Automatic control method and device for washing water volume of microbubble flotation machine

Abstract

The present application relates to the field of intelligent control technology, and specifically discloses a method and device for automatically controlling the amount of washing water of a micro-bubble flotation machine, which uses artificial intelligence technology based on deep learning to perform time-series analysis on the level and concentration of the ore pulp in the flotation tank, so as to capture the time-series correlation fusion characteristics of the ore pulp level and concentration, and to dig out the time-series correlation influence relationship between the ore pulp level and concentration on the thickness of the foam layer, so as to adaptively adjust the flow rate of the washing water. In this way, the automation level of the micro-bubble flotation machine can be improved, the precise control of the amount of washing water of the micro-bubble flotation machine can be achieved, the flotation efficiency and mineral recovery rate can be improved, and the waste of resources can be reduced.

Description

Technical Field

[0001] The present application relates to the field of intelligent control technology, and more specifically, to a method and device for automatically controlling the amount of cleaning water of a microbubble flotation machine. Background Art

[0002] Microbubble flotation machine is a kind of equipment widely used in the field of mineral sorting. Its basic working principle is to use the adsorption effect between bubbles and mineral particles to achieve effective separation of minerals. Specifically, when the microbubble flotation machine is running, the feed slurry is pumped into the high-pressure feed pipeline by a high-pressure pump, and the pressure provided by the high-pressure pump forms a high-speed jet at the outlet. The high-speed downward impacting slurry jet is mixed with the air provided by the self-priming inflation valve in the downpipe. The mineral particles with strong hydrophobicity will adhere to the bubbles to form mineralized bubbles. The mineralized bubbles float to the liquid surface under the action of buoyancy to form a foam layer. The bubbles and particles in the foam layer are further cleaned to improve the grade of the concentrate. In this process, by accurately controlling the amount of cleaning water, the thickness of the foam layer can be controlled to avoid the obstruction of foam overflow caused by the excessive thickness of the foam layer, and maintain the appropriate pulp concentration in the flotation tank, thereby improving the flotation efficiency and selectivity, so that valuable minerals can be separated from worthless minerals more effectively.

[0003] In actual production, the existing methods for controlling the amount of water used to clean microbubble flotation machines mostly rely on manual experience settings. However, since the properties of minerals, the diameter and density of bubbles, the rotation speed of the flotation machine and other conditions may change during the mineral sorting process, the demand for the amount of clean water also changes accordingly. This control method based on manual experience often cannot adapt to these changes in a timely and accurate manner, resulting in unstable flotation effects and low mineral recovery rates. Therefore, a method and device for automatically controlling the amount of clean water used in a microbubble flotation machine is desired. Summary of the invention

[0004] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a method and device for automatically controlling the amount of cleaning water of a micro-bubble flotation machine, which uses artificial intelligence technology based on deep learning to perform time series analysis on the liquid level and concentration of the ore pulp in the flotation tank, so as to capture the time series correlation fusion characteristics of the ore pulp liquid level and concentration, and to mine the time series correlation influence relationship of the ore pulp liquid level and concentration on the thickness of the foam layer, so as to adaptively adjust the flow rate of the cleaning water. In this way, the automation level of the micro-bubble flotation machine can be improved, the precise control of the amount of cleaning water of the micro-bubble flotation machine can be achieved, the flotation efficiency and mineral recovery rate can be improved, and the waste of resources can be reduced.

[0005] Accordingly, according to one aspect of the present application, a method for automatically controlling the amount of cleaning water for a microbubble flotation machine is provided, comprising: a liquid level measurement step: measuring the liquid level of the slurry in the flotation cell to obtain a time series of the slurry level; a data acquisition step: acquiring a time series of the slurry concentration and a time series of the foam layer thickness through a sensor group; a control algorithm step: determining an adjustment strategy for the amount of cleaning water based on the time series of the slurry level, the time series of the slurry concentration and the time series of the foam layer thickness, wherein the adjustment strategy for the amount of cleaning water is a recommended flow rate value for the amount of cleaning water; an actuator step: specifying a valve as an actuator for adjusting the amount of cleaning water; an automatic adjustment step: adjusting the opening value of the valve based on the adjustment strategy for the amount of cleaning water.

[0006] In the above-mentioned automatic control method for the cleaning water volume of the microbubble flotation machine, the control algorithm steps include: respectively extracting the time series pattern characteristics of the time series of the slurry level, the time series of the slurry concentration and the time series of the foam layer thickness to obtain the slurry level time series correlation feature vector, the slurry concentration time series correlation feature vector and the foam layer thickness time series correlation feature vector; using a dynamic element-level feature fusion module to perform feature element-by-element fusion on the slurry level time series correlation feature vector and the slurry concentration time series correlation feature vector to obtain a slurry multi-dimensional time series correlation feature vector; inputting the slurry multi-dimensional time series correlation feature vector and the foam layer thickness time series correlation feature vector into a responsive gamma correction fusion module to obtain a slurry parameter-foam layer thickness time series responsiveness correlation feature vector; and determining the adjustment strategy of the cleaning water volume based on the slurry parameter-foam layer thickness time series responsiveness correlation feature vector.

[0007] In the above-mentioned automatic control method for the amount of cleaning water of the microbubble flotation machine, the time series pattern features of the time series of the slurry level, the time series of the slurry concentration and the time series of the foam layer thickness are respectively extracted to obtain the slurry level time series associated feature vector, the slurry concentration time series associated feature vector and the foam layer thickness time series associated feature vector, including: arranging the time series of the slurry level, the time series of the slurry concentration and the time series of the foam layer thickness into a slurry level time series input vector, a slurry concentration time series input vector and a foam layer thickness time series input vector according to the time dimension respectively; passing the slurry level time series input vector, the slurry concentration time series input vector and the foam layer thickness time series input vector respectively through a time series pattern feature extractor based on a one-dimensional convolutional layer to obtain the slurry level time series associated feature vector, the slurry concentration time series associated feature vector and the foam layer thickness time series associated feature vector.

[0008] In the above-mentioned automatic control method for the amount of cleaning water of the microbubble flotation machine, a dynamic element-level feature fusion module is used to perform feature element-by-element fusion on the slurry level time series correlation feature vector and the slurry concentration time series correlation feature vector to obtain a slurry multi-dimensional time series correlation feature vector, including: calculating the information fusion balance threshold value between the slurry level time series correlation feature vector and the slurry concentration time series correlation feature vector; based on the information fusion balance threshold value, performing feature element-by-element fusion on the slurry level time series correlation feature vector and the slurry concentration time series correlation feature vector to obtain the slurry multi-dimensional time series correlation feature vector.

[0009] In the above-mentioned automatic control method for the amount of cleaning water of the microbubble flotation machine, the information fusion balance threshold value between the slurry level time series correlation feature vector and the slurry concentration time series correlation feature vector is calculated, including: cascading the slurry level time series correlation feature vector and the slurry concentration time series correlation feature vector to obtain a cascade feature vector; multiplying the cascade feature vector and the transformation vector to obtain a threshold scoring coefficient; adding the threshold scoring coefficient and the bias parameter to obtain a bias threshold scoring coefficient; and passing the bias threshold scoring coefficient through a sigmoid activation function to obtain the information fusion balance threshold value.

[0010] In the above-mentioned automatic control method for the amount of cleaning water of the microbubble flotation machine, based on the information fusion balance threshold value, the slurry level time series correlation feature vector and the slurry concentration time series correlation feature vector are feature-by-element fused to obtain the slurry multi-dimensional time series correlation feature vector, including: respectively calculating the natural exponential function values ​​of each characteristic value in the slurry level time series correlation feature vector and the slurry concentration time series correlation feature vector to obtain an indexed slurry level time series correlation feature vector and an indexed slurry concentration time series correlation feature vector; determining a first weight and a second weight based on the information fusion balance threshold value, and weighting the indexed slurry level time series correlation feature vector position by position with the first weight to obtain a weighted slurry level time series correlation feature vector, and weighting the indexed slurry concentration time series correlation feature vector position by position with the second weight to obtain a weighted slurry concentration time series correlation feature vector; and adding the corresponding elements of the weighted slurry level time series correlation feature vector and the weighted slurry concentration time series correlation feature vector to obtain the slurry multi-dimensional time series correlation feature vector.

[0011] In the above-mentioned automatic control method for the amount of cleaning water of the microbubble flotation machine, the multi-dimensional time-series correlation feature vector of the ore pulp and the time-series correlation feature vector of the foam layer thickness are input into a responsive gamma correction fusion module to obtain a ore pulp parameter-foam layer thickness time-series responsiveness correlation feature vector, including: calculating the position-by-position division of the multi-dimensional time-series correlation feature vector of the ore pulp and the time-series correlation feature vector of the foam layer thickness to obtain a ore pulp parameter-foam layer thickness response feature vector; respectively calculating the first nonlinear response factor and the second nonlinear response factor of each eigenvalue in the ore pulp parameter-foam layer thickness response feature vector, and dividing the first nonlinear response factor by the second nonlinear response factor to obtain the eigenvalue of a predetermined position of the ore pulp parameter-foam layer thickness time-series responsiveness correlation feature vector.

[0012] In the above-mentioned automatic control method of the cleaning water volume of the microbubble flotation machine, the adjustment strategy of the cleaning water volume is determined based on the slurry parameter-foam layer thickness time series responsiveness association characteristic vector, including: passing the slurry parameter-foam layer thickness time series responsiveness association characteristic vector through a classifier-based adjustment strategy generator to obtain the adjustment strategy of the cleaning water volume.

[0013] In the above-mentioned automatic control method for the amount of cleaning water of the microbubble flotation machine, a training step is also included: training the temporal pattern feature extractor based on the one-dimensional convolution layer, the dynamic element-level feature fusion module, the responsive gamma correction fusion module and the classifier-based adjustment strategy generator; wherein the training step includes: acquiring training data, the training data including the time series of training slurry level, the time series of training slurry concentration and the time series of training foam layer thickness, as well as the true value of the adjustment strategy for the amount of cleaning water; arranging the time series of the training slurry level, the time series of the training slurry concentration and the time series of the training foam layer thickness respectively according to the time dimension as a training slurry level time series input vector, a training slurry concentration time series input vector and a training foam layer thickness time series input vector; passing the training slurry level time series input vector, the training slurry concentration time series input vector and the training foam layer thickness time series input vector respectively through the temporal pattern feature extractor based on the one-dimensional convolution layer to obtain a training slurry level time series associated feature vector, a training slurry level time series associated feature vector, a training slurry level time series associated feature vector and a training slurry level time series associated feature vector. The invention relates to a time series correlation feature vector of slurry concentration and a time series correlation feature vector of training foam layer thickness; the dynamic element-level feature fusion module is used to perform feature element-by-element fusion on the time series correlation feature vector of training slurry level and the time series correlation feature vector of training slurry concentration to obtain a multi-dimensional time series correlation feature vector of training slurry; the multi-dimensional time series correlation feature vector of training slurry and the time series correlation feature vector of training foam layer thickness are input into the responsive gamma correction fusion module to obtain a time series responsiveness correlation feature vector of training slurry parameter-foam layer thickness; the time series responsiveness correlation feature vector of training slurry parameter-foam layer thickness is passed through the adjustment strategy generator based on the classifier to obtain a classification loss function value; the time series pattern feature extractor based on one-dimensional convolutional layer, the dynamic element-level feature fusion module, the responsive gamma correction fusion module and the adjustment strategy generator based on the classifier are trained with the classification loss function value, wherein, in each round of iteration of the training, the time series responsiveness correlation feature vector of training slurry parameter-foam layer thickness is optimized.

[0014] According to another aspect of the present application, there is provided an automatic control device for the amount of washing water of a micro-bubble flotation machine, which includes: a data time series association analysis module, which is used to extract the time series pattern characteristics of the time series of the slurry level, the time series of the slurry concentration and the time series of the foam layer thickness respectively to obtain the slurry level time series association feature vector, the slurry concentration time series association feature vector and the foam layer thickness time series association feature vector; a slurry multi-dimensional time series association analysis module, which is used to use a dynamic element-level feature fusion module to perform feature element-by-element fusion on the slurry level time series association feature vector and the slurry concentration time series association feature vector to obtain the slurry multi-dimensional time series association feature vector; a time series responsiveness analysis module, which is used to input the slurry multi-dimensional time series association feature vector and the foam layer thickness time series association feature vector into a responsiveness gamma correction fusion module to obtain the slurry parameter-foam layer thickness time series responsiveness association feature vector; and an adjustment strategy generation module, which is used to determine the adjustment strategy of the washing water amount based on the slurry parameter-foam layer thickness time series responsiveness association feature vector.

[0015] Compared with the prior art, the automatic control method and device for the washing water volume of the micro-bubble flotation machine provided by the present application adopts artificial intelligence technology based on deep learning to perform time series analysis on the liquid level and concentration of the ore pulp in the flotation tank, so as to capture the time series correlation fusion characteristics of the ore pulp liquid level and concentration, and to dig out the time series correlation influence relationship between the ore pulp liquid level and concentration and the foam layer thickness, so as to adaptively adjust the flow rate of the washing water volume. In this way, the automation level of the micro-bubble flotation machine can be improved, the precise control of the washing water volume of the micro-bubble flotation machine can be achieved, the flotation efficiency and mineral recovery rate can be improved, and the waste of resources can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0017] Figure 1 The present invention is a flowchart of the control algorithm steps in the method for automatically controlling the amount of cleaning water for a microbubble flotation machine according to an embodiment of the present application.

[0018] Figure 2 It is a schematic diagram of the architecture of the control algorithm steps in the method for automatically controlling the amount of washing water for a microbubble flotation machine according to an embodiment of the present application.

[0019] Figure 3A flowchart for extracting the time series pattern characteristics of the slurry level time series, the slurry concentration time series and the foam layer thickness time series in the automatic control method of the cleaning water volume of the microbubble flotation machine according to an embodiment of the present application to obtain the slurry level time series correlation feature vector, the slurry concentration time series correlation feature vector and the foam layer thickness time series correlation feature vector.

[0020] Figure 4 This is a flow chart of using a dynamic element-level feature fusion module in the automatic control method of the cleaning water volume of a microbubble flotation machine according to an embodiment of the present application to perform feature element-by-element fusion on the slurry level time series correlation feature vector and the slurry concentration time series correlation feature vector to obtain a slurry multi-dimensional time series correlation feature vector.

[0021] Figure 5 The present invention is a flowchart of the training steps in the method for automatically controlling the amount of cleaning water for a microbubble flotation machine according to an embodiment of the present application.

[0022] Figure 6 It is a block diagram of an automatic control device for the amount of cleaning water of a microbubble flotation machine according to an embodiment of the present application. DETAILED DESCRIPTION

[0023] Below, the embodiments of the present application will be described in more detail in conjunction with the accompanying drawings, and the above and other purposes, features and advantages of the present application will become more apparent. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the example embodiments described herein.

[0024] As mentioned in the above background technology, precise control of the amount of washing water in the microbubble flotation machine is of great significance for improving flotation efficiency and stabilizing the flotation process. If the amount of washing water is too large, the contact time between bubbles and mineral particles will be shortened, the adsorption effect will be insufficient, and the flotation effect will be reduced; on the contrary, if the amount of washing water is too small, the foam layer may be too thick, resulting in obstruction of foam overflow and impurities cannot be effectively removed, which will also affect the flotation effect. Therefore, in the actual application of the microbubble flotation machine, the amount of washing water should be dynamically adjusted according to the thickness of the foam layer in the microbubble flotation machine to ensure the stability and efficiency of the flotation process. At the same time, since the thickness of the foam layer is affected by factors such as the slurry level and slurry concentration, therefore, in the process of controlling the amount of washing water, it is necessary to comprehensively consider factors such as the slurry level, slurry concentration and foam layer thickness to achieve more accurate control of the amount of washing water.

[0025] However, most of the existing methods for controlling the amount of water used to clean the microbubble flotation machine rely on manual experience settings, and it is often difficult to adapt to changes in the thickness of the foam layer in a timely and accurate manner, resulting in unstable flotation effects and low mineral recovery rates. In response to the above technical problems, the technical concept of this application is to use artificial intelligence technology based on deep learning to perform time-series analysis on the level and concentration of the slurry in the flotation tank, so as to capture the time-series correlation fusion characteristics of the slurry level and concentration, and to dig out the time-series correlation influence relationship between the slurry level and concentration and the thickness of the foam layer, so as to adaptively adjust the flow rate of the cleaning water. In this way, the automation level of the microbubble flotation machine can be improved, precise control of the amount of water used to clean the microbubble flotation machine can be achieved, flotation efficiency and mineral recovery rate can be improved, and resource waste can be reduced.

[0026] Based on this, the present application proposes a method for automatically controlling the amount of cleaning water for a microbubble flotation machine, which includes: a liquid level measurement step: measuring the liquid level of the slurry in the flotation cell to obtain a time series of the slurry level; a data acquisition step: acquiring the time series of the slurry concentration and the time series of the foam layer thickness through a sensor group; a control algorithm step: determining an adjustment strategy for the amount of cleaning water based on the time series of the slurry level, the time series of the slurry concentration and the time series of the foam layer thickness, wherein the adjustment strategy for the amount of cleaning water is a recommended flow rate value for the amount of cleaning water; an actuator step: specifying a valve as an actuator for adjusting the amount of cleaning water; an automatic adjustment step: adjusting the opening value of the valve based on the adjustment strategy for the amount of cleaning water.

[0027] That is, considering that in the actual application of the microbubble flotation machine, as the slurry concentration increases, the solid particle content in the slurry increases, which leads to a corresponding increase in the solid content in the foam, making the foam layer thicker. However, too high a slurry concentration may cause a sharp deterioration in the slurry aeration, so that the ore particles and bubbles cannot flow freely, and may increase the friction between the ore particles, resulting in coarse particles. The mud, the probability of coarse particles falling off the bubble surface increases, thereby affecting the thickness of the foam layer. At the same time, when the slurry level increases, the thickness of the foam layer usually becomes thinner; conversely, when the slurry level decreases, the foam layer becomes thicker. Therefore, in the control algorithm step, it is expected to use deep learning technology to analyze the temporal correlation effect of the slurry level and slurry concentration on the thickness of the foam layer to determine the recommended flow rate value of the cleaning water.

[0028] Figure 1 The present invention is a flowchart of the control algorithm steps in the method for automatically controlling the amount of cleaning water for a microbubble flotation machine according to an embodiment of the present application. Figure 2 FIG. 1 is a schematic diagram of the control algorithm steps in the automatic control method of the washing water volume of the microbubble flotation machine according to the embodiment of the present application. Figure 1 and Figure 2As shown, the control algorithm steps of the automatic control method of the cleaning water volume of the microbubble flotation machine according to the embodiment of the present application include: S110, respectively extracting the time series pattern characteristics of the time series of the slurry level, the time series of the slurry concentration and the time series of the foam layer thickness to obtain the slurry level time series correlation feature vector, the slurry concentration time series correlation feature vector and the foam layer thickness time series correlation feature vector; S120, using a dynamic element-level feature fusion module to perform feature element-by-element fusion on the slurry level time series correlation feature vector and the slurry concentration time series correlation feature vector to obtain a slurry multi-dimensional time series correlation feature vector; S130, inputting the slurry multi-dimensional time series correlation feature vector and the foam layer thickness time series correlation feature vector into a responsive gamma correction fusion module to obtain a slurry parameter-foam layer thickness time series responsiveness correlation feature vector; S140, determining the adjustment strategy of the cleaning water volume based on the slurry parameter-foam layer thickness time series responsiveness correlation feature vector.

[0029] In the above-mentioned automatic control method for the amount of washing water of the micro-bubble flotation machine, the step S110 extracts the time series pattern characteristics of the time series of the slurry level, the time series of the slurry concentration and the time series of the foam layer thickness respectively to obtain the slurry level time series correlation feature vector, the slurry concentration time series correlation feature vector and the foam layer thickness time series correlation feature vector. Among them, Figure 3 The present invention is a flowchart of extracting the time series pattern characteristics of the slurry level, the slurry concentration and the foam layer thickness in the automatic control method of the washing water volume of the micro-bubble flotation machine according to the embodiment of the present application to obtain the slurry level time series correlation feature vector, the slurry concentration time series correlation feature vector and the foam layer thickness time series correlation feature vector. Figure 3 As shown, the step S110 includes: S111, arranging the time series of the slurry level, the time series of the slurry concentration and the time series of the foam layer thickness into a slurry level time series input vector, a slurry concentration time series input vector and a foam layer thickness time series input vector according to the time dimension; S112, passing the slurry level time series input vector, the slurry concentration time series input vector and the foam layer thickness time series input vector through a time series pattern feature extractor based on a one-dimensional convolutional layer to obtain the slurry level time series associated feature vector, the slurry concentration time series associated feature vector and the foam layer thickness time series associated feature vector.

[0030] Specifically, in step S111, the time series of the slurry level, the time series of the slurry concentration, and the time series of the foam layer thickness are arranged according to the time dimension into a slurry level time series input vector, a slurry concentration time series input vector, and a foam layer thickness time series input vector. It should be understood that, considering that the slurry level, concentration, and foam layer thickness are all dynamically changing over time, therefore, in order to better capture the changing trends of the slurry level, slurry concentration, and foam layer thickness over time, the time series of the slurry level, the time series of the slurry concentration, and the time series of the foam layer thickness are further arranged according to the time dimension to maintain the time order of the data and integrate the time series distribution information of the data, so as to better understand the dynamic behavior of the flotation process.

[0031] Specifically, in step S112, the slurry level time series input vector, the slurry concentration time series input vector and the foam layer thickness time series input vector are respectively passed through a time series pattern feature extractor based on a one-dimensional convolutional layer to obtain the slurry level time series associated feature vector, the slurry concentration time series associated feature vector and the foam layer thickness time series associated feature vector. It should be understood that the one-dimensional convolutional layer is a deep learning network layer, which is particularly suitable for processing one-dimensional time series data. In the technical solution of the present application, the time series pattern feature extractor based on the one-dimensional convolution layer can capture the local feature representation of the input vector, learn the dependency of the data on the local time step, and effectively extract the time series pattern characteristics of the slurry level, slurry concentration and foam layer thickness in the time dimension by performing one-dimensional convolution processing on the slurry level time series input vector, the slurry concentration time series input vector and the foam layer thickness time series input vector respectively, thereby revealing the trend and periodicity of the three changes over time, and thus more fully understanding the time series characteristics of the slurry level, slurry concentration and foam layer thickness.

[0032] In the above-mentioned automatic control method for the amount of washing water of the micro-bubble flotation machine, in the above-mentioned automatic control method for the amount of washing water of the micro-bubble flotation machine, the step S120 uses a dynamic element-level feature fusion module to perform feature element-by-element fusion on the slurry level time series correlation feature vector and the slurry concentration time series correlation feature vector to obtain a multi-dimensional slurry time series correlation feature vector. It should be understood that the slurry level and slurry concentration jointly affect the thickness of the foam layer. Therefore, in order to further explore the correlation between the two on the thickness of the foam layer, in the technical solution of the present application, the slurry level time series correlation feature vector and the slurry concentration time series correlation feature vector are fused element by element through a dynamic element-level feature fusion module. Specifically, the dynamic element-level feature fusion module adopts an adaptive weight allocation mechanism, dynamically adjusts the fusion weight according to the importance of the slurry level time series correlation feature vector and the slurry concentration time series correlation feature vector, and performs element-level fusion operations in a dynamic manner. Compared with simple feature splicing, this dynamic fusion method fully explores the potential correlation and complementary information between the two, which can reduce information loss, realize effective integration of features, and thus form a more comprehensive description of the slurry state.

[0033] Figure 4 This is a flow chart of using a dynamic element-level feature fusion module in the automatic control method of the washing water volume of a micro-bubble flotation machine according to an embodiment of the present application to perform feature element-by-element fusion on the slurry level time series correlation feature vector and the slurry concentration time series correlation feature vector to obtain a multi-dimensional slurry time series correlation feature vector. Figure 4 As shown, the step S120 includes: S121, calculating the information fusion balance threshold value between the slurry level time series correlation feature vector and the slurry concentration time series correlation feature vector; S122, based on the information fusion balance threshold value, performing feature element-by-element fusion on the slurry level time series correlation feature vector and the slurry concentration time series correlation feature vector to obtain the slurry multi-dimensional time series correlation feature vector.

[0034] More specifically, the step S121 includes: cascading the slurry level time series associated feature vector and the slurry concentration time series associated feature vector to obtain a cascaded feature vector; multiplying the cascaded feature vector and the transformation vector to obtain a threshold scoring coefficient; adding the threshold scoring coefficient and the bias parameter to obtain a bias threshold scoring coefficient; and passing the bias threshold scoring coefficient through a sigmoid activation function to obtain the information fusion balance threshold value.

[0035] More specifically, the step S122 includes: respectively calculating the natural exponential function values ​​of each characteristic value in the slurry level time series correlation feature vector and the slurry concentration time series correlation feature vector to obtain an indexed slurry level time series correlation feature vector and an indexed slurry concentration time series correlation feature vector; determining a first weight and a second weight based on the information fusion balance threshold value, and weighting the indexed slurry level time series correlation feature vector position by position with the first weight to obtain a weighted slurry level time series correlation feature vector, and weighting the indexed slurry concentration time series correlation feature vector position by position with the second weight to obtain a weighted slurry concentration time series correlation feature vector; adding corresponding elements of the weighted slurry level time series correlation feature vector and the weighted slurry concentration time series correlation feature vector to obtain the slurry multi-dimensional time series correlation feature vector.

[0036] That is, the slurry level time series correlation feature vector and the slurry concentration time series correlation feature vector are fused element by element using the following dynamic element-level feature fusion formula to obtain the slurry multi-dimensional time series correlation feature vector, wherein the dynamic element-level feature fusion formula is:

[0037]

[0038] in, represents the time series correlation feature vector of the slurry concentration, represents the temporal correlation feature vector of the slurry level, v g represents the multi-dimensional time series correlation feature vector of the slurry, concat(·,·) represents the connection operation, v c is the transformation vector, b is the bias parameter, sigmoid represents the activation function, t is the information fusion balance threshold, and t∈[0,1].

[0039] In the above-mentioned automatic control method for the amount of washing water of the micro-bubble flotation machine, the step S130 inputs the multi-dimensional time series correlation feature vector of the slurry and the time series correlation feature vector of the foam layer thickness into the responsive gamma correction fusion module to obtain the slurry parameter-foam layer thickness time series responsiveness correlation feature vector. That is, in order to more accurately capture the complex nonlinear relationship between the slurry parameters (liquid level and concentration) and the foam layer thickness, in the technical solution of the present application, a responsive gamma correction fusion module is introduced to perform responsive fusion processing on the multi-dimensional time series correlation feature vector of the slurry and the time series correlation feature vector of the foam layer thickness. Specifically, the responsive gamma correction fusion module compensates for the nonlinear responsiveness of the time series correlation relationship between the slurry parameters and the foam layer thickness, and can highlight more local response details to improve the fine-grained characterization strength of the response process without destroying the original characteristic distribution, strengthen the correlation and interactive response relationship between the two, and provide more accurate and reliable data support for the subsequent washing water adjustment strategy.

[0040] In a specific example of the present application, the step S130 includes: calculating the position-by-position division of the slurry multi-dimensional time series correlation characteristic vector and the foam layer thickness time series correlation characteristic vector to obtain a slurry parameter-foam layer thickness response characteristic vector; respectively calculating the first nonlinear response factor and the second nonlinear response factor of each characteristic value in the slurry parameter-foam layer thickness response characteristic vector, and dividing the first nonlinear response factor by the second nonlinear response factor to obtain the characteristic value of a predetermined position of the slurry parameter-foam layer thickness time series responsiveness correlation characteristic vector.

[0041] That is, the multi-dimensional time series correlation feature vector of the slurry and the time series correlation feature vector of the foam layer thickness are processed by the following responsive gamma correction fusion formula to obtain the slurry parameter-foam layer thickness time series responsiveness correlation feature vector; wherein the responsive gamma correction fusion formula is:

[0042]

[0043] Among them, v 1i is the ith eigenvalue of the multi-dimensional time series correlation eigenvector of the slurry, v 2i is the ith eigenvalue of the temporal correlation eigenvector of the foam layer thickness, α, β, γ and ε are adjustment parameters with different values, V i is the i-th eigenvalue of the slurry parameter-foam layer thickness time series responsiveness associated eigenvector.

[0044] In the above-mentioned automatic control method of the amount of washing water for the micro-bubble flotation machine, the step S140 determines the adjustment strategy of the amount of washing water based on the slurry parameter-foam layer thickness time series responsiveness associated characteristic vector. In a specific example of the present application, the implementation method of determining the adjustment strategy of the amount of washing water based on the slurry parameter-foam layer thickness time series responsiveness associated characteristic vector is to pass the slurry parameter-foam layer thickness time series responsiveness associated characteristic vector through a classifier-based adjustment strategy generator to obtain the adjustment strategy of the amount of washing water. That is, using the classifier as a decision module, according to the characteristic information in the slurry parameter-foam layer thickness time series responsiveness associated characteristic vector, a corresponding adjustment strategy of washing water is formulated, so that the amount of washing water can adapt to the current slurry state and foam layer thickness to optimize the flotation process. Specifically, the adjustment strategy generator based on the classifier establishes a mapping relationship between the slurry state and foam layer thickness and the adjustment strategy of washing water by learning and training historical data. Then, according to the learned classification rules, the slurry parameter-foam layer thickness time series responsiveness associated characteristic vector is classified and judged to obtain the recommended flow rate value of the washing water volume for different slurry states and foam layer thicknesses as the adjustment strategy of the washing water volume. In this way, in the actuator step, the opening value of the designated valve can be further controlled according to the adjustment strategy of the washing water volume to realize automatic adjustment of the washing water volume, thereby ensuring the stability and efficiency of the flotation process.

[0045] It should be understood that before using the above neural network model, it needs to be trained. That is to say, in the automatic control method of the washing water volume of the microbubble flotation machine of the present application, it also includes a training step of training the temporal pattern feature extractor based on the one-dimensional convolution layer, the dynamic element-level feature fusion module, the responsive gamma correction fusion module and the classifier-based adjustment strategy generator.

[0046] Figure 5 FIG. 1 is a flow chart of the training steps in the method for automatically controlling the amount of washing water for a microbubble flotation machine according to an embodiment of the present application. Figure 5As shown, the training step includes: S210, obtaining training data, the training data including a time series of training slurry level, a time series of training slurry concentration, and a time series of training foam layer thickness, as well as a true value of the adjustment strategy for the amount of cleaning water; S220, arranging the time series of the training slurry level, the time series of the training slurry concentration, and the time series of the training foam layer thickness respectively according to the time dimension as a training slurry level time series input vector, a training slurry concentration time series input vector, and a training foam layer thickness time series input vector; S230, passing the training slurry level time series input vector, the training slurry concentration time series input vector, and the training foam layer thickness time series input vector respectively through the time series pattern feature extractor based on the one-dimensional convolution layer to obtain a training slurry level time series associated feature vector, a training slurry concentration time series associated feature vector, and a training foam layer thickness time series associated feature vector; S240, using the dynamic element-level feature fusion module to The training slurry level time series associated feature vector and the training slurry concentration time series associated feature vector are feature-fused element by element to obtain the training slurry multi-dimensional time series associated feature vector; S250, the training slurry multi-dimensional time series associated feature vector and the training foam layer thickness time series associated feature vector are input into the responsive gamma correction fusion module to obtain the training slurry parameter-foam layer thickness time series responsiveness associated feature vector; S260, the training slurry parameter-foam layer thickness time series responsiveness associated feature vector is passed through the classifier-based adjustment strategy generator to obtain the classification loss function value; S270, the one-dimensional convolutional layer-based time series pattern feature extractor, the dynamic element-level feature fusion module, the responsive gamma correction fusion module and the classifier-based adjustment strategy generator are trained with the classification loss function value, wherein, in each round of iteration of the training, the training slurry parameter-foam layer thickness time series responsiveness associated feature vector is optimized.

[0047] In the technical solution described above, the training slurry level time series correlation feature vector, the training slurry concentration time series correlation feature vector and the training foam layer thickness time series correlation feature vector respectively express the time series correlation characteristics of the training slurry level, the training slurry concentration and the training foam layer thickness, and the training slurry multi-dimensional time series correlation feature vector expresses the time series correlation fusion characteristics of the training slurry level and the training slurry concentration. Therefore, after the training slurry multi-dimensional time series correlation feature vector and the training foam layer thickness time series correlation feature vector are input into the responsive gamma correction fusion module, the obtained training slurry parameter-foam layer thickness time series responsiveness correlation feature vector has a responsiveness time series feature representation mode based on multiple time series feature association representation dimensions based on different data source distribution patterns.

[0048] Therefore, considering the parallel model branch structure between the feature extractor of the training slurry multi-dimensional time series correlation feature vector, i.e., the time series pattern feature extractor based on the one-dimensional convolution layer plus the dynamic element-level feature fusion module, and the feature extractor of the training foam layer thickness time series correlation feature vector, i.e., the time series pattern feature extractor based on the one-dimensional convolution layer, there is a corresponding relationship between the time series feature representation of different source data under different time series correlation modes between the training slurry multi-dimensional time series correlation feature vector and the training foam layer thickness time series correlation feature vector. Therefore, it is expected to perform responsive gamma correction fusion on the basis of maintaining the class convergence iterative consistency of the training slurry multi-dimensional time series correlation feature vector and the training foam layer thickness time series correlation feature vector under different time series correlation modes, thereby improving the accuracy of the classification results.

[0049] Based on this, the present application optimizes the training slurry parameter-foam layer thickness time series responsiveness associated characteristic vector in each iteration of the training, including: for the first eigenvalue v of the training slurry parameter-foam layer thickness time series responsiveness associated characteristic vector i and the second eigenvalue v j , calculate the absolute value of the difference divided by the sum of the absolute values ​​and then correct the relative probability value, wherein the correction makes the relative probability value greater than zero and less than one; calculate the first product of the first eigenvalue multiplied by the relative probability value, the second product of the second eigenvalue multiplied by the absolute value of the difference between the relative probability value and one, the first quotient of the first eigenvalue divided by the absolute value of the difference between the relative probability value and one, and the second quotient of the second eigenvalue divided by the relative probability value; after adding the first product and the second product, further subtract the difference between the first quotient and the second quotient to obtain the first eigenvalue v i and the second eigenvalue v j The corresponding eigenvalue m of the corresponding optimization matrix i,j ; and, multiplying the optimization matrix with the training slurry parameter-foam layer thickness time series responsiveness associated characteristic vector to obtain an optimized training slurry parameter-foam layer thickness time series responsiveness associated characteristic vector.

[0050] Specifically, it is expressed as:

[0051]

[0052] m i,j ∈M,v i ∈V,v j(j≠i) ∈V

[0053]

[0054] Wherein, V represents the training slurry parameter-foam layer thickness time series responsiveness correlation feature vector, v i and v j represent the first eigenvalue and the second eigenvalue of the training slurry parameter-foam layer thickness time series responsiveness correlation eigenvector, p i,j Represents the relative probability value, m i,j Represents the first eigenvalue v i and the second eigenvalue v j The corresponding eigenvalue of the optimization matrix, M, represents the first eigenvalue v i and the second eigenvalue v j The corresponding optimization matrix is, represents matrix multiplication, V ′ represents the optimized training slurry parameter-foam layer thickness time series responsiveness correlation feature vector;

[0055] Among them, the first eigenvalue v i and the second eigenvalue v j The corresponding relative probability value p i,j It is expressed as:

[0056]

[0057] Among them, v i and v j denote the first eigenvalue and the second eigenvalue of the training slurry parameter-foam layer thickness time series responsiveness associated eigenvector, ε denotes a predetermined hyperparameter, and p i,j Represents the relative probability value.

[0058] Here, with regard to the regional-boundary integral relationship of the high-dimensional characteristic manifold of the training slurry parameter-foam layer thickness timing responsiveness associated feature vector, by defining the overall probability distribution representation of the feature set of the training slurry parameter-foam layer thickness timing responsiveness associated feature vector as the manifold constraint boundary, the high-dimensional manifold representation of the feature set of the training slurry parameter-foam layer thickness timing responsiveness associated feature vector, which is composed of eigenvalue pairs, is approximated to avoid the ambiguity in the mapping of the diversified feature representation of the training slurry parameter-foam layer thickness timing responsiveness associated feature vector to the local manifold representation in the high-dimensional feature classification regression convergence space, and improve the execution iteration consistency of each local feature distribution in the mapping task, so as to improve the feature classification regression convergence effect and the accuracy of the classification results.

[0059] In summary, the automatic control method of the washing water volume of the micro-bubble flotation machine according to the embodiment of the present application is explained, which uses artificial intelligence technology based on deep learning to perform time series analysis on the liquid level and concentration of the ore pulp in the flotation tank to capture the time series correlation fusion characteristics of the ore pulp liquid level and concentration, and dig out the time series correlation influence relationship between the ore pulp liquid level and concentration on the thickness of the foam layer, so as to adaptively adjust the flow rate of the washing water volume. In this way, the automation level of the micro-bubble flotation machine can be improved, the precise control of the washing water volume of the micro-bubble flotation machine can be achieved, the flotation efficiency and mineral recovery rate can be improved, and the waste of resources can be reduced.

[0060] Figure 6 FIG. 1 is a block diagram of an automatic control device for the amount of washing water of a microbubble flotation machine according to an embodiment of the present application. Figure 6 As shown, according to the embodiment of the present application, the automatic control device 100 for the amount of cleaning water of the micro-bubble flotation machine includes: a data time series association analysis module 110, which is used to extract the time series pattern characteristics of the time series of the slurry level, the time series of the slurry concentration and the time series of the foam layer thickness respectively to obtain the slurry level time series association feature vector, the slurry concentration time series association feature vector and the foam layer thickness time series association feature vector; a slurry multi-dimensional time series association analysis module 120, which is used to use a dynamic element-level feature fusion module to perform feature element-by-element fusion on the slurry level time series association feature vector and the slurry concentration time series association feature vector to obtain the slurry multi-dimensional time series association feature vector; a time series responsiveness analysis module 130, which is used to input the slurry multi-dimensional time series association feature vector and the foam layer thickness time series association feature vector into a responsiveness gamma correction fusion module to obtain the slurry parameter-foam layer thickness time series responsiveness association feature vector; an adjustment strategy generation module 140, which is used to determine the adjustment strategy of the amount of cleaning water based on the slurry parameter-foam layer thickness time series responsiveness association feature vector.

[0061] Here, those skilled in the art will appreciate that the specific operations of each module in the above-mentioned automatic control device for washing water volume of the microbubble flotation machine have been described in the above reference. Figures 1 to 5 The description of the automatic control method of the cleaning water amount of the microbubble flotation machine has been described in detail, and therefore, its repeated description will be omitted.

[0062] The basic principle of the present invention is described above in conjunction with specific embodiments. However, it should be pointed out that the advantages, strengths, effects, etc. mentioned in the present invention are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. must be possessed by each embodiment of the present invention. In addition, the specific details of the above embodiments are only for the purpose of illustration and facilitation of understanding, rather than limitation, and the above details do not limit the present invention to being implemented by adopting the above specific details.

[0063] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, please refer to the relevant description of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the module division is only a logical function division, and there may be other division methods in actual implementation. The modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0064] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference to a figure in a claim should not be considered as limiting the claim to which it relates.

[0065] In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. The multiple units stated in the device claim can also be implemented by one unit through software or hardware.

[0066] Finally, it should be noted that the above description has been given for the purpose of illustration and description. In addition, the above embodiments are only used to illustrate the technical solution of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for automatically controlling the amount of washing water for a microbubble flotation machine, characterized in that: include: Liquid level measurement step: measuring the liquid level of the slurry in the flotation tank to obtain a time series of the slurry level; Data collection steps: collecting the time series of slurry concentration and foam layer thickness through the sensor group; Control algorithm step: determining an adjustment strategy for the amount of cleaning water based on the time series of the slurry level, the time series of the slurry concentration and the time series of the foam layer thickness, wherein the adjustment strategy for the amount of cleaning water is a recommended flow rate value for the amount of cleaning water; Actuator step: designating a valve as an actuator for adjusting the amount of cleaning water; Automatic adjustment step: adjusting the opening value of the valve based on the adjustment strategy of the cleaning water volume; Wherein, the control algorithm steps include: Respectively extracting the time series pattern features of the time series of the slurry level, the time series of the slurry concentration, and the time series of the foam layer thickness to obtain a slurry level time series correlation feature vector, a slurry concentration time series correlation feature vector, and a foam layer thickness time series correlation feature vector; Using a dynamic element-level feature fusion module, the slurry level time series correlation feature vector and the slurry concentration time series correlation feature vector are subjected to feature element-by-element fusion to obtain a slurry multi-dimensional time series correlation feature vector; Inputting the slurry multi-dimensional time series correlation feature vector and the foam layer thickness time series correlation feature vector into a responsive gamma correction fusion module to obtain a slurry parameter-foam layer thickness time series responsiveness correlation feature vector; Based on the slurry parameter-foam layer thickness time series responsiveness correlation characteristic vector, the adjustment strategy of the cleaning water volume is determined.

2. The automatic control method for washing water volume of a microbubble flotation machine according to claim 1, characterized in that: Respectively extracting the time series pattern features of the time series of the slurry level, the time series of the slurry concentration, and the time series of the foam layer thickness to obtain a slurry level time series correlation feature vector, a slurry concentration time series correlation feature vector, and a foam layer thickness time series correlation feature vector, including: Arrange the time series of the slurry level, the time series of the slurry concentration and the time series of the foam layer thickness respectively according to the time dimension into a slurry level time series input vector, a slurry concentration time series input vector and a foam layer thickness time series input vector; The slurry level time series input vector, the slurry concentration time series input vector and the foam layer thickness time series input vector are respectively passed through a time series pattern feature extractor based on a one-dimensional convolutional layer to obtain the slurry level time series associated feature vector, the slurry concentration time series associated feature vector and the foam layer thickness time series associated feature vector.

3. The automatic control method of washing water volume of a microbubble flotation machine according to claim 2, characterized in that: Using a dynamic element-level feature fusion module to perform feature element-by-element fusion on the slurry level time series correlation feature vector and the slurry concentration time series correlation feature vector to obtain a slurry multi-dimensional time series correlation feature vector, including: Calculating an information fusion balance threshold value between the slurry level time series correlation feature vector and the slurry concentration time series correlation feature vector; Based on the information fusion balance threshold value, the slurry level time series correlation feature vector and the slurry concentration time series correlation feature vector are fused element by element to obtain the slurry multi-dimensional time series correlation feature vector.

4. The automatic control method of washing water volume of a microbubble flotation machine according to claim 3, characterized in that: Calculating the information fusion balance threshold between the slurry level time series correlation feature vector and the slurry concentration time series correlation feature vector, including: Cascading the slurry level time series associated feature vector and the slurry concentration time series associated feature vector to obtain a cascade feature vector; Multiplying the concatenated feature vector and the transformation vector to obtain a threshold scoring coefficient; Adding the threshold scoring coefficient and the bias parameter to obtain a bias threshold scoring coefficient; The bias threshold scoring coefficient is passed through a sigmoid activation function to obtain the information fusion balance threshold value.

5. The automatic control method for washing water volume of a microbubble flotation machine according to claim 4, characterized in that: Based on the information fusion balance threshold value, the slurry level time series correlation feature vector and the slurry concentration time series correlation feature vector are fused element by element to obtain the slurry multi-dimensional time series correlation feature vector, including: Respectively calculating the natural exponential function values ​​of each characteristic value in the slurry level time series correlation characteristic vector and the slurry concentration time series correlation characteristic vector to obtain an indexed slurry level time series correlation characteristic vector and an indexed slurry concentration time series correlation characteristic vector; Determine a first weight and a second weight based on the information fusion balance threshold value, and weight the indexed slurry level time series associated feature vector position by position with the first weight to obtain a weighted slurry level time series associated feature vector, and weight the indexed slurry concentration time series associated feature vector position by position with the second weight to obtain a weighted slurry concentration time series associated feature vector; The weighted slurry level time series correlation feature vector and the weighted slurry concentration time series correlation feature vector are added together to obtain the slurry multi-dimensional time series correlation feature vector.

6. The automatic control method of washing water volume of a microbubble flotation machine according to claim 5, characterized in that: Inputting the slurry multi-dimensional time series correlation feature vector and the foam layer thickness time series correlation feature vector into a responsive gamma correction fusion module to obtain a slurry parameter-foam layer thickness time series responsiveness correlation feature vector, including: Calculating the position-by-position division of the slurry multi-dimensional time series correlation feature vector and the foam layer thickness time series correlation feature vector to obtain a slurry parameter-foam layer thickness response feature vector; The first nonlinear response factor and the second nonlinear response factor of each eigenvalue in the slurry parameter-foam layer thickness response eigenvector are calculated respectively, and the first nonlinear response factor is divided by the second nonlinear response factor to obtain the eigenvalue of the predetermined position of the slurry parameter-foam layer thickness time series responsiveness associated eigenvector.

7. The automatic control method of washing water volume of a microbubble flotation machine according to claim 6, characterized in that: Based on the slurry parameter-foam layer thickness time series responsiveness correlation characteristic vector, the adjustment strategy of the cleaning water volume is determined, including: The slurry parameter-foam layer thickness time series responsiveness associated feature vector is passed through a classifier-based adjustment strategy generator to obtain the adjustment strategy for the cleaning water volume.

8. The automatic control method of washing water volume of a microbubble flotation machine according to claim 7, characterized in that: The method further includes a training step of: training the one-dimensional convolutional layer-based temporal pattern feature extractor, the dynamic element-level feature fusion module, the responsive gamma correction fusion module, and the classifier-based adjustment strategy generator; Wherein, the training step includes: Acquire training data, wherein the training data includes a time series of training slurry level, a time series of training slurry concentration, and a time series of training foam layer thickness, as well as a true value of a cleaning water volume adjustment strategy; Arrange the time series of the training slurry level, the time series of the training slurry concentration and the time series of the training foam layer thickness respectively according to the time dimension into a training slurry level time series input vector, a training slurry concentration time series input vector and a training foam layer thickness time series input vector; The training slurry level time series input vector, the training slurry concentration time series input vector and the training foam layer thickness time series input vector are respectively passed through the time series pattern feature extractor based on the one-dimensional convolution layer to obtain a training slurry level time series associated feature vector, a training slurry concentration time series associated feature vector and a training foam layer thickness time series associated feature vector; Using the dynamic element-level feature fusion module, the training slurry level time series correlation feature vector and the training slurry concentration time series correlation feature vector are subjected to feature element-by-element fusion to obtain a training slurry multi-dimensional time series correlation feature vector; Inputting the training slurry multi-dimensional time series correlation feature vector and the training foam layer thickness time series correlation feature vector into the responsiveness gamma correction fusion module to obtain the training slurry parameter-foam layer thickness time series responsiveness correlation feature vector; Passing the training slurry parameter-foam layer thickness time series responsiveness association feature vector through the classifier-based adjustment strategy generator to obtain a classification loss function value; The one-dimensional convolutional layer-based temporal pattern feature extractor, the dynamic element-level feature fusion module, the responsive gamma correction fusion module and the classifier-based adjustment strategy generator are trained with the classification loss function value, wherein in each round of iteration of the training, the training slurry parameter-foam layer thickness temporal responsiveness associated feature vector is optimized.

9. An automatic control device for the amount of washing water of a micro-bubble flotation machine, using the automatic control method for the amount of washing water of a micro-bubble flotation machine according to any one of claims 1 to 8, characterized in that: include: A data time series correlation analysis module is used to extract the time series pattern characteristics of the time series of slurry level, the time series of slurry concentration and the time series of foam layer thickness respectively to obtain a slurry level time series correlation feature vector, a slurry concentration time series correlation feature vector and a foam layer thickness time series correlation feature vector; A slurry multi-dimensional time series correlation analysis module, used to use a dynamic element-level feature fusion module to perform feature element-by-element fusion on the slurry level time series correlation feature vector and the slurry concentration time series correlation feature vector to obtain a slurry multi-dimensional time series correlation feature vector; A time series responsiveness analysis module, used for inputting the slurry multi-dimensional time series correlation feature vector and the foam layer thickness time series correlation feature vector into a responsiveness gamma correction fusion module to obtain a slurry parameter-foam layer thickness time series responsiveness correlation feature vector; The adjustment strategy generation module is used to determine the adjustment strategy of the cleaning water volume based on the slurry parameter-foam layer thickness time series responsiveness correlation characteristic vector.

Images