SCADA-based pulp grinding process control parameter optimization method and system

Abstract

The application relates to the technical field of SCADA, and particularly discloses a SCADA-based pulp grinding process control parameter optimization method and system. The pulp concentration, the grinding disc gap and the driving power are collected through a SCADA platform, and the water shortage state of the grinding disc gap is identified by using the high-frequency energy proportion of the grinding disc seat acoustic emission signal; when the water is short, the water replenishment closed-loop regulation is first performed, when the water is not short, the fiber length and the fanning rate are predicted through a process parameter-quality correlation model, and the process parameter correction amount is generated according to the prediction deviation, so that the linkage closed loop of the water replenishment control and the pulp grinding quality optimization is realized.

Description

Technical Field

[0001] This invention relates to the field of SCADA technology, and more specifically to a method and system for optimizing control parameters in the grinding process based on SCADA. Background Technology

[0002] The initial moisture content of wood chips varies depending on the season and storage conditions. After entering the pre-steaming and pulping process, the fiber cell walls absorb water and swell, and the fiber clumps undergo wetting, dispersion, and compressive shearing processes. For wood chips with low initial moisture content or uneven wetting, even if the SCADA system shows that the overall pulp concentration is within the set range, it does not mean that a stable and uniform free water distribution has been formed in the grinding disc gap.

[0003] Since the slurry concentration mainly reflects the overall ratio of fiber to water in the slurry, it is difficult to directly characterize the state of free water involved in lubrication, heat transfer and fiber dispersion in local areas between grinding discs. Therefore, there may still be local free water deficiency, which will increase the contact and friction between fibers and grinding discs and between fibers, causing fluctuations in grinding load, driving power and local temperature rise, which in turn leads to instability in fiber cutting, splitting and buffing processes, resulting in fluctuations in fiber length and buffing rate. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for optimizing control parameters of the pulping process based on SCADA, thereby solving the above-mentioned technical problems.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] The method for optimizing control parameters of the pulping process based on SCADA includes the following steps:

[0007] S1. The process parameters of the pulping process are collected in real time through the SCADA platform. The process parameters include pulp concentration, grinding disc gap and drive power.

[0008] S2. Collect the sound signal output by the acoustic emission sensor installed on the grinding disc base in real time within the preset time window, and calculate the high-frequency energy ratio based on the sound signal.

[0009] S3. Based on the high-frequency energy ratio and the pre-calibrated normal center and water-deficient center, determine whether the grinding disc is in a water-deficient state.

[0010] S4. When in a water shortage state, send a water replenishment command to the actuator and execute S2-S3 again; when not in a water shortage state, execute S5.

[0011] S5. Obtain the process parameters collected by the SCADA platform at the end of the time window, and output the predicted fiber length and predicted brooming rate by combining the process parameter-quality correlation model.

[0012] S6. Determine the process parameter correction amount based on the predicted fiber length and predicted buffing rate, and generate adjustment instructions.

[0013] As a further aspect of the present invention, the specific process for calculating the proportion of high-frequency energy is as follows:

[0014] The sound signal collected within the time window is subjected to a fast Fourier transform to obtain the sound signal spectrum;

[0015] Calculate the sum of the energy of all frequency components in the spectrum whose frequency is higher than a preset frequency threshold, and use this as the high-frequency energy; calculate the sum of the energy of all frequency components in the entire frequency band of the spectrum, and use this as the full-frequency energy.

[0016] Divide the high-frequency energy by the total frequency energy to obtain the high-frequency energy percentage.

[0017] As a further aspect of the present invention: the process of determining whether the water shortage state is as follows:

[0018] Multiple sets of pulping test data for wood chips with different initial moisture contents were obtained. The pulping test data included the proportions of the first and second high-frequency energy before and after the water replenishment command was executed. The absolute difference between the proportions of the first and second high-frequency energy was calculated as the screening value.

[0019] The pulping test data with a screening value < the first change threshold are marked as normal data, the pulping test data with a screening value > the second change threshold are marked as abnormal data, and the pulping test data with a screening value ≥ the first change threshold and ≤ the second change threshold are removed. The first change threshold is < the second change threshold and both the first change threshold and the second change threshold are preset.

[0020] Calculate the mean of the first high-frequency energy proportion in the normal and abnormal data respectively as the normal center and the water-deficient center, and calculate the Euclidean distance A1 and A2 between the current high-frequency energy proportion and the normal center and the water-deficient center respectively;

[0021] If A1 ≥ A2, then the water is in a state of shortage; otherwise, it is not in a state of shortage.

[0022] As a further aspect of the present invention: the water replenishment process is as follows:

[0023] Using the Euclidean distance A1 as the process variable and zero as the setpoint of the process variable, calculate the positive deviation obtained by subtracting the setpoint from the process variable;

[0024] The control output value is obtained by performing proportional-integral calculation on the positive deviation;

[0025] The control output value is limited to obtain the valve opening setpoint, and a water replenishment command containing the valve opening setpoint is generated and sent to the actuator.

[0026] As a further aspect of the present invention, the process of obtaining the predicted fiber length and predicted broom formation rate is as follows:

[0027] The slurry concentration, grinding disc gap, and drive power collected by the SCADA platform at the end of multiple historical time windows are obtained. The slurry concentration, grinding disc gap, and drive power are normalized respectively to obtain the normalized slurry concentration, grinding disc gap, and drive power. The normalized slurry concentration, grinding disc gap, and drive power are used as historical process parameter samples.

[0028] Fiber length and brooming rate detected after a preset lag time following the end of the historical time window are used as historical quality samples.

[0029] Historical process parameter samples and historical quality samples are paired according to time windows;

[0030] The regression model is structured as a quadratic polynomial regression function of fiber length and buffing rate with respect to normalized slurry concentration, grinding disc gap, and driving power. The quadratic polynomial regression function includes a first-order term, a quadratic term, and an interaction term.

[0031] Using paired historical process parameter samples as input variables and historical quality samples as output variables, the coefficients of the quadratic polynomial regression function are fitted using the least squares method with regularization terms to obtain the process parameter-quality correlation model.

[0032] Input the current process parameters into the process parameter-quality correlation model, and output the predicted fiber length and predicted brooming rate.

[0033] As a further aspect of the present invention: the process of generating adjustment instructions is as follows:

[0034] Calculate the length deviation between the predicted fiber length and the preset fiber length target value, and calculate the firification rate deviation between the predicted firification rate and the preset firification rate target value;

[0035] Using the sum of the squares of length deviation and brooming rate deviation as the loss function, the partial derivatives of the loss function with respect to the normalized slurry concentration, grinding disc gap and driving power are obtained respectively.

[0036] Each partial derivative is multiplied by a preset adjustment step size and inverted to obtain the normalized correction values ​​for slurry concentration, grinding disc gap, and drive power. The normalized correction values ​​are then converted back to the correction values ​​for slurry concentration, grinding disc gap, and drive power, and an adjustment command carrying the correction values ​​is generated.

[0037] The SCADA-based system for optimizing control parameters in the pulping process includes:

[0038] Data Acquisition Module: Real-time acquisition of process parameters during the pulping process via the SCADA platform. These parameters include pulp concentration, grinding disc clearance, and drive power.

[0039] Judgment module: Collects the sound signal output by the acoustic emission sensor installed on the grinding disc in real time within a preset time window, and calculates the high-frequency energy ratio based on the sound signal;

[0040] Based on the high-frequency energy ratio and the pre-calibrated normal center and water-deficient center, determine whether the grinding wheel is in a water-deficient state;

[0041] Branch module: When in a water shortage state, it sends a water replenishment command to the actuator and returns to the execution judgment module; when not in a water shortage state, it executes the control module.

[0042] Control module: Acquires process parameters collected by the SCADA platform at the end of the time window, and outputs predicted fiber length and predicted brooming rate by combining the process parameter-quality correlation model;

[0043] The process parameter correction amount is determined based on the predicted fiber length and predicted buffing rate, and adjustment instructions are generated.

[0044] The beneficial effects of this invention compared to the prior art are as follows:

[0045] This invention introduces the high-frequency energy ratio of the acoustic emission signal from the grinding disc to identify changes in friction state caused by insufficient free water between the grinding discs. This allows the control process to no longer rely solely on the overall pulp concentration to determine the grinding state, but to promptly trigger water replenishment adjustment when the overall concentration is within the set range but the local moisture distribution between the grinding discs is abnormal. Simultaneously, after the water shortage is resolved, this invention uses a correlation model between process parameters and fiber length and fibrillation rate to predict pulp quality. Based on the predicted quality deviation, it generates corrections for pulp concentration, grinding disc gap, and drive power, creating a closed-loop adjustment process that seamlessly integrates water replenishment control and quality control. This reduces the impact of local dry friction or increased frictional load on fiber cutting, splitting, and fibrillation stability, reduces pulp quality fluctuations caused by differences in initial moisture content and uneven wetting of wood chips, and improves the adaptability of the grinding process control to actual fiber state changes. Attached Figure Description

[0046] The invention will now be further described with reference to the accompanying drawings.

[0047] Figure 1 This is a schematic flowchart of the method for optimizing control parameters of the pulping process based on SCADA according to the present invention.

[0048] Figure 2 This is a schematic diagram of the process for obtaining the predicted fiber length and predicted brooming rate according to the present invention. Detailed Implementation

[0049] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] Please see Figures 1-2 As shown, this invention is a method for optimizing control parameters in a slurry grinding process based on SCADA, comprising the following steps:

[0051] S1. The process parameters of the pulping process are collected in real time through the SCADA platform. The process parameters include pulp concentration, grinding disc gap and drive power.

[0052] Specifically, the SCADA platform communicates with field devices via industrial Ethernet. The pulp concentration is measured by an online concentration transmitter installed in the feed pipe. The transmitter outputs a 4mA to 20mA current signal, which is converted into a digital value via an analog input module. This digital value corresponds to the mass concentration of oven-dry fibers in the pulp. The grinding disc gap is measured by a displacement sensor installed on the grinding disc adjustment mechanism. The displacement sensor outputs a 0V to 10V voltage signal, which is converted into a digital value via an analog input module. This digital value corresponds to the gap between the fixed and moving grinding discs. The drive power is uploaded by the main motor inverter of the refiner via the Modbus RTU protocol. The SCADA platform reads the value from the inverter's internal active power register, with the value measured in kilowatts. The SCADA platform polls each channel of the above three parameters every 2 seconds, binding the value obtained in each poll to the current timestamp and storing it in the corresponding tag point in the real-time database.

[0053] S2. Collect the sound signal output by the acoustic emission sensor installed on the grinding disc base in real time within the preset time window, and calculate the high-frequency energy ratio based on the sound signal.

[0054] In a preferred embodiment of the present invention, the specific process for calculating the proportion of high-frequency energy is as follows:

[0055] The acoustic emission sensor is fixed to the radial side of the grinding disc seat of the second-stage refiner using magnetic clamps. The voltage signal output by the acoustic emission sensor is amplified by a preamplifier and then converted from analog to digital by the data acquisition module at a sampling frequency of 1MHz. The data acquisition module stores the converted digital signal into a circular buffer in time sequence. Whenever the circular buffer is filled with a data frame of a preset duration of 0.5s, the data frame is sent to the upper computing unit. The upper computing unit divides the 0.5s data frame into multiple consecutive 2048-point subframes and performs a Fast Fourier Transform on each 2048-point subframe to generate the corresponding spectrum sequence.

[0056] The sum of the squares of the amplitudes of all frequency points in the spectrum sequence with frequencies higher than a preset frequency threshold, such as 100kHz, is calculated as the high-frequency energy. The sum of the squares of the amplitudes of all frequency points in the spectrum sequence is also calculated as the full-band energy. The upper-level computing unit calculates the ratio of the high-frequency energy to the full-band energy, rounds the ratio to four decimal places to obtain a high-frequency energy percentage, and averages the high-frequency energy percentages across multiple subframes. The average result is then used as the high-frequency energy percentage for the current time window.

[0057] S3. Based on the high-frequency energy ratio and the pre-calibrated normal center and water-deficient center, determine whether the grinding disc is in a water-deficient state.

[0058] In another preferred embodiment of the present invention, the process of determining whether the water shortage state is as follows:

[0059] During the calibration stage, at least three batches of wood chips are prepared. Take 500 grams of sample from each batch of wood chips and dry them in an oven at 105 degrees Celsius until constant weight. Calculate the initial moisture content by dividing the difference in mass before and after drying by the mass before drying. For example, the initial moisture contents of the three batches of wood chips are 15%, 25%, and 35%, respectively.

[0060] Multiple refining tests were conducted on each batch of wood chips. After each refining process reached stable operation, the high-frequency energy percentage was collected as the first high-frequency energy percentage. Then, a micro-water replenishment command was executed. After the water replenishment was completed, the high-frequency energy percentage was collected again as the second high-frequency energy percentage. Each test recorded one set of data, including one first high-frequency energy percentage and one second high-frequency energy percentage.

[0061] For each set of experimental data, the absolute value of the difference between the proportion of the first high-frequency energy and the proportion of the second high-frequency energy is calculated, and this absolute value is used as the screening value for that set of experimental data. The preset first change threshold is 0.02, and the preset second change threshold is 0.08. Experimental data with a screening value < 0.02 are marked as normal data, experimental data with a screening value > 0.08 are marked as abnormal data, and experimental data with a screening value ≥ 0.02 and ≤ 0.08 are discarded. When the number of normal and abnormal data both reach the preset number, the calculation of the normal center and the water-deficient center is performed; when the number of normal or abnormal data does not reach the preset number, grinding test data collection continues.

[0062] The arithmetic mean of the proportion of the first high-frequency energy in all normal data is calculated as the normal center, and the arithmetic mean of the proportion of the first high-frequency energy in all abnormal data is calculated as the water-deficient center. For example, the normal center is 0.0810 and the water-deficient center is 0.1760. These centers are stored in the online judgment module. During online judgment, the module receives the high-frequency energy proportion of the current time window, calculates the Euclidean distance A1 between the current high-frequency energy proportion and the normal center, and calculates the Euclidean distance A2 between the current high-frequency energy proportion and the water-deficient center. The Euclidean distance is calculated by subtracting the two values ​​and taking the absolute value. A1 and A2 are compared. If A1 is greater than or equal to A2, the water-deficient state judgment result is output; if A1 is less than A2, the non-water-deficient state judgment result is output.

[0063] It is understandable that using the change in the proportion of high-frequency energy before and after water replenishment to distinguish between normal and abnormal data is because when there is sufficient free water between the grinding discs, the water in the slurry can participate in fiber suspension, transmission, and buffering of the mechanical action between the grinding discs and fibers. The high-frequency components of the acoustic emission signal at the grinding disc seat change little before and after water replenishment. However, when there is insufficient free water between the grinding discs, localized dry friction or contact impact is more likely to occur between the fibers and the grinding discs, making the high-frequency acoustic emission components relatively prominent. After water replenishment, this contact state is altered, and the proportion of high-frequency energy changes significantly. Therefore, the absolute difference in the proportion of high-frequency energy before and after water replenishment can be used as a basis for sample selection.

[0064] Samples with smaller differences represent normal conditions that are not sensitive to water replenishment, while samples with larger differences represent water-deficient conditions that are sensitive to water replenishment. Removing data from the middle interval avoids boundary samples interfering with the determination of the central value.

[0065] The average of the high-frequency energy percentages before water replenishment in both normal and abnormal data is taken as the normal center and the water-deficient center, respectively. This is equivalent to establishing acoustic characteristic references for two types of free water states. During real-time operation, the distance between the current high-frequency energy percentage and each of the two centers is calculated. The center closer the current value is to indicates which historical state its acoustic characteristics are closer to. Therefore, when the distance from the current high-frequency energy percentage to the water-deficient center is not greater than the distance to the normal center, it can be determined that the grinding discs are in a water-deficient state; otherwise, it is determined that they are not in a water-deficient state. The judgment result obtained in this way does not rely solely on a fixed threshold, but uses the state reference formed by historical water replenishment responses to identify the current free water state between the grinding discs. This ensures that the water replenishment action is triggered only when the acoustic characteristics are close to the water-deficient state, and that subsequent process parameter optimization is based on a relatively normal free water state as much as possible. This avoids mistaking abnormal friction caused by water deficiency as a problem that requires adjustment of concentration, grinding disc gap, or drive power itself.

[0066] S4. When in a water shortage state, send a water replenishment command to the actuator and execute S2-S3 again; when not in a water shortage state, execute S5.

[0067] In a preferred embodiment of the present invention, the water replenishment process is as follows:

[0068] When a water shortage is detected, the water replenishment control module obtains the current value of the Euclidean distance A1, uses A1 as the process variable and 0 as the setpoint, and calculates the difference between the process variable and the setpoint; this difference is the positive deviation. The water replenishment control module has a built-in proportional-integral (PI) controller. The PI controller's proportional gain is preset to 0.8, and the integral time is preset to 3 seconds. When the positive deviation is input to the PI controller, the proportional term output equals the proportional gain multiplied by the current positive deviation, and the integral term output equals the proportional gain divided by the integral time and then multiplied by the cumulative positive deviation over time. The sum of the proportional and integral terms yields the control output value.

[0069] The control output value is limited, with an upper limit of 100 and a lower limit of 0. When the control output value is 110, the output after limiting is 100; when the control output value is -5, the output after limiting is 0; and when the control output value is 65, the output after limiting is 65. The value after limiting is used as the valve opening setpoint, in percentage.

[0070] The water replenishment control module generates a water replenishment command containing the valve opening setpoint and sends it to the water replenishment actuator via the Modbus TCP protocol. The actuator is an electric regulating valve. Upon receiving the water replenishment command, the electric regulating valve adjusts its opening to the position corresponding to the valve opening setpoint, injecting dilution water into the refiner inlet. During the water replenishment process, S2 and S3 are continuously executed, and A1 is updated in real time. The water replenishment control module cyclically executes the aforementioned proportional-integral calculation and amplitude limiting processing, adjusting the valve opening setpoint in real time. When S3 determines that there is no water shortage, the water replenishment control module sets the valve opening setpoint to 0, sends the water replenishment command to the electric regulating valve (valve opening setpoint 0), closes the electric regulating valve, and the water replenishment ends.

[0071] It is important to note that the Euclidean distance A1 is used as a process variable because A1 represents the degree to which the current high-frequency energy ratio deviates from the normal center. The larger A1 is, the further the acoustic emission characteristics between the grinding discs are from the normal state of free water, and the stronger the water replenishment requirement. Using zero as the set value and calculating the positive deviation is equivalent to using "returning to the vicinity of the normal center" as the control reference, so that the magnitude of the deviation can directly reflect the degree to which the current state needs to be corrected by water replenishment.

[0072] When performing proportional-integral (PI) calculations on positive deviations, the proportional component rapidly outputs the valve opening change based on the current deviation level; the larger the deviation, the greater the control output. The integral component accumulates continuously existing deviations, ensuring that even small individual deviations, as long as the grinding discs continue to deviate from their normal state, will gradually increase water replenishment, preventing prolonged water shortage due to insufficient replenishment. After amplitude limiting processing, the valve opening setpoint is obtained, converting the continuous control output obtained from the PI calculation into a safe opening range acceptable to the actuator, thus establishing a correspondence between the water replenishment amount and the degree of water shortage.

[0073] The essence of obtaining the valve opening setpoint in this way is to use the distance of the acoustic emission characteristics deviating from the normal state as a feedback quantity. Through closed-loop control, the acoustic anomaly is gradually pulled back to the vicinity of the normal state, so that the water replenishment action is not a simple on / off trigger, but is adjusted according to the degree of water shortage. This reduces the interference of dry friction or abnormal contact caused by insufficient free water on the grinding process, and provides a relatively stable process state for subsequent correction of slurry concentration, grinding disc gap and drive power based on process parameters and quality correlation models.

[0074] S5. Obtain the process parameters collected by the SCADA platform at the end of the time window, and output the predicted fiber length and predicted brooming rate by combining the process parameter-quality correlation model.

[0075] In another preferred embodiment of the present invention, the process of obtaining the predicted fiber length and the predicted broom formation rate is as follows:

[0076] During the model building phase, data corresponding to historical time windows within the past 30 days that were determined not to be in a water-scarce state are extracted from the SCADA platform's real-time database as sample sources. The slurry concentration, grinding disc gap, and drive power collected at the end of each historical time window are extracted. For example, a sample might have a slurry concentration of 28 g / L, a grinding disc gap of 0.3 mm, and a drive power of 180 kW. These data are recorded as historical process parameter samples. For the same historical time window corresponding to a sample, the fiber length and fissile rate detected online after a preset lag time (45 s) are extracted. For example, a detected fiber length of 1.05 mm and a fissile rate of 38% are recorded as historical quality samples. Historical process parameter samples and historical quality samples corresponding to the same historical time window are paired to form a modeling sample. This process is repeated to extract at least 100 sets of modeling samples to constitute a modeling sample set.

[0077] The regression model is constructed as follows:

[0078] L = X1 + X2Z1 + X3Z2 + X4Z3 + X5Z1 2 +X6Z2 2 +X7Z32 +X8Z1Z2+X9Z1Z3+X 10 Z2Z3;

[0079] R = Y1 + Y2Z1 + Y3Z2 + Y4Z3 + Y5Z1 2 +Y6Z2 2 +Y7Z3 2 +Y8Z1Z2+Y9Z1Z3+Y 10 Z2Z3;

[0080] Where L represents the predicted fiber length, R represents the predicted ablation rate, Z1 represents the normalized pulp concentration, Z2 represents the normalized grinding disc gap, Z3 represents the normalized drive power, X1 represents the constant term of the predicted fiber length regression function, X2 represents the fiber length coefficient corresponding to the first-order term of pulp concentration, X3 represents the fiber length coefficient corresponding to the first-order term of grinding disc gap, X4 represents the fiber length coefficient corresponding to the first-order term of drive power, X5 represents the fiber length coefficient corresponding to the second-order term of pulp concentration, X6 represents the fiber length coefficient corresponding to the second-order term of grinding disc gap, X7 represents the fiber length coefficient corresponding to the second-order term of drive power, X8 represents the fiber length coefficient corresponding to the interaction term between pulp concentration and grinding disc gap, and X9 represents the fiber length coefficient corresponding to the interaction term between pulp concentration and drive power. 10 Y1 represents the fiber length coefficient corresponding to the interaction term between the grinding disc gap and the driving power; Y2 represents the fibrillation rate coefficient corresponding to the first-order term of the slurry concentration; Y3 represents the fibrillation rate coefficient corresponding to the first-order term of the grinding disc gap; Y4 represents the fibrillation rate coefficient corresponding to the first-order term of the driving power; Y5 represents the fibrillation rate coefficient corresponding to the second-order term of the slurry concentration; Y6 represents the fibrillation rate coefficient corresponding to the second-order term of the grinding disc gap; Y7 represents the fibrillation rate coefficient corresponding to the second-order term of the driving power; Y8 represents the fibrillation rate coefficient corresponding to the interaction term between the slurry concentration and the grinding disc gap; and Y9 represents the fibrillation rate coefficient corresponding to the interaction term between the slurry concentration and the driving power. 10 This represents the brooming rate coefficient corresponding to the interaction term between the grinding disc clearance and the driving power.

[0081] Using Z1, Z2, Z3 and their quadratic and interaction terms as regression inputs, and fiber length as the first output variable, the coefficients of the regression model are fitted using the least squares method with a regularization term. Using the same input variables, and broomming rate as the second output variable, the coefficients of the regression model are fitted using the least squares method with a regularization term. The fiber length function and broomming rate function obtained by solving together constitute the process parameter-quality correlation model.

[0082] The objective functions with regularization terms are as follows:

[0083] minΣ(LQ) 2 +λ(X2)2 +X3 2 +X4 2 +X5 2 +X6 2 +X7 2 +X8 2 +X9 2 +X 10 2 );

[0084] minΣ(RW) 2 +μ(Y2 2 +Y3 2 +Y4 2 +Y5 2 +Y6 2 +Y7 2 +Y8 2 +Y9 2 +Y 10 2 );

[0085] Where Q represents the actual fiber length in the historical quality sample, W represents the actual brooming rate in the historical quality sample, Σ is the summation symbol, indicating that the error squares are calculated and summed for all modeling samples in the modeling sample set. Each sample corresponds to an LQ and RW, so summation is required. λ represents the preset regularization coefficient corresponding to the fiber length function, μ represents the preset regularization coefficient corresponding to the brooming rate function, and the constant terms X1 and Y1 do not participate in the regularization constraint. min indicates that by adjusting these coefficients, the total error square sum is minimized.

[0086] During online prediction, the pulp concentration, grinding disc gap, and drive power collected at the end of the current time window are obtained from the SCADA platform and normalized according to the same normalization relationship as in the model building stage to obtain the normalized pulp concentration, grinding disc gap, and drive power corresponding to the current time window. The normalized pulp concentration, grinding disc gap, and drive power, as well as their quadratic and interaction terms, are substituted into the fiber length function and the buffing rate function, respectively, to calculate the predicted fiber length and predicted buffing rate.

[0087] It is important to note that the regression model establishes a correspondence between process parameters and quality indicators by using pulp concentration, grinding disc gap, and drive power collected at the end of the historical time window as inputs, and fiber length and fibrillation rate detected after a preset lag time as outputs. This is because the mechanical intensity, residence, and pressure state of fibers during pulping change with concentration, gap, and power, and fiber length and fibrillation rate are the quality manifestations resulting from the accumulation of these mechanical effects. Since the fiber quality detected online is not formed immediately at the moment of process parameter acquisition, pairing the quality sample after lag with the current process parameters allows the input and output data to correspond causally in the process. After fitting the regression model with historical samples, the process parameter-quality correlation model can convert the current process parameters into predicted fiber length and predicted fibrillation rate, enabling the control process to obtain quality prediction results based on the current process state, and generating process parameter correction amounts based on the deviation between the quality prediction results and the target values. This allows the pulping process after water replenishment to continue closed-loop regulation around the target fiber quality.

[0088] The regression model is set as a function containing linear, quadratic, and interaction terms because the effects of pulp concentration, grinding disc gap, and drive power on fiber quality are usually not isolated and proportional. Changes in concentration affect fiber dispersion and force transmission within the grinding zone, changes in gap affect the degree of compression and shearing experienced by the fibers, and changes in power reflect the intensity of mechanical energy input. Furthermore, the effect of any parameter is constrained by the state of other parameters. The linear term represents the direct impact of a single process parameter on fiber length or fibrillation rate. The quadratic term represents the trend of the influence of the same parameter changing across different value ranges. The interaction term represents the coupling effect on quality indicators when two process parameters change together. With this setup, the model no longer simplifies the pulping process to a single linear relationship but can more fully characterize the correlation between process parameter combinations and fiber quality within the historical sample coverage. This allows subsequent calculations of corrections based on predicted quality deviations to simultaneously consider the effects of single-parameter changes and multi-parameter linkages on fiber length and fibrillation rate, reducing discrepancies between adjustment directions and actual quality changes.

[0089] S6. Determine the process parameter correction amount based on the predicted fiber length and predicted buffing rate, and generate adjustment instructions.

[0090] In another preferred embodiment of the present invention, the process of generating adjustment instructions is as follows:

[0091] Read the preset fiber length target value A and the preset fibrillation rate target value B, calculate the length deviation LA and fibrillation rate deviation RB, and then... 2 +(RB) 2As loss functions, L and R represent the predicted fiber length and predicted brooming rate obtained in the preceding steps, respectively. The partial derivative of the loss function with respect to Z1 is calculated as 2(LA)(X2+2X5Z1+X8Z2+X9Z3)+2(RB)(Y2+2Y5Z1+Y8Z2+Y9Z3), and the partial derivative of the loss function with respect to Z2 is calculated as 2(LA)(X3+2X6Z2+X8Z1+X... 10 Z3)+2(RB)(Y3+2Y6Z2+Y8Z1+Y 10 Z3), the partial derivative of the loss function with respect to Z3 is calculated as 2(LA)(X4+2X7Z3+X9Z1+X). 10 Z2)+2(RB)(Y4+2Y7Z3+Y9Z1+Y 10 Z2). Multiply the above three partial derivatives by the preset adjustment step size and invert them to obtain the normalized slurry concentration correction, normalized grinding disc gap correction, and normalized drive power correction. Then, according to the same normalization conversion relationship in the modeling stage, restore the normalized correction to the actual correction corresponding to the slurry concentration, grinding disc gap, and drive power, and generate an adjustment command carrying the actual correction.

[0092] After the adjustment command is generated, the control module distributes the correction amounts for slurry concentration, grinding disc gap, and drive power to the corresponding actuators according to the instructions. Each correction amount is first limited to ensure it remains within the equipment's allowable adjustment range. For the slurry concentration correction, if the correction indicates a need to reduce the slurry concentration, an instruction to increase the dilution water flow or decrease the proportion of concentrated slurry is sent to the concentration adjustment actuator; if the correction indicates a need to increase the slurry concentration, an instruction to decrease the dilution water flow or increase the proportion of concentrated slurry is sent. This concentration adjustment changes the overall concentration of the slurry entering the grinding process, distinct from the micro-water replenishment under water shortage conditions. For the grinding disc gap correction, it is converted into a displacement setpoint for the grinding disc adjustment mechanism. An electric or hydraulic adjustment mechanism drives the grinding disc seat to feed or retract, causing the grinding disc gap to decrease or increase according to the correction amount. For the drive power correction, it is converted into a power setpoint or load setpoint for the main motor inverter. The inverter adjusts the main motor output, causing the drive power to change in the correction direction. After each actuator completes its action, the SCADA platform continues to collect data on slurry concentration, grinding disc clearance, and drive power. In the next time window, it recalculates the high-frequency energy ratio and predicts quality indicators, enabling the adjusted process state to enter the next round of closed-loop judgment and parameter correction.

[0093] This invention addresses the problem that insufficient free water between grinding discs can easily interfere with the adjustment of process parameters during pulping, leading to inaccurate fiber quality prediction and parameter optimization. It constructs a closed-loop control approach based on SCADA data, acoustic emission monitoring, and a quality correlation model. First, process parameters such as pulp concentration, grinding disc gap, and drive power are collected in real time through the SCADA platform. Sound signals output from acoustic emission sensors at the grinding disc base are also collected. The high-frequency energy ratio is calculated using Fast Fourier Transform (FFT). Then, using the normal and water-deficient centers established from historical water replenishment tests, the current high-frequency energy ratio is assessed to determine if there is insufficient free water between the grinding discs. If a water-deficient state is determined, a water replenishment command is generated based on the degree of deviation from the normal center to restore the state between the grinding discs to a suitable range for pulping control. If a water-deficient state is not determined, the current process parameters are input into the quality correlation model to predict fiber length and fibrillation rate. The correction amount for each process parameter is calculated based on the deviation between the predicted and target values, and an adjustment command is generated. In this way, the present invention first eliminates the interference of abnormal friction caused by insufficient free water on the judgment of pulping quality, and then optimizes the process parameters that truly affect fiber length and fibrillation rate, so that water replenishment control and pulping parameter adjustment form a continuous closed loop, thereby solving the problems in traditional pulping control that it is difficult to distinguish between water shortage abnormalities and process parameter deviations in a timely manner, and that parameter adjustment is lagging and prone to misadjustment.

[0094] The SCADA-based system for optimizing control parameters in the pulping process includes:

[0095] Data Acquisition Module: Real-time acquisition of process parameters during the pulping process via the SCADA platform. These parameters include pulp concentration, grinding disc clearance, and drive power.

[0096] Judgment module: Collects the sound signal output by the acoustic emission sensor installed on the grinding disc in real time within a preset time window, and calculates the high-frequency energy ratio based on the sound signal;

[0097] Based on the high-frequency energy ratio and the pre-calibrated normal center and water-deficient center, determine whether the grinding wheel is in a water-deficient state;

[0098] Branch module: When in a water shortage state, it sends a water replenishment command to the actuator and returns to the execution judgment module; when not in a water shortage state, it executes the control module.

[0099] Control module: Acquires process parameters collected by the SCADA platform at the end of the time window, and outputs predicted fiber length and predicted brooming rate by combining the process parameter-quality correlation model;

[0100] The process parameter correction amount is determined based on the predicted fiber length and predicted buffing rate, and adjustment instructions are generated.

[0101] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the present invention should still fall within the scope of the present invention.

Claims

1. A method for optimizing control parameters in a pulping process based on SCADA, characterized in that, Includes the following steps: S1. The process parameters of the pulping process are collected in real time through the SCADA platform. The process parameters include pulp concentration, grinding disc gap and drive power. S2. Collect the sound signal output by the acoustic emission sensor installed on the grinding disc base in real time within the preset time window, and calculate the high-frequency energy ratio based on the sound signal. S3. Based on the high-frequency energy ratio and the pre-calibrated normal center and water-deficient center, determine whether the grinding disc is in a water-deficient state. S4. When in a water shortage state, send a water replenishment command to the actuator and execute S2-S3 again; when not in a water shortage state, execute S5. S5. Obtain the process parameters collected by the SCADA platform at the end of the time window, and output the predicted fiber length and predicted brooming rate by combining the process parameter-quality correlation model. S6. Determine the process parameter correction amount based on the predicted fiber length and predicted buffing rate, and generate adjustment instructions.

2. The method for optimizing control parameters of the pulping process based on SCADA according to claim 1, characterized in that, The specific process for calculating the proportion of high-frequency energy is as follows: The sound signal collected within the time window is subjected to a fast Fourier transform to obtain the sound signal spectrum; Calculate the sum of the energy of all frequency components in the spectrum whose frequency is higher than a preset frequency threshold, and use this as the high-frequency energy; calculate the sum of the energy of all frequency components in the entire frequency band of the spectrum, and use this as the full-frequency energy. Divide the high-frequency energy by the total frequency energy to obtain the high-frequency energy percentage.

3. The method for optimizing control parameters of the pulping process based on SCADA according to claim 1, characterized in that, The process of determining whether there is a water shortage is as follows: Multiple sets of pulping test data for wood chips with different initial moisture contents were obtained. The pulping test data included the proportions of the first and second high-frequency energy before and after the water replenishment command was executed. The absolute difference between the proportions of the first and second high-frequency energy was calculated as the screening value. The pulping test data with a screening value < the first change threshold are marked as normal data, the pulping test data with a screening value > the second change threshold are marked as abnormal data, and the pulping test data with a screening value ≥ the first change threshold and ≤ the second change threshold are removed. The first change threshold is < the second change threshold and both the first change threshold and the second change threshold are preset. Calculate the mean of the first high-frequency energy proportion in the normal and abnormal data respectively as the normal center and the water-deficient center, and calculate the Euclidean distance A1 and A2 between the current high-frequency energy proportion and the normal center and the water-deficient center respectively; If A1 ≥ A2, then the water is in a state of shortage; otherwise, it is not in a state of shortage.

4. The method for optimizing control parameters of the pulping process based on SCADA according to claim 1, characterized in that, The process of replenishing moisture is as follows: Using the Euclidean distance A1 as the process variable and zero as the setpoint of the process variable, calculate the positive deviation obtained by subtracting the setpoint from the process variable; The control output value is obtained by performing proportional-integral calculation on the positive deviation; The control output value is limited to obtain the valve opening setpoint, and a water replenishment command containing the valve opening setpoint is generated and sent to the actuator.

5. The method for optimizing control parameters of the pulping process based on SCADA according to claim 1, characterized in that, The process of obtaining the predicted fiber length and predicted broom rate is as follows: The slurry concentration, grinding disc gap, and drive power collected by the SCADA platform at the end of multiple historical time windows are obtained. The slurry concentration, grinding disc gap, and drive power are normalized respectively to obtain the normalized slurry concentration, grinding disc gap, and drive power. The normalized slurry concentration, grinding disc gap, and drive power are used as historical process parameter samples. Fiber length and brooming rate detected after a preset lag time following the end of the historical time window are used as historical quality samples. Historical process parameter samples and historical quality samples are paired according to time windows; The regression model is structured as a quadratic polynomial regression function of fiber length and buffing rate with respect to normalized slurry concentration, grinding disc gap, and driving power. The quadratic polynomial regression function includes a first-order term, a quadratic term, and an interaction term. Using paired historical process parameter samples as input variables and historical quality samples as output variables, the coefficients of the quadratic polynomial regression function are fitted using the least squares method with regularization terms to obtain the process parameter-quality correlation model. Input the current process parameters into the process parameter-quality correlation model, and output the predicted fiber length and predicted brooming rate.

6. The method for optimizing control parameters of the pulping process based on SCADA according to claim 1, characterized in that, The process of generating adjustment instructions is as follows: Calculate the length deviation between the predicted fiber length and the preset fiber length target value, and calculate the firification rate deviation between the predicted firification rate and the preset firification rate target value; Using the sum of the squares of length deviation and brooming rate deviation as the loss function, the partial derivatives of the loss function with respect to the normalized slurry concentration, grinding disc gap and driving power are obtained respectively. Each partial derivative is multiplied by a preset adjustment step size and inverted to obtain the normalized correction values ​​for slurry concentration, grinding disc gap, and drive power. The normalized correction values ​​are then converted back to the correction values ​​for slurry concentration, grinding disc gap, and drive power, and an adjustment command carrying the correction values ​​is generated.

7. A SCADA-based system for optimizing control parameters in a pulping process, characterized in that, include: Data Acquisition Module: Real-time acquisition of process parameters during the pulping process via the SCADA platform. These parameters include pulp concentration, grinding disc clearance, and drive power. Judgment module: Collects the sound signal output by the acoustic emission sensor installed on the grinding disc in real time within a preset time window, and calculates the high-frequency energy ratio based on the sound signal; Based on the high-frequency energy ratio and the pre-calibrated normal center and water-deficient center, determine whether the grinding wheel is in a water-deficient state; Branch module: When in a water shortage state, it sends a water replenishment command to the actuator and returns to the execution judgment module; when not in a water shortage state, it executes the control module. Control module: Acquires process parameters collected by the SCADA platform at the end of the time window, and outputs predicted fiber length and predicted brooming rate by combining the process parameter-quality correlation model; The process parameter correction amount is determined based on the predicted fiber length and predicted buffing rate, and adjustment instructions are generated.

Images