Thickening agent adding control system based on online viscosity and salinity double-parameter feedback

Abstract

The invention provides a thickening agent adding control system based on online viscosity and salt double parameter feedback, and relates to the technical field of industrial automation, the method comprises the following steps: collecting a slurry conductivity signal in real time, establishing a complex nonlinear mapping relation between the conductivity signal and preset mineral ion interference, dynamically taking a current conductivity value as an input, and calculating the viscosity and salt content of the slurry; outputting the real-time corrected salinity concentration; using real-time correction salinity concentration as a trigger condition, and activating an online rotary viscometer to collect a viscosity original time sequence signal; kalman filtering processing of mechanical damping lag errors is carried out on the time sequence signals, three characteristic measurement points are dynamically calibrated in a continuous time sequence window to construct a covariance matrix, characteristic values of the covariance matrix are extracted to generate viscosity compensation factors, and after original viscosity data are corrected, a real-time compensation viscosity value is output. According to the invention, dynamic and accurate regulation and control of addition of the thickening agent are realized.

Description

Technical Field

[0001] This invention relates to the field of industrial automation technology, and in particular to a thickener addition control system based on online viscosity and salinity dual-parameter feedback. Background Technology

[0002] In the field of thickener addition control, the coordinated regulation of viscosity and salinity of the slurry system is a core element in ensuring product quality stability. Traditional technical solutions are gradually revealing significant limitations when dealing with complex working conditions, especially in the application of neural network algorithms, which have multiple defects: The integration of neural networks with other sensor data is insufficient. In traditional methods, viscosity signal acquisition is often disconnected from salinity monitoring, and neural networks are only used for static fitting of a single parameter, without considering the coupling effect between salinity and viscosity. This means that when salinity changes drastically, the hysteresis of the viscosity response cannot be corrected in real time by the neural network, leading to deviations between the thickener addition amount and the actual requirements. Furthermore, traditional network structures lack a dynamic adjustment mechanism for the timing window, resulting in limited effectiveness in filtering out mechanical damping hysteresis errors in the raw viscosity signal, further reducing the reliability of parameter feedback. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a thickener addition control system based on online viscosity and salt content dual-parameter feedback, so as to realize dynamic and precise control of thickener addition.

[0004] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: A first aspect is a thickener addition control method based on online viscosity and salt content dual-parameter feedback, the method comprising: Step 1: Real-time acquisition of slurry conductivity signal, establishment of complex nonlinear mapping relationship between conductivity signal and preset mineral ion interference, dynamic use of current conductivity value as input, output of real-time corrected salinity concentration; Step 2: Using real-time correction of salinity concentration as the trigger condition, activate the online rotating viscometer to acquire the original viscosity time-series signal; perform Kalman filtering on the time-series signal to compensate for mechanical damping hysteresis error; dynamically calibrate three feature measurement points within a continuous time window to construct a covariance matrix; extract its eigenvalues ​​to generate a viscosity compensation factor; and output the real-time compensated viscosity value after correcting the original viscosity data. Step 3: Calculate the salinity-viscosity coupling coefficient based on the corrected salinity, dynamically solve the viscosity setting threshold, input the compensated viscosity value and the solved threshold into the controller for comparison, and generate the thickener basic addition amount instruction; Step 4: Respond to basic commands, collect the temperature of the thickener solution in real time, construct a two-dimensional feasible domain with temperature effect boundary constraints based on historical data, and calculate the Euclidean distance between the real-time temperature point and the boundary of each domain; when the Euclidean distance reaches the activation threshold, perform boundary recognition to generate corner compensation coefficients. If there are at least two kinds of interference among temperature, concentration and pressure, the two-dimensional feasible domain is expanded into a high-dimensional manifold. Step 5: Based on the high-dimensional manifold, mutually exclusive subdomains are divided by Delaunay triangulation and an adaptive compensation strategy is configured. The basic addition amount command is dynamically corrected using the corner compensation coefficient, and finally the metering pump control signal is output.

[0005] Furthermore, the slurry conductivity signal is acquired in real time, and a complex nonlinear mapping relationship is established between the conductivity signal and preset mineral ion interference. The current conductivity value is dynamically used as input to output real-time corrected salinity concentration, including: Step 1.1: Continuously acquire the raw conductivity signal of the slurry using an embedded conductivity sensor; Step 1.2: Receive the raw conductivity signal and perform sliding window normalization on it to eliminate sensor drift error and generate standardized conductivity time series data; Step 1.3: Based on the standardized conductivity time series data, call the preset mineral ion interference mode library, input the current conductivity value into the pre-trained deep neural network, and simultaneously analyze the conductivity interference components of calcium ions, magnesium ions and iron ions in the hidden layer of the neural network. Calculate the weight allocation of each ion interference through multi-layer nonlinear transformation, and generate an interference compensation vector containing the compensation coefficient of each ion in the output layer. Step 1.4: Perform dynamic correction based on the interference compensation vector. Subtract the interference components of calcium ions, magnesium ions and iron ions from the original conductivity value according to the weight ratio to obtain the pure conductivity value. According to the salinity-conductivity conversion curve, map the pure conductivity value to the initial salinity concentration. Perform moving average filtering on the initial salinity concentration within the time window to output the real-time corrected salinity concentration.

[0006] Furthermore, using real-time salinity concentration correction as a trigger condition, the online rotating viscometer is activated to acquire the raw viscosity time-series signal. The time-series signal is then processed using Kalman filtering to account for mechanical damping hysteresis error. Within a continuous time window, three characteristic measurement points are dynamically calibrated to construct a covariance matrix. Its eigenvalues ​​are extracted to generate a viscosity compensation factor. After correcting the raw viscosity data, the real-time compensated viscosity value is output, including: Step 2.1: When the real-time calibrated salinity concentration reaches the preset trigger threshold, activate the online rotational viscometer to collect the original time-series signal of the slurry viscosity; Step 2.2: Perform Kalman filtering on the original viscosity time series signal to eliminate the hysteresis error caused by the mechanical damping of the rotating viscometer and generate the filtered time series signal. Step 2.3: Dynamically calibrate the three characteristic measurement points of the filtered time-series signal within a continuous time window: peak point, valley point, and inflection point. Step 2.4: Construct a covariance matrix based on the three feature measurement points, and extract the largest eigenvalue as the viscosity compensation factor through eigenvalue decomposition. Step 2.5: Multiply the viscosity compensation factor by the original viscosity time-series signal to output the real-time compensated viscosity value.

[0007] Furthermore, based on the salinity correction, the salinity-viscosity coupling coefficient is calculated, and a viscosity threshold is dynamically determined. The compensated viscosity is then compared with the determined threshold and input to the controller to generate a thickener base addition instruction, including: Step 3.1: Obtain the real-time corrected salinity concentration. Using the salinity-viscosity coupling coefficient calculator, output the coupling coefficient based on the proportional relationship between the salinity change gradient and the viscosity response rate. Step 3.2: Input the coupling coefficient into the viscosity threshold solver, add the preset reference viscosity value and multiply it by the dynamic gain factor to generate the viscosity setting threshold. Step 3.3: Input the real-time compensated viscosity value and the viscosity setting threshold into the proportional-integral-derivative controller for real-time comparison, and generate a thickener basic addition amount instruction based on the comparison result.

[0008] Furthermore, in response to basic commands, the temperature of the thickener solution is collected in real time. Based on historical data, a two-dimensional parametric feasible domain of temperature effect boundary constraints is constructed, and the Euclidean distance between the real-time temperature point and the boundary of each domain is calculated, including: Step 4.1: Receive the instruction for the basic amount of thickener to be added, and collect the temperature value of the thickener solution in real time; Step 4.2: Call the temperature and concentration dataset stored in the historical database, and perform the following operations to construct a two-dimensional parameter feasible region: extract historical temperature values ​​as the x-axis set; extract the corresponding historical concentration values ​​as the y-axis set; calculate the joint probability distribution of the sets, and determine the confidence level greater than the boundary of the distribution area; connect the boundary vertices to form a closed polygon, which is defined as the two-dimensional parameter feasible region. Step 4.3: Map the real-time temperature value to coordinate points and perform the following positioning calculations to calculate the vertical distance from the coordinate point to each boundary of the feasible region. When the boundary is a curve, use the Newton-Raphson iteration method to solve for the shortest distance from the coordinate point to the curve.

[0009] Furthermore, when the Euclidean distance reaches the activation threshold, boundary identification is performed to generate corner compensation coefficients. If at least two of the disturbances—temperature, concentration, and pressure—exist, the two-dimensional feasible region is extended to a high-dimensional manifold, including: Step 4.4: Receive the geometric distance calculation result. When any boundary distance value is less than or equal to the preset activation threshold, trigger the boundary recognizer. Step 4.5: Determine the relative positional relationship between the current temperature coordinate point and the feasible domain boundary through the boundary identifier, generate the corner compensation coefficient, and collect real-time signals from the temperature sensor, concentration sensor, and pressure sensor in parallel. Step 4.6: If at least two sensor signals exceed the preset fluctuation range, the spatial dimension expander is activated. The constructed two-dimensional parameter feasible domain is used as the basic structure, and interference parameters are added as new dimensions to construct a three-dimensional or higher parameter space.

[0010] Furthermore, based on a high-dimensional manifold, mutually exclusive subdomains are partitioned using Delaunay triangulation and an adaptive compensation strategy is configured. Corner compensation coefficients are used to dynamically correct the basic addition command, ultimately outputting a metering pump control signal, including: Step 5.1: Based on the high-dimensional manifold parameter space, the topological connectivity of discrete data points in the space is calculated using the Delaunay triangulation algorithm to generate a set of mutually exclusive subdomains covering the entire parameter space. Step 5.2: Based on the geometric characteristics of the set of mutually exclusive subdomains, match the compensation parameter weight coefficients corresponding to each subdomain from the pre-built adaptive strategy library to complete the loading of the compensation strategy; Step 5.3: Receive the corner compensation coefficient and thickener basic addition amount instructions, and perform the following dynamic correction process to locate the coordinate point in the high-dimensional manifold based on the currently collected temperature, concentration, and pressure values; The corresponding compensation parameter weighting coefficient is called according to the subdomain to which the coordinate point belongs; the corner compensation coefficient and the compensation parameter weighting coefficient are weighted and fused to generate a comprehensive compensation coefficient; the comprehensive compensation coefficient is applied to the thickener basic addition amount instruction to calculate the corrected addition amount and output the pulse width modulation signal corresponding to the corrected addition amount.

[0011] Secondly, a thickener addition control system based on online viscosity and salt content dual-parameter feedback includes: The acquisition module is used to collect the slurry conductivity signal in real time, establish a complex nonlinear mapping relationship between the conductivity signal and preset mineral ion interference, dynamically take the current conductivity value as input, and output the real-time corrected salinity concentration. The calculation module is used to activate the online rotating viscometer to acquire the original time-series viscosity signal, triggered by real-time correction of salinity concentration. It performs Kalman filtering on the time-series signal to compensate for mechanical damping hysteresis error, dynamically calibrates three feature measurement points within a continuous time window to construct a covariance matrix, extracts its eigenvalues ​​to generate a viscosity compensation factor, corrects the original viscosity data, and outputs the real-time compensated viscosity value. Based on the corrected salinity, it calculates the salinity-viscosity coupling coefficient, dynamically solves the viscosity setting threshold, compares the compensated viscosity value with the solved threshold input to the controller, and generates a thickener basic addition amount instruction. The extension module is used to respond to basic commands, collect the temperature of the thickener solution in real time, construct a two-dimensional feasible domain with temperature effect boundary constraints based on historical data, and calculate the Euclidean distance between the real-time temperature point and the boundary of each domain. When the Euclidean distance reaches the activation threshold, boundary recognition is performed to generate corner compensation coefficients. If there are at least two kinds of interferences among temperature, concentration and pressure, the two-dimensional feasible domain is extended into a high-dimensional manifold. The output module is used to divide mutually exclusive subdomains based on a high-dimensional manifold through Delaunay triangulation and configure an adaptive compensation strategy. It uses corner compensation coefficients to dynamically correct the basic addition command and finally outputs the metering pump control signal.

[0012] Thirdly, a computing device includes: One or more processors; A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to implement the method.

[0013] Fourthly, a computer-readable storage medium storing a program that, when executed by a processor, implements the method.

[0014] The above-described solution of the present invention has at least the following beneficial effects: By eliminating the interference of mineral ions on salinity monitoring through neural network dynamic compensation and optimizing viscosity signal accuracy through Kalman filtering and eigenvalue analysis, accurate real-time feedback of both salinity and viscosity parameters is achieved. By dynamically calculating the control threshold based on the salinity-viscosity coupling relationship, the limitations of single-parameter control are overcome. Through the dynamic expansion from the two-dimensional feasible domain to the high-dimensional manifold and the adaptive compensation strategy of Delaunay triangulation, the system effectively copes with multi-source composite interferences such as temperature, concentration, and pressure, improving the real-time performance and accuracy of thickener addition control, and ensuring the stability of the slurry system and the consistency of product quality. Attached Figure Description

[0015] Figure 1 This is a schematic flowchart of a thickener addition control method based on online viscosity and salt content dual-parameter feedback provided in an embodiment of the present invention.

[0016] Figure 2 This is a schematic diagram of a thickener addition control system based on online viscosity and salt content dual-parameter feedback provided in an embodiment of the present invention. Detailed Implementation

[0017] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0018] like Figure 1 As shown, embodiments of the present invention propose a thickener addition control method based on online viscosity and salinity dual-parameter feedback, the method comprising the following steps: Step 1: Real-time acquisition of slurry conductivity signal, establishment of complex nonlinear mapping relationship between conductivity signal and preset mineral ion interference, dynamic use of current conductivity value as input, output of real-time corrected salinity concentration; Step 2: Using real-time correction of salinity concentration as the trigger condition, activate the online rotating viscometer to acquire the original viscosity time-series signal; perform Kalman filtering on the time-series signal to compensate for mechanical damping hysteresis error; dynamically calibrate three feature measurement points within a continuous time window to construct a covariance matrix; extract its eigenvalues ​​to generate a viscosity compensation factor; and output the real-time compensated viscosity value after correcting the original viscosity data. Step 3: Calculate the salinity-viscosity coupling coefficient based on the corrected salinity, dynamically solve the viscosity setting threshold, input the compensated viscosity value and the solved threshold into the controller for comparison, and generate the thickener basic addition amount instruction; Step 4: Respond to basic commands, collect the temperature of the thickener solution in real time, construct a two-dimensional feasible domain with temperature effect boundary constraints based on historical data, and calculate the Euclidean distance between the real-time temperature point and the boundary of each domain; when the Euclidean distance reaches the activation threshold, perform boundary recognition to generate corner compensation coefficients. If there are at least two kinds of interference among temperature, concentration and pressure, the two-dimensional feasible domain is expanded into a high-dimensional manifold. Step 5: Based on the high-dimensional manifold, mutually exclusive subdomains are divided by Delaunay triangulation and an adaptive compensation strategy is configured. The basic addition amount command is dynamically corrected using the corner compensation coefficient, and finally the metering pump control signal is output.

[0019] In this embodiment of the invention, the slurry conductivity is monitored in real time and mineral ion interference is eliminated by using a neural network for dynamic compensation, resulting in accurate output of corrected salinity. Viscosity acquisition is activated by salinity as a trigger, and the accuracy of the viscosity signal is optimized by combining Kalman filtering and eigenvalue analysis, achieving reliable feedback of both salinity and viscosity parameters. Based on the dynamic calculation of the salinity-viscosity coupling relationship, the control threshold is generated to generate the basic amount of thickener to be added, overcoming the limitations of single-parameter control. By constructing a two-dimensional feasible domain and extending it to a high-dimensional manifold under multi-source interference, and combining Delaunay triangulation to achieve adaptive compensation correction, the invention effectively addresses the combined interference of temperature, concentration, and pressure, ultimately improving the real-time performance and accuracy of thickener addition control, ensuring the stability of the slurry system and the consistency of product quality.

[0020] In a preferred embodiment of the present invention, step 1 above may include: Step 1.1: Continuously acquire the raw conductivity signal of the slurry using an embedded conductivity sensor; Step 1.2: Receive the raw conductivity signal and perform sliding window normalization on it to eliminate sensor drift error and generate standardized conductivity time series data; Step 1.3: Based on the standardized conductivity time series data, call the preset mineral ion interference mode library, input the current conductivity value into the pre-trained deep neural network, and simultaneously analyze the conductivity interference components of calcium ions, magnesium ions and iron ions in the hidden layer of the neural network. Calculate the weight allocation of each ion interference through multi-layer nonlinear transformation, and generate an interference compensation vector containing the compensation coefficient of each ion in the output layer. Step 1.4: Perform dynamic correction based on the interference compensation vector. Subtract the interference components of calcium ions, magnesium ions and iron ions from the original conductivity value according to the weight ratio to obtain the pure conductivity value. According to the salinity-conductivity conversion curve, map the pure conductivity value to the initial salinity concentration. Perform moving average filtering on the initial salinity concentration within the time window to output the real-time corrected salinity concentration.

[0021] In this embodiment of the invention, continuous acquisition of conductivity signals is achieved through embedded sensors, and sensor drift errors are effectively eliminated by sliding window normalization processing. A neural network algorithm based on a multilayer perceptron architecture is used to dynamically learn the characteristics of mineral ion interference based on historical data and calculate compensation values ​​in real time, accurately deducting interference to generate purified conductivity signals, thus solving the problem that mineral ion interference is difficult to dynamically eliminate in traditional methods. Finally, real-time corrected salinity concentration is obtained through polynomial regression calculation, improving the accuracy and timeliness of salinity monitoring.

[0022] In this embodiment of the invention, the specific steps include: Step 1.1: Integrate the embedded conductivity sensor into the detection section of the slurry conveying pipeline (the sensor probe is directly immersed in the slurry to ensure full contact with the measured medium), and continuously collect the raw conductivity signal of the slurry at a preset sampling frequency (e.g., 30 times per second). This signal is output in voltage form (e.g., 0-10V for a conductivity range of 0-2000μS / cm), and sent to the data processing unit via a signal transmission line as the initial data source for subsequent signal processing.

[0023] In a specific embodiment: the sensor adopts a probe-type structure and is installed in the middle section of the pipeline (where the slurry flows stably and mixes evenly). The original signal collected at a certain moment corresponds to a voltage value of 5V, which, after preliminary conversion, corresponds to a conductivity of 1000μS / cm.

[0024] Step 1.2: After receiving the original conductivity signal, perform sliding window normalization on it: Set the sliding window length to 60 consecutive sampling points (corresponding to a time span of 2 seconds), calculate the mean and range (difference between the maximum and minimum values) of the original signal in each window; then, subtract the window mean from the original value of each sampling point in the window, and then divide by the range to complete the normalization transformation.

[0025] This operation effectively eliminates drift errors caused by long-term sensor operation (such as slow signal baseline shift due to electrode polarization) and generates standardized conductivity time-series data. For example, if the original signal shows a gradual upward trend due to drift within a certain window (from 950 μS / cm to 1050 μS / cm), after normalization, the data fluctuation range is standardized to [-0.5, 0.5], eliminating the systematic bias caused by drift.

[0026] In a specific embodiment: the original signal sequence is [950, 970, 990, 1010, 1030, 1050, ...], with a window mean of 1000 μS / cm and a range of 100 μS / cm. After normalization, it is converted into [-0.5, -0.3, -0.1, 0.1, 0.3, 0.5, ...], forming stable and standardized time series data.

[0027] Step 1.3: Based on the standardized conductivity time series data, call the system's pre-stored preset mineral ion interference pattern library (this library stores the interference feature samples of calcium ions, magnesium ions, and iron ions at different concentrations on conductivity, such as the corresponding relationship curves between the concentration of each ion and the conductivity deviation), and input the current standardized conductivity value into the pre-trained deep neural network (this network has been trained with a large amount of experimental data and has learned the nonlinear mapping law of the three ion interferences).

[0028] When the neural network is working, the hidden layer synchronously analyzes the conductive interference components of the three ions through multi-layer neuron operations: for example, the first layer of neurons extracts the high-frequency fluctuation features of the signal (corresponding to the instantaneous interference of iron ions), the second layer analyzes the low-frequency trend (corresponding to the continuous interference of calcium and magnesium ions), and determines the proportion of each ion interference in the total deviation through multi-layer nonlinear transformation (such as adjusting the weight allocation through activation functions) (e.g., calcium ions account for 40%, magnesium ions for 30%, and iron ions for 30%); finally, the output layer generates an interference compensation vector, which contains the compensation coefficients of the three ions (e.g., calcium ion compensation coefficient 0.02, magnesium ion 0.015, and iron ion 0.01).

[0029] In a specific embodiment: the standardized conductivity value is 0.3, and after neural network calculation, the output interference compensation vector is [0.02, 0.015, 0.01], which correspond to the interference compensation coefficients of calcium, magnesium and iron ions, respectively.

[0030] Step 1.4, Perform dynamic correction based on the interference compensation vector: Subtract the corresponding interference components from the original conductivity value (e.g., 1000 μS / cm) according to the weight ratio of each ion in the vector (40%, 30%, 30%) (calcium ion interference 1000×0.02×40%=8μS / cm, magnesium ion 1000×0.015×30%=4.5μS / cm, iron ion 1000×0.01×30%=3μS / cm), the total subtraction is 15.5μS / cm, and the pure conductivity value is 984.5μS / cm.

[0031] Subsequently, based on the salinity-conductivity conversion curve (which was experimentally calibrated by preparing standard solutions of different salinities, measuring their pure conductivity values, and plotting the corresponding relationships, such as 1.5% salinity corresponding to 900 μS / cm and 2.0% corresponding to 1200 μS / cm), the pure conductivity value of 984.5 μS / cm was mapped to an initial salinity concentration of 1.65%. Then, a moving average filter was applied to the initial salinity concentration within a time window (e.g., 5 seconds) (calculating the average of 5 initial values ​​within the window), and finally, a real-time corrected salinity concentration of 1.64% was output.

[0032] In a specific embodiment: the initial salinity concentration sequence [1.66%, 1.65%, 1.64%, 1.63%, 1.65%], after being filtered by moving average, has a real-time corrected salinity concentration of (1.66+1.65+1.64+1.63+1.65)÷5=1.64%.

[0033] In a preferred embodiment of the present invention, step 2 above may include: Step 2.1: When the real-time calibrated salinity concentration reaches the preset trigger threshold, activate the online rotational viscometer to collect the original time-series signal of the slurry viscosity; Step 2.2: Perform Kalman filtering on the original viscosity time series signal to eliminate the hysteresis error caused by the mechanical damping of the rotating viscometer and generate the filtered time series signal. Step 2.3: Dynamically calibrate the three characteristic measurement points of the filtered time-series signal within a continuous time window: peak point, valley point, and inflection point. Step 2.4: Construct a covariance matrix based on the three feature measurement points, and extract the largest eigenvalue as the viscosity compensation factor through eigenvalue decomposition. Step 2.5: Multiply the viscosity compensation factor by the original viscosity time-series signal to output the real-time compensated viscosity value.

[0034] In this embodiment of the invention, by setting a trigger threshold for real-time correction of salinity concentration, viscosity acquisition is activated on demand, avoiding unnecessary energy consumption. Kalman filtering is used to specifically eliminate the hysteresis error of the rotating viscometer's mechanical damping, purifying the original time-series signal. Peak, valley, and inflection point characteristic measurement points are dynamically calibrated within a continuous window to capture the signal's dynamic characteristics. A covariance matrix is ​​constructed based on these characteristic points, and the maximum eigenvalue is extracted through eigenvalue decomposition as a viscosity compensation factor, transforming the signal's statistical regularity into a quantitative correction basis. Finally, the real-time compensated viscosity value is output through the product of the compensation factor and the original signal. The entire process, from triggering logic and signal purification to characteristic quantification correction, progressively addresses the problems of "mechanical error interference and insufficient utilization of signal dynamics," enabling viscosity measurement to overcome inherent hardware limitations while adapting to the temporal variation characteristics of slurry.

[0035] In this embodiment of the invention, the specific steps include: Step 2.1: Pre-configure a salinity trigger threshold in the system (e.g., for a certain type of slurry, set the real-time corrected salinity concentration ≥5% to trigger acquisition); when the output real-time corrected salinity concentration meets the threshold, activate the online rotary viscometer adapted to the slurry conveying pipeline (its rotor is immersed in the slurry and generates resistance torque with the flow of the slurry), and continuously acquire the original viscosity time sequence signal within 10 seconds at a preset sampling frequency (e.g., 100Hz) to form a continuous signal sequence.

[0036] Step 2.2: For the original viscosity time-series signal, due to the mechanical damping of the rotary viscometer rotor bearing and transmission mechanism (obtained experimentally by introducing a transient viscosity change into a standard viscous liquid, recording the time difference between the instrument output signal and the actual change, and measuring an average lag time of 0.5 seconds), a time-axis reverse compensation strategy is adopted to eliminate the lag error: For each sampling point, based on the changing trend of adjacent data, the equivalent signal value corresponding to the lag time is traced back (or the signal at the leading moment is predicted by linear interpolation), and the lag measurement value is mapped to the actual occurrence time. For example, the sampling value of 155 mPa·s at the 10th second in the original signal is corrected by time alignment and returned to the time sequence position of the 9.5th second, generating a filtered time-series signal in the form of [149 mPa·s, 151 mPa·s, 153 mPa·s, ...], so that the signal change is synchronized with the actual viscosity fluctuation of the slurry.

[0037] Step 2.3: Set the continuous time window length to 20 sampling periods (corresponding to a time span of 0.2 seconds), and dynamically identify three types of feature points within the sliding window for the filtered time signal from Step 2.2: Peak point: Traverse the signal values ​​within the window and select the point with the largest value (e.g., for signals [149, 151, 153, 155, 152, ...] within the window, the peak point is the time corresponding to 155 mPa·s). Valley point: Select the point with the smallest value (e.g., the time corresponding to the minimum value of 149 mPa·s within the window). Inflection point: Identify the point where the derivative changes from positive to negative (or from negative to positive) by calculating the first derivative of the signal (e.g., the derivative is positive before a certain point and negative after a certain point, corresponding to a signal value of 152 mPa·s).

[0038] This allows for dynamic calibration of peak values, valley values, and inflection points within each window, capturing the dynamic characteristics of the signal.

[0039] Step 2.4: Extract three feature measurement points (e.g., peak v1 = 155 mPa·s, valley v2 = 149 mPa·s, inflection point v3 = 152 mPa·s). Based on their viscosity values ​​and timestamps, construct a three-dimensional data sample, calculate the covariance between samples, and generate a covariance matrix (reflecting the fluctuation correlation of feature points in viscosity and time dimensions). Perform eigenvalue decomposition on this matrix to obtain multiple eigenvalues ​​(e.g., λ1 = 8.5, λ2 = 2.3, λ3 = 1.1 after decomposition). Select the largest eigenvalue (e.g., λ1 = 8.5) as the viscosity compensation factor. This factor quantifies the signal fluctuation pattern implied by the feature points, providing a quantitative basis for viscosity correction.

[0040] Step 2.5 involves multiplying the viscosity compensation factor (e.g., 8.5) point by point with the original viscosity time-series signal (the baseline value needs to be calibrated experimentally; for example, if the baseline value is 10, the compensation calculation is the original value × (8.5 ÷ 10)). For instance, if the original signal is 150 mPa·s at a certain point, the compensated value is 150 × 0.85 = 127.5 mPa·s. After point-by-point correction, a real-time compensated viscosity value sequence is output, providing accurate and dynamically adaptive viscosity parameters for subsequent thickener addition control.

[0041] In a preferred embodiment of the present invention, step 3 above may include: Step 3.1: Obtain the real-time corrected salinity concentration. Using the salinity-viscosity coupling coefficient calculator, output the coupling coefficient based on the proportional relationship between the salinity change gradient and the viscosity response rate. Step 3.2: Input the coupling coefficient into the viscosity threshold solver, add the preset reference viscosity value and multiply it by the dynamic gain factor to generate the viscosity setting threshold. Step 3.3: Input the real-time compensated viscosity value and the viscosity setting threshold into the proportional-integral-derivative controller for real-time comparison, and generate a thickener basic addition amount instruction based on the comparison result.

[0042] In this embodiment of the invention, a dynamic calculation mechanism for the salinity-viscosity coupling coefficient is constructed to overcome the limitations of the traditional fixed proportional relationship and accurately quantify the correlation between salinity change gradient and viscosity response rate. Then, using the coupling coefficient as the core, a preset benchmark viscosity is superimposed and a dynamic gain factor is introduced to generate a viscosity setting threshold adapted to real-time salinity conditions, solving the problem that a single threshold cannot cover complex fluctuations. Finally, relying on the closed-loop comparison of the proportional-integral-derivative controller, the deviation between the real-time compensated viscosity and the dynamic threshold is converted into a thickener basic addition command, achieving real-time response and stable convergence of the control logic. The entire process, from coupling relationship modeling and threshold dynamic adaptation to closed-loop control execution, progressively overcomes the problem of insufficient control accuracy under salinity-viscosity cross-interference, laying the core logical foundation for precise control of thickener addition.

[0043] In this embodiment of the invention, the specific steps include: Step 3.1: Obtain the output real-time corrected salinity concentration and transmit it to the salinity-viscosity coupling coefficient calculator. First, calculate the salinity change gradient (the ratio of the salinity difference between two consecutive moments to the time interval), and simultaneously extract the real-time compensated viscosity value from Step 2. Calculate the viscosity response rate (the ratio of the viscosity difference within the same time interval to the time). Through pre-experimental calibration, simulate the viscosity response under different salinity gradients in the laboratory, record multiple sets of proportional relationships between "salinity change gradient and viscosity response rate," and store them as a built-in mapping table in the calculator. Real-time matching of the proportional values ​​corresponding to the gradient and rate is used as the coupling coefficient output.

[0044] In a specific calculation example: if the salinity was 5.0% at the previous moment and is currently 5.2% with a 10-second interval, then the salinity change gradient is (5.2% - 5.0%) ÷ 10s = 0.02% / s; if the viscosity was 100 mPa·s at the previous moment and is currently 105 mPa·s, then the viscosity response rate is (105 - 100) ÷ 10s = 0.5 mPa·s / s. If the ratio of this gradient to the rate in the mapping table is 0.8, then the output coupling coefficient is 0.8.

[0045] Step 3.2: Input the coupling coefficient into the viscosity threshold solver. The solver has a preset reference viscosity value (calibrated according to process requirements, such as 100 mPa·s for a certain slurry). First, the coupling coefficient is superimposed with the reference viscosity, and then a dynamic gain factor is introduced (by dividing the slurry concentration range through experiments, such as 0.9-1.1 for high concentration and 1.1-1.3 for low concentration). The superposition result is multiplied by the gain factor matched to the current working condition to generate the viscosity setting threshold.

[0046] In a specific calculation example: with a coupling coefficient of 0.8 and a reference viscosity of 100 mPa·s, the superposition yields 100 + 0.8 = 100.8 mPa·s; if the current slurry is of low concentration and the matching gain factor is 1.0, then the threshold is 100.8 × 1.0 = 100.8 mPa·s.

[0047] Step 3.3: Input the real-time compensated viscosity value and the viscosity setting threshold into the proportional-integral-derivative (PID) controller; first, compare the two to calculate the deviation value, and then process the deviation through the preset proportional coefficient, integral time, and derivative time (determined through debugging, such as proportional coefficient 2, integral time 5s, and derivative time 1s) respectively through the proportional link (directly outputting the adjustment amount according to the deviation), the integral link (accumulating past deviations to eliminate steady-state errors), and the derivative link (predicting the trend based on the deviation growth rate). Finally, the outputs of the three links are superimposed to generate the thickener basic addition amount instruction.

[0048] In a specific calculation example: the real-time compensated viscosity is 95 mPa·s, the threshold is 100.8 mPa·s, and the deviation is 100.8 - 95 = 5.8 mPa·s; the proportional loop outputs 5.8 × 2 = 11.6, the integral loop accumulates 3 deviations and adds up to 2.4, and the derivative loop outputs 1 × 0.5 = 0.5 according to the deviation increase rate of 0.5 mPa·s / s. The sum of the three loops is 11.6 + 2.4 + 0.5 = 14.5, which corresponds to the generation of a thickener basic addition instruction of 14.5 units.

[0049] In a preferred embodiment of the present invention, step 4 above may include: Step 4.1: Receive the instruction for the basic amount of thickener to be added, and collect the temperature value of the thickener solution in real time; Step 4.2: Call the temperature and concentration dataset stored in the historical database, and perform the following operations to construct a two-dimensional parameter feasible region: extract historical temperature values ​​as the x-axis set; extract the corresponding historical concentration values ​​as the y-axis set; calculate the joint probability distribution of the sets, and determine the confidence level greater than the boundary of the distribution area; connect the boundary vertices to form a closed polygon, which is defined as the two-dimensional parameter feasible region. Step 4.3: Map the real-time temperature value to coordinate points and perform the following positioning calculations to calculate the vertical distance from the coordinate point to each boundary of the feasible region. When the boundary is a curve, use the Newton-Raphson iteration method to solve for the shortest distance from the coordinate point to the curve.

[0050] In this embodiment of the invention, by receiving the thickener basic addition amount instruction and simultaneously collecting temperature values, a correlation link between the control logic and environmental interference parameters is constructed. Based on the historical temperature-concentration dataset, by extracting the coordinate set, calculating the joint probability distribution and confidence boundary, the effective correlation rules of past operating conditions are transformed into a two-dimensional parameter feasible domain (closed polygon constraint), breaking through the static limitations of a single threshold. For the real-time temperature mapping point, the deviation between the current operating condition and the historical effective domain is accurately quantified by calculating the vertical distance (using the curve boundary adaptation Newton iteration method to solve the shortest distance). Real-time deviation quantification solves the problem of "control boundary ambiguity under multi-parameter coupling interference of temperature and concentration" layer by layer.

[0051] In this embodiment of the invention, the specific steps include: Step 4.1: The control system receives the thickener basic addition command (e.g., 15 units) output in Step 3 through the hardware signal interface; it synchronously triggers the temperature sensor (using a platinum resistance sensor with an accuracy of ±0.1℃, ensuring the probe is submerged in the solution flow channel during installation) deployed on the outlet pipeline of the thickener solution storage tank, and collects the solution temperature value in real time at a preset sampling interval (e.g., 2 seconds / time).

[0052] In a specific embodiment: if the temperature of the thickener solution is 24.5℃ at a certain moment, the sensor converts this physical quantity into a 4-20mA standard electrical signal (corresponding to 16.8mA for 24.5℃ when the temperature range is 0-50℃), and transmits it to the system data processing module as a real-time input for subsequent parameter domain analysis.

[0053] Step 4.2: The system retrieves the temperature-concentration correlation dataset stored in the historical database (this dataset originates from past production records, accumulating 8000 sets of valid data, covering temperatures from 16-34℃ and concentrations from 5-19%), and constructs the two-dimensional parameter feasible region according to the following process: Extract coordinate sets: Extract all historical temperature values ​​from the dataset to form a horizontal axis set (e.g., {16.3℃, 17.8℃, ..., 33.7℃}); Simultaneously extract the thickener concentration values ​​corresponding to each temperature group to form a vertical axis set (e.g., {5.2%, 6.5%, ..., 18.9%}).

[0054] Valid data filtering: Data points are processed using density clustering algorithms to remove outliers caused by equipment malfunctions (such as sensor drift, or invalid points where the temperature suddenly jumps to 40℃), retaining valid data from qualified production conditions. Statistical analysis shows that 90% of the valid data is concentrated in the temperature range of 19-31℃ and the concentration range of 6-17%. Edge data points in this range (such as (19℃, 6%), (31℃, 6%), (31℃, 17%), and (19℃, 17%)) are selected as the boundary vertices of the distribution area.

[0055] Constructing the feasible region: Connect the boundary vertices in clockwise order to form a closed polygon (such as a rectangle or trapezoid), which is defined as a two-dimensional parametric feasible region, representing the effective working range of "temperature-concentration" verified in history.

[0056] In a specific embodiment: if the boundary vertices of the effective data after clustering are (19℃, 6%), (31℃, 6%), (31℃, 17%), and (19℃, 17%), they are connected to form a rectangular feasible region. The temperature-concentration of the thickener solution must fall within this range to ensure stability.

[0057] Step 4.3: Map the collected real-time temperature value (e.g., 24.5℃) and the synchronously collected thickener concentration value (13.0% measured by an online concentration sensor) to a coordinate point (24.5℃, 13.0%) in a two-dimensional coordinate system. Then calculate the distance from this point to each boundary of the feasible region constructed in Step 4.2: Straight line boundary (below the lower boundary: horizontal line with concentration = 6%): The distance is calculated directly by geometric difference, i.e., 13.0% - 6% = 7.0%, which represents the vertical distance from the point to the lower boundary.

[0058] Curve boundary (e.g., the boundary near 24℃ is arc-shaped due to data distribution; assuming the boundary curve is a quadratic function fit): The shortest distance is found using a successive approximation method. Initial distance estimation: Based on the curve trend, assume that the initial distance from this point to the curve is 3.0% (e.g., if the upper limit of the concentration at 24.5℃ is estimated to be 15.7%, 15.7%-13.0%≈2.7%, the example here is 3.0% for demonstration purposes). Fine-tuning verification: Gradually adjust the concentration value towards the curve (0.1% each time), and calculate the distance from the adjusted point to the curve (e.g., if adjusted to 13.5%, the distance is 2.5%; if adjusted to 14.0%, the distance is 2.2%). Determine the minimum value: when the distance no longer decreases (e.g., when adjusted to 14.3%, the distance is 2.7%; when adjusted in the opposite direction to 14.2%, the distance is 2.6%; the shortest distance is finally determined to be 2.7%).

[0059] In a specific embodiment: through stepwise approximation calculation, the shortest distance from the coordinate point (24.5℃, 13.0%) to the upper boundary curve is 2.7%. Combined with the 7.0% distance of the straight line boundary, the deviation of this point from the historical feasible region is fully quantified.

[0060] In a preferred embodiment of the present invention, step 4 above may include: Step 4.4: Receive the geometric distance calculation result. When any boundary distance value is less than or equal to the preset activation threshold, trigger the boundary recognizer. Step 4.5: Determine the relative positional relationship between the current temperature coordinate point and the feasible domain boundary through the boundary identifier, generate the corner compensation coefficient, and collect real-time signals from the temperature sensor, concentration sensor, and pressure sensor in parallel. Step 4.6: If at least two sensor signals exceed the preset fluctuation range, the spatial dimension expander is activated. The constructed two-dimensional parameter feasible domain is used as the basic structure, and interference parameters are added as new dimensions to construct a three-dimensional or higher parameter space.

[0061] In this embodiment of the invention, a parameter domain adaptive evolution mechanism under multi-interference scenarios is constructed through a progressive logic of "distance triggering → boundary identification → multi-parameter diagnosis → high-dimensional expansion": boundary responses are precisely triggered using geometric distance thresholds to avoid ineffective activation of compensation strategies, achieving intelligent triggering of "response upon interference approach"; in boundary identification, the relative position of the operating point and feasible domain is located to generate corner compensation coefficients, and real-time signals of temperature, concentration, and pressure are collected in parallel to construct diagnostic criteria for interference types; for complex interferences with fluctuations in at least two parameters, the limitations of the two-dimensional parameter domain are overcome, expanding it into a high-dimensional manifold space, making parameter constraints more closely fit the dynamic operating conditions of multi-interference coupling. The entire process effectively solves the technical pain point that "a single two-dimensional domain cannot cover multi-parameter dynamic interference," providing a more comprehensive spatial analysis framework for the accurate correction of thickener addition, and supporting the improvement of the control system's adaptability under complex operating conditions.

[0062] In this embodiment of the invention, the specific steps include: Step 4.4: Receive the output geometric distance calculation results (i.e., the distance values ​​from the real-time coordinate point to each boundary of the two-dimensional parameter feasible domain, such as 6.2℃ to the left boundary, 5.8℃ to the right boundary, 2.3% to the upper boundary, and 4.7% to the lower boundary); at the same time, the system pre-stores preset activation thresholds (through experimental calibration, the temperature boundary activation threshold is set to 3.0℃, and the concentration boundary activation threshold is set to 2.0%, which are used to determine whether the working point is close to the boundary and compensation needs to be activated).

[0063] In a specific embodiment: each distance value is compared with the corresponding threshold. If the distance to the upper boundary is calculated to be 1.9% at a certain moment (less than the concentration activation threshold of 2.0%), the system’s built-in boundary identifier is triggered, and the subsequent position judgment and compensation coefficient generation process is started to ensure that the compensation mechanism is activated only when necessary, thus avoiding resource waste.

[0064] Step 4.5: After the boundary identifier is triggered, the relative position of the real-time coordinate point and the feasible domain boundary is first determined by coordinate comparison: For example, if the real-time coordinate point is (26.5℃, 17.8%), its concentration value is only 0.2% away from the upper boundary (concentration 18%) and is located within the feasible domain, it is determined to be "located within the feasible domain and adjacent to the upper boundary"; if the coordinate point is (33℃, 12%), the temperature exceeds the right boundary (32℃) by 1℃, it is determined to be "located outside the feasible domain and deviating from the right boundary".

[0065] Based on this, a corner compensation coefficient is generated. The coefficient is positively correlated with the distance from the boundary (the closer the distance, the larger the coefficient): for example, when it is close to the upper boundary by 0.2%, the corresponding compensation coefficient is 0.18; when it is 1° away from the right boundary, the corresponding compensation coefficient is 0.15. This is used to quantify the correction requirements of the boundary effect.

[0066] Meanwhile, the system collects data in parallel through multiple signal acquisition channels: real-time value of 26.5℃ from the temperature sensor, real-time value of 17.8% from the concentration sensor, and real-time value of 0.32MPa from the pressure sensor (deployed in the slurry conveying pipeline), providing real-time data support for subsequent interference type determination.

[0067] Step 4.6: The system first retrieves the preset fluctuation range of each sensor (based on historical normal operating condition statistics, temperature fluctuation range ±1℃, concentration ±0.5%, pressure ±0.05MPa), and compares the collected real-time signals with the corresponding ranges.

[0068] In a specific embodiment: if the temperature is 26.5℃ (0.5℃ above the upper limit of 26℃) and the pressure is 0.32MPa (0.02MPa above the upper limit of 0.3MPa), but the concentration is 17.8% within the range (17%-18%), then because the two parameters are out of range, it is determined to be "multi-parameter coupling interference", and the spatial dimension expander is then activated.

[0069] The expander, based on the original two-dimensional feasible domain (temperature-concentration), adds pressure as a new dimension to construct a three-dimensional parameter space (temperature-concentration-pressure). Its feasible domain range is expanded to: temperature 18-32℃, concentration 7%-18%, and pressure 0.25-0.35MPa. If subsequent flow rate fluctuations exceeding the range are detected (e.g., flow rate preset range ±5m),... 3 / h, real-time over 8m 3If / h), then a flow dimension is added to form a four-dimensional space, making the parameter constraint system adaptable to complex working conditions with multiple coupled factors.

[0070] In a preferred embodiment of the present invention, step 5 above may include: Step 5.1: Based on the high-dimensional manifold parameter space, the topological connectivity of discrete data points in the space is calculated using the Delaunay triangulation algorithm to generate a set of mutually exclusive subdomains covering the entire parameter space. Step 5.2: Based on the geometric characteristics of the set of mutually exclusive subdomains, match the compensation parameter weight coefficients corresponding to each subdomain from the pre-built adaptive strategy library to complete the loading of the compensation strategy; Step 5.3: Receive the corner compensation coefficient and thickener basic addition amount instructions, and perform the following dynamic correction process to locate the coordinate point in the high-dimensional manifold based on the currently collected temperature, concentration, and pressure values; The corresponding compensation parameter weighting coefficient is called according to the subdomain to which the coordinate point belongs; the corner compensation coefficient and the compensation parameter weighting coefficient are weighted and fused to generate a comprehensive compensation coefficient; the comprehensive compensation coefficient is applied to the thickener basic addition amount instruction to calculate the corrected addition amount and output the pulse width modulation signal corresponding to the corrected addition amount.

[0071] In this embodiment of the invention, the Delaunay triangulation algorithm is used to divide the high-dimensional manifold parameter space into mutually exclusive subdomains, accurately capturing the topological relationships of discrete data points and constructing a structured analysis framework for complex parameter spaces. Based on the geometric features of the subdomains, pre-stored compensation parameter weight coefficients are matched to achieve precise adaptation between the compensation strategy and the characteristics of the subdomains. Then, combined with the subdomain to which the real-time coordinate point belongs, the corner compensation coefficients and weight coefficients are weighted and fused to generate a comprehensive compensation coefficient, which is applied to the thickener basic addition amount command and outputs a corresponding pulse width modulation signal. This forms a complete link from high-dimensional space partitioning and strategy matching to dynamic correction, solving the problem of insufficient targeting of compensation strategies under multi-parameter coupling, improving the accuracy of thickener addition control and adaptability to complex working conditions, and ensuring that the control signal highly matches actual needs.

[0072] In this embodiment of the invention, the specific steps include: Step 5.1: Based on the constructed high-dimensional manifold parameter space (such as a three-dimensional temperature-concentration-pressure space containing discrete data points such as (20℃, 8%, 0.2MPa), (25℃, 12%, 0.3MPa), etc.), the Delaunay triangulation method is used to process the discrete data points in the space: Specifically, by traversing all data points and calculating the Euclidean distance between them, the system prioritizes connecting the closest points to form triangular units (tetrahedrons in 3D space). This ensures that the circumsphere of each unit does not contain other data points, thus avoiding subdomain overlap. In this way, the entire high-dimensional parameter space is divided into multiple mutually exclusive subdomains (e.g., generating 20 tetrahedral subdomains in 3D space). Each subdomain is defined by its vertex coordinates (e.g., the vertices of a subdomain are (20℃, 8%, 0.2MPa), (22℃, 10%, 0.25MPa), (25℃, 9%, 0.22MPa), (23℃, 11%, 0.28MPa)). The subdomains do not overlap and completely cover the entire parameter space, providing structured units for subsequent policy matching.

[0073] In a specific embodiment: if there are 100 discrete data points in a three-dimensional space, after partitioning, 80 mutually exclusive tetrahedral subdomains are generated. Subdomain A covers the range of temperature 20-22℃, concentration 8-10%, and pressure 0.2-0.25MPa. Subdomain B is adjacent to subdomain A and covers the range of temperature 22-25℃, concentration 9-11%, and pressure 0.22-0.28MPa. This process is repeated to form a continuous spatial division.

[0074] Step 5.2: For the generated set of mutually exclusive subdomains, firstly extract the geometric features of each subdomain (such as the center coordinates, vertex extrema, and spatial volume of the subdomain; for example, the center of subdomain A is (21℃, 9%, 0.23MPa), and the volume corresponds to the product of the parameter ranges of 2℃×2%×0.05MPa); then call the pre-built adaptive strategy library, which pre-stores the correspondence between the geometric features of the subdomains and the weight coefficients of the compensation parameters (through experimental calibration, for example, subdomains with center coordinates closer to the upper temperature limit have higher temperature compensation weights).

[0075] By matching features (such as matching the geometric features of subdomain A with the "low temperature, low concentration, low pressure" type in the library), corresponding compensation parameter weight coefficients are assigned to each subdomain (such as temperature weight of 0.3, concentration weight of 0.4, and pressure weight of 0.3 for subdomain A), thus completing the loading of the compensation strategy.

[0076] In a specific embodiment: Subdomain B, with its center coordinates (23.5℃, 10%, 0.25MPa) in the middle of the parameter space, is matched with weighting coefficients of 0.2 for temperature, 0.3 for concentration, and 0.5 for pressure. This means that the impact of pressure fluctuations on the amount of thickener added in this subdomain needs to be compensated.

[0077] Step 5.3: Receive the output corner compensation coefficient (e.g., 0.12) and the output thickener base addition command (e.g., 15 units), and perform the following dynamic correction: Coordinate positioning: Based on the real-time collected temperature (24℃), concentration (11%), and pressure (0.26MPa), the corresponding coordinate point (24℃, 11%, 0.26MPa) is located in the high-dimensional manifold space. By comparing the range of the subdomain vertices, it is determined that the point belongs to the subdomain B generated in step 5.1.

[0078] Weighting Invocation: Based on subdomain B, invoke the compensation parameter weighting coefficients matched in step 5.2 (temperature 0.2, concentration 0.3, pressure 0.5).

[0079] Coefficient fusion: The corner compensation coefficient (0.12) is weighted and fused with the subdomain weight coefficient. For example, the fusion weight is allocated proportionally to the weight coefficient, and the comprehensive compensation coefficient is calculated as follows: 0.12×(0.2+0.3+0.5)=0.12 (or according to the experimentally calibrated fusion formula, such as 0.12×0.8+subdomain baseline coefficient 0.05=0.146).

[0080] Calibration and Output: Apply the comprehensive compensation coefficient (e.g., 0.146) to the basic addition command and calculate the addition amount after calibration: 15 × (1 + 0.146) = 17.19 units; then convert the addition amount into a pulse width modulation signal (e.g., 1 unit corresponds to a 10ms pulse width, 17.19 units correspond to a 171.9ms pulse) and output it to the metering pump actuator to complete the dynamic calibration.

[0081] In a specific embodiment: if the comprehensive compensation coefficient is 0.146, the basic addition amount is 15 units, and the corrected amount is 17.19 units, the corresponding pulse width is 171.9ms, ensuring that the amount of thickener added is adapted to the current high-dimensional parameter conditions.

[0082] like Figure 2 As shown, embodiments of the present invention also provide a thickener addition control system based on online viscosity and salinity dual-parameter feedback, including: The acquisition module is used to collect the slurry conductivity signal in real time, establish a complex nonlinear mapping relationship between the conductivity signal and preset mineral ion interference, dynamically take the current conductivity value as input, and output the real-time corrected salinity concentration. The calculation module is used to activate the online rotating viscometer to acquire the original time-series viscosity signal, triggered by real-time correction of salinity concentration. It performs Kalman filtering on the time-series signal to compensate for mechanical damping hysteresis error, dynamically calibrates three feature measurement points within a continuous time window to construct a covariance matrix, extracts its eigenvalues ​​to generate a viscosity compensation factor, corrects the original viscosity data, and outputs the real-time compensated viscosity value. Based on the corrected salinity, it calculates the salinity-viscosity coupling coefficient, dynamically solves the viscosity setting threshold, compares the compensated viscosity value with the solved threshold input to the controller, and generates a thickener basic addition amount instruction. The extension module is used to respond to basic commands, collect the temperature of the thickener solution in real time, construct a two-dimensional feasible domain with temperature effect boundary constraints based on historical data, and calculate the Euclidean distance between the real-time temperature point and the boundary of each domain. When the Euclidean distance reaches the activation threshold, boundary recognition is performed to generate corner compensation coefficients. If there are at least two kinds of interferences among temperature, concentration and pressure, the two-dimensional feasible domain is extended into a high-dimensional manifold. The output module is used to divide mutually exclusive subdomains based on a high-dimensional manifold through Delaunay triangulation and configure an adaptive compensation strategy. It uses corner compensation coefficients to dynamically correct the basic addition command and finally outputs the metering pump control signal.

[0083] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for controlling the addition of thickeners based on online viscosity and salt content dual-parameter feedback, characterized in that, The method includes: Step 1: Real-time acquisition of slurry conductivity signal, establishment of complex nonlinear mapping relationship between conductivity signal and preset mineral ion interference, dynamic use of current conductivity value as input, output of real-time corrected salinity concentration; Step 2: Using real-time correction of salinity concentration as the trigger condition, activate the online rotating viscometer to acquire the original viscosity time-series signal; perform Kalman filtering on the time-series signal to compensate for mechanical damping hysteresis error; dynamically calibrate three feature measurement points within a continuous time window to construct a covariance matrix; extract its eigenvalues ​​to generate a viscosity compensation factor; and output the real-time compensated viscosity value after correcting the original viscosity data. Step 3: Calculate the salinity-viscosity coupling coefficient based on the corrected salinity, dynamically solve the viscosity setting threshold, input the compensated viscosity value and the solved threshold into the controller for comparison, and generate the thickener basic addition amount instruction; Step 4: Respond to basic commands, collect the temperature of the thickener solution in real time, construct a two-dimensional feasible domain with temperature effect boundary constraints based on historical data, and calculate the Euclidean distance between the real-time temperature point and the boundary of each domain; when the Euclidean distance reaches the activation threshold, perform boundary recognition to generate corner compensation coefficients. If there are at least two kinds of interference among temperature, concentration and pressure, the two-dimensional feasible domain is expanded into a high-dimensional manifold. Step 5: Based on the high-dimensional manifold, mutually exclusive subdomains are divided by Delaunay triangulation and an adaptive compensation strategy is configured. The basic addition amount command is dynamically corrected using the corner compensation coefficient, and finally the metering pump control signal is output.

2. The thickener addition control method based on online viscosity and salt content dual-parameter feedback according to claim 1, characterized in that, Real-time acquisition of slurry conductivity signals; establishment of a complex nonlinear mapping relationship between conductivity signals and preset mineral ion interference; dynamic use of the current conductivity value as input; and output of real-time corrected salinity concentration, including: Step 1.1: Continuously acquire the raw conductivity signal of the slurry using an embedded conductivity sensor; Step 1.2: Receive the raw conductivity signal and perform sliding window normalization on it to eliminate sensor drift error and generate standardized conductivity time series data; Step 1.3: Based on the standardized conductivity time series data, call the preset mineral ion interference mode library, input the current conductivity value into the pre-trained deep neural network, and simultaneously analyze the conductivity interference components of calcium ions, magnesium ions and iron ions in the hidden layer of the neural network. Calculate the weight allocation of each ion interference through multi-layer nonlinear transformation, and generate an interference compensation vector containing the compensation coefficient of each ion in the output layer. Step 1.4: Perform dynamic correction based on the interference compensation vector. Subtract the interference components of calcium ions, magnesium ions and iron ions from the original conductivity value according to the weight ratio to obtain the pure conductivity value. According to the salinity-conductivity conversion curve, map the pure conductivity value to the initial salinity concentration. Perform moving average filtering on the initial salinity concentration within the time window to output the real-time corrected salinity concentration.

3. The thickener addition control method based on online viscosity and salt content dual-parameter feedback according to claim 2, characterized in that, Using real-time salinity concentration correction as a trigger condition, an online rotating viscometer is activated to acquire the raw viscosity time-series signal. The time-series signal is processed using Kalman filtering to address mechanical damping hysteresis. Within a continuous time window, three characteristic measurement points are dynamically calibrated to construct a covariance matrix. Its eigenvalues ​​are extracted to generate a viscosity compensation factor. After correcting the raw viscosity data, a real-time compensated viscosity value is output, including: Step 2.1: When the real-time calibrated salinity concentration reaches the preset trigger threshold, activate the online rotational viscometer to collect the original time-series signal of the slurry viscosity; Step 2.2: Perform Kalman filtering on the original viscosity time series signal to eliminate the hysteresis error caused by the mechanical damping of the rotating viscometer and generate the filtered time series signal. Step 2.3: Dynamically calibrate the three characteristic measurement points of the filtered time-series signal within a continuous time window: peak point, valley point, and inflection point. Step 2.4: Construct a covariance matrix based on the three feature measurement points, and extract the largest eigenvalue as the viscosity compensation factor through eigenvalue decomposition. Step 2.5: Multiply the viscosity compensation factor by the original viscosity time-series signal to output the real-time compensated viscosity value.

4. The thickener addition control method based on online viscosity and salt content dual-parameter feedback according to claim 3, characterized in that, Based on the salinity-viscosity coupling coefficient calculated using salinity correction, a viscosity threshold is dynamically calculated. The compensated viscosity and the calculated threshold are then compared with the controller to generate a basic thickener addition instruction, including: Step 3.1: Obtain the real-time corrected salinity concentration. Using the salinity-viscosity coupling coefficient calculator, output the coupling coefficient based on the proportional relationship between the salinity change gradient and the viscosity response rate. Step 3.2: Input the coupling coefficient into the viscosity threshold solver, add the preset reference viscosity value and multiply it by the dynamic gain factor to generate the viscosity setting threshold. Step 3.3: Input the real-time compensated viscosity value and the viscosity setting threshold into the proportional-integral-derivative controller for real-time comparison, and generate a thickener basic addition amount instruction based on the comparison result.

5. The thickener addition control method based on online viscosity and salt content dual-parameter feedback according to claim 4, characterized in that, Responding to basic commands, the system collects the temperature of the thickener solution in real time, constructs a two-dimensional parametric feasible region based on historical data to constrain the temperature effect boundary, and calculates the Euclidean distance between the real-time temperature point and the boundary of each region, including: Step 4.1: Receive the instruction for the basic amount of thickener to be added, and collect the temperature value of the thickener solution in real time; Step 4.2: Call the temperature and concentration dataset stored in the historical database, and perform the following operations to construct a two-dimensional parameter feasible region: extract historical temperature values ​​as the x-axis set; extract the corresponding historical concentration values ​​as the y-axis set; calculate the joint probability distribution of the sets, and determine the confidence level greater than the boundary of the distribution area; connect the boundary vertices to form a closed polygon, which is defined as the two-dimensional parameter feasible region. Step 4.3: Map the real-time temperature value to coordinate points and perform the following positioning calculations to calculate the vertical distance from the coordinate point to each boundary of the feasible region. When the boundary is a curve, use the Newton-Raphson iteration method to solve for the shortest distance from the coordinate point to the curve.

6. The thickener addition control method based on online viscosity and salt content dual-parameter feedback according to claim 5, characterized in that, When the Euclidean distance reaches the activation threshold, boundary identification is performed to generate corner compensation coefficients. If at least two of the disturbances—temperature, concentration, and pressure—exist, the two-dimensional feasible region is extended to a high-dimensional manifold, including: Step 4.4: Receive the geometric distance calculation result. When any boundary distance value is less than or equal to the preset activation threshold, trigger the boundary recognizer. Step 4.5: Determine the relative positional relationship between the current temperature coordinate point and the feasible domain boundary through the boundary identifier, generate the corner compensation coefficient, and collect real-time signals from the temperature sensor, concentration sensor, and pressure sensor in parallel. Step 4.6: If at least two sensor signals exceed the preset fluctuation range, the spatial dimension expander is activated. The constructed two-dimensional parameter feasible domain is used as the basic structure, and interference parameters are added as new dimensions to construct a three-dimensional or higher parameter space.

7. The thickener addition control method based on online viscosity and salt content dual-parameter feedback according to claim 6, characterized in that, Based on a high-dimensional manifold, mutually exclusive subdomains are partitioned using Delaunay triangulation and an adaptive compensation strategy is configured. Corner compensation coefficients are used to dynamically correct the basic addition command, ultimately outputting a metering pump control signal, including: Step 5.1: Based on the high-dimensional manifold parameter space, the topological connectivity of discrete data points in the space is calculated using the Delaunay triangulation algorithm to generate a set of mutually exclusive subdomains covering the entire parameter space. Step 5.2: Based on the geometric characteristics of the set of mutually exclusive subdomains, match the compensation parameter weight coefficients corresponding to each subdomain from the pre-built adaptive strategy library to complete the loading of the compensation strategy; Step 5.3: Receive the corner compensation coefficient and thickener basic addition amount instructions, and perform the following dynamic correction process to locate the coordinate point in the high-dimensional manifold based on the currently collected temperature, concentration, and pressure values; The corresponding compensation parameter weighting coefficient is called according to the subdomain to which the coordinate point belongs; the corner compensation coefficient and the compensation parameter weighting coefficient are weighted and fused to generate a comprehensive compensation coefficient; the comprehensive compensation coefficient is applied to the thickener basic addition amount instruction to calculate the corrected addition amount and output the pulse width modulation signal corresponding to the corrected addition amount.

8. A thickener addition control system based on online viscosity and salinity dual-parameter feedback, wherein the system implements the method as described in any one of claims 1 to 7, characterized in that, include: The acquisition module is used to collect the slurry conductivity signal in real time, establish a complex nonlinear mapping relationship between the conductivity signal and preset mineral ion interference, dynamically take the current conductivity value as input, and output the real-time corrected salinity concentration. The calculation module is used to activate the online rotating viscometer to acquire the original time-series viscosity signal, triggered by real-time correction of salinity concentration. It performs Kalman filtering on the time-series signal to compensate for mechanical damping hysteresis error, dynamically calibrates three feature measurement points within a continuous time window to construct a covariance matrix, extracts its eigenvalues ​​to generate a viscosity compensation factor, corrects the original viscosity data, and outputs the real-time compensated viscosity value. Based on the corrected salinity, it calculates the salinity-viscosity coupling coefficient, dynamically solves the viscosity setting threshold, compares the compensated viscosity value with the solved threshold input to the controller, and generates a thickener basic addition amount instruction. The extension module is used to respond to basic commands, collect the temperature of the thickener solution in real time, construct a two-dimensional feasible domain with temperature effect boundary constraints based on historical data, and calculate the Euclidean distance between the real-time temperature point and the boundary of each domain. When the Euclidean distance reaches the activation threshold, boundary recognition is performed to generate corner compensation coefficients. If there are at least two kinds of interferences among temperature, concentration and pressure, the two-dimensional feasible domain is extended into a high-dimensional manifold. The output module is used to divide mutually exclusive subdomains based on a high-dimensional manifold through Delaunay triangulation and configure an adaptive compensation strategy. It uses corner compensation coefficients to dynamically correct the basic addition command and finally outputs the metering pump control signal.

9. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the method as described in any one of claims 1 to 7.

Images