An automated control system and method for spice compounding production line

Abstract

The present invention discloses an automated control system and method for the batching of spice production lines, belonging to the technical field of production line control. The present invention acquires sensor data, defines a state vector, and fuses the sensor data; establishes an anti-caking control model, performs adaptive model correction, and outputs a caking prevention control strategy; identifies agglomerated particles, outputs a particle size distribution, and classifies them into particle size grades; implements a particle size grade-based disassembly strategy, verifies the disassembly effect, and obtains an agglomeration verification result; when it is confirmed that there are no agglomerates in the feeder, updates the lag time caused by the viscous ingredients; adds fuzzy adaptive compensation to obtain a compensation time; establishes a lag compensation effect evaluation index, performs no-load and full-load lag tests on each spice based on the lag time, and verifies the compensated coupling error; establishes a cross-compensation matrix for the strongly coupled pairs through relative gain matrix analysis, and generates a compensated feeder speed command.

Description

Technical Field

[0001] The present invention relates to the technical field of production line control, and in particular to an automated control system and method for controlling ingredients in a production line for compounding spices. Background Art

[0002] Spice blending is a key factor in determining product flavor consistency, and automated control of spice blending directly impacts production efficiency and product quality. Spices are mostly powdered or granular materials with strong hygroscopicity and poor fluidity. During storage, transportation, and blending, they are prone to forming lumps due to factors such as electrostatic adsorption and deliquescence. This can lead to uneven feeder delivery and large deviations in blending accuracy. Furthermore, the viscosity of different spices varies significantly, and the hysteresis effect of the blending system can further exacerbate coupling errors in multi-component blending, causing blending ratios to deviate from the preset formula and impacting product flavor stability.

[0003] Traditional anti-caking methods rely on mechanical stirring or fixed-frequency vibration for loosening, without incorporating dynamic adjustments to real-time material conditions. For viscous spices like chili powder and ginger powder, existing hysteresis compensation models are often based on fixed parameters and fail to track changes in material properties in real time. Traditional control systems utilize a single-loop, independent control mode, failing to account for the dynamic coupling between multiple feeders. Summary of the Invention

[0004] The object of the present invention is to provide an automated control system and method for the production line of spice compounding to solve the problems raised in the prior art.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] In a first aspect, the present application provides a method for automated control of ingredients in a production line for spice compounding, comprising the following steps:

[0007] Acquire sensor data, including spice status and wall adhesion layer thickness, define a state vector, and fuse the sensor data using an extended Kalman filter. Establish an anti-caking control model based on the fused sensor data and perform adaptive model correction. The anti-caking control model outputs a caking prevention control strategy for graded anti-caking control.

[0008] When there are agglomerates in the feeder, the YOLOv8 model is used to identify agglomerated particles, output the particle size distribution, and classify the particle size grades. A particle size grade-based breakup strategy is implemented to verify the breakup effect and obtain the agglomeration verification results.

[0009] Based on the agglomeration verification results, when it is confirmed that there is no agglomeration in the feeder, the double exponential smoothing method is used to dynamically update the lag time caused by viscous ingredients; based on the lag time, fuzzy adaptive compensation is added to obtain the compensation time;

[0010] An evaluation index for the lag compensation effect was established. Based on the lag time, lag tests were carried out on each spice at both empty and full loads to verify the coupling error after compensation. Based on the coupling error and compensation time, a cross-compensation matrix was established for the strong coupling pair through relative gain matrix analysis to generate the compensated feeder speed command.

[0011] In combination with the first aspect, in a first implementation of the first aspect of the present application, acquiring sensor data, including the spice state and the wall adhesion layer thickness, defining a state vector, and using an extended Kalman filter to fuse the sensor data, includes:

[0012] The spice state includes moisture, particle size, viscosity, and density. Sensor data is defined as a state vector. A random noise term is set for each state vector to reflect unmodeled dynamics. A state equation is established based on the physical laws of state vector changes. A mapping relationship between sensor data and the actual state is established, and a covariance matrix is ​​set for each sensor data to reflect measurement uncertainty.

[0013] Based on the state estimate at the previous moment, the state at the current moment is predicted based on the state equation, and the covariance matrix of the predicted state at the current moment is calculated; the difference between the actual sensor value and the predicted value is compared, the Kalman gain is calculated, and the predicted state is corrected based on the Kalman gain to obtain the optimal estimate; the covariance matrix is ​​updated to complete a filtering cycle; the iterative extended Kalman filter is used to fuse the sensor data.

[0014] In combination with the first aspect, in a second implementation of the first aspect of the present application, establishing an anti-caking control model based on the fused sensor data and performing adaptive correction of the model include:

[0015] The anti-caking control model consists of a data layer, a model layer, and a strategy layer. The data layer receives fused sensor data and performs normalization processing. In the model layer, a fuzzy logic controller is used to construct nonlinear mapping relationships, define language variables, and establish an IF-THEN rule base (a production rule base). In the strategy layer, the model output is converted into specific equipment control instructions. The data source for training the model is historical production data, specifically sensor data and manual processing records corresponding to caking events of different spices. The training goal is to ensure that the strategy level output by the model is negatively correlated with the actual caking rate.

[0016] Define the correction trigger conditions, increase the rule priority of high-frequency scenarios based on real-time agglomeration data, and dynamically adjust the membership function.

[0017] In combination with the first aspect, in a third embodiment of the first aspect of the present application, when there is agglomeration in the feeder, identifying agglomerated particles by using a YOLOv8 model, outputting a particle size distribution, and dividing the particle size grades includes:

[0018] Typical agglomerate samples were collected and images were taken in the laboratory under simulated lighting conditions, spice states, and background interference. The agglomerated particles were labeled using the LabelImg tool as a training set. The training set was rotated, scaled, and brightness adjusted to expand the training set. For the YOLOv8 model, the COCO pre-trained weights were loaded, the training parameters were set, and the prior box size was recalculated using K-means clustering to train the YOLOv8 model. The pre-processed images were input into the trained YOLOv8 model to output the bounding box coordinates and category confidence of the agglomerated particles. A mapping relationship between pixels and actual size was established using a calibration plate. For each detection box, the geometric mean of its width and height was taken as the equivalent diameter. The detection box in the edge area was perspective corrected based on the camera installation angle. For the adhered particles within the detection box, the watershed algorithm was used for secondary segmentation to improve the accuracy of particle size calculation.

[0019] Count the particle sizes of all agglomerated particles in the current frame, generate a particle size histogram at intervals, and output the particle size distribution, including the average particle size, D10, D50, and D90. According to the predefined particle size grades, count the number and mass proportions of particles in each grade and divide the particle size grades.

[0020] In combination with the first aspect, in a fourth implementation of the first aspect of the present application, executing a particle size-based disintegration strategy, performing disintegration effect verification, and obtaining an agglomeration verification result include:

[0021] The particle size grades include primary agglomeration, secondary agglomeration, tertiary agglomeration, quaternary agglomeration and quintuple agglomeration. For primary agglomeration, the low-frequency vibrating screen at the outlet of the feeder is started to loosen and separate the particle clusters through vibration, and the ultrasonic vibration device on the inner wall of the pipe is turned on to emit pulses to destroy the weak adhesion between the particles. For secondary agglomeration, the double-roll crusher at the inlet of the feeder is started to shear the agglomerated particles through differential rotation, and the pneumatic vibrator at the bottom of the silo is turned on simultaneously to prevent the accumulation of subsequent ingredients that affect the crushing effect. For tertiary agglomeration, the treatment measures based on the secondary agglomeration are used. For the fourth-level agglomeration, the high-pressure air knife is turned on to blow the crushed particles away from the blades to prevent secondary adhesion. For the fourth-level agglomeration, the double-roll crusher is switched to reverse rotation mode to crush the hard agglomerates through rubbing and grinding. The centrifugal breakers at the end of the pipeline are started to use centrifugal force to throw large particles toward the wear-resistant liner and crush them through impact. For the fifth-level agglomeration, the feeder is stopped, the silo outlet valve is closed, the sound and light alarm is triggered, and the 360-degree rotating cleaning arm inside the silo is activated to try to physically remove the oversized agglomerates. If they are not removed within the set time, the manual maintenance process is triggered.

[0022] Get the current particle size distribution. When it meets the qualification criteria, determine that there is no agglomeration in the feeder and output the agglomeration verification result.

[0023] In combination with the first aspect, in a fifth embodiment of the first aspect of the present application, based on the agglomeration verification result, when it is confirmed that there is no agglomeration in the feeder, the double exponential smoothing method is used to dynamically update the lag time caused by the viscous ingredients, including:

[0024] The lag time for viscous ingredients is the time it takes for the actual flow rate of the ingredient to reach the target value after the feeder speed adjustment command is issued, reflecting the transmission delay of the viscous ingredient in the pipeline. With the feeder unloaded, the speed is increased from 0 to A% of the rated value. The pressure rise curve is recorded using the pressure sensor at the end of the pipeline. The time it takes for the pressure to first reach B% of the steady-state value is calculated as the initial lag time, where A and B are positive integers set by the user and A>B. Repeat the test for different spices, and take the average value as the baseline lag time for the ingredient, which is stored in the ingredient characteristics database.

[0025] The lag time is considered as a dynamic sequence including the current lag value and the lag time change rate; the state is defined, including the horizontal component and the trend component, the horizontal component is the smoothed value of the current lag time, reflecting the immediate lag level, and the trend component is the changing trend of the lag time, reflecting the lag time increase or decrease rate caused by the viscosity change; the horizontal component weight and the trend component weight are set according to the viscosity coefficient of the ingredient. Dynamic adjustment, where , T represents the thickness of the wall adhesion layer, H represents the humidity of the spices, and D represents the D50 in the particle size distribution; the current lag time measured value, viscosity coefficient, horizontal component of the previous moment, and trend component of the previous moment are collected; when the trend remains unchanged, the horizontal prediction value is the previous moment horizontal component plus the previous moment trend component, and the current trend component is set to the previous moment trend component; the current horizontal component is obtained according to the difference between the current lag time measured value and the horizontal prediction value; the current trend component is adjusted based on the difference between the current horizontal component and the previous moment horizontal value; the lag time is the sum of the current horizontal component and the current trend component.

[0026] In combination with the first aspect, in a sixth implementation of the first aspect of the present application, the adding of fuzzy adaptive compensation based on the lag time to obtain the compensation time includes:

[0027] Fuzzy linguistic values ​​are defined for input variables and output variables, including negative large, negative small, zero, positive small, and positive large. The input variables use a trapezoidal membership function, and the output variables use a single-valued fuzzy set. Based on the experience of viscous batching control and the fuzzy linguistic values, fuzzy rules are established to adjust the compensation gain and compensation time. The input variables are converted into the membership of each fuzzy set, and the corresponding fuzzy rules are activated according to the membership. The outputs of the activated rules are weighted averaged to obtain the fuzzy output. The compensation gain is calculated based on the fuzzy output using the center of gravity method.

[0028] The basic compensation time is calculated based on the lag time and the compensation gain, and the basic compensation time is corrected by adding the deviation change rate correction amount. The boundary constraint is performed to obtain the compensation time.

[0029] In combination with the first aspect, in a seventh implementation of the first aspect of the present application, establishing a hysteresis compensation effect evaluation index, performing no-load and full-load hysteresis tests on each spice based on the hysteresis time, and verifying the compensated coupling error, includes:

[0030] The evaluation indicators for hysteresis compensation effectiveness include hysteresis time estimation error, batching accuracy error, coupling error, and system response time. The hysteresis time estimation error is the percentage deviation between the hysteresis time obtained by the double exponential smoothing method and the actual hysteresis time. The batching accuracy error is the percentage deviation between the actual discharge amount and the target discharge amount. The coupling error is the percentage deviation of the flow rate of another ingredient caused by the adjustment of one ingredient in a strongly coupled batching alignment. The system response time is the time from the viscosity mutation to the effectiveness of the compensation strategy.

[0031] Empty the silo and close all feed inlets to ensure that there is no residual feed in the pipeline. Step the feeder speed from 0 to A% of the rated speed, synchronously record the timestamp of the feeder control signal and the response timestamp of the pressure sensor at the end of the pipeline, and calculate the inherent lag time. Record the inherent lag time at different speeds to form a comparison table of speed and inherent lag time. Group by viscosity coefficient, perform benchmark tests and compensation tests, compare data, and verify the coupling error after compensation.

[0032] In combination with the first aspect, in an eighth implementation of the first aspect of the present application, based on the coupling error and the compensation time, a cross-compensation matrix is ​​established for the strong coupling pair through relative gain matrix analysis to generate a compensated feeder speed instruction, including:

[0033] The feeder system is abstracted as a MIMO system, a unit step input is applied to the MIMO system, the steady-state response data is recorded, and the strong coupling pairs are determined; an n×n cross-compensation matrix C is constructed, and the elements in the cross-compensation matrix C are represents the compensation coefficient of the j-th feeder to the i-th feeder, and n represents the number of mutually coupled feeders. The compensation intensity factors of the non-diagonal elements in the cross-compensation matrix C are adjusted in real time according to the coupling error. For the strongly coupled pair (i, j), the compensation time ratio is introduced as a weight to dynamically update the compensation coefficient.

[0034] Based on the original speed command vector output by the PID controller, cross compensation is calculated and physical constraints are imposed; the compensation effect is verified in real time and matrix optimization is performed to generate the compensated feeder speed command.

[0035] In a second aspect, the present application provides an automated control system for ingredients in a production line for spice compounding, comprising:

[0036] The hierarchical anti-caking control module includes a data fusion unit, an anti-caking control model construction unit, and a hierarchical anti-caking control unit. The data fusion unit acquires sensor data, including the state of spices and the thickness of the wall adhesion layer, defines a state vector, and fuses the sensor data using an extended Kalman filter. The anti-caking control model construction unit establishes an anti-caking control model based on the fused sensor data and performs adaptive model correction. The anti-caking control model in the hierarchical anti-caking control unit outputs a caking prevention control strategy to perform hierarchical anti-caking control.

[0037] Agglomeration Verification Module: This module includes a particle size classification unit and an agglomeration verification unit. When agglomerates are found in the feeder, the particle size classification unit uses the YOLOv8 model to identify agglomerated particles, outputs the particle size distribution, and classifies the particles into size classes. The agglomeration verification unit implements a particle size class-based decomposition strategy, verifies the decomposition effect, and obtains an agglomeration verification result.

[0038] Lag and compensation calculation module: includes: lag time calculation unit and compensation time calculation unit; among them, the lag time calculation unit is based on the agglomeration verification result. When it is confirmed that there is no agglomeration in the feeder, the double exponential smoothing method is used to dynamically update the lag time caused by viscous ingredients; the compensation time calculation unit adds fuzzy adaptive compensation based on the lag time to obtain the compensation time;

[0039] Feeder speed instruction generation module: includes: a coupling error verification unit and an instruction generation unit; wherein, the coupling error verification unit establishes a hysteresis compensation effect evaluation index, performs no-load and full-load hysteresis tests on each spice based on the hysteresis time, and verifies the compensated coupling error; the instruction generation unit establishes a cross-compensation matrix for the strong coupling pair based on the coupling error and compensation time through relative gain matrix analysis, and generates the compensated feeder speed instruction.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] 1. This application establishes an adaptive anti-caking control model by fusing multi-source sensor data through extended Kalman filtering, which can dynamically adjust parameters such as stirring frequency and vibration amplitude in real time to reduce the incidence of agglomeration; agglomerated particle identification and particle size classification based on the YOLOv8 model can quickly complete agglomeration detection and output particle size distribution, and implement differentiated breakup strategies for different levels of agglomerates.

[0042] 2. This application adopts the double exponential smoothing method to update the viscous ingredient lag time in real time, combined with the fuzzy adaptive compensation algorithm to control the lag time prediction error; through the relative gain matrix analysis of the multivariable coupling strength, a cross-compensation matrix is ​​established for the strongly coupled feeder pair, so that the coupling error of multiple groups of ingredients is reduced and the consistency of the compounding ratio is improved.

[0043] 3. This application integrates sensor fusion, visual recognition, adaptive control and multivariable decoupling technology to achieve full process automation from agglomeration prevention, real-time detection, dynamic compensation to multi-machine collaborative control, reducing manual intervention and improving batching efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a schematic diagram of the steps of an automated control method for ingredients in a production line for spice compounding according to the present invention;

[0045] Figure 2 The present invention is a system structure diagram of an automatic control system for controlling ingredients in a production line for compounding spices. DETAILED DESCRIPTION

[0046] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0047] Example 1: Figure 1 As shown in the schematic diagram of the steps of a method for automatically controlling the ingredients of a production line for compounding spices, the present application provides a method for automatically controlling the ingredients of a production line for compounding spices, comprising the following steps:

[0048] Step S100: Acquire sensor data, including the state of spices and the thickness of the wall adhesion layer, define a state vector, and fuse the sensor data using an extended Kalman filter; establish an anti-caking control model based on the fused sensor data and perform adaptive correction on the model; the anti-caking control model outputs a caking prevention control strategy to perform graded anti-caking control;

[0049] Specifically, the spice state includes the moisture, particle size, viscosity, and density of the spice. The sensor data is defined as a state vector. A random noise term is set for each state vector to reflect the unmodeled dynamics. A state equation is established based on the physical law of the state vector change. A mapping relationship between sensor data and the actual state is established. A covariance matrix is ​​set for each sensor data to reflect the measurement uncertainty.

[0050] Based on the state estimate at the previous moment, the state at the current moment is predicted based on the state equation, and the covariance matrix of the predicted state at the current moment is calculated; the difference between the actual sensor value and the predicted value is compared, the Kalman gain is calculated, and the predicted state is corrected based on the Kalman gain to obtain the optimal estimate; the covariance matrix is ​​updated to complete a filtering cycle; the iterative extended Kalman filter is used to fuse the sensor data.

[0051] Furthermore, the anti-caking control model includes a data layer, a model layer, and a strategy layer. In the data layer, fused sensor data is received and normalized. In the model layer, a fuzzy logic controller is used to construct a nonlinear mapping relationship, define language variables, and establish an IF-THEN rule base. In the strategy layer, the model output is converted into specific equipment control instructions. The data source for training the model is historical production data, specifically sensor data and manual processing records corresponding to caking events of different spices. The training goal is to make the strategy level output by the model negatively correlated with the actual caking rate.

[0052] Define the correction trigger conditions, increase the rule priority of high-frequency scenarios based on real-time agglomeration data, and dynamically adjust the membership function.

[0053] Example 2: Historical data covers five spices (chili powder, ginger powder, black pepper powder, Sichuan pepper powder, and cumin powder), including more than 2,000 agglomeration event samples. Among them, mild agglomeration (particle size ≤ 2 mm) accounts for 45%, corresponding to sensor characteristics of humidity ≤ 15% RH and adhesion layer thickness ≤ 0.8 mm; moderate agglomeration (2-5 mm) accounts for 35%, corresponding to humidity 15%-20% RH and adhesion layer thickness 0.8-1.5 mm; severe agglomeration (> 5 mm) accounts for 20%, corresponding to humidity > 20% RH and adhesion layer thickness > 1.5 mm.

[0054] In the fuzzy logic controller, the input variables include humidity (linguistic values: low, medium, high) with a domain of [0, 30% RH], wall adhesion layer thickness (linguistic values: thin, medium, thick) with a domain of [0, 3 mm], and viscosity (linguistic values: low, medium, high) with a domain of [0, 800 mPa·s]. The output variable is the control strategy level (linguistic values: Level 1 - mild agitation, Level 2 - moderate vibration, Level 3 - strong vibration and agitation) with a domain of [1, 3].

[0055] The following is an example of a rule base:

[0056] IF humidity = HIGH AND adhesion layer thickness = THICK THEN strategy level = 3;

[0057] IF Humidity = Medium AND Viscosity = High THEN Strategy Level = 2;

[0058] IF Adhesion Layer Thickness = Thin AND Viscosity = Low THEN Strategy Level = 1.

[0059] After training, the model achieved an 88% accuracy rate in predicting the caking rate, and the negative correlation coefficient between the policy level and the actual caking rate increased from -0.52 to -0.79. When caking still occurred despite humidity levels exceeding 22%RH for three consecutive times and a policy level of 3, the system automatically shifted the "Humidity = High" membership function left by 2%RH (initiating the higher-level policy earlier). This correction reduced the caking rate in this scenario from 15% to 4%.

[0060] Step S200: When there are agglomerates in the feeder, the agglomerated particles are identified using the YOLOv8 model, the particle size distribution is output, and the particle size grades are divided; a breakup strategy based on the particle size grades is executed to verify the breakup effect and obtain an agglomeration verification result;

[0061] Specifically, typical agglomerate samples were collected, and images were taken in the laboratory under simulated lighting conditions, spice states, and background interference. The agglomerated particles were labeled using the LabelImg tool as a training set. The training set was rotated, scaled, and brightness adjusted to expand the training set. For the YOLOv8 model, the COCO pre-trained weights were loaded, the training parameters were set, and the prior box size was recalculated using K-means clustering to train the YOLOv8 model. The pre-processed images were input into the trained YOLOv8 model to output the bounding box coordinates and category confidence of the agglomerated particles. A mapping relationship between pixels and actual sizes was established using a calibration plate, and for each detection frame, the geometric mean of its width and height was taken as the equivalent diameter. Based on the camera installation angle, the detection frame in the edge area was perspective corrected. For the adhered particles within the detection frame, the watershed algorithm was used for secondary segmentation to improve the accuracy of particle size calculation.

[0062] Count the particle sizes of all agglomerated particles in the current frame, generate a particle size histogram at intervals, and output the particle size distribution, including the average particle size, D10, D50, and D90. According to the predefined particle size grades, count the number and mass proportions of particles in each grade and divide the particle size grades.

[0063] Furthermore, the particle size grades include primary agglomeration, secondary agglomeration, tertiary agglomeration, quaternary agglomeration and quintuple agglomeration. For primary agglomeration, the low-frequency vibrating screen at the outlet of the feeder is started to loosen and separate the particle clusters through vibration, and the ultrasonic vibration device on the inner wall of the pipe is turned on to emit pulses to destroy the weak adhesion between the particles. For secondary agglomeration, the double-roll crusher at the inlet of the feeder is started to shear the agglomerated particles through differential rotation, and the pneumatic vibrator at the bottom of the silo is turned on simultaneously to prevent the subsequent accumulation of ingredients from affecting the crushing effect. For tertiary agglomeration, based on the treatment of secondary agglomeration, The solution is to start the high-pressure air knife to blow the crushed particles away from the blades to prevent secondary adhesion. For the fourth-level agglomeration, the double-roll crusher is switched to reverse rotation mode to crush the hard agglomerates through rubbing and grinding. The centrifugal breakers at the end of the pipeline are started to use centrifugal force to throw large particles toward the wear-resistant liner and crush them through impact. For the fifth-level agglomeration, the feeder is stopped, the silo outlet valve is closed, the sound and light alarm is triggered, and the 360-degree rotating cleaning arm inside the silo is activated to try to physically remove the oversized agglomerates. If they are not removed within the set time, the manual maintenance process is triggered.

[0064] Get the current particle size distribution. When it meets the qualification criteria, determine that there is no agglomeration in the feeder and output the agglomeration verification result.

[0065] Example 3: The agglomeration levels are divided as follows: first-level agglomeration: equivalent diameter ≤ 1 mm; second-level agglomeration: 1 mm < equivalent diameter ≤ 3 mm; third-level agglomeration: 3 mm < equivalent diameter ≤ 5 mm; fourth-level agglomeration: 5 mm < equivalent diameter ≤ 10 mm; fifth-level agglomeration: equivalent diameter > 10 mm.

[0066] 3,000 images of black pepper clumps were collected, covering humidity levels of 8%-22% RH and particle sizes of 50-150 μm. Over 12,000 agglomerated particles were annotated. Data augmentation expanded the training set to 12,000 images, including samples with variations of ±15° rotation, 0.8-1.2x scaling, and ±20% brightness. The input image size was 640×640, the batch size was 16, the training epochs were 100, the initial learning rate was 0.001, and cosine annealing was used for decay. K-means clustering yielded five prior box sizes: (12, 16), (24, 32), (40, 56), (72, 96), and (128, 160). The test set achieved a mAP@0.5 of 95.2%, with an average detection time of 28 ms per frame.

[0067] There is a cluster of adhesions in the input image. Before perspective correction, it is detected as a fourth-level cluster with an equivalent diameter of 8 mm. After segmentation using the watershed algorithm, it is identified as three particles (with diameters of 3 mm, 4 mm, and 5 mm, respectively). It is then corrected to two third-level clusters and one fourth-level cluster.

[0068] The initial particle size distribution is as follows: average particle size 6.8mm, D10=2.1mm, D50=5.5mm, D90=11.2mm; grade proportion: third-level agglomerates (3-5mm) account for 35%, fourth-level agglomerates (5-10mm) account for 50%, and fifth-level agglomerates (>10mm) account for 15%.

[0069] When a fifth-grade agglomerate with an equivalent diameter of 12 mm is detected, the cleaning arm is triggered to rotate (rotation speed 10 rpm), with a removal rate of 90% within 30 seconds. The remaining agglomerate with a diameter of 8 mm is downgraded to the fourth grade. The double-roll crusher is switched to reverse rotation (rotation speed difference 300 rpm), and the centrifugal disperser rotates at 2000 rpm. After treatment, 90% of the particles are less than 5 mm, and the remaining 10% are third-grade agglomerates of 3-5 mm. The double-roll crusher is started in forward shear (rotation speed 1500 rpm) + high-pressure air knife (air pressure 0.6 MPa). After treatment, 85% of the particles are less than 3 mm, and 15% are second-grade agglomerates of 1-3 mm. The double-roll crusher maintains forward rotation, and the pneumatic vibrator is started (frequency 50 Hz). After treatment, 98% of the particles are less than 1 mm, meeting the standard of the first-grade agglomeration range.

[0070] Particle size distribution after treatment: average particle size 0.8mm, D10=0.3mm, D50=0.7mm, D90=1.1mm; grade ratio: first-level agglomerates (≤1mm) account for 99%, second-level agglomerates account for 1%, and there are no third to fifth-level agglomerates; dismantling efficiency: it takes 120 seconds from detection to complete processing, which is 50% more efficient than the traditional single crushing method.

[0071] Step S300: Based on the agglomeration verification result, when it is confirmed that there is no agglomeration in the feeder, a double exponential smoothing method is used to dynamically update the lag time caused by the viscous ingredients; fuzzy adaptive compensation is added based on the lag time to obtain the compensation time;

[0072] Specifically, the lag time for viscous ingredients is the time it takes for the actual flow rate of the ingredient to reach the target value after the feeder speed adjustment command is issued, reflecting the transmission delay of the viscous ingredient in the pipeline. With the feeder unloaded, the speed is increased from 0 to A% of the rated value. The pressure rise curve is recorded by the pressure sensor at the end of the pipeline. The time it takes for the pressure to first reach B% of the steady-state value is calculated as the initial lag time, where A and B are positive integers set by the user and A>B. Repeat the test for different spices, and take the average value as the benchmark lag time for the ingredient, which is stored in the ingredient characteristics database.

[0073] The lag time is considered as a dynamic sequence including the current lag value and the lag time change rate; the state is defined, including the horizontal component and the trend component, the horizontal component is the smoothed value of the current lag time, reflecting the immediate lag level, and the trend component is the changing trend of the lag time, reflecting the lag time increase or decrease rate caused by the viscosity change; the horizontal component weight and the trend component weight are set according to the viscosity coefficient of the ingredient. Dynamic adjustment, where , T represents the thickness of the wall adhesion layer, H represents the humidity of the spices, and D represents the D50 in the particle size distribution; the current lag time measured value, viscosity coefficient, horizontal component of the previous moment, and trend component of the previous moment are collected; when the trend remains unchanged, the horizontal prediction value is the previous moment horizontal component plus the previous moment trend component, and the current trend component is set to the previous moment trend component; the current horizontal component is obtained according to the difference between the current lag time measured value and the horizontal prediction value; the current trend component is adjusted based on the difference between the current horizontal component and the previous moment horizontal value; the lag time is the sum of the current horizontal component and the current trend component.

[0074] Furthermore, fuzzy linguistic values ​​are defined for input variables and output variables, including: negative large, negative small, zero, positive small, and positive large. The input variables use a trapezoidal membership function, and the output variables use a single-valued fuzzy set. Based on the experience of viscous batching control and the fuzzy linguistic values, fuzzy rules are established to adjust the compensation gain and compensation time. The input variables are converted into the membership of each fuzzy set, and the corresponding fuzzy rules are activated according to the membership. The outputs of the activated rules are weighted averaged to obtain the fuzzy output. The compensation gain is calculated based on the fuzzy output using the center of gravity method.

[0075] The basic compensation time is calculated based on the lag time and the compensation gain, and the basic compensation time is corrected by adding the deviation change rate correction amount. The boundary constraint is performed to obtain the compensation time.

[0076] Example 4: Feeder parameters: rated speed 200 rpm, pipe length 5 m, inner diameter 50 mm. Initial test parameters: A = 80 (speed increased from 0 to 80% of rated value), B = 90 (pressure reached 90% of steady-state value). Viscosity coefficient calculation formula: viscosity × humidity / D50 (viscosity in mPa·s, humidity in %RH, D50 in μm).

[0077] The no-load test data is as follows (repeated 5 times): Test 1: lag time 4.2s; Test 2: 4.5s; Test 3: 4.0s; Test 4: 4.3s; Test 5: 4.4s; average value: 4.28s, which is stored in the database as the benchmark lag time.

[0078] The full-load test data is as follows (humidity 18% RH, D50 = 60 μm, viscosity 450 mPa s): Actual lag time: 6.8 s (because the viscosity of the material is significantly higher than that of the no-load test, the transmission delay increases).

[0079] Double exponential smoothing is used to dynamically update the lag time. For input 1: lag time deviation (measured value - baseline value), the current deviation is 7.2 - 4.28 = 2.92 seconds, with a language value of "positive large." For input 2: rate of change of deviation (current deviation - previous deviation), the rate of change is 2.92 - 2.72 (the previous deviation was 7.0 - 4.28 = 2.72) = 0.2, with a language value of "positive small." Output: compensation gain K, fuzzy rule: IF deviation = positive large AND rate of change = positive small THENK = 1.2 (enhanced compensation).

[0080] Basic compensation time: estimated lag time × compensation gain = 7.26 × 1.2 = 8.71 seconds; deviation change rate correction: +0.5 seconds (due to the positive change rate, the compensation is appropriately extended); boundary constraint (maximum value 10 seconds): final compensation time = min(8.71+0.5,10) = 9.21 seconds.

[0081] Step S400: Establish a hysteresis compensation effect evaluation index, perform no-load and full-load hysteresis tests on each spice based on the hysteresis time, and verify the compensated coupling error; based on the coupling error and compensation time, establish a cross-compensation matrix for the strong coupling pair through relative gain matrix analysis to generate a compensated feeder speed instruction.

[0082] Specifically, the evaluation indicators of hysteresis compensation effect include hysteresis time estimation error, batching accuracy error, coupling error and system response time. The hysteresis time estimation error is the percentage deviation between the hysteresis time obtained by double exponential smoothing method and the actual hysteresis time. The batching accuracy error is the percentage deviation between the actual discharge amount and the target discharge amount. The coupling error is the percentage deviation of the flow rate of another ingredient caused by the adjustment of a certain ingredient in the strong coupling batching. The system response time is the time from the viscosity mutation to the effectiveness of the compensation strategy.

[0083] Empty the silo and close all feed inlets to ensure that there is no residual feed in the pipeline. Step the feeder speed from 0 to A% of the rated speed, synchronously record the timestamp of the feeder control signal and the response timestamp of the pressure sensor at the end of the pipeline, and calculate the inherent lag time. Record the inherent lag time at different speeds to form a comparison table of speed and inherent lag time. Group by viscosity coefficient, perform benchmark tests and compensation tests, compare data, and verify the coupling error after compensation.

[0084] Furthermore, the feeder system is abstracted as a MIMO system, a unit step input is applied to the MIMO system, the steady-state response data is recorded, and the strong coupling pairs are determined; an n×n cross-compensation matrix C is constructed, and the elements in the cross-compensation matrix C are represents the compensation coefficient of the j-th feeder to the i-th feeder, and n represents the number of mutually coupled feeders. The compensation intensity factors of the non-diagonal elements in the cross-compensation matrix C are adjusted in real time according to the coupling error. For the strongly coupled pair (i, j), the compensation time ratio is introduced as a weight to dynamically update the compensation coefficient.

[0085] Based on the original speed command vector output by the PID controller, cross compensation is calculated and physical constraints are imposed; the compensation effect is verified in real time and matrix optimization is performed to generate the compensated feeder speed command.

[0086] Example 5: Using a two-component dispensing system for chili powder (low viscosity) and ginger powder (high viscosity) as an example, the following experimental conditions were established: Two feeders (numbered M1 and M2), each delivering chili powder and ginger powder, respectively, with a rated speed of 300 rpm. Viscosity coefficient groups were chili powder (viscosity coefficient <200) and ginger powder (viscosity coefficient ≥200). Strong coupling threshold: A coupling error >10% was considered a strong coupling pair. Hysteresis testing and compensation effectiveness evaluation were performed.

[0087] When a step signal is input to M1 (speed from 150 rpm to 200 rpm), in steady state: M1's actual flow rate changes by +20% (target value). M2's actual flow rate changes by -12% (coupling interference). When a step signal is input to M2 (speed from 150 rpm to 200 rpm), in steady state: M2's actual flow rate changes by +18% (target value). M1's actual flow rate changes by +8% (coupling interference).

[0088] Calculate the RGA matrix. If the absolute value of the off-diagonal elements is greater than 0.1, M1 and M2 are considered to be a strongly coupled pair. Initialize the cross-compensation matrix. The element C[1,2] represents the compensation coefficient of M2 for M1, and C[2,1] represents the compensation coefficient of M1 for M2.

[0089] When it is detected that the M1 adjustment causes the M2 flow deviation to be 4.2% (target ≤ 5%), the current compensation coefficient is maintained; if the deviation is greater than 5%, the compensation coefficient is updated.

[0090] Generate compensated speed command:

[0091] Original PID instructions: M1=200rpm, M2=200rpm.

[0092] Cross compensation calculation:

[0093] M1 compensation command: 200+(-0.3)×200=140rpm (to offset M2 coupling interference).

[0094] M2 compensation command: 200+(-0.5)×200=100rpm (to offset M1 coupling interference).

[0095] Physical constraint correction: the speed must be ≥50rpm, the final instruction is: M1=140rpm, M2=100rpm.

[0096] Example 6: Figure 2 As shown in the system structure diagram of an automatic control system for the batching of spice production lines, the present application provides an automatic control system for the batching of spice production lines, including:

[0097] The hierarchical anti-caking control module includes a data fusion unit, an anti-caking control model construction unit, and a hierarchical anti-caking control unit. The data fusion unit acquires sensor data, including the state of spices and the thickness of the wall adhesion layer, defines a state vector, and fuses the sensor data using an extended Kalman filter. The anti-caking control model construction unit establishes an anti-caking control model based on the fused sensor data and performs adaptive model correction. The anti-caking control model in the hierarchical anti-caking control unit outputs a caking prevention control strategy to perform hierarchical anti-caking control.

[0098] Agglomeration Verification Module: This module includes a particle size classification unit and an agglomeration verification unit. When agglomerates are found in the feeder, the particle size classification unit uses the YOLOv8 model to identify agglomerated particles, outputs the particle size distribution, and classifies the particles into size classes. The agglomeration verification unit implements a particle size class-based decomposition strategy, verifies the decomposition effect, and obtains an agglomeration verification result.

[0099] Lag and compensation calculation module: includes: lag time calculation unit and compensation time calculation unit; among them, the lag time calculation unit is based on the agglomeration verification result. When it is confirmed that there is no agglomeration in the feeder, the double exponential smoothing method is used to dynamically update the lag time caused by viscous ingredients; the compensation time calculation unit adds fuzzy adaptive compensation based on the lag time to obtain the compensation time;

[0100] Feeder speed instruction generation module: includes: a coupling error verification unit and an instruction generation unit; wherein, the coupling error verification unit establishes a hysteresis compensation effect evaluation index, performs no-load and full-load hysteresis tests on each spice based on the hysteresis time, and verifies the compensated coupling error; the instruction generation unit establishes a cross-compensation matrix for the strong coupling pair based on the coupling error and compensation time through relative gain matrix analysis, and generates the compensated feeder speed instruction.

[0101] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. A method for controlling the automatic batching of spice production lines, characterized in that: The following steps are involved: Acquire sensor data, including spice status and wall adhesion layer thickness, define a state vector, and fuse the sensor data using an extended Kalman filter. Establish an anti-caking control model based on the fused sensor data and perform adaptive model correction. The anti-caking control model outputs a caking prevention control strategy for graded anti-caking control. When there are agglomerates in the feeder, the YOLOv8 model is used to identify agglomerated particles, output the particle size distribution, and classify the particle size grades. A particle size grade-based breakup strategy is implemented to verify the breakup effect and obtain the agglomeration verification results. Based on the agglomeration verification results, when it is confirmed that there is no agglomeration in the feeder, the double exponential smoothing method is used to dynamically update the lag time caused by viscous ingredients; Add fuzzy adaptive compensation based on the lag time to obtain compensation time; An evaluation index for hysteresis compensation was established. Based on the hysteresis time, hysteresis tests were conducted on each spice with no load and full load to verify the coupling error after compensation. Based on the coupling error and compensation time, a cross compensation matrix is ​​established for the strong coupling pair through relative gain matrix analysis to generate the compensated feeder speed command.

2. The method for controlling the automatic batching of spice production lines according to claim 1, wherein: The sensor data is obtained, including the state of spices and the thickness of the wall adhesion layer, and a state vector is defined. The sensor data is fused using an extended Kalman filter, including: The spice state includes moisture, particle size, viscosity, and density. Sensor data is defined as a state vector. A random noise term is set for each state vector to reflect unmodeled dynamics. A state equation is established based on the physical laws of state vector changes. A mapping relationship between sensor data and the actual state is established, and a covariance matrix is ​​set for each sensor data to reflect measurement uncertainty. Based on the state estimate at the previous moment, the state at the current moment is predicted based on the state equation, and the covariance matrix of the predicted state at the current moment is calculated; the difference between the actual sensor value and the predicted value is compared, the Kalman gain is calculated, and the predicted state is corrected based on the Kalman gain to obtain the optimal estimate; the covariance matrix is ​​updated to complete a filtering cycle; the iterative extended Kalman filter is used to fuse the sensor data.

3. The method for controlling the automatic batching of spice production lines according to claim 1, wherein: The anti-caking control model is established based on the fused sensor data, and the model is adaptively corrected, including: The anti-caking control model consists of a data layer, a model layer, and a strategy layer. The data layer receives fused sensor data and performs normalization processing. In the model layer, a fuzzy logic controller is used to construct nonlinear mapping relationships, define language variables, and establish an if-then rule base. In the strategy layer, the model output is converted into specific equipment control instructions. The data source for training the model is historical production data, specifically sensor data and manual processing records corresponding to caking events of different spices. The training goal is to ensure that the strategy level output by the model is negatively correlated with the actual caking rate. Define the correction trigger conditions, increase the rule priority of high-frequency scenarios based on real-time agglomeration data, and dynamically adjust the membership function.

4. The method for controlling the automatic batching of spice production lines according to claim 1, wherein: When there is agglomeration in the feeder, the YOLOv8 model is used to identify the agglomerated particles, output the particle size distribution, and classify the particle size grades, including: Typical agglomerate samples were collected and images were taken in the laboratory under simulated lighting conditions, spice states, and background interference. The agglomerated particles were labeled using the LabelImg tool as a training set. The training set was rotated, scaled, and brightness adjusted to expand the training set. For the YOLOv8 model, the COCO pre-trained weights were loaded, the training parameters were set, and the prior box size was recalculated using K-means clustering to train the YOLOv8 model. The pre-processed images were input into the trained YOLOv8 model to output the bounding box coordinates and category confidence of the agglomerated particles. A mapping relationship between pixels and actual size was established using a calibration plate. For each detection box, the geometric mean of its width and height was taken as the equivalent diameter. The detection box in the edge area was perspective corrected based on the camera installation angle. For the adhered particles within the detection box, the watershed algorithm was used for secondary segmentation to improve the accuracy of particle size calculation. Count the particle sizes of all agglomerated particles in the current frame, generate a particle size histogram at intervals, and output the particle size distribution, including the average particle size, D10, D50, and D90. According to the predefined particle size grades, count the number and mass proportions of particles in each grade and divide the particle size grades.

5. The method for controlling the automatic batching of spice production lines according to claim 1, wherein: The method of executing the particle size-based disintegration strategy, verifying the disintegration effect, and obtaining the agglomeration verification result includes: The particle size grades include primary agglomeration, secondary agglomeration, tertiary agglomeration, quaternary agglomeration and quintuple agglomeration. For primary agglomeration, the low-frequency vibrating screen at the outlet of the feeder is started to loosen and separate the particle clusters through vibration, and the ultrasonic vibration device on the inner wall of the pipe is turned on to emit pulses to destroy the weak adhesion between the particles. For secondary agglomeration, the double-roll crusher at the inlet of the feeder is started to shear the agglomerated particles through differential rotation, and the pneumatic vibrator at the bottom of the silo is turned on simultaneously to prevent the accumulation of subsequent ingredients that affect the crushing effect. For tertiary agglomeration, the treatment measures based on the secondary agglomeration are used. For the fourth-level agglomeration, the high-pressure air knife is turned on to blow the crushed particles away from the blades to prevent secondary adhesion. For the fourth-level agglomeration, the double-roll crusher is switched to reverse rotation mode to crush the hard agglomerates through rubbing and grinding. The centrifugal breakers at the end of the pipeline are started to use centrifugal force to throw large particles toward the wear-resistant liner and crush them through impact. For the fifth-level agglomeration, the feeder is stopped, the silo outlet valve is closed, the sound and light alarm is triggered, and the 360-degree rotating cleaning arm inside the silo is activated to try to physically remove the oversized agglomerates. If they are not removed within the set time, the manual maintenance process is triggered. Get the current particle size distribution. When it meets the qualification criteria, determine that there is no agglomeration in the feeder and output the agglomeration verification result.

6. The method for controlling the automatic batching of spice production lines according to claim 1, wherein: Based on the agglomeration verification result, when it is confirmed that there is no agglomeration in the feeder, the double exponential smoothing method is used to dynamically update the lag time caused by the viscous ingredients, including: The lag time for viscous ingredients is the time it takes for the actual flow rate of the ingredient to reach the target value after the feeder speed adjustment command is issued, reflecting the transmission delay of the viscous ingredient in the pipeline. With the feeder unloaded, the speed is increased from 0 to A% of the rated value. The pressure rise curve is recorded using the pressure sensor at the end of the pipeline. The time it takes for the pressure to first reach B% of the steady-state value is calculated as the initial lag time, where A and B are positive integers set by the user and A>B. Repeat the test for different spices, and take the average value as the baseline lag time for the ingredient, which is stored in the ingredient characteristics database. The lag time is considered as a dynamic sequence including the current lag value and the lag time change rate; the state is defined, including the horizontal component and the trend component, the horizontal component is the smoothed value of the current lag time, reflecting the immediate lag level, and the trend component is the changing trend of the lag time, reflecting the lag time increase or decrease rate caused by the viscosity change; the horizontal component weight and the trend component weight are set according to the viscosity coefficient of the ingredient. Dynamic adjustment, where , T represents the thickness of the wall adhesion layer, H represents the humidity of the spices, and D represents the D50 in the particle size distribution; the current lag time measured value, viscosity coefficient, horizontal component of the previous moment, and trend component of the previous moment are collected; when the trend remains unchanged, the horizontal prediction value is the previous moment horizontal component plus the previous moment trend component, and the current trend component is set to the previous moment trend component; the current horizontal component is obtained according to the difference between the current lag time measured value and the horizontal prediction value; the current trend component is adjusted based on the difference between the current horizontal component and the previous moment horizontal value; the lag time is the sum of the current horizontal component and the current trend component.

7. The method for controlling the automatic batching of spice production lines according to claim 1, wherein: The step of adding fuzzy adaptive compensation based on the lag time to obtain compensation time includes: Fuzzy linguistic values ​​are defined for input variables and output variables, including negative large, negative small, zero, positive small, and positive large. The input variables use a trapezoidal membership function, and the output variables use a single-valued fuzzy set. Based on the experience of viscous batching control and the fuzzy linguistic values, fuzzy rules are established to adjust the compensation gain and compensation time. The input variables are converted into the membership of each fuzzy set, and the corresponding fuzzy rules are activated according to the membership. The outputs of the activated rules are weighted averaged to obtain the fuzzy output. The compensation gain is calculated based on the fuzzy output using the center of gravity method. The basic compensation time is calculated based on the lag time and the compensation gain, and the basic compensation time is corrected by adding the deviation change rate correction amount. The boundary constraint is performed to obtain the compensation time.

8. The method for controlling the automatic batching of spice production lines according to claim 1, wherein: The method of establishing a hysteresis compensation effect evaluation index and performing no-load and full-load hysteresis tests on each spice based on the hysteresis time to verify the coupling error after compensation includes: The evaluation indicators for hysteresis compensation effectiveness include hysteresis time estimation error, batching accuracy error, coupling error, and system response time. The hysteresis time estimation error is the percentage deviation between the hysteresis time obtained by the double exponential smoothing method and the actual hysteresis time. The batching accuracy error is the percentage deviation between the actual discharge amount and the target discharge amount. The coupling error is the percentage deviation of the flow rate of another ingredient caused by the adjustment of one ingredient in a strongly coupled batching alignment. The system response time is the time from the viscosity mutation to the effectiveness of the compensation strategy. Empty the silo and close all feed inlets to ensure that there is no residual feed in the pipeline. Step the feeder speed from 0 to A% of the rated speed, synchronously record the timestamp of the feeder control signal and the response timestamp of the pressure sensor at the end of the pipeline, and calculate the inherent lag time. Record the inherent lag time at different speeds to form a comparison table of speed and inherent lag time. Group by viscosity coefficient, perform benchmark tests and compensation tests, compare data, and verify the coupling error after compensation.

9. The method for controlling the automatic batching of spice production lines according to claim 1, wherein: The method is based on the coupling error and the compensation time, and establishes a cross compensation matrix for the strong coupling pair through relative gain matrix analysis to generate a compensated feeder speed instruction, including: The feeder system is abstracted as a MIMO system, a unit step input is applied to the MIMO system, the steady-state response data is recorded, and the strong coupling pairs are determined; an n×n cross-compensation matrix C is constructed, and the elements in the cross-compensation matrix C are represents the compensation coefficient of the j-th feeder to the i-th feeder, and n represents the number of mutually coupled feeders. The compensation intensity factors of the non-diagonal elements in the cross-compensation matrix C are adjusted in real time according to the coupling error. For the strongly coupled pair (i, j), the compensation time ratio is introduced as a weight to dynamically update the compensation coefficient. Based on the original speed command vector output by the PID controller, cross compensation is calculated and physical constraints are imposed; the compensation effect is verified in real time and matrix optimization is performed to generate the compensated feeder speed command.

10. An automated control system for ingredients in a production line for spice compounding, using the automated control method for ingredients in a production line for spice compounding according to any one of claims 1 to 9, characterized in that: include: The hierarchical anti-caking control module includes a data fusion unit, an anti-caking control model construction unit, and a hierarchical anti-caking control unit. The data fusion unit acquires sensor data, including the state of spices and the thickness of the wall adhesion layer, defines a state vector, and fuses the sensor data using an extended Kalman filter. The anti-caking control model construction unit establishes an anti-caking control model based on the fused sensor data and performs adaptive model correction. The anti-caking control model in the hierarchical anti-caking control unit outputs a caking prevention control strategy to perform hierarchical anti-caking control. Agglomeration Verification Module: This module includes a particle size classification unit and an agglomeration verification unit. When agglomerates are found in the feeder, the particle size classification unit uses the YOLOv8 model to identify agglomerated particles, outputs the particle size distribution, and classifies the particles into size classes. The agglomeration verification unit implements a particle size class-based decomposition strategy, verifies the decomposition effect, and obtains an agglomeration verification result. Lag and compensation calculation module: includes: lag time calculation unit and compensation time calculation unit; among them, the lag time calculation unit is based on the agglomeration verification result. When it is confirmed that there is no agglomeration in the feeder, the double exponential smoothing method is used to dynamically update the lag time caused by viscous ingredients; the compensation time calculation unit adds fuzzy adaptive compensation based on the lag time to obtain the compensation time; Feeder speed instruction generation module: includes: a coupling error verification unit and an instruction generation unit; wherein, the coupling error verification unit establishes a hysteresis compensation effect evaluation index, performs no-load and full-load hysteresis tests on each spice based on the hysteresis time, and verifies the compensated coupling error; the instruction generation unit establishes a cross-compensation matrix for the strong coupling pair based on the coupling error and compensation time through relative gain matrix analysis, and generates the compensated feeder speed instruction.

Images