Tortoise shell glue bead whole process parameter control method based on data algorithm

Abstract

The application discloses a kind of based on data algorithm's tortoise shell glue pearl whole process parameter control method.First, the initial moisture content of tortoise shell glue Ding wet base and cutting Ding particle size distribution are detected, and the roasting adaptation index is generated by weighting;Second, the type of adjuvant is automatically selected according to the index and the accurate dosage is determined;Then, the temperature and stirring frequency setting value are output and executed by querying the lookup table;Then, the feed rate time series distribution is estimated by using the feedforward prediction model, and the parameters of the packaging equipment are preset;Finally, the packaging machine parameters are updated by the feedforward-feedback adaptive control strategy by fusing the actual discharge measurement value, so that the packaging process and the roasting discharge rate are dynamically matched.The application realizes the parameter accurate adaptive matching and whole line linkage control closed loop, and improves the qualified bead rate and the whole line automation production efficiency.

Description

Technical Field

[0001] This invention relates to the technical field of tortoise shell glue bead production control, and in particular to a method for controlling the parameters of the entire tortoise shell glue bead process based on data algorithms. Background Technology

[0002] Tortoise shell glue, a precious traditional Chinese medicine, possesses significant medicinal effects such as nourishing yin and suppressing yang, tonifying the kidneys and strengthening bones, and is widely used in clinical and health care fields. In both traditional and modern processing of tortoise shell glue, the glue blocks are typically cut into cubes and then stir-fried at high temperatures with excipients such as clam powder and talc powder. This causes the glue to expand and form brittle tortoise shell glue beads, facilitating subsequent pulverization, decoction, and the dissolution of active ingredients. With the advancement of modernization in traditional Chinese medicine production, equipment such as electromagnetic stir-frying machines and automated granule packaging machines have been widely used in tortoise shell glue bead production lines. However, in actual large-scale industrial production, existing processing control methods and production line coordination mechanisms still face many technical bottlenecks that urgently need to be addressed.

[0003] First, existing technologies lack scientific quantitative assessment and adaptive adjustment mechanisms when dealing with batch variations in raw materials. The initial moisture content and particle size distribution of tortoise shell gelatin cubes are key factors affecting the frying quality. Fluctuations in moisture content directly affect the uniformity of heat transfer in the pan and the rate of thermal expansion of the gelatin cubes; while uneven particle size easily leads to the phenomenon of "small particles burning and large particles undercooked" during frying, severely restricting the improvement of the qualified bead formation rate. Currently, most manufacturers still rely on fixed process parameters or manual experience for production, failing to establish comprehensive evaluation indicators based on the multidimensional characteristics of raw materials (such as moisture content and particle size variation coefficient). This results in the selection and dosage of auxiliary materials (such as clam powder or talc) often being fixed or based solely on subjective judgment, failing to achieve precise matching for the characteristics of the current batch of materials, thus causing fluctuations in frying quality.

[0004] Secondly, existing roasting equipment lacks data-driven dynamic optimization capabilities in parameter settings. In the control of electromagnetic roasting machines, roasting temperature and stirring frequency are the core control variables determining the quality of the beads. Traditional control systems typically employ open-loop control or simple fixed threshold settings, failing to effectively combine the adaptability characteristics of the upstream materials with process data from historical high-quality batches. Due to the lack of intelligent optimization algorithms for offline calibration and online interpolation, existing methods cannot automatically recommend and distribute the optimal combination of temperature and stirring frequency based on the actual conditions of the current batch of materials. This leads to a disconnect between equipment operating parameters and actual material requirements, increasing energy consumption and making it difficult to guarantee the uniformity of each batch of tortoise shell glue beads.

[0005] Furthermore, existing production lines suffer from severe "information silos" between the roasting and packaging processes, lacking a closed-loop, integrated control system for the entire process. The discharge rate after roasting is often a non-stationary process that fluctuates over time, while existing granule packaging machines (such as the DXDK granule packaging machine) typically operate with fixed target weighing settings and gate opening / closing rhythms. Because the discharge rhythm of the roasting process cannot be predicted in advance, the packaging equipment cannot preset parameters before the material actually arrives. This mismatch between the upstream and downstream processes easily leads to large accumulations and blockages of material at the packaging station, or equipment idling and waiting, severely reducing the overall operating efficiency of the production line.

[0006] Finally, regarding dynamic control strategies, existing technologies struggle to handle complex disturbances during the material discharge process. Although some advanced production lines have introduced sensors for monitoring, these are mostly limited to single feedback control, exhibiting significant lag. When the discharge rate fluctuates drastically, simple feedback adjustment often leads to frequent gate actuation of the packaging machine or increased weighing errors. Existing methods generally lack adaptive error-driven control strategies that integrate feedforward prediction with real-time feedback adjustment, making it difficult to maintain a high-precision dynamic match between the packaging process's processing rate and the front-end roasting discharge rate.

[0007] In summary, to address the industry pain points such as the large differences in raw material characteristics, low accuracy in matching auxiliary materials and frying parameters, and poor coordination between front-end and back-end processes in the production of tortoise shell glue beads, there is an urgent need for a data algorithm-based full-process parameter control method for tortoise shell glue beads. This method would enable intelligent linkage control of the entire process from material preparation and frying to packaging, thereby effectively improving the yield of qualified beads and enhancing the automated production efficiency of the entire line. Summary of the Invention

[0008] In view of this, the present invention provides a data algorithm-based method for full-process parameter control of tortoise shell glue beads. The purpose is to achieve precise adaptive matching of auxiliary material selection and frying parameters through in-depth analysis and quantitative evaluation of material characteristics, and to establish a linkage control mechanism based on feedforward prediction and feedback adaptive fusion, thereby eliminating the cycle time barrier between the front and back processes and comprehensively improving the qualified bead formation rate and the overall automated production efficiency of tortoise shell glue beads.

[0009] To achieve the above objectives, this invention provides a method for controlling the entire process parameters of tortoise shell glue beads based on data algorithms, comprising the following steps: B1: The initial moisture content and particle size distribution of the tortoise shell glue cubes in the current batch are detected. Based on the influence of the initial moisture content of the wet base on the heat transfer uniformity of the auxiliary materials in the pot and the influence of the non-uniformity of the particle size on the qualified bead formation rate, the two types of characteristics are assigned corresponding weights and weighted fusion is performed to generate the stir-frying compatibility index of the current batch. B2: Taking the stir-fry compatibility index as input, the auxiliary material type is automatically selected based on the partition threshold rules determined by the historical batch auxiliary material selection rules, and the precise auxiliary material dosage is determined based on the piecewise linear interpolation lookup table of the auxiliary material dosage calibrated offline for each auxiliary material type. The auxiliary material type selection result and the precise auxiliary material dosage are output. B3: Using the excipient type selection result and the stir-frying compatibility index as a joint index, query the offline calibrated stir-frying parameter recommendation lookup table, output the quality verification qualified temperature setting value and stirring drive frequency setting value of the current batch through piecewise linear interpolation, and send the parameter combination to the electromagnetic stir-frying machine control system for execution; B4: Use a feedforward prediction model of discharge rhythm trained based on historical batch roasting cycle data to estimate the time sequence distribution of discharge rate in the current batch roasting process in advance, and provide parameter preset basis for packaging equipment before the material actually arrives at the packaging station. B5: Using the time-series prediction value of the discharge rate as the feedforward benchmark, the actual discharge rate measurement value collected by the packaging station sensor is synchronously integrated. Through the feedforward-feedback fusion error-driven adaptive control strategy, the target weighing set value and the opening and closing rhythm parameter vector of the DXDK granule packaging machine are continuously updated. This ensures that the processing rate of the packaging process and the discharge rate of the front-end roasting process are dynamically matched, realizing a linkage control closed loop from material preparation parameters, roasting parameters to packaging parameters with the discharge rate matching as the goal.

[0010] As a further improvement of the present invention: Optionally, step B1 further includes: The current batch of tortoise shell glue cubes was sampled and tested using a halogen moisture analyzer. The arithmetic mean of the moisture analyzer readings at each sampling point was taken as the initial wet basis moisture content of the current batch of tortoise shell glue cubes. The units are percentages; images of the cut pieces of the current batch of tortoise shell glue samples are acquired using an image acquisition device. The equivalent circular diameter of each particle is extracted using an image processing algorithm as the particle size measurement value. The particle size measurements of each particle are statistically analyzed, and the coefficient of variation of the particle size distribution of the current batch of cut pieces is calculated. The coefficient of variation The standard deviation of the particle size measurement values ​​of each diced particle within the batch is divided by the sample mean of the particle size measurement values. Two independent univariate standardized regression analyses were conducted on historical batch data. The initial moisture content and particle size variation coefficient of each historical batch (standardized by z-score) were used as independent variables, and the qualified bead formation rate of the corresponding batch was used as the dependent variable. Two standardized linear regression models were established, and the absolute values ​​of the standardized regression coefficients were defined as the absolute sensitivity of the initial moisture content of the wet substrate. absolute sensitivity to particle size variation coefficient ;when and When the sum is greater than zero, and The characteristic weighting coefficients of water content were obtained after linear normalization. Weighting coefficient of particle size inhomogeneity ,in , ;when and When the sum equals zero, let The and satisfy ; Based on the initial moisture content of the wet basis With particle size variation coefficient The raw material characteristic input is the minimum initial moisture content on a wet basis from historical batch records. With the maximum value right Range normalization was performed using the minimum value of the particle size variation coefficient from historical batch records. With the maximum value right Perform range normalization; when At that time, The range normalization result is taken as 0.5; when At that time, The range normalization result is taken as 0.5; for the range normalization result... and Cut off to An interval is used to handle situations where the characteristics of the current batch exceed the historical range. The normalized value of the initial moisture content of the wet basis after truncation is denoted as... The normalized value of the truncated particle size variation coefficient is denoted as ; Weighting coefficient based on water content characteristics Weighting coefficient of particle size inhomogeneity characteristics Weighted fusion generates the current batch's trading suitability index. The stir-frying suitability index The calculation formula is: ; in, This indicates the stir-frying suitability index for the current batch. .

[0011] Optionally, step B2 further includes: Based on historical batch data, historical batches are categorized according to the stir-frying suitability index. Sort in ascending order and extract batches that have used clam powder and passed quality verification. Numerical set and batches that use talc powder and have passed quality verification Numerical set; select qualified batches using talcum powder. The maximum value in the set of numbers is used as the threshold for the next partition. Batch of products using qualified clam powder were sampled. The minimum value of the set of values ​​is used as the threshold for the upper partition. ; Before officially enabling the threshold partitioning rule, and Perform a one-time configuration validity check: If If the threshold configuration is valid, then the threshold configuration is effective; if If the threshold configuration is invalid, a configuration error alarm will be issued, prompting the operator to manually specify the threshold based on current process knowledge. Valid threshold pairs; After confirmation After its establishment, the operational period will be used to adapt the current batch of stir-frying index. The following three-zone judgment rules will be uniformly applied, without distinguishing any additional circumstances: If ,but Talc; if ,but Clam powder; if If the system sends a confirmation request for the type of artificial auxiliary material to the operation console, it will pause the automatic decision-making process and wait for operator confirmation. Then proceed with the next steps; Separate piecewise linear interpolation lookup tables were created for the dosage of clam powder and talcum powder; each lookup table was based on a stir-fry compatibility index. As an index variable, store The calibration index point, the first One calibration index point, Includes the calibrated stir-frying adaptation index Corresponding auxiliary material dosage calibration values ; Calibration values ​​of auxiliary material usage at calibration index points The data is determined from historical batch records of batches that used the corresponding auxiliary material type and passed quality verification, using the arithmetic mean of the actual auxiliary material usage of qualified batches within the range of each calibration compatibility index; each calibration index point is determined according to... Strictly ascending in order of increasing value; if duplicate values ​​appear in the initially determined calibration and cooking compatibility index, the average value of the actual auxiliary material usage of qualified batches within the corresponding interval will be calculated and replaced by a single calibration index point. The matching index for the current batch of stir-frying ,exist Locating the corresponding auxiliary material type in the lookup table satisfies two adjacent calibration index points and Piecewise linear interpolation is used to calculate the accurate amount of excipients. : ; when Below the minimum index point of the corresponding lookup table At that time, with As Output; when Higher than the maximum index point of the corresponding lookup table At that time, with As The output, i.e., performs boundary saturation processing on the index values ​​of the lookup table; after interpolation and boundary saturation processing are completed, the output is... Cut off the mixing ratio range; in, This indicates the precise amount of excipients used in the current batch. Indicates the first The calibration values ​​of auxiliary material usage at each calibration index point. Indicates the first The calibration values ​​of auxiliary material usage at each calibration index point. Indicates the first The calibration and matching index at each calibration index point. Indicates the first The calibration and cooking adaptation index at each calibration index point.

[0012] Optionally, step B3 further includes: Separate lookup tables for recommended stir-frying parameters were created for clam powder and talcum powder; each lookup table is based on a stir-frying suitability index. As an index variable, store The calibration index point, the first One calibration index point, Includes calibrated stir-frying compatibility index Recommended temperature calibration values Recommended calibration value for stirring drive frequency Recommended calibration value for temperature at the calibration index point. Recommended calibration value for stirring drive frequency From historical batch records of batches using the corresponding auxiliary material type and confirmed as qualified through quality verification, the arithmetic mean of the actual frying temperature records and actual stirring drive frequency records of qualified batches within the range near each calibrated frying compatibility index is used to determine the index points; each calibration index point is determined according to... The values ​​are arranged in a strictly ascending order. If duplicate values ​​appear in the initially determined calibration roasting compatibility index, the average of the actual temperature and frequency records of qualified batches within the corresponding interval is calculated and replaced by a single calibration index point. The distribution range of the roasting compatibility index of the calibration index point should cover the usage of the corresponding auxiliary material type in historical batches. Numerical range; The matching index for the current batch of stir-frying ,in accordance with Select the recommended lookup table for the stir-frying parameters corresponding to the type of auxiliary material, and locate the parameter that meets the requirements in the selected lookup table. two adjacent calibration index points and The temperature setpoint was calculated using piecewise linear interpolation. With stirring drive frequency setting value : ; ; when Below the minimum index point of the selected lookup table At that time, with and As respectively and Output; when Higher than the maximum index point of the selected lookup table At that time, with and As respectively and The output, i.e., performs boundary saturation processing on the index values ​​of the lookup table; after interpolation and boundary saturation processing are completed, the output is... Implement the recommended temperature cutoff range for stir-frying to ensure Falling Within Celsius; for Perform a recommended range cutoff for the stirring drive frequency to ensure Falling Within Hz; in, This indicates the temperature setting value for the current batch. This indicates the stirring drive frequency setting for the current batch. and They represent the first With the Recommended calibration values ​​for temperature at each calibration index point. and They represent the first With the Recommended calibration values ​​for the stirring drive frequency at each calibration index point. and They represent the first With the The calibration cooking adaptation index at each calibration index point; Temperature setting With stirring drive frequency setting value The parameters are combined into a roasting parameter package, which is written into the parameter register of the electromagnetic roasting machine control system via an industrial communication protocol. The electromagnetic roasting machine control system then adjusts the heating power and stirring drive frequency according to the roasting parameter package. The control value for the stirring speed is sent to the frequency converter drive unit of the electromagnetic stir-frying machine.

[0013] Optionally, step B4 further includes: Construct a feedforward prediction model for the discharge rhythm, wherein the feedforward prediction model for the discharge rhythm uses a temperature setpoint. Stirring drive frequency setting value Precise dosage of excipients Binary coding of auxiliary material types Talc powder corresponds to code 0, clam powder corresponds to code 1, initial moisture content on wet basis. With particle size variation coefficient Stir-frying suitability index and the actual amount of material fed into the current batch. The concatenated 8-dimensional feature vector serves as input. A one-dimensional convolutional layer extracts local correlation features between the components of the input parameters. A max-pooling layer compresses these local correlation features. A fully connected layer maps the compressed feature vector to the temporal output space, outputting a sequence of predicted output rates arranged by time step index. The temporal prediction sequence contains a total of... The time step, the first The feedforward prediction of the discharge rate at each time step is denoted as . , ; The time-series predicted value sequence , ,..., When the current batch of roasting process starts, the feedforward prediction calculation is performed and a complete time-series prediction value sequence is generated. Before the material arrives at the packaging station, the prediction value sequence is output to the DXDK granule packaging machine control unit. The control unit then uses the time-series prediction value sequence to pre-set the initial parameters of the target weighing set value and the opening and closing rhythm of the feeding gate.

[0014] Optionally, step B5 further includes: A discharge rate sensor is deployed at the packaging station. This sensor combines a weighing sensor and a timing module. The weighing sensor collects the weight of material flowing through the packaging station per unit time, and the timing module records the corresponding time interval. The two sensors are combined to calculate the actual discharge rate measurement. The actual discharge rate measurement is continuously collected at fixed sampling intervals. The actual discharge rate measured at each time step is recorded as follows: ; At each time step Perform feedforward-feedback fusion computation: when At that time, the packaging station had not yet obtained a valid measured value of the discharge rate, so the feedforward weighting coefficient was set to... Feedback weighting coefficient Feedforward prediction value Directly used as the fusion driving rate ;when At that time, the first Feedforward prediction of discharge rate at each time step Compared with the actual discharge rate measurement value Feedforward weight coefficients updated adaptively based on error With feedback weight coefficient Perform weighted fusion to obtain the fusion-driven rate. The fusion drive rate The calculation formula is: ; in, Indicates the first Fusion driving rate at each time step Indicates the first Adaptive feedforward weight coefficients at each time step Indicates the first The adaptive feedback weight coefficients at each time step, and ; The error-driven adaptive weight update mechanism uses the normalized absolute error of the difference between the measured value and the predicted value at the current time step. Drive the update of the feedback weights at the next time step. The calculation formula is: ; No. Adaptive feedback weights at each time step Calculate using the following formula: ; ; in, Indicates the first Normalized absolute prediction error at each time step; To prevent small quantities with a denominator of zero; Indicates will Cut off to interval ; This represents the initial value of the feedback weights; Indicates the adaptive step size coefficient; and These are the lower and upper limits of the feedback weight, respectively; Indicates the first Adaptive feedforward weight coefficients for each time step; Driven by fusion rate As the driving force, the target weighing setpoint of the DXDK granule packaging machine is updated according to the offline calibrated weighing setpoint-rate mapping function. Update the opening and closing rhythm parameter vector of the feeding gate according to the offline calibrated feeding gate rhythm-rate mapping function. ,in Indicates the first The duration of the gate opening within each feeding cycle at each time step. Indicates the first The duration of gate closure within each feeding cycle at each time step; the offline calibrated feeding gate rhythm-rate mapping function is established by recording the corresponding feeding gate opening and closing rhythm parameter vectors under 5 different feeding rate conditions, and stored as a gate calibration lookup table using piecewise linear interpolation; fused drive rate The index value is input into the corresponding calibration lookup table, and the result is obtained through piecewise linear interpolation. and During the batch roasting process, the fusion calculation and parameter update are repeated at each time step to keep the processing rate of the DXDK granule packaging machine dynamically matched with the front-end roasting output rate.

[0015] Compared with the prior art, the present invention has at least the following beneficial effects: This invention innovatively constructs a frying adaptation index generation mechanism based on dynamic weighting of multi-dimensional material characteristics by deeply mining historical batch data. Unlike conventional fixed-weight evaluation or simple multivariate linear superposition, this invention utilizes univariate standardized regression analysis to extract the absolute sensitivity of the initial moisture content and particle size variation coefficient of the wet basis. Based on this, feature weights are adaptively allocated, and range normalization and truncation are performed. This improved algorithm accurately quantifies the comprehensive impact of the current batch of materials on heat transfer uniformity and bead formation rate, overcoming evaluation biases caused by extreme material characteristics. It provides a highly customized quantitative benchmark for subsequent automatic selection and dosage calculation of auxiliary materials. Furthermore, combined with a three-zone decision rule with an anomaly alarm mechanism and a piecewise linear interpolation algorithm, this invention can smoothly and safely map precise auxiliary material dosages, frying temperatures, and stirring frequencies. This interpolation optimization strategy based on boundary saturation processing and range truncation effectively avoids the risk of control quantities exceeding limits under unknown operating conditions, significantly improving the stability of equipment operation and the uniformity of bead formation quality.

[0016] Furthermore, this invention constructs a feedforward prediction and error-driven adaptive feedback control architecture for the discharge rhythm based on multi-dimensional feature fusion. Instead of simply applying conventional time-series prediction models, this invention deeply concatenates the front-end generated roasting adaptation index, auxiliary material parameters, and equipment settings into a multi-dimensional feature vector. It then uses convolution and pooling operations to extract local correlation features between parameters, thereby accurately depicting the temporal distribution of the discharge rate in advance. More importantly, in the front-end and back-end rate matching stage, this invention designs a dynamic adaptive update mechanism for the feedforward and feedback weights, using real-time normalized absolute error to drive the smooth evolution of the feedback weights. This improved fusion control strategy effectively overcomes the shortcomings of a single feedforward model being susceptible to sudden disturbances and a single feedback control exhibiting severe lag, enabling high-precision dynamic matching between the weighing settings and the discharge gate rhythm of the packaging equipment as the front-end discharge rate fluctuates.

[0017] From an overall perspective, this invention breaks through the limitations of isolated control of each process in traditional production lines, forming a highly collaborative closed-loop system. The precise quantification of material characteristics at the front end not only determines the roasting process parameters in the middle stage, but these process parameters, along with the material characteristics, serve as core prerequisites, directly driving the prediction and dynamic adjustment of the cycle time in the back-end packaging process. Each innovative link is interconnected; the output data from the previous stage is deeply integrated and guides the algorithm execution in the next stage, enabling true global optimization of material preparation, roasting, and packaging through data flow. This deep process coupling and algorithmic collaboration not only significantly reduces the risk of material accumulation or equipment idling, but also fundamentally improves the intelligence level and overall efficiency of the fully automated production of tortoise shell glue beads. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a data algorithm-based method for controlling the parameters of tortoise shell glue beads according to an embodiment of the present invention. Figure 2 This is a schematic diagram illustrating the effect of the dynamic adjustment process of adaptive feedforward weights and feedback weights with time steps under error-driven conditions according to an embodiment of the present invention. Detailed Implementation

[0019] The present invention will be further described below with reference to the accompanying drawings, but this is not intended to limit the present invention in any way. Any modifications or substitutions made based on the teachings of the present invention shall fall within the protection scope of the present invention.

[0020] Example 1: A method for controlling the entire process parameters of tortoise shell glue beads based on data algorithms, such as... Figure 1 As shown, it includes the following steps: B1: The initial moisture content and particle size distribution of the tortoise shell glue cubes in the current batch are tested. Based on the influence of the initial moisture content on the heat transfer uniformity of the auxiliary materials in the pan and the influence of the particle size non-uniformity on the qualified bead formation rate, corresponding weights are assigned to the two types of characteristics and weighted fusion is performed to generate the stir-frying suitability index for the current batch, including: The current batch of tortoise shell glue cubes was sampled and tested using a halogen moisture analyzer. In this embodiment, the sampling quantity was set to be no less than 5% of the total batch size. The arithmetic mean of the moisture analyzer readings at each sampling point was taken as the initial wet basis moisture content of the current batch of tortoise shell glue cubes. The units are percentages. Images of the cut pieces of the current batch of tortoise shell glue samples are acquired using an image acquisition device. The equivalent circular diameter of each particle is extracted using an image processing algorithm as the particle size measurement value. In this embodiment, the number of sample particles collected is set to be no less than 200. The particle size measurements of each particle are statistically analyzed, and the coefficient of variation of the particle size distribution of the current batch of cut pieces is calculated. The coefficient of variation The standard deviation of the particle size measurement values ​​of each diced particle within the batch is divided by the sample mean of the particle size measurement values. Two independent univariate standardized regression analyses were conducted on historical batch data. The initial moisture content and particle size variation coefficient of each historical batch (standardized by z-score) were used as independent variables, and the qualified bead formation rate of the corresponding batch was used as the dependent variable. Two standardized linear regression models were established, and the absolute values ​​of the standardized regression coefficients were defined as the absolute sensitivity of the initial moisture content of the wet substrate. absolute sensitivity to particle size variation coefficient ;when and When the sum is greater than zero, and The characteristic weighting coefficients of water content were obtained after linear normalization. Weighting coefficient of particle size inhomogeneity characteristics ,in , ;when and When the sum equals zero, let The and satisfy In this embodiment, the historical batch dataset contains no fewer than 50 batches of processed products with recorded qualified pearl formation rates, and both sets of standardized regression analyses are established using the least squares linear regression method. Based on the initial moisture content of the wet basis With particle size variation coefficient The raw material characteristic input is the minimum initial moisture content on a wet basis from historical batch records. With the maximum value right Range normalization was performed using the minimum value of the particle size variation coefficient from historical batch records. With the maximum value right Perform range normalization; when At that time, The range normalization result is taken as 0.5; when At that time, The range normalization result is taken as 0.5; for the range normalization result... and Cut off to An interval is used to handle situations where the characteristics of the current batch exceed the historical range. The normalized value of the initial moisture content of the wet basis after truncation is denoted as... The normalized value of the truncated particle size variation coefficient is denoted as ; Weighting coefficient based on water content characteristics Weighting coefficient of particle size inhomogeneity characteristics Weighted fusion generates the current batch's trading suitability index. The stir-frying suitability index The calculation formula is: ; in, This indicates the stir-frying suitability index for the current batch. In this embodiment, Set to 6.5%. Set to 12.0%. Set to 0.05, Set it to 0.25.

[0021] Optionally, in order to make the weighting coefficients and It can continuously optimize as production data accumulates, and when the number of newly added valid records in the batch historical dataset reaches the preset update trigger threshold. At that time, online recalibration was performed on the weight coefficients: two sets of standardized regression analyses were re-performed based on the latest full historical batch dataset, and the absolute sensitivity was updated. and And recalculate according to the above normalization formula. and The updated and This information is immediately used for calculating the cooking compatibility index for subsequent batches; in this embodiment, Setting it to 20 means that an online weight recalibration is triggered every time 20 new valid historical records of processing batches are added.

[0022] B2: Using the stir-fry compatibility index as input, it automatically selects the excipient type based on the partition threshold rules determined by the historical batch excipient selection patterns. It then determines the precise excipient dosage based on a piecewise linear interpolation lookup table of excipient dosages calibrated offline for each excipient type. The output includes the excipient type selection result and the precise excipient dosage, including: Based on historical batch data, historical batches are categorized according to the stir-frying suitability index. Sort in ascending order and extract batches that have used clam powder and passed quality verification. Numerical set and batches that use talc powder and have passed quality verification Numerical set; select qualified batches using talcum powder. The maximum value in the set of numbers is used as the threshold for the next partition. Batch of products using qualified clam powder were sampled. The minimum value of the set of values ​​is used as the threshold for the upper partition. ; Before officially enabling the threshold partitioning rule, and Perform a one-time configuration validity check: If If the threshold configuration is valid, then the threshold configuration is effective; if If the threshold configuration is invalid, a configuration error alarm will be issued, prompting the operator to manually specify the threshold based on current process knowledge. Valid threshold pairs; After confirmation After its establishment, the operational period will be used to adapt the current batch of stir-frying index. The following three-zone judgment rules will be uniformly applied, without distinguishing any additional circumstances: If ,but Talc; if ,but Clam powder; if If the system sends a confirmation request for the type of artificial auxiliary material to the operation console, it will pause the automatic decision-making process and wait for operator confirmation. Then proceed with the next steps; In this embodiment, the historical batch dataset contains no fewer than 50 processing batches with recorded auxiliary material types and qualified pearl formation rates; Set to 0.35. Setting it to 0.55 means that batches with a compatibility index of no more than 0.35 will automatically use talcum powder, batches with a compatibility index of no less than 0.55 will automatically use clam powder, and batches with a compatibility index between 0.35 and 0.55 will trigger a manual confirmation process. Separate piecewise linear interpolation lookup tables were created for the dosage of clam powder and talcum powder; each lookup table was based on a stir-fry compatibility index. As an index variable, store The calibration index point, the first One calibration index point, Includes the calibrated stir-frying adaptation index Corresponding auxiliary material dosage calibration values ; Calibration values ​​of auxiliary material usage at calibration index points The data is determined from historical batch records of batches that used the corresponding auxiliary material type and passed quality verification, using the arithmetic mean of the actual auxiliary material usage of qualified batches within the range of each calibration compatibility index; each calibration index point is determined according to... Strictly ascending in order of increasing value; if duplicate values ​​appear in the initially determined calibration and cooking compatibility index, the average value of the actual auxiliary material usage of qualified batches within the corresponding interval will be calculated and replaced by a single calibration index point. The matching index for the current batch of stir-frying ,exist Locating the corresponding auxiliary material type in the lookup table satisfies two adjacent calibration index points and Piecewise linear interpolation is used to calculate the accurate amount of excipients. : ; when Below the minimum index point of the corresponding lookup table At that time, with As Output; when Higher than the maximum index point of the corresponding lookup table At that time, with As The output, i.e., performs boundary saturation processing on the index values ​​of the lookup table; after interpolation and boundary saturation processing are completed, the output is... Execution ratio range cutoff: In this embodiment, for clam powder, Cut off to kg / 100 kg tortoise shell glue wet base; for talc powder Cut off to kg / 100 kg tortoise shell glue wet base; in, This indicates the precise amount of excipients used in the current batch. Indicates the first The calibration values ​​of auxiliary material usage at each calibration index point. Indicates the first The calibration values ​​of auxiliary material usage at each calibration index point. Indicates the first The calibration and matching index at each calibration index point. Indicates the first The calibration cooking adaptation index at each calibration index point; In this embodiment, both the clam powder dosage lookup table and the talc powder dosage lookup table are set with 5 calibration index points. The calibration and roasting compatibility index of the clam powder dosage lookup table is set to 0.55, 0.65, 0.75, 0.85, and 0.95, respectively, and the corresponding auxiliary material dosage calibration values ​​are set to 30, 35, 40, 45, and 50 kg / 100 kg of tortoise shell glue wet base, respectively. The calibration and roasting compatibility index of the talc powder dosage lookup table is set to 0.05, 0.15, 0.25, 0.30, and 0.35, respectively, and the corresponding auxiliary material dosage calibration values ​​are set to 40, 42, 44, 46, and 48 kg / 100 kg of tortoise shell glue wet base, respectively. The above calibration values ​​are determined based on the statistical average of historical batch data in this embodiment, and should be recalibrated according to the historical batch data of each production unit during actual deployment.

[0023] B3: Using the selected excipient type and the stir-frying compatibility index as a joint index, query the offline calibrated stir-frying parameter recommendation lookup table, output the quality verification qualified temperature setpoint and stirring drive frequency setpoint for the current batch through piecewise linear interpolation, and send the parameter combination to the electromagnetic stir-frying machine control system for execution, including: Separate lookup tables for recommended stir-frying parameters were created for clam powder and talcum powder; each lookup table is based on a stir-frying suitability index. As an index variable, store The calibration index point, the first One calibration index point, Includes calibrated stir-frying compatibility index Recommended temperature calibration values Recommended calibration value for stirring drive frequency Recommended calibration value for temperature at the calibration index point. Recommended calibration value for stirring drive frequency From historical batch records of batches using the corresponding auxiliary material type and confirmed as qualified through quality verification, the arithmetic mean of the actual frying temperature records and actual stirring drive frequency records of qualified batches within the range near each calibrated frying compatibility index is used to determine the index points; each calibration index point is determined according to... The values ​​are arranged in a strictly ascending order. If duplicate values ​​appear in the initially determined calibration roasting compatibility index, the average of the actual temperature and frequency records of qualified batches within the corresponding interval is calculated and replaced by a single calibration index point. The distribution range of the roasting compatibility index of the calibration index point should cover the usage of the corresponding auxiliary material type in historical batches. Numerical range; The matching index for the current batch of stir-frying ,in accordance with Select the recommended lookup table for the stir-frying parameters corresponding to the type of auxiliary material, and locate the parameter that meets the requirements in the selected lookup table. two adjacent calibration index points and The temperature setpoint was calculated using piecewise linear interpolation. With stirring drive frequency setting value : ; ; when Below the minimum index point of the selected lookup table At that time, with and As respectively and Output; when Higher than the maximum index point of the selected lookup table At that time, with and As respectively and The output, i.e., performs boundary saturation processing on the index values ​​of the lookup table; after interpolation and boundary saturation processing are completed, the output is... Implement the recommended temperature cutoff range for stir-frying to ensure Falling Within Celsius; for Perform a recommended range cutoff for the stirring drive frequency to ensure Falling Within Hz; in, This indicates the temperature setting value for the current batch. This indicates the stirring drive frequency setting for the current batch. and They represent the first With the Recommended calibration values ​​for temperature at each calibration index point. and They represent the first With the Recommended calibration values ​​for the stirring drive frequency at each calibration index point. and They represent the first With the The calibration cooking adaptation index at each calibration index point; In this embodiment, both the clam powder stir-frying parameter recommendation lookup table and the talc powder stir-frying parameter recommendation lookup table are set with 5 calibration index points. The calibration stir-frying adaptation index of the clam powder stir-frying parameter recommendation lookup table is set to 0.55, 0.65, 0.75, 0.85, and 0.95, respectively, with corresponding recommended temperature calibration values ​​of 170, 180, 190, 205, and 215 degrees Celsius, and corresponding recommended stirring drive frequency calibration values ​​of 12, 13, 14, 16, and 17 Hz. The calibration stir-frying adaptation index of the talc powder stir-frying parameter recommendation lookup table is set to 0.05, 0.15, 0.25, 0.30, and 0.35, respectively, with corresponding recommended temperature calibration values ​​of 165, 168, 172, 176, and 178 degrees Celsius, and corresponding recommended stirring drive frequency calibration values ​​of 11, 12, 13, 14, and 15 Hz. Hz; The above calibration values ​​are determined based on the statistical average of historical batch data in this embodiment. In actual deployment, they should be recalibrated according to the historical batch data of each production unit. Temperature setting With stirring drive frequency setting value The parameters are combined into a roasting parameter package, which is written into the parameter register of the electromagnetic roasting machine control system via an industrial communication protocol. The electromagnetic roasting machine control system then adjusts the heating power and stirring drive frequency according to the roasting parameter package. The control value for the stirring speed is sent to the frequency converter drive unit of the electromagnetic stir-frying machine.

[0024] B4: Using a feedforward prediction model of discharge rhythm trained on historical batch roasting cycle data, the model estimates the time-series distribution of discharge rate during the current batch roasting process in advance, providing parameter presets for packaging equipment before the material actually arrives at the packaging station, including: Construct a feedforward prediction model for the discharge rhythm, wherein the feedforward prediction model for the discharge rhythm uses a temperature setpoint. Stirring drive frequency setting value Precise dosage of excipients Binary coding of auxiliary material types Talc powder corresponds to code 0, clam powder corresponds to code 1, initial moisture content on wet basis. With particle size variation coefficient Stir-frying suitability index and the actual amount of material fed into the current batch. The concatenated 8-dimensional feature vector serves as input. A one-dimensional convolutional layer extracts local correlation features between the components of the input parameters. A max-pooling layer compresses these local correlation features. A fully connected layer maps the compressed feature vector to the temporal output space, outputting a sequence of predicted output rates arranged by time step index. The temporal prediction sequence contains a total of... The time step, the first The feedforward prediction of the discharge rate at each time step is denoted as . , In this embodiment, the kernel size of the one-dimensional convolutional layer is set to 3, with same padding, and the number of output channels is set to 16; the pooling window size of the max pooling layer is set to 2, the stride is set to 2, and same padding is used, resulting in a 64-dimensional vector after flattening; the fully connected layer consists of two layers with the number of neurons set to 64 and 32 respectively, and each fully connected layer is followed by a rectified linear unit activation function; the output layer uses a rectified linear unit activation function and has the following number of neurons. In this embodiment, Set to 10, with the interval between adjacent time steps set to 1 minute; where, The actual wet basis feed amount for the current batch, in kg, is read and transmitted to the prediction model input terminal by the feeding and weighing module when the feeding is completed; In this embodiment, the feedforward prediction model for the discharge rhythm uses the time-series data of the measured discharge rate for the corresponding batch. To supervise the labeling, the mean squared error of the difference between the predicted values ​​and the corresponding historical measured values ​​at all time steps is used as the loss function; the Adam optimizer is used for parameter updates, with the learning rate set to 0.001, the number of training epochs set to 200, and the batch size set to 8. The time-series predicted value sequence , ,..., When the current batch of roasting process starts, the feedforward prediction calculation is performed and a complete time-series prediction value sequence is generated. Before the material arrives at the packaging station, the prediction value sequence is output to the DXDK granule packaging machine control unit. The control unit then uses the time-series prediction value sequence to pre-set the initial parameters of the target weighing set value and the opening and closing rhythm of the feeding gate.

[0025] Optionally, to continuously correct the deviation between the feedforward predicted value and the actual output during the roasting process, a rolling-triggered online prediction correction mechanism is adopted: whenever the packaging station sensor accumulates data since the last correction trigger moment... When a new valid measured discharge rate reading is obtained, an online prediction correction is triggered; based on the most recent Measured discharge rate values ​​at each time step arithmetic mean Corresponding Feedforward predictions at each time step arithmetic mean difference As a correction bias, this correction bias Updated using a complete replacement method, without superimposing on previous bias corrections; for all feedforward predictions not yet reached after the current trigger time, according to... Perform translation correction. The function selects the larger of the two numbers as the corrected prediction. Replace the original feedforward prediction The parameters used for subsequent preset updates of the DXDK granule packaging machine control unit; historical time step predictions that have already been reached or applied are not affected by this correction; in this embodiment... Set it to 5.

[0026] B5: Using the predicted discharge rate as a feedforward benchmark, the actual discharge rate measurements collected by the packaging station sensors are synchronously fused. Through a feedforward-feedback fusion error-driven adaptive control strategy, the target weighing setpoint and the gate opening / closing rhythm parameter vector of the DXDK granule packaging machine are continuously updated. This ensures dynamic matching between the packaging process speed and the front-end roasting discharge rate, achieving a closed-loop linkage control from material preparation parameters, roasting parameters to packaging parameters, with discharge rate matching as the objective. This includes: A discharge rate sensor is deployed at the packaging station. This sensor combines a weighing sensor and a timing module. The weighing sensor collects the weight of material flowing through the packaging station per unit time, and the timing module records the corresponding time interval. The two sensors are combined to calculate the actual discharge rate measurement. The actual discharge rate measurement is continuously collected at a fixed sampling interval. In this embodiment, the fixed sampling interval is set to 1 minute. The actual discharge rate measured at each time step is recorded as follows: ; At each time step Perform feedforward-feedback fusion computation: when At that time, the packaging station had not yet obtained a valid measured value of the discharge rate, so the feedforward weighting coefficient was set to... Feedback weighting coefficient Feedforward prediction value Directly used as the fusion driving rate ;when At that time, the first Feedforward prediction of discharge rate at each time step Compared with the actual discharge rate measurement value Feedforward weight coefficients updated adaptively based on error With feedback weight coefficient Perform weighted fusion to obtain the fusion-driven rate. The fusion drive rate The calculation formula is: ; in, Indicates the first Fusion driving rate at each time step Indicates the first Adaptive feedforward weight coefficients at each time step Indicates the first The adaptive feedback weight coefficients at each time step, and ; The error-driven adaptive weight update mechanism uses the normalized absolute error of the difference between the measured value and the predicted value at the current time step. Drive the update of the feedback weights at the next time step. The calculation formula is: ; No. Adaptive feedback weights at each time step Calculate using the following formula, such as Figure 2 As shown: ; ; in, Indicates the first Normalized absolute prediction error at each time step; To prevent small quantities with a denominator of zero; Indicates will Cut off to interval ; This represents the initial value of the feedback weights; Indicates the adaptive step size coefficient; and These are the lower and upper limits of the feedback weight, respectively; Indicates the first The adaptive feedforward weight coefficients for each time step; in this embodiment... Set to 0.6, Set to 1.0, Set to 0.3, Set to 0.9, Set to 0.01 kg per minute; Driven by fusion rate As the driving force, the target weighing setpoint of the DXDK granule packaging machine is updated according to the offline calibrated weighing setpoint-rate mapping function. Update the opening and closing rhythm parameter vector of the feeding gate according to the offline calibrated feeding gate rhythm-rate mapping function. ,in Indicates the first The duration of the gate opening within each feeding cycle at each time step. Indicates the first The duration of gate closure within each feeding cycle at each time step; in this embodiment, the offline calibrated weighing setpoint-rate mapping function is established by recording the corresponding target weighing setpoints under 5 different discharge rate conditions, and stored as a weighing calibration lookup table using piecewise linear interpolation; the offline calibrated feeding gate rhythm-rate mapping function is established by recording the corresponding feeding gate opening and closing rhythm parameter vectors under 5 different discharge rate conditions, and stored as a gate calibration lookup table using piecewise linear interpolation; fused drive rate The index value is input into the corresponding calibration lookup table, and the result is obtained through piecewise linear interpolation. and During the batch roasting process, the fusion calculation and parameter update are repeated at each time step to keep the processing rate of the DXDK granule packaging machine dynamically matched with the front-end roasting output rate.

[0027] It should be noted that the sequence numbers of the above embodiments of the present invention are merely for descriptive purposes and do not represent the superiority or inferiority of the embodiments. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, apparatus, article, or method. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0028] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0029] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for controlling the entire process parameters of tortoise shell glue beads based on data algorithms, characterized in that, Includes the following steps: B1: The initial moisture content and particle size distribution of the tortoise shell glue cubes in the current batch are detected. Based on the influence of the initial moisture content of the wet base on the heat transfer uniformity of the auxiliary materials in the pot and the influence of the non-uniformity of the particle size on the qualified bead formation rate, the two types of characteristics are assigned corresponding weights and weighted fusion is performed to generate the stir-frying compatibility index of the current batch. B2: Taking the stir-fry compatibility index as input, the auxiliary material type is automatically selected based on the partition threshold rules determined by the historical batch auxiliary material selection rules, and the precise auxiliary material dosage is determined based on the piecewise linear interpolation lookup table of the auxiliary material dosage calibrated offline for each auxiliary material type. The auxiliary material type selection result and the precise auxiliary material dosage are output. B3: Using the excipient type selection result and the stir-frying compatibility index as a joint index, query the offline calibrated stir-frying parameter recommendation lookup table, output the quality verification qualified temperature setting value and stirring drive frequency setting value of the current batch through piecewise linear interpolation, and send the parameter combination to the electromagnetic stir-frying machine control system for execution; B4: Use a feedforward prediction model of discharge rhythm trained based on historical batch roasting cycle data to estimate the time sequence distribution of discharge rate in the current batch roasting process in advance, and provide parameter preset basis for packaging equipment before the material actually arrives at the packaging station. B5: Using the time-series prediction value of the discharge rate as the feedforward benchmark, the actual discharge rate measurement value collected by the packaging station sensor is synchronously integrated. Through the feedforward-feedback fusion error-driven adaptive control strategy, the target weighing set value and the opening and closing rhythm parameter vector of the DXDK granule packaging machine are continuously updated. This ensures that the processing rate of the packaging process and the discharge rate of the front-end roasting process are dynamically matched, realizing a linkage control closed loop from material preparation parameters, roasting parameters to packaging parameters with the discharge rate matching as the goal.

2. The method for controlling the entire process parameters of tortoise shell glue beads based on data algorithms according to claim 1, characterized in that, Step B1 includes: The current batch of tortoise shell glue cubes was sampled and tested using a halogen moisture analyzer. The arithmetic mean of the moisture analyzer readings at each sampling point was taken as the initial wet basis moisture content of the current batch of tortoise shell glue cubes. The units are percentages; images of the cut pieces of the current batch of tortoise shell glue samples are acquired using an image acquisition device. The equivalent circular diameter of each particle is extracted using an image processing algorithm as the particle size measurement value. The particle size measurements of each particle are statistically analyzed, and the coefficient of variation of the particle size distribution of the current batch of cut pieces is calculated. The coefficient of variation The standard deviation of the particle size measurement values ​​of each diced particle within the batch is divided by the sample mean of the particle size measurement values. Two independent univariate standardized regression analyses were conducted on historical batch data. The initial moisture content and particle size variation coefficient of each historical batch (standardized by z-score) were used as independent variables, and the qualified bead formation rate of the corresponding batch was used as the dependent variable. Two standardized linear regression models were established, and the absolute values ​​of the standardized regression coefficients were defined as the absolute sensitivity of the initial moisture content of the wet substrate. absolute sensitivity to particle size variation coefficient ;when and When the sum is greater than zero, and The characteristic weighting coefficients of water content were obtained after linear normalization. Weighting coefficient of particle size inhomogeneity ,in , ;when and When the sum equals zero, let The and satisfy ; Based on the initial moisture content of the wet basis With particle size variation coefficient The raw material characteristic input is the minimum initial moisture content on a wet basis from historical batch records. With the maximum value right Range normalization was performed using the minimum value of the particle size variation coefficient from historical batch records. With the maximum value right Perform range normalization; when At that time, The range normalization result is taken as 0.5; when At that time, The range normalization result is taken as 0.5; for the range normalization result... and Cut off to An interval is used to handle situations where the characteristics of the current batch exceed the historical range. The normalized value of the initial moisture content of the wet basis after truncation is denoted as... The normalized value of the truncated particle size variation coefficient is denoted as ; Weighting coefficient based on water content characteristics Weighting coefficient of particle size inhomogeneity Weighted fusion generates the current batch's trading suitability index. The stir-frying suitability index The calculation formula is: ； in, This indicates the stir-frying suitability index for the current batch. .

3. The method for controlling the entire process parameters of tortoise shell glue beads based on data algorithms according to claim 1, characterized in that, Step B2 includes: Based on historical batch data, historical batches are categorized according to the stir-frying suitability index. Sort in ascending order and extract batches that have used clam powder and passed quality verification. Numerical set and batches that use talc powder and have passed quality verification Numerical set; select qualified batches using talcum powder. The maximum value in the set of values ​​is used as the threshold for the next partition. Batch of products using qualified clam powder were sampled. The minimum value of the set of values ​​is used as the threshold for the upper partition. ; Before officially enabling the threshold partitioning rule, and Perform a one-time configuration validity check: If If the threshold configuration is valid, then the threshold configuration is effective; if If the threshold configuration is invalid, a configuration error alarm will be issued, prompting the operator to manually specify the threshold based on current process knowledge. Valid threshold pairs; After confirmation After its establishment, the operational period will be used to adapt the current batch of stir-frying index. The following three-zone judgment rules will be uniformly applied, without distinguishing any additional circumstances: If ,but Talc; if ,but Clam powder; if If the system sends a confirmation request for the type of artificial auxiliary material to the operation console, it will pause the automatic decision-making process and wait for operator confirmation. Then proceed with the next steps.

4. The method for controlling the entire process parameters of tortoise shell glue beads based on data algorithms according to claim 3, characterized in that, Step B2 further includes: Separate piecewise linear interpolation lookup tables were created for the dosage of clam powder and talcum powder; each lookup table was based on a stir-fry compatibility index. As an index variable, store The calibration index point, the first One calibration index point, Includes the calibrated stir-fry adaptation index Corresponding auxiliary material dosage calibration values Calibration values ​​of auxiliary material usage at calibration index points The data is determined from historical batch records of batches that used the corresponding auxiliary material type and passed quality verification, using the arithmetic mean of the actual auxiliary material usage of qualified batches within the range of each calibration compatibility index; each calibration index point is determined according to... Strictly ascending in order of increasing value; if duplicate values ​​appear in the initially determined calibration and cooking compatibility index, the average value of the actual auxiliary material usage of qualified batches within the corresponding interval will be calculated and replaced by a single calibration index point. The matching index for the current batch of stir-frying ,exist Locating the corresponding auxiliary material type in the lookup table satisfies two adjacent calibration index points and Piecewise linear interpolation is used to calculate the accurate amount of excipients. : ； when Below the minimum index point of the corresponding lookup table At that time, with As Output; when Higher than the maximum index point of the corresponding lookup table At that time, with As The output, i.e., performs boundary saturation processing on the index values ​​of the lookup table; after interpolation and boundary saturation processing are completed, the output is... Cut off the mixing ratio range; in, This indicates the precise amount of excipients used in the current batch. Indicates the first The calibration values ​​of auxiliary material usage at each calibration index point. Indicates the first The calibration values ​​of auxiliary material usage at each calibration index point. Indicates the first The calibration and matching index at each calibration index point. Indicates the first The calibration and matching index at each calibration index point.

5. The method for controlling the entire process parameters of tortoise shell glue beads based on data algorithms according to claim 1, characterized in that, Step B3 includes: Separate lookup tables for recommended stir-frying parameters were created for clam powder and talcum powder; each lookup table is based on a stir-frying suitability index. As an index variable, store The calibration index point, the first One calibration index point, Includes calibrated stir-frying compatibility index Recommended temperature calibration values Recommended calibration value for stirring drive frequency Recommended calibration value for temperature at the calibration index point. Recommended calibration value for stirring drive frequency From historical batch records of batches using the corresponding auxiliary material type and confirmed as qualified through quality verification, the arithmetic mean of the actual frying temperature records and actual stirring drive frequency records of qualified batches within the range near each calibrated frying compatibility index is used to determine the index points; each calibration index point is determined according to... The values ​​are arranged in a strictly ascending order. If duplicate values ​​appear in the initially determined calibration roasting compatibility index, the average of the actual temperature and frequency records of qualified batches within the corresponding interval is calculated and replaced by a single calibration index point. The distribution range of the roasting compatibility index of the calibration index point should cover the usage of the corresponding auxiliary material type in historical batches. Numerical range; The matching index for the current batch of stir-frying ,in accordance with Select the recommended lookup table for the stir-frying parameters corresponding to the type of auxiliary material, and locate the parameter that meets the requirements in the selected lookup table. two adjacent calibration index points and The temperature setpoint was calculated using piecewise linear interpolation. With stirring drive frequency setting value : ； ； when Below the minimum index point of the selected lookup table At that time, with and As respectively and Output; when Higher than the maximum index point of the selected lookup table At that time, with and As respectively and The output, i.e., performs boundary saturation processing on the index values ​​of the lookup table; after interpolation and boundary saturation processing are completed, the output is... Implement the recommended temperature cutoff range for stir-frying to ensure Falling Within Celsius; for Perform a recommended range cutoff for the stirring drive frequency to ensure Falling Within Hz; in, This indicates the temperature setting value for the current batch. This indicates the stirring drive frequency setting for the current batch. and They represent the first With the Recommended calibration values ​​for temperature at each calibration index point. and They represent the first With the Recommended calibration values ​​for the stirring drive frequency at each calibration index point. and They represent the first With the The calibration cooking adaptation index at each calibration index point; Temperature setting With stirring drive frequency setting value The parameters are combined into a roasting parameter package, which is written into the parameter register of the electromagnetic roasting machine control system via an industrial communication protocol. The electromagnetic roasting machine control system then adjusts the heating power and stirring drive frequency according to the roasting parameter package. The control value for the stirring speed is sent to the frequency converter drive unit of the electromagnetic stir-frying machine.

6. The method for controlling the entire process parameters of tortoise shell glue beads based on data algorithms according to claim 1, characterized in that, Step B4 includes: Construct a feedforward prediction model for the discharge rhythm, wherein the feedforward prediction model for the discharge rhythm uses a temperature setpoint. Stirring drive frequency setting value Precise dosage of excipients Binary coding of auxiliary material types Talc powder corresponds to code 0, clam powder corresponds to code 1, initial moisture content on wet basis. With particle size variation coefficient Stir-frying suitability index and the actual amount of material fed into the current batch. The concatenated 8-dimensional feature vector serves as input. A one-dimensional convolutional layer extracts local correlation features between the components of the input parameters. A max-pooling layer compresses these local correlation features. A fully connected layer maps the compressed feature vector to the temporal output space, outputting a sequence of predicted output rates arranged by time step index. The temporal prediction sequence contains a total of... The time step, the first The feedforward prediction of the discharge rate at each time step is denoted as . , ; The time-series predicted value sequence , ,..., When the current batch of roasting process starts, feedforward prediction calculation is performed and a complete time-series prediction value sequence is generated. Before the material arrives at the packaging station, the prediction value sequence is output to the DXDK granule packaging machine control unit. The DXDK granule packaging machine control unit pre-sets the initial parameters of the target weighing set value and the opening and closing rhythm of the feeding gate based on the time-series prediction value sequence.

7. The method for controlling the entire process parameters of tortoise shell glue beads based on data algorithms according to claim 1, characterized in that, Step B5 includes: A discharge rate sensor is deployed at the packaging station. This sensor combines a weighing sensor and a timing module. The weighing sensor collects the weight of material flowing through the packaging station per unit time, and the timing module records the corresponding time interval. The two sensors are combined to calculate the actual discharge rate measurement. The actual discharge rate measurement is continuously collected at fixed sampling intervals. The actual discharge rate measured at each time step is recorded as follows: ; At each time step Perform feedforward-feedback fusion computation: when At that time, the packaging station had not yet obtained a valid measured value of the discharge rate, so the feedforward weighting coefficient was set to... Feedback weighting coefficient Feedforward prediction value Directly used as the fusion driving rate ;when At that time, the first Feedforward prediction of discharge rate at each time step Compared with the actual discharge rate measurement value Feedforward weight coefficients updated adaptively based on error With feedback weight coefficient Perform weighted fusion to obtain the fusion-driven rate. The fusion drive rate The calculation formula is: ； in, Indicates the first Fusion driving rate at each time step Indicates the first Adaptive feedforward weight coefficients at each time step Indicates the first The adaptive feedback weight coefficients at each time step, and ; The error-driven adaptive weight update mechanism uses the normalized absolute error of the difference between the measured value and the predicted value at the current time step. Drive the update of the feedback weights at the next time step. The calculation formula is: ； No. Adaptive feedback weights at each time step Calculate using the following formula: ； ； in, Indicates the first Normalized absolute prediction error at each time step; To prevent small quantities with a denominator of zero; Indicates will Cut off to interval ; This represents the initial value of the feedback weights; Indicates the adaptive step size coefficient; and These are the lower and upper limits of the feedback weight, respectively; Indicates the first The adaptive feedforward weight coefficients for each time step.

8. The method for controlling the entire process parameters of tortoise shell glue beads based on data algorithms according to claim 7, characterized in that, Step B5 further includes: Driven by fusion rate As the driving force, the target weighing setpoint of the DXDK granule packaging machine is updated according to the offline calibrated weighing setpoint-rate mapping function. Update the opening and closing rhythm parameter vector of the feeding gate according to the offline calibrated feeding gate rhythm-rate mapping function. ,in Indicates the first The duration of the gate opening within each feeding cycle at each time step. Indicates the first The duration of gate closure within each feeding cycle at each time step; the offline calibrated feeding gate rhythm-rate mapping function is established by recording the corresponding feeding gate opening and closing rhythm parameter vectors under 5 different feeding rate conditions, and stored as a gate calibration lookup table using piecewise linear interpolation; fused drive rate The index value is input into the corresponding calibration lookup table, and the result is obtained through piecewise linear interpolation. and During the batch roasting process, the fusion calculation and parameter update are repeated at each time step to keep the processing rate of the DXDK granule packaging machine dynamically matched with the front-end roasting output rate.

Images