A dynamic pressure control intelligent regulation method for filter cake shaping

Abstract

The application provides a filter cake forming dynamic pressure control intelligent regulation method, and belongs to the technical field of intelligent regulation based on computer data processing; first, the characteristics of a solid-liquid mixture are extracted through the arrangement of pressure and flow sensors, and the state characteristics of the solid-liquid mixture are recognized through a camera; multi-dimensional information is comprehensively used to construct a pattern recognition model of the filter cake state, so that the model is prevented from being misjudged due to a single parameter. After confirming the filter cake forming stage, a long short-term memory network is used to dynamically model time sequence data, a nonlinear correlation between pressure fluctuation and time dimension in the filter cake forming process is established, and a pressure control target value of the filter press is calculated. The pressure sensor is used to monitor the real value of the pressure control of the filter press, and the pressure difference between the real value and the target value is used as the input of an improved segmented PID controller to finally realize the dynamic regulation of the pressure of the filter press. The application is widely applicable, improves the detection accuracy in the filter cake forming stage, and improves the automation degree of the filter press.

Description

Technical Field

[0001] This invention belongs to the field of intelligent control technology based on computer data processing, and particularly relates to an intelligent control method for dynamic pressure control of filter cake forming. Background Technology

[0002] A filter press is a mechanical device that uses a special filter medium to apply pressure to a material, causing the liquid to seep out. It is a commonly used solid-liquid separation device and is widely used in many fields such as mining, metallurgy, food, pharmaceuticals, and wastewater treatment.

[0003] Taking a mining scenario as an example, filter presses are mainly used in mineral processing for concentrate dewatering, tailings treatment, and wastewater recycling, directly impacting beneficiation efficiency and resource recovery rates. Concentrate dewatering reduces the moisture content of the concentrate, thereby improving smelting efficiency while avoiding the high costs associated with drying and stockpiling. Tailings are discharged dry, achieving solid-liquid separation and reducing environmental pollution. In the recycling of mineral processing wastewater, filter presses recover water resources through solid-liquid separation, reducing the pressure on wastewater discharge.

[0004] The pressure of a filter press is not always better the higher it is; it needs to be adjusted comprehensively based on the material characteristics, process parameters, and equipment performance. Pressure is a key parameter for dewatering in a filter press, but excessively high or low pressure will affect the effect. Insufficient pressure results in more residual moisture in the filter cake, leading to incomplete dewatering; excessive pressure may damage the pore structure of the filter cake, hindering water discharge and even damaging the equipment.

[0005] The filter cake forming process is divided into three stages, and the control methods used for each stage are different.

[0006] 1. Filter cake formation stage: Under vacuum, water is extracted from the slurry, and solid particles gradually accumulate to form a filter cake. Low-pressure control and rapid feeding are used at this stage to prevent filter cloth clogging. The lower pressure and faster feeding rate at this stage help reduce unevenness.

[0007] 2. Filter cake dewatering stage: As the vacuum continues to operate, the moisture content in the filter cake further decreases, reaching the desired moisture content. At this point, the pressure should be gradually increased while maintaining a stable feed rate. High pressure promotes the removal of moisture, forming a dense filter cake. During this stage, pressure changes must be monitored in real time to avoid overpressure damage to the equipment.

[0008] 3. Filter cake removal stage: After dewatering, the filter cake is removed from the filter media and proceeds to the next process. This stage employs high-pressure steady-state control, using a pressure sensor to adjust the feed pump speed or valve opening to precisely control the pressure fluctuation range.

[0009] Currently, the filter cake forming stage of filter presses mainly relies on manual observation and adjustment of pressure control strategies to match the filter cake forming stage. Due to the long duration and monotonous operation of the filter pressing process, the physiological fatigue and mental slackness of the operators are the main factors affecting production efficiency (such as improper operation leading to material pressing and overflow). Therefore, the traditional pressure control system that relies on manual operation urgently needs to be optimized. Summary of the Invention

[0010] To address the above problems, this invention designs a dynamic pressure control and intelligent regulation method for filter cake forming, comprising the following steps:

[0011] S1, by installing a first pressure sensor in the filtrate outlet pipe to monitor the filtration resistance, installing a flow sensor in the feed straight pipe section to monitor the flow rate of the feed fluid, and installing a camera above the filtrate water collection tank to identify the filtrate color;

[0012] A second pressure sensor is installed at the location of the pressure gauge built into the filter press to detect the system pressure and transmit the signal to the controller. After data preprocessing, the true pressure control value of the filter press is obtained.

[0013] S2, the real-time obtained filtration resistance, feed flow rate and filtrate color are used to establish a multimodal array, which is then preprocessed and input into the stage identification model to indicate the stage of filter cake formation based on the principle of relative majority, including the filter cake formation stage, filter cake dehydration stage and filter cake peeling stage.

[0014] S3 uses the pre-processed filtration resistance, feed flow rate, filtrate color, and actual filter press pressure as inputs to a Long Short-Term Memory (LSTM) network to dynamically model the time series data, establish a nonlinear relationship between pressure fluctuations and the time dimension during the filter cake forming process, and calculate the target value for filter press pressure control.

[0015] S4. Based on the different characteristics of the filter cake forming stage, a corresponding segmented PID controller is established. The pressure difference between the actual pressure control value and the pressure control target value is used as the input of the segmented PID controller. The output of the segmented PID controller controls the working pressure of the filter press, thereby realizing the dynamic regulation of the filter press pressure.

[0016] Preferably, the pretreatment process for filtration resistance, feed flow rate, filtrate color, and the actual value of filter press pressure includes:

[0017] S11, Create an average value array of length 10 to perform smoothing filtering on the sampled data. The data after average filtering is: and ; express The smoothing and filtering results of the feed fluid flow sensor at all times. express The smoothing filter result of the filtrate color camera at any given time. express The result of smoothing and filtering the actual pressure control value of the filter press at any time; express The result of smoothing and filtering the real-time filtration pressure data (pressure sensor data monitoring filtration resistance by the first pressure sensor installed in the filtrate outlet pipeline);

[0018] S12, distribute the filter pressure data, flow data, and actual pressure control data into a 0-1 distribution and normalize them to the same coordinate axis; thus obtaining... and ;

[0019] S13. Perform HSL transformation on the image data that has undergone average filtering. First, normalize the three components R, G, and B to the range of [0, 1]. Calculate the hue H, color saturation S, and intensity component L. The range of the H component is [0, 360], while the ranges of the S and L components are [0, 1].

[0020] Preferably, the specific data processing procedure for the stage identification model in S2 is as follows:

[0021] By setting a first threshold and a second threshold, the relationship between filtration resistance and the filter cake formation, dewatering, and peeling stages is defined. When the input value of filtration resistance is less than the first threshold, it is defined as being in the filter cake formation stage; when the input value of filtration resistance is greater than or equal to the first threshold and less than or equal to the second threshold, it is defined as being in the filter cake dewatering stage; when the input value of filtration resistance is greater than the second threshold, it is defined as being in the filter cake peeling stage.

[0022] By setting a third and a fourth threshold, the relationship between the feed fluid flow rate and the filter cake formation, dewatering, and peeling stages is defined. When the feed fluid flow rate input value is less than the third threshold, it is defined as being in the filter cake formation stage; when the feed fluid flow rate input value is greater than or equal to the third threshold and less than or equal to the fourth threshold, it is defined as being in the filter cake dewatering stage; when the feed fluid flow rate input value is greater than the fourth threshold, it is defined as being in the filter cake peeling stage.

[0023] After performing HSL transformation on the filtrate color image data, a first set of judgment conditions and a second set of judgment conditions are set based on the values ​​of the S and L components, thereby defining the relationship between filtrate color and the filter cake formation, dehydration, and peeling stages. The constant thresholds in the first set of judgment conditions and the second set of judgment conditions are all taken from empirical values. When the filtrate color meets the first set of judgment conditions, it is defined as being in the filter cake formation stage; when the filtrate color meets the second set of judgment conditions, it is defined as being in the filter cake peeling stage; when the filtrate color does not meet either the first set of judgment conditions or the second set of judgment conditions, it is defined as being in the filter cake dehydration stage.

[0024] Preferably, based on preprocessed filter pressure data :

[0025] like At this point, the filter cake is forming.

[0026] like At this point, the filter cake is in the dehydration stage;

[0027] like At this point, the filter cake is being peeled off.

[0028] Preferably, based on preprocessed traffic data :

[0029] like At this point, the filter cake is forming.

[0030] like At this point, the filter cake is in the dehydration stage;

[0031] like At this point, the filter cake is being peeled off.

[0032] Preferably, based on the transformed S and L components:

[0033] If at least one of the following three conditions is met, it is determined that the filter cake is currently in the forming stage:

[0034] ; )&( ; )&( If at least one of the following three conditions is met, it is determined that the filter cake is currently in the peeling stage:

[0035] ; )&( ; )&( ;

[0036] If neither of the above two judgment conditions is met at the same time, it is determined that the filter cake is in a dehydration state.

[0037] Preferably, S3 specifically includes:

[0038] Filtering resistance from preprocessed structured data Feed flow rate Filtrate color as well as , defined as the input parameter matrix of the LSTM model; sets the expressions for the system forget gate, input gate, and output gate, as well as the weight parameter matrices, to satisfy the preconditions for LSTM model calculation; establishes the nonlinear relationship between pressure fluctuation and time dimension during filter cake forming process; calculates the target value for filter press pressure control; and outputs the results for the LSTM model. .

[0039] Preferably, in step S4, the filter press pressure is controlled to the actual value. Pressure control calculation value output by LSTM The difference is used as the input of the position-type ID control algorithm, and the output of the PID algorithm model is used as the given value for the pressure control of the filter press. If it is in the filter cake forming stage, a PID controller is used; if it is in the filter cake dewatering stage, an incomplete derivative PID controller is selected; if it is in the filter cake peeling stage, a PI controller is selected. Control is achieved by accumulating errors, and the actual position of the corresponding actuator is directly output.

[0040] Preferably, the specific data processing procedure of S4 is as follows:

[0041] The error at time t is denoted by e(t), and its calculation expression is as follows:

[0042] ;

[0043] If the filter cake is forming, a PID controller is used, and the expression is as follows:

[0044] ;

[0045] In the formula This represents the proportional gain of the PID controller; This represents the integral coefficient of the PID controller. Represents the derivative coefficients of the PID controller;

[0046] If the filter cake is in the dewatering stage, an incomplete derivative PID controller should be selected:

[0047] A first-order inertial element is connected in series with the output of the PID controller to form an incomplete differential PID controller. u1(t) The transfer function of the first-order inertial element D(s), representing the output of the PID controller, is as follows:

[0048] ;

[0049] In the formula Let be the step response coefficient of the first-order inertial element. The expression for the incomplete differential controller is as follows:

[0050] ;

[0051] In the formula , This indicates the calculation result. This represents the output of the PID controller;

[0052] If the filter cake is in the removal stage, a PI controller is selected, and the expression is as follows:

[0053] ;

[0054] In the formula This represents the proportional gain of the PI controller; This represents the derivative coefficient of the PI controller.

[0055] Compared with the prior art, the innovative points of this invention include:

[0056] (1) Using multimodal data fusion analysis to achieve pattern recognition in the filter cake stage: combine sensor and image data to extract features of the solid-liquid mixture state, and construct a pattern recognition model of the filter cake state by integrating multi-dimensional information;

[0057] (2) Constructing a deep model to predict the target value of pressure control: The target value of pressure control of the filter press is predicted by deep modeling. The filter resistance, feed flow rate, filtrate color and the actual value of filter press pressure in the preprocessed structured dataset are used as inputs to a Long Short-Term Memory (LSTM) network to dynamically model the time series data, establish the nonlinear relationship between pressure fluctuation and time dimension during the filter cake forming process, and calculate the target value of pressure control of the filter press.

[0058] (3) Establish an adaptive pressure control strategy: Based on the different characteristics of the three stages of filter cake formation, dehydration and peeling, establish a corresponding improved segmented PID control segmented control model. Use the pressure difference between the actual pressure control value and the pressure control target value as the input of the improved segmented PID controller. The output of the improved segmented PID controller controls the working pressure of the filter press to achieve dynamic regulation of the filter press pressure.

[0059] The beneficial effects brought about by the innovation of this invention include:

[0060] Improve the accuracy of filter cake forming stage detection: This method performs multi-dimensional information comprehensive judgment through multi-modal data, which can accurately identify the filter cake forming stage, avoid the lag of traditional manual judgment and reduce the misjudgment rate caused by mental fatigue due to long working hours;

[0061] Improve the automation level of filter press operation: Adaptive pressure control strategy replaces the traditional manual judgment and adjustment of the pressure control system, thereby improving the level of production automation.

[0062] Wide range of applications: Through systematic judgment, this method is applicable to filter presses in various scenarios such as mining, metallurgy, food, and pharmaceuticals. This method is for the automatic identification of the physical properties of the input material, rather than for parameter optimization of a single device. Attached Figure Description

[0063] Figure 1 This is a flowchart illustrating the overall technical route of the present invention.

[0064] Figure 2 This is a diagram showing the relationship between the filter cake forming stage and the controller selection.

[0065] Figure 3 This is a block diagram of an incomplete differential controller.

[0066] Figure 4 This is a response curve of a traditional controller to a step response signal.

[0067] Figure 5 To improve the response curve of the segmented PID controller to the step response signal.

[0068] Figure 6 The figure shows the simulation results of EMC interference for a traditional PID controller.

[0069] Figure 7 Simulation results of EMC interference for improving segmented PID controller. Detailed Implementation

[0070] This invention discloses a dynamic pressure control intelligent regulation method for filter cake forming based on multimodal detection and depth modeling. The overall process is as follows: Figure 1 As shown:

[0071] S1. Construct a multimodal, multi-dimensional structured dataset. This involves monitoring filtration resistance by installing pressure sensors in the filtrate outlet pipe, monitoring feed flow rate by installing flow sensors in the feed straight pipe section, and identifying filtrate color by installing a camera above the filtrate water collection tank. Filtration resistance, feed flow rate, and filtrate color are used as raw data to establish a multimodal array. After data preprocessing, this data forms a structured dataset for pattern recognition that jointly indicates the filter cake forming state. Pressure sensors are installed at the location of the pressure gauge on the filter press to detect system pressure and transmit the signal to the controller. After data preprocessing, the true pressure control value of the filter press is obtained.

[0072] S2, combining the characteristics of the filter cake forming stage of the filter press, establish a pattern recognition model based on multi-dimensional information; explain the logical relationship between the filtration resistance, feed flow rate, and filtrate color in the preprocessed structured dataset and the three stages of filter cake formation, dehydration, and peeling, and indicate the stage of filter cake formation based on the principle of relative majority.

[0073] S3 uses deep modeling to predict the target value of filter press pressure control. The filter resistance, feed flow rate, filtrate color and the actual value of filter press pressure in the preprocessed structured dataset are used as inputs to a Long Short-Term Memory (LSTM) network to dynamically model the time series data, establish the nonlinear relationship between pressure fluctuation and time dimension during the filter cake forming process, and calculate the target value of filter press pressure control.

[0074] S4. Establish an adaptive pressure control strategy; based on the different characteristics of the three stages of filter cake formation, dewatering, and peeling, establish a corresponding improved segmented PID control model. Use the pressure difference between the actual pressure control value and the pressure control target value as the input of the improved segmented PID controller, and use the output of the improved segmented PID controller to control the working pressure of the filter press to achieve dynamic regulation of the filter press pressure.

[0075] The specific implementation process of the present invention will be described in detail below with reference to specific embodiments.

[0076] I. Data Acquisition and Preprocessing

[0077] To ensure the integrity of the perceived information, this invention extracts multimodal data to establish a multi-dimensional structured dataset. It monitors filtration resistance using a pressure sensor, feed fluid flow rate using a flow sensor, filtrate color using a camera, and the actual pressure control value of the filter press using a pressure sensor. A time-series structured array aligned with the time axis is established through average value filtering and serial normalization.

[0078] First, set the sampling interval of all sensing devices (including input parameters such as pressure sensors - filter pressure, flow sensors, and cameras; and output parameters such as pressure sensors - the actual value of filter press pressure control) to 500ms, and collect data from pressure sensor P, flow sensor Q, and camera image F to form a structured array. , This represents an array of input parameters. This represents the actual pressure control data for the filter press.

[0079] To balance the volatility of the acquired data, the sampled data undergoes smoothing filtering. Taking the filtering of time-series data acquired by a pressure sensor as an example, the specific process includes:

[0080] (1) Let The pressure sensor value is constantly filtered. Collect and record In the array This represents the pressure sensor values ​​from the previous 10 moments. This represents the pressure sensor values ​​from the previous 9 time points.

[0081] (2) After smoothing filtering The filter pressure sensor value at each step of the calculation process is... It satisfies the following formula.

[0082] ;

[0083] Similarly, we can obtain Therefore, the data after average filtering is , This represents the array of input parameters after average filtering. express The smoothing and filtering results of the feed fluid flow sensor at all times. express The smoothing filter result of the filtrate color camera at any given time. express The result of smoothing and filtering the actual pressure control value of the filter press at any time.

[0084] Pressure sensor that has undergone average filtering and flow sensor To implement a 0-1 distribution and normalize to the same coordinate axis, the specific process is as follows:

[0085] (1) Let The pressure sensor value after undergoing average filtering is: ,Record In the array This represents the pressure sensor values ​​from the previous 10 time points. This represents the pressure sensor values ​​from the previous 9 time points.

[0086] (2) Perform 0-1 distribution calculation to obtain the filter pressure pressure sensor data processing results Proceed to the next step, the formula is as follows.

[0087] ;

[0088] In the formula express The minimum value in the array; express The maximum value in the array.

[0089] Similarly, we can obtain and . express The result of the feed fluid flow sensor at any time is processed by smoothing and filtering and then distributed into a 0-1 distribution. express The result of applying a 0-1 distribution to the actual pressure control value of the filter press after smoothing and filtering.

[0090] HSL transformation is performed on the image data that has undergone average filtering. Since the images captured by the camera are in RGB color format, the RGB color space is converted to the HSL (Hue, Saturation, Intensity) color space, which is closer to the human eye's perception. Specifically, the R, G, and B components are first normalized to the range [0,1]. The conversion relationship between hue (H) and RGB is as follows:

[0091] ;

[0092] Here The calculation relationships are as follows:

[0093] ;

[0094] The conversion relationship between color saturation S and RGB is as follows:

[0095] ;

[0096] The conversion relationship between the intensity component L and RGB is as follows:

[0097] ;

[0098] From the above formula, we can conclude that after normalizing R, G, and B to the interval [0,1], the range of the H component is [0,360], while the ranges of the S and L components are [0,1].

[0099] Based on the above data acquisition and processing flow, a structured dataset of input parameters is obtained. and output parameters .

[0100] II. Data Processing for Stage Identification Model

[0101] Filter cake formation consists of three stages: formation, dehydration, and peeling. In the initial stage of filter press operation, the filtrate flow rate is typically high and the liquid is turbid. As the filter cake gradually forms, the filtration resistance gradually increases, the filtrate flow rate decreases significantly, and the color gradually becomes clearer. This indicates that the filter cake has initially formed and has entered the pressing stage. For accurate pattern recognition, a pressure sensor reflects the filtration resistance; a flow sensor monitors the feed fluid flow rate; and a camera identifies the filtrate color. The relationships between filtration resistance, feed fluid flow rate, filtrate color, and the filter cake formation, dehydration, and peeling stages are defined. A voting mechanism is used to determine the current stage of filter cake formation.

[0102] 1. Define the relationship between filtration resistance and the filter cake formation, dewatering, and peeling stages. Based on the principle that filtration resistance gradually increases as the filter cake gradually forms, the following judgment is made:

[0103] like At this point, the filter cake is forming.

[0104] like At this point, the filter cake is in the dehydration stage;

[0105] like At this point, the filter cake is being peeled off.

[0106] 2. Define the relationship between feed fluid flow rate and the filter cake formation, dewatering, and peeling stages. Initially, the filtrate flow rate is typically high; as the filter cake gradually forms, the filtrate flow rate decreases significantly. The following criteria are applied:

[0107] like At this point, the filter cake is forming.

[0108] like At this point, the filter cake is in the dehydration stage;

[0109] like At this point, the filter cake is being peeled off.

[0110] 3. Define the relationship between filtrate color and the stages of filter cake formation, dehydration, and peeling. Initially, the filtrate is turbid; as the filter cake gradually forms, the filtrate color gradually becomes clearer. Although there is no direct relationship between H, S, and L, based on their color distribution characteristics, the following relationship exists when dividing the color regions:

[0111] When the lightness (L) component is extremely high or extremely low, the hue (H) will have no effect on color recognition.

[0112] When the saturation S, which represents color saturation, is extremely low or even close to 0, the hue H will fluctuate within a certain range and cannot be stable;

[0113] When the lightness value (L) is extremely high or extremely low, the color saturation (S) will have no effect on color recognition.

[0114] The color is expressed by splitting it into three components: HSL. Since the color subdivision in the colored area is more complex, only the threshold range of black (completely cloudy, corresponding to the filter cake formation stage) and white (completely clear, corresponding to the filter cake peeling stage) in the non-colored area is judged. The constant data empirical value used as the judgment threshold is selected in the formula.

[0115] If at least one of the following three conditions is met, it is determined that the filter cake is currently in the formation stage.

[0116] ; )&( ; )&( ;

[0117] If at least one of the following three conditions is met, it is determined that the filter cake is currently in the peeling stage.

[0118] ; )&( ; )&( ;

[0119] If neither of the above two judgment conditions is met at the same time, it is determined that the filter cake is in a dehydration state.

[0120] 4. The stage of filter cake formation is determined through a voting mechanism, using the principle of relative majority vote. A state with more than 2 votes indicates the stage at which the filter cake is forming.

[0121] III. Predicting Target Pressure Control Values ​​for Filter Presses Using Deep Modeling

[0122] The filter resistance, feed flow rate, filtrate color, and actual filter press pressure values ​​from the preprocessed structured dataset are used as inputs to a Long Short-Term Memory (LSTM) network. This establishes a nonlinear relationship between pressure fluctuations and the time dimension during filter cake formation, and calculates the target pressure control value for the filter press. The input parameter matrix of the LSTM model is defined, and the expressions for the forget gate, input gate, and output gate, as well as the weight parameter matrices, are set to satisfy the preconditions for LSTM model calculation. The specific steps are as follows:

[0123] Let the current time step be t, and the input be... for The hidden state at the previous moment was Cell state is The calculation process of LSTM is as follows:

[0124] The expression for the forget gate is as follows:

[0125] ;

[0126] In the formula The weight matrix representing the forget gate; The bias term representing the forget gate. This represents the Sigmoid function. .

[0127] Input gates are used to control the generation and storage of new information. The activation values ​​of the input gates are as follows:

[0128] ;

[0129] In the formula Indicates the output of the input gate; This represents the weight matrix of the input gate; This represents the bias term of the input gate.

[0130] The expression for the candidate new information is as follows:

[0131] ;

[0132] In the formula This represents the next state predicted by the cellular algorithm. The weight matrix representing cell renewal; The bias term representing cell renewal.

[0133] The cell state is updated by combining the results of the forget gate and the input gate, as shown in the following expression:

[0134] ;

[0135] In the formula Indicates cell renewal output; This represents element-wise multiplication (Hadamard product).

[0136] The expression for the output gate is as follows:

[0137] ;

[0138] In the formula Indicates the output of the output gate; This represents the weight matrix of the output gate; This represents the bias term of the output gate.

[0139] The hidden state update expression is as follows:

[0140] ;

[0141] The target value for pressure control of the filter press is calculated as the output value of the LSTM model. .

[0142] IV. Establishing an Adaptive Pressure Control Strategy

[0143] The pressure control strategy for the filter press is related to the filter cake forming stage. The filter cake forming stage requires low-pressure, fast-response control, so a PID controller is selected. The filter cake dewatering stage requires gradually increasing pressure control, so a non-incomplete derivative PID controller is selected. The filter cake peeling stage requires high-pressure, steady-state control, so a PI controller is selected. Based on this, an improved piecewise PID control algorithm is established, such as... Figure 2 As shown.

[0144] By selecting a positional PID controller, which controls based on accumulated error, high control accuracy and stability can be achieved. It directly outputs the actual position of the corresponding actuator without the need for manual / automatic switching, making the process simple.

[0145] Control the filter press pressure to the actual value. Pressure control calculation value output by LSTM The difference is used as the input to the positional PID control algorithm, and the output of the PID algorithm model is used as the setpoint for the pressure control of the filter press.

[0146] The error at time t is denoted by e(t), and its calculation expression is as follows:

[0147] ;

[0148] If the filter cake is forming, a PID controller is used, with the following expression:

[0149] ;

[0150] In the formula This represents the proportional gain of the PID controller; This represents the integral coefficient of the PID controller. Represents the derivative coefficients of the PID controller;

[0151] If the filter cake is in the dewatering stage, a partial derivative PID controller should be selected, such as... Figure 3 As shown.

[0152] A first-order inertial element is connected in series with the output of the PID controller to form an incomplete differential PID controller. u1(t) represents the output of the PID controller. The transfer function of the first-order inertial element D(s) is as follows:

[0153] ;

[0154] In the formula The step response coefficient of a first-order inertial element, expressed by an incompletely differential controller, is as follows:

[0155] ;

[0156] In the formula , This indicates the calculation result. This indicates the output of the PID controller.

[0157] If the filter cake is in the removal stage, a PI controller should be used, with the following expression:

[0158] ;

[0159] In the formula This represents the proportional gain of the PI controller; This represents the derivative coefficient of the PI controller.

[0160] V. Experimental Analysis

[0161] The improved piecewise PID was simulated in Matlab, as follows: Figure 4 and Figure 5 As shown in the experimental results, under the condition of the same step signal superimposed with the same random signal interference as input, the improved piecewise PID controller has a significantly better control effect than the traditional PID controller. The step response of the PID controller is relatively oscillating in the first 1-3 seconds, and although it is relatively stable thereafter, there are still small oscillations; while the step response of the improved piecewise PID controller is significantly smoother, with a faster response speed, and superior stability and anti-interference performance.

[0162] Table 1 shows the parameter configurations and simulation results for PID and the improved piecewise PID controller. The experimental results demonstrate that the improved piecewise PID controller significantly improves performance in terms of overshoot, settling time, and response time compared to the traditional PID controller.

[0163] Table 1 Parameter Configuration and Simulation Results

[0164]

[0165] like Figure 6 and Figure 7 As shown, the purple line represents the input, the yellow line represents the noise, and the green line represents the output. The horizontal axis represents time, and the vertical axis represents the EMC noise response. The experimental results show that, compared with the traditional PID controller, the improved segmented PID controller can quickly recover the balance state after being subjected to electromagnetic interference, and has better anti-interference and robustness.

[0166] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0167] While the specific embodiments of the present invention have been described above, they are not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for intelligent control of dynamic pressure during filter cake forming, characterized in that, Includes the following steps: S1, by installing a first pressure sensor in the filtrate outlet pipe to monitor the filtration resistance, installing a flow sensor in the feed straight pipe section to monitor the flow rate of the feed fluid, and installing a camera above the filtrate water collection tank to identify the filtrate color; A second pressure sensor is installed at the location of the pressure gauge built into the filter press to detect the system pressure and transmit the signal to the controller. After data preprocessing, the true pressure control value of the filter press is obtained. S2, the real-time obtained filtration resistance, feed flow rate and filtrate color are used to establish a multimodal array, which is then preprocessed and input into the stage identification model to indicate the stage of filter cake formation based on the principle of relative majority, including the filter cake formation stage, filter cake dehydration stage and filter cake peeling stage. S3 uses the pre-processed filtration resistance, feed flow rate, filtrate color, and actual filter press pressure as inputs to a Long Short-Term Memory (LSTM) network to dynamically model the time series data, establish a nonlinear relationship between pressure fluctuations and the time dimension during the filter cake forming process, and calculate the target value for filter press pressure control. S4. Based on the different characteristics of the filter cake forming stage, a corresponding segmented PID controller is established. The pressure difference between the actual pressure control value and the pressure control target value is used as the input of the segmented PID controller. The output of the segmented PID controller controls the working pressure of the filter press, thereby realizing the dynamic regulation of the filter press pressure.

2. The intelligent control method for dynamic pressure control in filter cake forming as described in claim 1, characterized in that, The pretreatment process for filtration resistance, feed flow rate, filtrate color, and actual filter press pressure includes: S11, Create an average value array of length 10 to perform smoothing filtering on the sampled data. The data after average filtering is: and ; express Filter pressure data continuously; express The smoothing and filtering results of the feed fluid flow sensor at all times. express The smoothing filter result of the filtrate color camera at any given time. express The result of smoothing and filtering the actual pressure control value of the filter press at any time; S12, distribute the filter pressure data, flow data, and actual pressure control data into a 0-1 distribution and normalize them to the same coordinate axis; thus obtaining... and ; S13. Perform HSL transformation on the image data that has undergone average filtering. First, normalize the three components R, G, and B to the range of [0, 1]. Calculate the hue H, color saturation S, and intensity component L. The range of the H component is [0, 360], while the ranges of the S and L components are [0, 1].

3. The intelligent control method for dynamic pressure control of filter cake forming as described in claim 2, characterized in that: The specific data processing procedure for the stage identification model in S2 is as follows: By setting a first threshold and a second threshold, the relationship between filtration resistance and the filter cake formation, dewatering, and peeling stages is defined. When the input value of filtration resistance is less than the first threshold, it is defined as being in the filter cake formation stage; when the input value of filtration resistance is greater than or equal to the first threshold and less than or equal to the second threshold, it is defined as being in the filter cake dewatering stage; when the input value of filtration resistance is greater than the second threshold, it is defined as being in the filter cake peeling stage. By setting a third and a fourth threshold, the relationship between the feed fluid flow rate and the filter cake formation, dewatering, and peeling stages is defined. When the feed fluid flow rate input value is less than the third threshold, it is defined as being in the filter cake formation stage; when the feed fluid flow rate input value is greater than or equal to the third threshold and less than or equal to the fourth threshold, it is defined as being in the filter cake dewatering stage. When the feed fluid flow rate input value is greater than the fourth threshold, it is defined as being in the filter cake stripping stage; After performing HSL transformation on the filtrate color image data, a first set of judgment conditions and a second set of judgment conditions are set based on the values ​​of the S and L components, thereby defining the relationship between filtrate color and the filter cake formation, dehydration, and peeling stages. The constant thresholds in the first set of judgment conditions and the second set of judgment conditions are all taken from empirical values. When the filtrate color meets the first set of judgment conditions, it is defined as being in the filter cake formation stage; when the filtrate color meets the second set of judgment conditions, it is defined as being in the filter cake peeling stage; when the filtrate color does not meet either the first set of judgment conditions or the second set of judgment conditions, it is defined as being in the filter cake dehydration stage.

4. The intelligent control method for dynamic pressure control of filter cake forming as described in claim 3, characterized in that: Based on preprocessed filtration pressure data : like At this point, the filter cake is forming. like At this point, the filter cake is in the dehydration stage; like At this point, the filter cake is being peeled off.

5. The intelligent control method for dynamic pressure control of filter cake forming as described in claim 3, characterized in that: Based on preprocessed traffic data : like At this point, the filter cake is forming. like At this point, the filter cake is in the dehydration stage; like At this point, the filter cake is being peeled off.

6. The intelligent control method for dynamic pressure control of filter cake forming as described in claim 3, characterized in that: Based on the transformed S and L components: If at least one of the following three conditions is met, it is determined that the filter cake is currently in the forming stage: If at least one of the following three conditions is met, it is determined that the filter cake is currently in the peeling stage: ; If neither of the above two judgment conditions is met at the same time, it is determined that the filter cake is in a dehydration state.

7. The intelligent control method for dynamic pressure control of filter cake forming as described in claim 2, characterized in that: Specifically, S3 is: Filtering resistance from preprocessed structured data Feed flow rate Filtrate color as well as , is defined as the input parameter matrix of the LSTM model; The expressions for the system's forget gate, input gate, and output gate, as well as the weight parameter matrices, are defined to satisfy the preconditions for LSTM model calculation. A nonlinear relationship between pressure fluctuations and the time dimension during filter cake forming is established, and the target pressure control value for the filter press is calculated to provide the output results for the LSTM model. .

8. The intelligent control method for dynamic pressure control of filter cake forming as described in claim 2, characterized in that: In step S4, the filter press pressure is controlled to the actual value. Pressure control calculation value output by LSTM The difference is used as the input of the positional PID control algorithm, and the output of the PID algorithm model is used as the given value for the pressure control of the filter press. If it is in the filter cake forming stage, a PID controller is used; if it is in the filter cake dewatering stage, an incomplete derivative PID controller is selected; if it is in the filter cake peeling stage, a PI controller is selected. Control is achieved by accumulating the error, and the actual position of the corresponding actuator is directly output.

Images