process is based on the correction weight obtained previously and is calculated according to certain mathematical operation rules, so as to obtain a cone angle ratio that better meets the current classification requirements. For example, assume that the original cone angle ratio of a certain section of the cyclone is α'' i , the calculated correction weight is , and the corrected cone angle ratio is obtained through a specific calculation method (such as multiplying the two or other operations determined according to the actual situation). Such calculation can optimize the cone angle ratio targeted according to the actual state of the classification system to improve the flow field distribution inside the cyclone, enhance the particle separation effect, and further improve the classification accuracy. Since the correction weight comprehensively considers factors such as particle size distribution balance and actual flow rate, the corrected cone angle ratio can better adapt to the operating conditions of the classification system and reduce the classification error caused by unreasonable cone angle ratio.

[0235] Sub-step S2340 integrates the corrected cone angle ratios of each section of the cyclone to form a complete sequence. This sequence comprehensively reflects the optimization results of the cone angle ratios of each section of the cyclone in the entire classification system, providing the key structural parameter part for constructing the third optimization parameter set later. Through the collaborative optimization of the cone angle ratios of each section of the cyclone, the entire classification system can better adjust the internal flow field when facing three-level fluctuations, realize more efficient particle separation, and improve the overall classification effect. For example, in a classification system with five sections of cyclones, the corrected cone angle ratios of each section of the cyclone obtained through the previous steps are combined into a corrected cone angle ratio sequence.

[0236] Sub-step S2350 integrates each key parameter optimized for the three-level fluctuation. The calibrated overflow pipe diameter ratio sequence takes into account the adjustment of the overflow pipe diameter ratio under different working conditions to adapt to the overflow characteristics of particles; the corrected cone angle ratio sequence is the result of optimizing the cone angle ratio based on factors such as the cross-section particle size balance factor and actual flow rate, which can improve the internal flow field of the cyclone; the pressure gradient reference value ΔP0, the ore concentration reference value C0, and the flow rate ratio reference value Q' are important working condition parameters affecting the classification process. Combining these parameters to construct the third optimized parameter set can provide a comprehensive and adaptable set of operating parameters for the cyclone under the three-level fluctuation condition. For example, in the classification system of a certain concentrator, through the previous calculations and analyses, the calibrated overflow pipe diameter ratio, corrected cone angle ratio, and other related parameters suitable for the current three-level fluctuation condition are obtained.

[0237] In step S2400, the first optimized parameter set, the second optimized parameter set, and the third optimized parameter set are combined to form an optimized parameter set, and the cyclone classification process is precisely controlled according to the optimized parameter set.

[0238] Specifically, the first optimized parameter set is obtained through methods such as transfer learning optimization for the first-level fluctuation situation, and it is targeted at solving the problem of particle size mutual feedback between different cyclones; the second optimized parameter set is optimized for the second-level fluctuation by constructing a pressure-flow coupling equation, designing a decoupling control strategy, etc., and is mainly used to solve the problem of pressure-flow coupling instability; the third optimized parameter set is obtained by calculating the cross-section particle size balance factor, adjusting the cone angle ratio, etc. for the third-level fluctuation, and focuses on solving the problem of continuous deviation of the overflow particle size distribution. Integrating these three optimized parameter sets together to form a comprehensive optimized parameter set can enable precise control of the cyclone classification process in the face of different types and degrees of fluctuations. In actual production, before a certain concentrator did not adopt this method of integrating optimized parameter sets, the classification system was often affected by various fluctuations, resulting in unstable product quality and low production efficiency. After adopting the integrated optimized parameter set for precise control, the system can quickly respond to different types of fluctuations. When the first-level fluctuation occurs, the system adjusts according to the parameters in the first optimized parameter set, effectively suppressing the particle size mutual feedback and ensuring the classification effect; when the second-level fluctuation occurs, the pressure-flow is decoupled and controlled according to the second optimized parameter set, making the operation of the feed pump more stable; when the third-level fluctuation occurs, the third optimized parameter set is used to adjust parameters such as the cone angle ratio, improving the overflow particle size distribution. Different optimized parameter sets are specifically optimized for different fluctuation situations, and integrating them together can cover various fluctuation scenarios that may occur during the classification process. Through precise control, the operating parameters of the cyclone can be adjusted in real time according to the actual situation, improving the adaptability and stability of the classification system, reducing production losses caused by fluctuations, improving product quality and production efficiency, and at the same time reducing energy consumption, achieving a double improvement in economic benefits and resource utilization efficiency.

[0239] Example 2

[0240] Based on Example 1, this example provides an intelligent control system for classification fineness based on a mathematical model, as Figure 8 shown, including:

[0241] Structure parameter correction module: Obtain the initial classification scheme, and parse the first structure parameter set and the first operating condition parameter set from the initial classification scheme; based on the first structure parameter set and the first operating condition parameter set, perform adaptive calibration on the first structure parameter set to generate the second structure parameter set;

[0242] Fluctuation evaluation module: Construct a distributed particle size analysis network, and obtain the second particle size feature set through the distributed particle size analysis network; measure the multi-source data during the operation of the n-stage cyclone in real time, and obtain the second operating condition parameter set according to the multi-source data; perform classification fluctuation evaluation based on the second particle size feature set and the second operating condition parameter set to obtain the classification fluctuation evaluation result;

[0243] Hierarchical control module: According to the hierarchical fluctuation evaluation results, optimize the first operating condition parameter set and the second structural parameter set to obtain an optimized parameter set, where the optimized parameter set includes a first optimized parameter set, a second optimized parameter set, and a third optimized parameter set; precisely control the cyclone classification process according to the optimized parameter set.

[0244] In the structural parameter correction module, obtaining the initial classification scheme and parsing the first structural parameter set and the first operating condition parameter set from the initial classification scheme includes:

[0245] Step S1111, the first structural parameter set includes the cone angle ratio sequence [α'1, α'2,..., α' n and the overflow pipe diameter ratio sequence [D'1, D'2,..., D' n ; where α' i is the cone angle ratio of the i-th stage cyclone, specifically the ratio of the cone angle α i of the i-th stage cyclone to the reference cone angle α0, and D' i is the overflow pipe diameter ratio of the i-th stage cyclone, specifically the ratio of the overflow pipe diameter D i of the i-th stage cyclone to the feed port diameter D0, 1 ≤ i ≤ n;

[0246] Step S1112, the first operating condition parameter set includes the feed concentration reference value C0, the pressure gradient reference value ΔP0, and the flow ratio reference value Q'.

[0247] The step S1112 includes:

[0248] Step S11121, the feed concentration reference value C0 is dynamically calibrated according to the material density distribution histogram;

[0249] Step S11122, the pressure gradient reference value ΔP0 is calculated by the Reynolds number correction formula;

[0250] Step S11123, the flow ratio reference value Q' = Q0 / Q1, where Q0 is the designed flow rate and Q1 is the actual flow rate; the flow ratio reference value Q' is dynamically adjusted by the Bernoulli equation.

[0251] In the structural parameter correction module, adaptively calibrating the first structural parameter set based on the first structural parameter set and the first operating condition parameters to generate the second structural parameter set includes:

[0252] According to the first structural parameter set and the first operating condition parameter set, obtain the structural error compensation coefficient matrix k;

[0253] Based on the structural error compensation coefficient matrix k, adaptively calibrate the first set of structural parameters to generate a second set of structural parameters, and iteratively optimize the second set of structural parameters.

[0254] Obtaining the structural error compensation coefficient matrix k according to the first set of structural parameters and the first set of operating condition parameters includes:

[0255] Step S1121, under standard material conditions, using the cone angle ratio sequence [α'1, α'2,..., α' n and the overflow pipe diameter ratio sequence [D'1, D'2,..., D' n as a reference, conduct a baseline test on the n-stage cyclone; the standard material is a material with a particle size distribution conforming to a Gaussian distribution;

[0256] Step S1122, under the operating conditions of the first set of operating condition parameters, collect the overflow particle size distribution curve and the underflow concentration curve of the n-stage cyclone, and extract the actual overflow particle size median sequence and the actual underflow concentration sequence ; where is the actual overflow particle size median of the i-th stage cyclone, is the actual underflow concentration of the i-th stage cyclone;

[0257] Step S1123, obtain the theoretical overflow particle size median sequence of the n-stage cyclone, where is the theoretical overflow particle size median of the i-th stage cyclone; according to the theoretical overflow particle size median sequence and the actual overflow particle size median sequence , calculate the overflow particle size median deviation sequence of the n-stage cyclone, where is the overflow particle size median deviation of the i-th stage cyclone;

[0258] The method for calculating includes:

[0259]

[0260] Step S1124, calculate the standard deviation sequence of the actual underflow concentration of the n-stage cyclone, and calculate the underflow concentration fluctuation coefficient sequence of the n-stage cyclone according to the standard deviation sequence and the actual underflow concentration sequence , where is the underflow concentration fluctuation coefficient of the i-th stage cyclone;

[0261] The method for calculating includes:

[0262]

[0263] Among them, is the standard deviation of the actual underflow concentration of the i-th stage cyclone;

[0264] Step S1125, based on the overflow particle size median deviation sequence and the underflow concentration fluctuation coefficient sequence, construct a structural error compensation coefficient matrix , where is the structural error compensation coefficient of the i-th stage cyclone.

[0265] The method for calculating includes:

[0266]

[0267] Among them, is the non-linear weight coefficient, used to capture the non-linear relationship in error compensation, and is obtained by fitting experimental data; is the high-order non-linear coefficient, used to further capture the complex relationship in error compensation, and is also obtained by fitting experimental data. and are non-linear terms, used to capture more complex error compensation relationships.

[0268] The adaptive calibration of the first set of structural parameters based on the structural error compensation coefficient matrix k to generate a second set of structural parameters and the iterative optimization of the second set of structural parameters include:

[0269] Step S1131, according to the structural error compensation coefficient matrix k, calibrate the cone angle ratio sequence [α'1, α'2,..., α' n in the first set of structural parameters to obtain the calibrated cone angle ratio sequence [α''1, α''2,..., α'' n ;

[0270] The calibration of the cone angle ratio sequence [α'1, α'2,..., α' n in the first set of structural parameters includes:

[0271]

[0272] where α'' i is the cone angle ratio of the calibrated i-th stage cyclone, is the hyperbolic tangent function, used to smooth the correction amplitude of the cone angle ratio and avoid excessive parameter jumps;

[0273] Step S1132, according to the structural error compensation coefficient matrix k, for the overflow pipe diameter ratio sequence [D'1, D'2,..., D' nCalibration is performed to obtain the calibrated overflow pipe diameter ratio sequence [D''1, D''2,..., D'' n ;

[0274] The calibration of the overflow pipe diameter ratio sequence [D'1, D'2,..., D' n in the first set of structural parameters includes:

[0275]

[0276] where is the overflow pipe diameter ratio of the i-th cyclone after calibration, The item reflects the non-linear correction characteristic of the overflow pipe diameter ratio, that is, when the overflow pipe diameter ratio is larger, the larger the correction amount of the overflow pipe diameter ratio caused by the same overflow particle size median deviation;

[0277] Step S1133, based on the calibrated cone angle ratio sequence [α''1, α''2,..., α'' n and the calibrated overflow pipe diameter ratio sequence [D''1, D''2,..., D'' n , generate the second set of structural parameters;

[0278] Step S1134, according to the overflow particle size distribution curve and the underflow concentration curve, recalculate the overflow particle size median deviation sequence and the underflow concentration fluctuation coefficient sequence under the second set of structural parameters, where is the overflow particle size median deviation of the i-th cyclone under the second set of structural parameters, is the underflow concentration fluctuation coefficient of the i-th cyclone under the second set of structural parameters;

[0279] Step S1135, preset the deviation convergence accuracy and the fluctuation convergence accuracy , traverse the overflow particle size median deviation sequence and the underflow concentration fluctuation coefficient sequence under the second set of structural parameters, and judge whether it satisfies and , if not satisfied, return to step S1131 for repeated iteration until satisfied.

[0280] In the fluctuation evaluation module, the construction of the distributed particle size analysis network to obtain the second set of particle size characteristics through the distributed particle size analysis network includes:

[0281] Step S1211, deploy on-line particle size analyzers at the overflow and underflow ports of the n-stage cyclone to form a distributed particle size analysis network;

[0282] Step S1212, collect the particle size distribution data of the overflow and underflow of the n-stage hydrocyclone in real time through a distributed particle size analysis network to obtain a second particle size feature set; the second particle size feature set includes an overflow particle size skewness index sequence S = [S1, S2,..., S n , an underflow particle size kurtosis index sequence K = [K1, K2,..., K n , and a cross-stage particle size covariance matrix ; where S i is the overflow particle size skewness index of the i-th stage hydrocyclone, and K i is the underflow particle size kurtosis index of the i-th stage hydrocyclone.

[0283] In the fluctuation evaluation module, the multi-source data of the n-stage hydrocyclone during operation is measured in real time, and the second working condition parameter set obtained from the multi-source data includes:

[0284] Step S1221, deploy pressure sensors, flow meters and concentration meters at corresponding positions of the n-stage hydrocyclone to measure the multi-source data of the n-stage hydrocyclone during operation in real time. The multi-source data includes feed pressure, ore volume flow and feed concentration;

[0285] Step S1222, calculate the dynamic feed concentration gradient according to the feed concentration ;

[0286]

[0287] where C(t) is the feed concentration at the current sampling time t, C(t - Δt) is the feed concentration at the previous sampling time t - Δt, and Δt is the sampling period;

[0288] Step S1223, calculate the pressure pulsation amplitude according to the feed pressure ;

[0289] Step S1224, calculate the flow variation coefficient according to the ore volume flow 。

[0290] In the fluctuation evaluation module, perform hierarchical fluctuation evaluation based on the second particle size feature set and the second working condition parameter set, and the obtained hierarchical fluctuation evaluation results include:

[0291] Step S1231, judge whether it is a first-level fluctuation based on the cross-stage particle size covariance matrix in the second particle size feature set;

[0292] Judging whether it is a first-level fluctuation based on the cross-stage particle size covariance matrix in the second particle size feature set includes:

[0293] Calculate the cross-stage particle size covariance matrix The maximum eigenvalue λ max , if λ max >λ α , it is a first-level fluctuation, where λ α is the preset first-level fluctuation threshold.

[0294] Step S1232, determine whether it is a second-level fluctuation based on the second set of operating condition parameters;

[0295] The determination of whether it is a second-level fluctuation based on the second set of operating condition parameters includes:

[0296] Preset the second-level fluctuation threshold λ β , if the product of the pressure pulsation amplitude and the flow variation coefficient exceeds λ β , it is a second-level fluctuation.

[0297] Step S1233, determine whether it is a third-level fluctuation based on the overflow particle size skewness index sequence S in the second particle size feature set.

[0298] In the hierarchical control module, the obtaining of the optimized parameter set includes:

[0299] Step S2100, if the hierarchical fluctuation evaluation result is a first-level fluctuation, perform transfer learning optimization on the first set of operating condition parameters and the second set of structural parameters to obtain the first optimized parameter set;

[0300] Step S2200, if the hierarchical fluctuation evaluation result is a second-level fluctuation, optimize the first set of operating condition parameters to obtain the second optimized parameter set;

[0301] Step S2300, if the hierarchical fluctuation evaluation result is a third-level fluctuation, optimize the first set of operating condition parameters to obtain the third optimized parameter set;

[0302] Step S2400, construct the optimized parameter set from the first optimized parameter set, the second optimized parameter set, and the third optimized parameter set, and precisely control the cyclone classification process according to the optimized parameter set.

[0303] The said Step S2100 includes:

[0304] Step S2110, retrieve the historical optimization cases similar to the current operating condition from the historical operating condition database according to the preset matching conditions;

[0305] Step S2120, extract the historical cone angle ratio sequence corresponding to the historical optimization cases, and optimize the calibrated cone angle ratio sequence [α''1, α''2,..., α'' n in the second set of structural parameters to obtain the optimized cone angle ratio sequence [α'''1, α'''2,..., α''' n; where, α''' i is the optimized cone angle ratio of the i-th stage cyclone;

[0306] Step S2130, according to λ max and the first-stage fluctuation threshold λ α calculate the over-limit amplitude of the eigenvalue, and optimize the pressure gradient reference value ΔP0 in the first set of operating parameters according to the over-limit amplitude of the eigenvalue and the structural error compensation coefficient matrix , to obtain the first compensated pressure gradient ΔP c ;

[0307] Step S2140, according to the calibrated overflow pipe diameter ratio sequence [D''1, D''2,..., D'' n , the optimized cone angle ratio sequence [α'''1, α'''2,..., α''' n , the first compensated pressure gradient ΔP c , the ore concentration reference value C0 and the flow ratio reference value Q', construct the first optimized parameter set.

[0308] The said step S2200 includes:

[0309] Step S2210, construct a pressure-flow coupling equation according to the second set of operating parameters;

[0310] Step S2220, design a pressure-flow decoupling control strategy based on the pressure-flow coupling equation; according to the pressure-flow decoupling control strategy, obtain the adjusted feed ore concentration reference value C'0;

[0311] Step S2230, according to the calibrated overflow pipe diameter ratio sequence [D''1, D''2,..., D'' n , the calibrated cone angle ratio sequence [α''1, α''2,..., α'' n , the adjusted feed ore concentration reference value C'0, the pressure gradient reference value ΔP0 and the flow ratio reference value Q', construct the second optimized parameter set.

[0312] The said step S2300 includes:

[0313] Step S2310, calculate the cross-stage particle size equilibrium factor ξ according to the second particle size feature set;

[0314] Step S2320, calculate the correction weight of the cone angle ratio of the i-th stage cyclone based on the cross-stage particle size equilibrium factor ξ ;

[0315] Step S2330, according to the correction weight , calculate the corrected cone angle ratio of the -th stage cyclone ;

[0316] Step S2340, from the correction cone angle ratio of the stage cyclone, to obtain a correction cone angle ratio sequence ;

[0317] Step S2350, according to the calibrated overflow pipe diameter ratio sequence [D''1, D''2,..., D'' n , the correction cone angle ratio sequence , the pressure gradient reference value ΔP0, the ore concentration reference value C0, and the flow ratio reference value Q', construct a third optimization parameter set.

[0318] The methods and systems of the present application can be implemented in many ways. For example, the methods and systems of the present application can be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is only for illustration, and the steps of the method of the present application are not limited to the specific order described above, unless otherwise specifically stated.

[0319] In addition, parts of the above technical solutions provided in the embodiments of the present application that are consistent with the corresponding technical solutions in the prior art in terms of implementation principles are not described in detail to avoid excessive elaboration.

[0320] As described in the above specific embodiments, the purpose, technical solutions, and beneficial effects of the present invention are further described in detail. It should be understood that the above is only the specific embodiment of the present invention and is not used to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An intelligent regulation method for grading fineness based on a mathematical model, characterized in that The method includes: Obtain an initial classification scheme, and parse a first set of structural parameters and a first set of operating parameters from the initial classification scheme; based on the first set of structural parameters and the first set of operating parameters, perform adaptive calibration on the first set of structural parameters to generate a second set of structural parameters; construct a distributed particle size analysis network, and obtain a second set of particle size characteristics through the distributed particle size analysis network; measure multi-source data during the operation of the n-stage cyclone in real time, and obtain a second set of operating parameters based on the multi-source data; perform classification fluctuation evaluation based on the second set of particle size characteristics and the second set of operating parameters to obtain a classification fluctuation evaluation result; According to the classification fluctuation evaluation result, optimize the first set of operating parameters and the second set of structural parameters to obtain an optimized parameter set, where the optimized parameter set includes a first optimized parameter set, a second optimized parameter set, and a third optimized parameter set; precisely control the cyclone classification process according to the optimized parameter set.

2. The intelligent regulation method for grading fineness based on a mathematical model according to claim 1, wherein The first set of structural parameters includes the cone angle ratio sequence [α'1, α'2,..., α' n and the overflow pipe diameter ratio sequence [D'1, D'2,..., D' n ; where α' i is the cone angle ratio of the i-th cyclone, specifically the ratio of the cone angle α i of the i-th cyclone to the reference cone angle α0, and D' i is the overflow pipe diameter ratio of the i-th cyclone, specifically the ratio of the overflow pipe diameter D i of the i-th cyclone to the feed inlet diameter D0, 1 ≤ i ≤ n; the first set of operating parameters includes the reference feed concentration C0, the reference pressure gradient ΔP0, and the reference flow ratio Q'.

3. The intelligent regulation method for grading fineness based on a mathematical model according to claim 2, characterized in that, The performing adaptive calibration on the first set of structural parameters to generate a second set of structural parameters includes: According to the first set of structural parameters and the first set of operating parameters, obtain a structural error compensation coefficient matrix k; based on the structural error compensation coefficient matrix k, perform adaptive calibration on the first set of structural parameters to generate a second set of structural parameters, and perform iterative optimization on the second set of structural parameters.

4. The intelligent regulation method for grading fineness based on a mathematical model according to claim 3, characterized in that, The obtaining the structural error compensation coefficient matrix k includes: Under standard material conditions, taking the cone angle ratio sequence [α'1, α'2,..., α' n and the overflow pipe diameter ratio sequence [D'1, D'2,..., D' n as the reference, under the working conditions of the first set of working condition parameters, collect the overflow particle size distribution curve and the underflow concentration curve of the n-stage cyclone, and extract the actual overflow particle size median sequence and the actual underflow concentration sequence ; where is the actual overflow particle size median of the -stage cyclone, and is the actual underflow concentration of the i-stage cyclone; Obtain the theoretical overflow particle size median sequence of the n-stage cyclone , where is the theoretical overflow particle size median of the -th stage cyclone; According to the theoretical overflow particle size median sequence and the actual overflow particle size median sequence , calculate the overflow particle size median deviation sequence of the n-stage cyclone , where is the overflow particle size median deviation of the i-th stage cyclone; Calculate the standard deviation sequence of the actual underflow concentration of the n-stage cyclone. According to the standard deviation sequence and the actual underflow concentration sequence Calculate the underflow concentration fluctuation coefficient sequence of the n-stage cyclone , where is the underflow concentration fluctuation coefficient of the i-th stage cyclone; Construct a structural error compensation coefficient matrix based on the overflow particle size median deviation sequence and the underflow concentration fluctuation coefficient sequence , where is the structural error compensation coefficient of the i-th cyclone stage 5. The intelligent regulation method for grading fineness based on a mathematical model according to claim 4, characterized in that The based on the structural error compensation coefficient matrix k, performing adaptive calibration on the first set of structural parameters to generate a second set of structural parameters, and performing iterative optimization on the second set of structural parameters includes: Step S1131, according to the structural error compensation coefficient matrix k, calibrate the cone angle ratio sequence [α'1, α'2,..., α' n in the first set of structural parameters to obtain the calibrated cone angle ratio sequence [α''1, α''2,..., α'' n ; Step S1132, according to the structural error compensation coefficient matrix k, calibrate the overflow pipe diameter ratio sequence [D'1, D'2,..., D' n to obtain the calibrated overflow pipe diameter ratio sequence [D''1, D''2,..., D'' n ; Step S1133, generate a second set of structural parameters based on the calibrated cone angle ratio sequence [α''1, α''2,..., α'' n and the calibrated overflow pipe diameter ratio sequence [D''1, D''2,..., D'' n ; Step S1134, according to the overflow particle size distribution curve and the underflow concentration curve, recalculate the overflow particle size median deviation sequence under the second set of structural parameters and the underflow concentration fluctuation coefficient sequence , where is the overflow particle size median deviation of the i-th stage cyclone under the second set of structural parameters, is the underflow concentration fluctuation coefficient of the i-th stage cyclone under the second set of structural parameters; Step S1135, preset deviation convergence accuracy and fluctuation convergence accuracy , traverse the overflow particle size median deviation sequence and underflow concentration fluctuation coefficient sequence under the second structural parameter set, and judge whether it satisfies and . If not satisfied, return to step S1131 for repeated iteration until satisfied.

6. The intelligent regulation method for grading fineness based on a mathematical model according to claim 1, wherein The constructing a distributed particle size analysis network and obtaining a second set of particle size characteristics through the distributed particle size analysis network includes: Deploy on-line particle size analyzers at the overflow port and underflow port of the n-stage cyclone to form a distributed particle size analysis network; Collect the particle size distribution data of the overflow and underflow of the n-stage cyclone in real time through a distributed particle size analysis network to obtain a second particle size feature set; the second particle size feature set includes an overflow particle size skewness index sequence S = [S1, S2,..., S n , an underflow particle size kurtosis index sequence K = [K1, K2,..., K n , and a cross-stage particle size covariance matrix ; where S i is the overflow particle size skewness index of the i-th stage cyclone, and K i is the underflow particle size kurtosis index of the i-th stage cyclone.

7. The intelligent regulation method for grading fineness based on a mathematical model according to claim 6, wherein, The measuring multi-source data during the operation of the n-stage cyclone in real time and obtaining a second set of operating parameters based on the multi-source data includes: Deploy pressure sensors, flow meters, and concentration meters at corresponding positions of the n-stage cyclone to measure multi-source data during the operation of the n-stage cyclone in real time, where the multi-source data includes feed pressure, ore volume flow, and feed concentration; Calculate the dynamic feed concentration gradient based on the feed concentration ; Calculate the amplitude of pressure pulsation based on the feed pressure ; Calculate the coefficient of variation of flow rate based on the ore volume flow rate .

8. The intelligent regulation method for grading fineness based on a mathematical model according to claim 7, characterized in that The classification fluctuation evaluation result includes first-level fluctuation, second-level fluctuation, and third-level fluctuation; The obtaining the classification fluctuation evaluation result includes: Based on the cross-segment granularity covariance matrix in the second granularity feature set Determine whether it is a first-level fluctuation; Based on the second set of operating parameters, determine whether it is a second-level fluctuation; Based on the overflow particle size skewness index sequence S in the second set of particle size characteristics, determine whether it is a third-level fluctuation.

9. The intelligent regulation method for grading fineness based on a mathematical model according to claim 8, wherein The cross-segment granularity covariance matrix based on the second granularity feature set Determining whether it is a first-level fluctuation includes: Calculate the cross-segment granularity covariance matrix in the second granularity feature set for the maximum eigenvalue λ max . If λ max > λ α , it is a first-level fluctuation, where λ α is the preset first-level fluctuation threshold.

10. The intelligent regulation method for grading fineness based on a mathematical model according to claim 8, characterized in that, The based on the second set of operating parameters to determine whether it is a second-level fluctuation includes: Preset secondary fluctuation threshold λ β , if the product of the pressure pulsation amplitude and the flow variation coefficient exceeds λ β , it is a secondary fluctuation.

11. The intelligent regulation method for grading fineness based on a mathematical model according to claim 8, characterized in that, The based on the overflow particle size skewness index sequence S in the second set of particle size characteristics to determine whether it is a third-level fluctuation includes: Set the normal range [L0, H0] of the overflow particle size skewness index. If S continuously exceeds the normal range [L0, H0] for λ N times, it is a third-level fluctuation, where L0 is the lower limit of the normal range of the overflow particle size skewness index, H0 is the upper limit of the normal range of the overflow particle size skewness index, and λ N is the preset third-level fluctuation threshold.

12. The intelligent regulation method for grading fineness based on a mathematical model according to claim 8, characterized in that, The optimizing the first set of operating parameters and the second set of structural parameters to obtain an optimized parameter set includes: If the classification fluctuation evaluation result is a first-level fluctuation, perform transfer learning optimization on the first set of operating parameters and the second set of structural parameters to obtain a first optimized parameter set; If the classification fluctuation evaluation result is a second-level fluctuation, optimize the first set of operating parameters to obtain a second optimized parameter set; If the classification fluctuation evaluation result is a third-level fluctuation, optimize the first set of operating parameters to obtain a third optimized parameter set.

13. The intelligent regulation method for grading fineness based on a mathematical model according to claim 12, wherein The obtaining the third optimized parameter set includes: Calculate the cross-segment granularity balance factor ξ according to the second granularity feature set; Calculate the corrected weight of the cone angle ratio of the i-th cyclone section based on the cross-section granularity balance factor ξ ; According to the correction weight , calculate the correction cone angle ratio of the -th stage cyclone ; From the modified cone angle ratio of the section cyclone being , the modified cone angle ratio sequence is obtained; According to the calibrated overflow pipe diameter ratio sequence [D''1, D''2,..., D'' n , the corrected cone angle ratio sequence , the pressure gradient reference value ΔP0, the ore concentration reference value C0, and the flow rate ratio reference value Q', construct the third optimization parameter set.

14. A grading fineness intelligent control system based on a mathematical model, which is used to implement the grading fineness intelligent control method based on the mathematical model described in any one of claims 1-13, and is characterized in that, The system includes: Structure parameter correction module: Obtain the initial classification scheme, and parse the first structure parameter set and the first working condition parameter set from the initial classification scheme; Based on the first structure parameter set and the first working condition parameters, adaptively calibrate the first structure parameter set to generate the second structure parameter set; Fluctuation evaluation module: Construct a distributed granularity analysis network, and obtain the second granularity feature set through the distributed granularity analysis network; Real-time measure the multi-source data during the operation of the n-stage cyclone, and obtain the second working condition parameter set according to the multi-source data; Based on the second granularity feature set and the second working condition parameter set, conduct a classification fluctuation evaluation to obtain a classification fluctuation evaluation result; Classification control module: According to the classification fluctuation evaluation result, optimize the first working condition parameter set and the second structure parameter set to obtain an optimized parameter set, where the optimized parameter set includes a first optimized parameter set, a second optimized parameter set, and a third optimized parameter set; Accurately control the cyclone classification process according to the optimized parameter set.

Images