A carton production dynamic adjustment control system and a carton production dynamic adjustment method thereof

Abstract

The present application relates to the technical field of intelligent industrial manufacturing and automation, in particular to a carton production dynamic adjustment control system and a carton production dynamic adjustment method thereof, which solves the problems of parameter lag, process coupling interference and quality traceability difficulty in traditional carton production through multi-source data sensing, edge intelligent calculation, hybrid model prediction and cross-process collaborative control.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent industrial manufacturing and automation, more particularly, it relates to a carton production dynamic adjustment control system and a carton production dynamic adjustment method thereof. BACKGROUND

[0002] Cartons play a key role in daily life and industrial production, such as express delivery and cargo transportation; in traditional carton production, raw paperboard is pressed and bent to form a carton; during the manufacturing process, errors may occur in carton production due to fluctuations in raw materials, such as grammage, moisture content, and environmental factors, resulting in defective products, so production parameters need to be adjusted to overcome these problems; however, in the traditional adjustment process, most rely on manual experience or fixed parameter control, which cannot respond in real time, and there is dynamic coupling between the processes of feeding, pressing, and forming, such as the influence of raw material changes on pressing pressure requirements and the influence of pressing temperature on forming mold deformation, so traditional single-process control cannot be decoupled, resulting in unsatisfactory adjustment results.

[0003] Therefore, a carton production dynamic adjustment control system and method based on multi-modal data fusion and hybrid model prediction are proposed, which are suitable for intelligent control of corrugated paperboard production lines, carton forming machines, and other packaging equipment. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application aims to provide a carton production dynamic adjustment control system and a carton production dynamic adjustment method thereof to solve the problems in the background art.

[0005] The above technical purpose of the present application is achieved by the following technical solution: a carton production dynamic adjustment control system, comprising:

[0006] A data acquisition module is provided in each process of the carton production process for collecting multi-modal production parameters in the feeding process, pressing process, forming process, drying process, and quality inspection process and transmitting them.

[0007] An edge computing module is used to receive the multi-modal production parameters collected by the data acquisition module, perform data preprocessing, and then use a feature extraction algorithm to sequentially obtain the feeding model input, pressing model input, forming model input, drying model input, and quality inspection model input and transmit them.

[0008] A hybrid model module is used to receive the model inputs transmitted by the edge computing module, input them into the corresponding process model, and then output the model prediction results and transmit them.

[0009] An instruction calculation module is configured to receive the multi-modal production parameters collected by the data collection module and the model prediction result output by the hybrid model module, dynamically generate an adjustment instruction corresponding to a process after comprehensive judgment, and transmit the adjustment instruction.

[0010] A cross-process execution module is configured to receive the adjustment instruction transmitted by the instruction calculation module, and dynamically adjust the production parameters of the corresponding process before executing the adjustment instruction to overcome the process coupling interference, and then execute the adjustment instruction for adjustment.

[0011] Optionally, the hybrid model module comprises:

[0012] A random forest multi-objective optimization model is configured to receive the feeding model input quantity, predict the influence trend of raw materials and environmental humidity in the feeding process, and generate a feeding prediction result.

[0013] An LSTM-PINN hybrid model is configured to receive the pressing model input quantity, predict the paperboard deformation trend in the pressing process, and generate a pressing prediction result.

[0014] A physics-guided reinforcement learning model is configured to receive the forming model input quantity, predict the influence trend of mold spacing and optimized conveying speed in the forming process, and generate a forming prediction result.

[0015] A multivariable predictive control model is configured to receive the drying model input quantity, predict the drying efficiency and energy consumption trend in the drying process, and generate a drying prediction result.

[0016] A graph neural network knowledge graph model is configured to receive the quality inspection model input quantity, perform defect root cause analysis and quality traceability in the quality inspection process, and generate a quality inspection prediction result.

[0017] The feeding prediction result, the pressing prediction result, the forming prediction result, the drying prediction result, and the quality inspection prediction result are fitted to generate the model prediction result.

[0018] Optionally, the hybrid model module comprises:

[0019] A random forest multi-objective optimization model is configured to receive the feeding model input quantity, predict the influence trend of raw materials and environmental humidity in the feeding process, and generate a feeding prediction result.

[0020] An LSTM-PINN hybrid model is configured to receive the pressing model input quantity, predict the paperboard deformation trend in the pressing process, and generate a pressing prediction result.

[0021] A physics-guided reinforcement learning model is configured to receive the forming model input quantity, predict the influence trend of mold spacing and optimized conveying speed in the forming process, and generate a forming prediction result.

[0022] a multivariable predictive control model, configured to receive the drying model input quantity, predict drying efficiency and energy consumption trend in the drying process, and generate a drying prediction result;

[0023] a graph neural network knowledge graph model, configured to receive the quality inspection model input quantity, perform defect root cause analysis and quality traceability in the quality inspection process, and generate a quality inspection prediction result;

[0024] The feeding prediction result, the pressing prediction result, the forming prediction result, the drying prediction result, and the quality inspection prediction result are fitted to generate the model prediction result.

[0025] Optionally, the instruction calculation module comprises:

[0026] an adjustment threshold calculation unit, configured to dynamically generate an adjustment threshold of each process based on the multi-modal production parameters collected by the data collection module;

[0027] an adjustment determination calculation module, configured to compare the feeding prediction result, the pressing prediction result, the forming prediction result, the drying prediction result, and the quality inspection prediction result with the adjustment threshold of the corresponding process, and then determine whether adjustment is needed;

[0028] an adjustment target value calculation module, configured to calculate a corresponding adjustment target value of the process that needs to be adjusted based on the determination result of the adjustment determination calculation module, and generate the adjustment instruction based on the generated adjustment target value.

[0029] Optionally, the cross-process execution module comprises:

[0030] an instruction receiving unit, configured to receive the adjustment instruction transmitted by the instruction calculation module and parse the adjustment target value of the corresponding process;

[0031] a multi-process coupling modeling unit, configured to establish a process dynamic model of the dynamic relationship between the multi-processes, and based on the adjustment target value, fine-tune the process parameters in advance according to the output result of the process dynamic model to offset the adjustment interference;

[0032] a program execution module, configured to adjust the corresponding process equipment based on the adjustment target value parsed by the instruction receiving unit;

[0033] a fine-tuning execution module, configured to detect the production parameters of the adjusted process in real time, and compare the production parameters of the process with a trigger target threshold to perform fine-tuning.

[0034] Optionally, the trigger target threshold comprises a prediction result deviation value, a process coupling interference trigger value, a process safety limit value, and a quality feedback trigger value.

[0035] Optionally, the data acquisition module comprises:

[0036] a data sensing unit arranged on each process for collecting multi-modal data;

[0037] a data alignment unit for binding and marking the multi-modal data collected by the data sensing unit with a time sequence for primary time synchronization;

[0038] a data transmission unit for transmitting the multi-modal data processed by the data alignment unit.

[0039] Optionally, the multi-modal production parameters comprise mechanical parameters, optical parameters, environmental parameters and material parameters.

[0040] Optionally, the edge computing module comprises:

[0041] a data receiving unit for receiving the multi-modal data collected by the data acquisition module;

[0042] a data preprocessing unit for data cleaning and preprocessing of the multi-modal data to remove noise and invalid data, and for time synchronization alignment to realize secondary time synchronization;

[0043] a feature extraction unit for extracting the feeding model input, the pressing model input, the forming model input, the drying model input and the quality inspection model input from the multi-modal data by using a feature extraction algorithm;

[0044] a feature input unit for inputting the model input extracted by the feature extraction unit to the hybrid model module.

[0045] Optionally, it further comprises a self-healing safety module for integrating a blockchain instruction storage unit and a three-level fault degradation control mechanism to respond to sudden accidents.

[0046] A carton production dynamic adjustment method based on the above-mentioned carton production dynamic adjustment control system, comprising:

[0047] Step A, multi-modal data acquisition: based on the data acquisition module arranged in the feeding process, pressing process, forming process, drying process and quality inspection process, multi-modal production parameters in normal production are collected;

[0048] Step B, model input extraction: based on the feature extraction algorithm, the feeding model input, the pressing model input, the forming model input, the drying model input and the quality inspection model input are extracted after the multi-modal production parameters are preprocessed by the edge computing module;

[0049] Step C, process prediction result acquisition: based on the loading model input, the pressing model input, the forming model input, the drying model input and the quality inspection model input, the corresponding inputs are input into the mixed model module, and then the model prediction result is obtained;

[0050] Step D, adjustment target value acquisition: the instruction calculation module based on the model prediction result of each process corresponding to step C is used for comparison with the adjustment threshold of each process dynamically generated, to confirm whether adjustment is needed, and then the adjustment target value of the process to be adjusted is calculated and generated, and the adjustment instruction is generated based on the corresponding generated adjustment target value;

[0051] Step E, adjustment instruction execution: based on the output result of the process dynamic model, the production parameter adjustment of the corresponding process equipment is performed to overcome the process coupling interference, and the adjustment instruction generated by the instruction calculation module is executed to adjust the production equipment of each process;

[0052] Step F, target perfecting fine tuning: the cross-process execution module monitors the adjusted production equipment, and when the trigger target threshold is reached, the production equipment is fine tuned again to perfect the adjustment effect.

[0053] In summary, the present application has the following beneficial effects:

[0054] 1. By setting the data acquisition module on each process, the production parameters in each process during the carton production process can be effectively collected, involving mechanical parameters, optical parameters, environmental parameters and material parameters, which are input into the mixed model module, and the model output result corresponding to different processes is output respectively to represent the carton production trend in the future period, and is transmitted to the instruction calculation module to calculate the adjustment target value of the process to be adjusted and generate the adjustment instruction to adjust the corresponding process equipment, so as to ensure the yield of the carton production.

[0055] 2. Before the adjustment process of each process by the cross-process execution module, in order to avoid mutual interference between processes during the adjustment process, online modeling and real-time decoupling control of multi-process dynamic coupling interference are adopted to further perfect the adjustment process and achieve better adjustment effect. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 is the system execution flowchart of the present application;

[0057] Figure 2 is the system execution method step flowchart of the present application. DETAILED DESCRIPTION

[0058] In order to make the objects, characteristics and advantages of the present application more apparent, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. The present application is shown in several embodiments in the drawings. However, the present application can be realized in many different forms and is not limited to the embodiments described herein.

[0059] In the present application, unless specifically defined and limited otherwise, the terms "mounting", "connection", "connecting", "fixed", and the like should be interpreted broadly, for example, can be fixedly connected, can be detachably connected, or integrally connected; can be mechanically connected, or electrically connected; can be directly connected, or indirectly connected through an intermediate medium, or can be internal communication of two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances. The terms "first", "second" are only for the purpose of description, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features limited by "first", "second" can explicitly or implicitly include one or more of the features.

[0060] In the present application, unless specifically defined and limited otherwise, "on" or "under" of the first feature to the second feature can include that the first and second features are in direct contact, or that the first and second features are not in direct contact but are in contact through another feature between them. Moreover, "on", "above" and "on" of the first feature to the second feature includes that the first feature is directly above and obliquely above the second feature, or only indicates that the horizontal height of the first feature is higher than that of the second feature. "Below", "under" and "under" of the first feature to the second feature includes that the first feature is directly below and obliquely below the second feature, or only indicates that the horizontal height of the first feature is less than that of the second feature. The terms "vertical", "horizontal", "left", "right", "up", "down" and similar expressions are only for the purpose of description, and are not indicative or suggestive of the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application.

[0061] The present application will be described in detail below with reference to the accompanying drawings and examples.

[0062] The present application provides a carton production dynamic adjustment control system, as shown in Figure 1 The present application provides a carton production dynamic adjustment control system, as shown in

[0063] The data acquisition module is arranged at each process in the carton production process, and is used for acquiring and transmitting multi-modal production parameters in the feeding process, the pressing process, the forming process, the drying process and the quality inspection process.

[0064] An edge computing module is configured to receive the multi-modal production parameters collected by the data collection module, perform data preprocessing, and then use a feature extraction algorithm to sequentially obtain the feeding model input, the pressing model input, the forming model input, the drying model input, and the quality inspection model input, and transmit them.

[0065] A hybrid model module is configured to receive the model inputs transmitted by the edge computing module, input them into the corresponding process model, and then output and transmit the model prediction results.

[0066] An instruction calculation module is configured to receive the multi-modal production parameters collected by the data collection module and the model prediction results output by the hybrid model module, dynamically generate adjustment instructions for the corresponding process after comprehensive judgment, and transmit them.

[0067] A cross-process execution module is configured to receive the adjustment instructions transmitted by the instruction calculation module, and according to the adjustment instructions, dynamically adjust the production parameters of the corresponding process before executing the adjustment instructions to overcome the process coupling interference, and then execute the adjustment instructions for adjustment.

[0068] Further, the hybrid model module comprises:

[0069] A random forest multi-objective optimization model is configured to receive the feeding model input, predict the influence trend of raw materials and environmental humidity in the feeding process, and generate a feeding prediction result.

[0070] An LSTM-PINN hybrid model is configured to receive the pressing model input, predict the paperboard deformation trend in the pressing process, and generate a pressing prediction result.

[0071] A physics-guided reinforcement learning model is configured to receive the forming model input, predict the influence trend of mold spacing and optimized conveying speed in the forming process, and generate a forming prediction result.

[0072] A multivariable predictive control model is configured to receive the drying model input, predict the drying efficiency and energy consumption trend in the drying process, and generate a drying prediction result.

[0073] A graph neural network knowledge graph model is configured to receive the quality inspection model input, perform defect root cause analysis and quality traceability in the quality inspection process, and generate a quality inspection prediction result.

[0074] The feeding prediction result, the pressing prediction result, the forming prediction result, the drying prediction result, and the quality inspection prediction result are fitted to generate the model prediction result.

[0075] In Example 1, the random forest multi-objective optimization model uses single-tree prediction and multi-objective output. The model inputs are represented as follows: raw material characteristics: raw material weight, raw material moisture content, fiber orientation angle; environmental parameters: environmental humidity, environmental temperature; historical operating conditions: historical average energy consumption, raw material switching frequency; the above model inputs are all aligned with data of different sampling frequencies through dynamic time warping (DTW).

[0076] The output is represented as: Single Tree Prediction Model This can be represented as follows: for the t-th tree, the input sample x i The predicted value, where R m Represented as the leaf node region, c m Represented as node output values, this model can simultaneously predict three target variables: raw material cost C, energy consumption E, and equipment wear W. At the same time, triples are stored on each leaf of each tree.

[0077] The calculated multi-objective variable values ​​are then optimized using a multi-objective optimization formula, expressed as follows: Where u = [u1, u2, u3] are control variables, where u1 represents the proportion of raw material A, u2 represents the humidity compensation coefficient, and u3 represents the feeding speed. Conditional constraints are also used to optimize the results, represented as: equipment safety constraints, 0.9P. 额定 ≤P(u1,u3)≤1.1P 额定 Mass constraint: predicted edge compression strength ≥ 7 kN / m³; Process constraint: |u²⁻¹| ≤ 0.2·(H) 实时 -50%);

[0078] Based on the above target values, the raw material cost trend, energy consumption trend, and equipment wear trend are output respectively. The raw material cost trend is represented as ΔC = 0.15u1 - 0.08u3, and the energy consumption trend is represented as ΔE = 1.2u3 + 0.05H. 2 The wear trend of the equipment is represented as This leads to the output of the material feeding prediction result;

[0079] The LSTM-PINN hybrid model uses an LSTM network to predict the shape variables of cardboard during future production cycles. The model input consists of multi-dimensional time series data, including real-time average pressure, pressure fluctuation variance, energy percentage of vibration signal from 1 to 5 kHz, temperature gradient, and historical deformation sequence.

[0080] The output is represented as h through the LSTM network structure. t =LSTM(x t ,h t-1 ;θLSTM ), where the input gate / forget gate / output gate are parameterized, the time window length = 60s, the hidden layer dimension = 128, and the physical information constraint is added simultaneously, denoted as where E represents the elastic modulus, μ represents the damping coefficient, and a hybrid loss function is used to optimize the prediction results, denoted as where represents the data fitting term, represents the physical residual term, and F represents the mechanical equation residual operator, where the elastic modulus E is updated by online maximum likelihood estimation, denoted as and then the curvature radius change ΔR and the thickness deviation prediction value in the future period are obtained where t1 represents the future period time, and then the pressing prediction result is output;

[0081] The physical guided reinforcement learning model uses state space-action space-reward function to predict the influence trend of the mold gap and optimize the conveying speed, the input is a multi-dimensional state vector, including the pressing mold gap, the real-time conveying speed, the paperboard stiffness coefficient, the environmental humidity and the historical loss;

[0082] The output is: inputting the multi-dimensional state vector into the model, the state space is denoted as s t = d 模具 ,v 传送 ,S 挺度 ,H 湿度 ,E 能耗 ], the action space is denoted as a t = [Δd 模具 ∈-0.1,0,+0.1}mm,Δv 传送 ∈{-0.5,0,+0.5}m / min], and the reward function is denoted as R t = 0.6·QualityScore + 0.3·Throughput - 0.1·EnergyCost, where the quality score is based on the size deviation and the crease angle; and then the optimal mold gap and the optimal conveying speed are calculated, and the pressing prediction result is output;

[0083] The multivariable predictive control model uses a state space model to predict the drying efficiency and energy consumption trend in the drying process, the input is a multi-dimensional process variable, including the hot air temperature, the exhaust fan speed, the paperboard moisture content, the instantaneous energy consumption, and the environmental dew point temperature.

[0084] The output is: inputting the multi-dimensional process variable into the state space model, denoted as x k+1 = Ax k + Bu k + w k, y k = Cx k + v k , where state variable x = [T 纸板 , H 含水 , P 能耗 ] T , control variable u = [Delta T 热风 , Delta RPM 风机 ] T , while using the optimization formula, expressed as a rolling horizon optimization: where the prediction horizon N p = 10, the control horizon N c = 5, the weight matrix Q = diag(0.7, 0.3), R = diag(0.2, 0.8); and further calculation obtains the optimal hot air temperature and the optimal exhaust fan speed, and outputs the drying prediction result;

[0085] The graph neural network knowledge graph model performs defect root cause analysis and quality traceability through a graph convolution network, and the input is multi-dimensional defect features, including: defect type, position coordinates, size deviation, edge pressure intensity, and associated process parameter snapshots;

[0086] The output is: inputting multi-dimensional defect features into a graph convolution network, expressed as an adjacency matrix including the association between equipment, processes, and defects, node features H (0) contain historical case data, while using a multi-task learning objective, expressed as wherein: classification task: defect type identification (cross-entropy loss), regression task: root cause probability prediction (mean square error), graph reconstruction task: knowledge graph completion (contrastive loss); and further outputting defect phenomenon and defect root cause probability distribution analysis, and performing quality traceability, associating past 30-day similar defect cases, generating an SPC control chart, identifying abnormal process parameters, and further outputting quality inspection prediction results;

[0087] Further, the instruction calculation module comprises:

[0088] An adjustment threshold calculation unit dynamically generates adjustment thresholds for each process based on the multi-modal production parameters collected by the data acquisition module;

[0089] An adjustment determination calculation module compares the feeding prediction result, the pressing prediction result, the forming prediction result, the drying prediction result, and the quality inspection prediction result with the adjustment thresholds of the corresponding processes, and further determines whether adjustment is needed;

[0090] An adjustment target value calculation module calculates a corresponding adjustment target value of the process to be adjusted based on the determination result of the adjustment determination calculation module, and generates the adjustment instruction based on the generated adjustment target value.

[0091] In Embodiment Two, the adjustment threshold calculation unit is based on the multi-modal production parameters collected by the data collection module in the current carton production; in the feeding process, the calculation of the adjustment threshold is represented as the grammage fluctuation threshold: δ = μ ± 3σ, and the adjustment of the feeding process is triggered when the real-time grammage deviation is greater than the grammage fluctuation threshold or the energy consumption exceeds the limit determined by the adjustment determination module.

[0092] In the pressing process, the adjustment threshold is the standard curvature radius change rate and the upper limit of the allowable thickness deviation, and the adjustment of the pressing process is triggered when the curvature radius change rate or the thickness deviation value is greater than the adjustment threshold determined by the adjustment determination module, which can be manually set or automatically generated by the system according to historical work records;

[0093] In the forming process, the adjustment threshold is the size deviation threshold, which is represented as the precision of the order product * the yield, and the adjustment of the forming process is triggered when the real-time size deviation is greater than the size deviation threshold determined by the adjustment determination module, or the model confidence calculated by the hybrid model module is less than 85%;

[0094] In the drying process, the adjustment threshold is the energy consumption safety threshold, which is represented as 0.9P 额定 (1-0.1·vibration level), and the adjustment of the drying process is triggered when the instantaneous energy consumption in the drying process is greater than the energy consumption safety threshold determined by the adjustment determination module, or the moisture content deviation is greater than 2% for five minutes;

[0095] In the quality inspection process, the SPC control line formula between processes is represented as UCL = μ + 3σ, LCL = μ - 3σ, and then the defect probability and the defective probability are obtained, and the adjustment of the quality inspection process is triggered when the adjustment determination module determines that there are two consecutive defective products or the defect probability is greater than 90%;

[0096] In the adjustment calculation of the feeding process, the feeding prediction result output by the hybrid model module is used to generate a Pareto solution set based on the NSGA-III algorithm, generate three different adjustment mode schemes of efficiency, economy, and comprehensiveness, and then generate the adjustment target value of the feeding process based on the results of system preset or manual selection and control;

[0097] In the adjustment calculation of the pressing process, the predicted curvature deviation is represented as ΔR 偏差 = ΔR - R set , and the thickness deviation is represented as Δh 偏差 = Δh t+10 -h max, the pressure compensation base value is expressed as ΔP based on PID regulation amount calculation base = K p · Δh 偏差 , the integral term (historical deviation accumulation) is expressed as The differential term (change trend suppression) is expressed as The total pressure compensation amount ΔP is calculated based on the above-mentioned comprehensive calculation ID , and the adjustment target value of the pressing process is generated;

[0098] In the adjustment calculation of the forming process, the optimal mold spacing and optimal conveying speed obtained by the physical guide reinforcement learning model are used to generate the adjustment target value of the forming process;

[0099] In the adjustment calculation of the drying process, the optimal hot air temperature and optimal exhaust fan speed obtained by the multivariable predictive control model are used to generate the adjustment target value of the drying process;

[0100] In the adjustment calculation of the quality inspection process, the defect phenomenon and defect root cause probability distribution analysis are output by the graph neural network knowledge graph model, and quality traceability is performed. The same defect cases in the past 30 days are associated to generate an SPC control chart to identify abnormal process parameters. The type of adjustment target value is a strategy distribution, for example: the defect phenomenon is insufficient edge pressure, and the root cause probability distribution is 72% for low pressing pressure, 72% for low pressing pressure, and 10% for raw material weight deviation. The corresponding adjustment strategy is to increase the pressure compensation by 0.2 MPa, increase the drying temperature by 8°C, and adjust the raw material ratio. The management personnel adjusts the corresponding adjustment strategy distribution to improve the finished product.

[0101] Further, the cross-process execution module comprises:

[0102] An instruction receiving unit configured to receive the adjustment instruction transmitted by the instruction calculation module and parse the adjustment target value of the corresponding process;

[0103] A multi-process coupling modeling unit configured to establish a process dynamic model of the dynamic relationship between multiple processes. Based on the adjustment target value, the output result of the process dynamic model is used to fine-tune the process parameters in advance to offset the adjustment disturbance;

[0104] A program execution module configured to adjust the corresponding process equipment based on the adjustment target value parsed by the instruction receiving unit;

[0105] A fine-tuning execution module configured to detect the production parameters of the adjusted process in real time, and compare the production parameters of the process with the trigger target threshold to perform fine-tuning.

[0106] In the embodiment three, during the execution of each process, the multi-process coupling modeling unit first establishes a process dynamic model of the dynamic relationship between the multi-processes, which is expressed by a transfer function matrix as Wherein G ij (s) represents the transfer function from the jth process to the ith process, i, j = 1, 2, ΔQ i represents the output quantity deviation (such as the pressing thickness, forming accuracy) of the ith process, ΔP j represents the input quantity change (such as the feeding pressure, conveying speed) of the jth process. In the specific implementation process, when the interference source process is the feeding process and the interfered process is the pressing process, the interference represents the change of the pressing pressure demand caused by the fluctuation of the raw material gram weight, and the transfer function is When the interference source process is the pressing process and the interfered process is the forming process, the interference represents the influence of the pressing temperature change on the thermal deformation of the forming mold, and the transfer function is When the interference source process is the forming process and the interfered process is the quality inspection process, the interference represents the deviation of the crease angle caused by the excessive forming speed, and the transfer function is

[0107] The decoupling controller is designed by using inverse model feedforward compensation, which is expressed as Wherein T f = 0.5s is the filter time constant, the parameter is adjusted, the stability is ensured by frequency domain response matching, the gain margin is > 6db, and the phase margin is > 45°, and a feedback correction loop is used, a Smith predictor is used to compensate for the large time delay system, which is expressed as τ = 0.2s represents the time delay time, and T = 1.8s represents the time constant, and the decoupling command is obtained by the above steps;

[0108] In the specific embodiment, taking feeding and pressing as an example, the interference prediction is expressed as ΔQ pred = G ij @Δu j , which represents that the feeding speed is increased by 0.5m / min, and the predicted pressing pressure needs to be increased by 0.075MPa, the decoupling command is generated, the feedforward compensation amount is calculated: Δu 补偿 = C ff (s)·ΔQ pred , the calculation result generates the decoupling command, which is executed in the corresponding process, and then the process parameters are fine-tuned in advance to offset the adjustment interference.

[0109] Further, the trigger target threshold value comprises a prediction result deviation value, a process coupling interference trigger value, a process safety limit value and a quality feedback trigger value.

[0110] In other embodiments, after the adjustment is completed, a judgment is made, and the judgment result is compared with the predicted result deviation value, the process coupling interference trigger value, the process safety limit value and the quality feedback trigger value, and the process safety limit value, the quality feedback trigger value, the predicted result deviation value and the process coupling interference trigger value are in turn prioritized from high to low; when the process safety limit value is triggered, the machine is immediately stopped; when the quality feedback trigger value is triggered, the production is suspended and an alarm is given; when the predicted result deviation value is triggered, parameter compensation is started to make fine adjustments; when the process coupling interference trigger value is triggered, feedforward decoupling is performed; when multiple thresholds are triggered at the same time, the highest priority action is performed;

[0111] The predicted result deviation value is obtained by analyzing historical data, counting the production data of the past 6 months, calculating the 3σ range of the model prediction error (confidence of 99.7%), and dynamically setting according to the confidence interval of the predicted result (such as the probability value output by LSTM). If the model confidence is ≥ 90%, the threshold is relaxed by 10%; if the confidence is ≤ 70%, the threshold is tightened by 20%, to calibrate the model confidence;

[0112] A dynamic adjustment strategy and parameter fine-tuning are used to deal with threshold errors. The error statistics are recalculated every 24 hours (window size = recent 1000 batches of data); when the prediction error exceeds the threshold for 10 consecutive times, the model is triggered for rapid fine-tuning, and the threshold is updated simultaneously;

[0113] The process coupling interference trigger threshold is obtained by transfer function identification, and the inter-process coupling gain is measured by step response experiment. The dynamic coupling degree is calculated and represented as When C ij ≥ 30% is determined as significant interference;

[0114] The threshold is optimized and adjusted by a dynamic adjustment strategy, including online transfer function update: recursive least squares (RLS) is used to re-identify G ij every ten minutes, and the coupling degree threshold is updated, represented as At the same time, during the changeover stage, the threshold is temporarily reduced to 20% to improve sensitivity and realize production mode adaptation to adjust the normal operation between processes affected by coupling interference;

[0115] The process safety limit threshold is obtained by setting a hard safety value through the rated parameters of the equipment, determining the limit temperature threshold that the material can tolerate through the paperboard material convention, and adjusting the threshold to better meet the actual requirements of actual work through equipment health compensation and environmental adaptability adjustment;

[0116] The method for obtaining the quality feedback triggering threshold value sets the size deviation threshold value and the edge pressure strength threshold value through the process standard of the actual production specification, and controls the control line (UCL / LCL) through statistical process control calculation, UCL=μ+3σ; when optimizing the threshold value, the order strictness is matched with the high-precision order (such as a medicine packaging), the threshold value is tightened by 50%, defect mode recognition is adopted, and if a specific defect (such as glue opening) is detected, the related threshold value is temporarily adjusted by 30% to realize fine adjustment and matching of the threshold value.

[0117] Further, the data acquisition module comprises:

[0118] A data sensing unit is arranged on each process for collecting multi-modal data;

[0119] A data alignment unit is used to bind and mark the multi-modal data collected by the data sensing unit with a time sequence for primary time synchronization;

[0120] A data transmission unit is used to transmit the multi-modal data processed by the data alignment unit.

[0121] Further, the multi-modal production parameters comprise mechanical parameters, optical parameters, environmental parameters and material parameters.

[0122] Further, the edge computing module comprises:

[0123] A data receiving unit is used to receive the multi-modal data collected by the data acquisition module;

[0124] A data preprocessing unit is used to perform data cleaning and preprocessing on the multi-modal data to remove noise and invalid data, and to perform time synchronization alignment, thereby realizing secondary time synchronization;

[0125] A feature extraction unit is used to extract the feeding model input, the pressing model input, the forming model input, the drying model input and the quality inspection model input from the multi-modal data by using a feature extraction algorithm;

[0126] A feature input unit is used to input the model input extracted by the feature extraction unit to the hybrid model module.

[0127] In Embodiment Four, the model input of the feeding process is the raw material gram weight (g / m2), moisture content (%), fiber orientation angle (°), environmental temperature and humidity, and historical energy consumption (kWh) extracted by the data acquisition module, which are subjected to data cleaning and preprocessing, feature extraction and time sequence synchronization, and the feature vector is extracted as the model input of the feeding process;

[0128] The model inputs for the pressing process, obtained from the data acquisition module, include: pressure timing, vibration spectrum, and temperature gradient. A sliding window segmentation method is used with a 60-second window and a 1-second step size to generate a 60×15 matrix (15 dimensions / second). In feature extraction, time-domain statistics are performed to calculate the mean, variance, and kurtosis, followed by frequency-domain energy analysis. Temperature gradient calculations are then used to fit the above data to derive a feature vector, which serves as the input to the LSTM-PINN model, represented as [μ_P, σ_P]. 2 [,K_P,E_1-5kHz,VT];

[0129] The model inputs for the molding process, obtained from the data acquisition module, include mold spacing, conveyor speed, and cardboard stiffness. In feature extraction, the original data is smoothed using Kalman filtering, and then the spacing change rate is calculated and expressed as... By utilizing the velocity-stiffness coupling characteristics, the dynamic stiffness coefficient and the velocity-stiffness product are calculated, and expressed as follows: The above data is fitted to obtain the feature vector [d,Δd,v,S_dynamic,H_compensation], which is used as the input of the physics-guided reinforcement learning model.

[0130] The model inputs for the drying process, obtained from the data acquisition module, include hot air temperature, exhaust fan speed, and moisture content. In feature extraction, the moving average over a certain time period is calculated by analyzing hot air temperature fluctuations; in this embodiment, ten minutes is selected. Simultaneously calculate the temperature gradient, expressed as: Calculate the energy efficiency index, expressed as: The main frequency amplitude of the fan speed frequency is extracted by frequency domain features. main =max(|FFT(RPM)|); and then obtain the feature vector [T] by fitting the above data. a vg,VT,n,A m [ain] serves as the input to the multivariate predictive control model;

[0131] The model input for the quality inspection process acquires defect images, edge crush strength, and process parameters from the data acquisition module. In feature extraction, HOG and LBP texture features are extracted through image feature extraction. In process parameter correlation, a defect-parameter correlation matrix is ​​constructed to extract frequently occurring defect phenomena within the process parameters. The slope of the edge crush strength for the most recent 10 boxes is calculated to obtain the edge crush strength trend, represented as follows: The above data is fitted to obtain feature vectors [HOG features (64 dimensions), LBP features (256 dimensions), M_ij, k], which are then used as inputs to the graph neural network.

[0132] Further, a self-healing safety module is included for integrating the blockchain instruction storage unit and the three-level fault degradation control mechanism to respond to sudden accidents.

[0133] In embodiment five, the first-level fault trigger condition is a predicted model deviation continuously exceeding the limit (>15%, for 10 seconds), and the processing solution is to switch to a backup model and send a warning to the MES system.

[0134] The second-level fault trigger condition is that a key parameter exceeds a safety threshold (such as pressure >2.5 MPa) or the device vibration level is ≥3, and the processing solution is to run at a reduced load (power limited to 70%) and start online diagnosis.

[0135] The third-level fault trigger condition is a hardware fault (such as motor overcurrent) or three consecutive regulation failures, and the processing solution is to shut down the device and generate a blockchain maintenance order.

[0136] The self-healing safety module includes a degradation execution mechanism, including: a safety relay (cut off the power supply, response time <50 ms); a redundant control channel (main and backup PLC switching time <200 ms); an emergency ventilation system (forced start when the temperature rises >10℃ / min);

[0137] In other embodiments, the self-healing safety module can operate in coordination with other modules to increase protection of the device.

[0138] In one embodiment, the self-healing safety module interacts with the data acquisition module. When the data acquisition module detects a sudden increase in vibration level (from level 2 to level 4), it pushes an alarm to the self-healing safety module. The self-healing module checks the historical blockchain record and confirms that it is not a false alarm, triggering a second-level degradation. The degradation instruction is written to the PLC through OPC UA, limiting the pressing pressure to 1.8 MPa.

[0139] In one embodiment, the self-healing safety module cooperates with the edge computing module. When the edge module detects five consecutive prediction deviations >20%, it marks a first-level fault. The self-healing module calls an intelligent contract, records the fault context to the blockchain, switches to a backup random forest model, and updates the edge computing feature weights.

[0140] In one embodiment, the self-healing safety module is linked to the cross-process execution module. When the execution module reports a mold jam fault (third-level fault), the self-healing module triggers an emergency shutdown and stores the operation records 10 minutes before the fault. Through the blockchain work order system, maintenance tasks are assigned and the device authority is locked synchronously.

[0141] In one embodiment, the self-healing security module and the hybrid model module perform data verification, and after the physically guided reinforcement learning model generates a control instruction, an SM3 hash is calculated; the instruction content and the hash value are packaged and chained; before execution, the digital twin verification is performed, and if the success rate is less than 90%, the instruction is frozen and an alarm is given.

[0142] A carton production dynamic adjustment method based on the above-mentioned carton production dynamic adjustment control system, comprising:

[0143] Step A, multi-modal data acquisition: based on the data acquisition modules arranged in the feeding process, the pressing process, the forming process, the drying process and the quality inspection process, multi-modal production parameters in normal production are acquired;

[0144] Step B, model input extraction: after the multi-modal production parameters are preprocessed by the edge computing module, based on the feature extraction algorithm, the feeding model input, the pressing model input, the forming model input, the drying model input and the quality inspection model input are extracted;

[0145] Step C, process prediction result acquisition: based on the feeding model input, the pressing model input, the forming model input, the drying model input and the quality inspection model input, they are input into the hybrid model module, and then the model prediction result is obtained;

[0146] Step D, adjustment target value acquisition: the instruction calculation module based on the model prediction result corresponding to each process output by step C is used to compare with the dynamically generated adjustment threshold value of each process to confirm whether adjustment is needed, and then the adjustment target value of the process to be adjusted is calculated and generated adjustment instruction based on the corresponding generated adjustment target value;

[0147] Step E, adjustment instruction execution: based on the output result of the process dynamic model, the production parameter adjustment of the corresponding process equipment is performed to overcome the process coupling interference, and then the adjustment instruction generated by the instruction calculation module is executed to adjust the production equipment of each process;

[0148] Step F, target fine-tuning: the cross-process execution module monitors the adjusted production equipment, and when the trigger target threshold is reached, the production equipment is fine-tuned again to improve the adjustment effect.

[0149] The application can effectively collect production parameters in each process in the carton production process by setting data collection modules on each process, and the production parameters involve mechanical parameters, optical parameters, environmental parameters and material parameters. When the production parameters are input into a hybrid model module, model output results corresponding to different processes are output respectively to represent carton production trends in a future period, and the model output results are transmitted to an instruction calculation module to calculate adjustment target values of processes to be adjusted and generate adjustment instructions to adjust corresponding process equipment to ensure the yield of carton production. Before the adjustment process of each process is executed by a cross-process execution module, in order to avoid mutual interference between processes during the adjustment process, online modeling and real-time decoupling control of multi-process dynamic coupling interference are adopted to further improve the adjustment process and achieve better adjustment effect.

[0150] The above is only the preferred embodiment of the present application, and the protection scope of the present application is not limited to the above-mentioned embodiments. Any technical solution falling within the concept of the present application shall fall within the protection scope of the present application. It should be noted that, for ordinary skilled persons in the art, some improvements and refinements without departing from the principles of the present application shall also be considered as falling within the protection scope of the present application.

Claims

1. A dynamic adjustment and control system for cardboard box production, characterized in that, include: The data acquisition module is installed at each stage of the carton production process to collect and transmit multimodal production parameters in the feeding, pressing, forming, drying and quality inspection processes. The edge computing module is used to receive the multimodal production parameters collected by the data acquisition module, and after data preprocessing, it uses a feature extraction algorithm to sequentially obtain and transmit the input quantities of the feeding model, pressing model, molding model, drying model, and quality inspection model. The hybrid model module is used to receive the model input transmitted by the edge computing module, input it into the model of the corresponding process, and then output and transmit the model prediction results. The instruction calculation module is used to receive the multimodal production parameters collected by the data acquisition module and the model prediction results output by the hybrid model module, and after comprehensive judgment, dynamically generate and transmit the corresponding process adjustment instructions; The cross-process execution module is used to receive adjustment instructions transmitted by the instruction calculation module, and dynamically adjust the production parameters of the corresponding process before executing the adjustment instructions to overcome process coupling interference, and then execute the adjustment instructions for adjustment.

2. The dynamic adjustment and control system for carton production according to claim 1, characterized in that, The hybrid model module includes: A random forest multi-objective optimization model is used to receive the input from the feeding model, predict the influence trend of raw materials and environmental humidity during the feeding process, and generate feeding prediction results. The LSTM-PINN hybrid model is used to receive the input from the pressing model, predict the deformation trend of the cardboard during the pressing process, and generate pressing prediction results. A physics-guided reinforcement learning model is used to receive the input from the molding model, predict the influence trend of mold spacing and optimized conveying speed in the molding process, and generate molding prediction results. A multivariate predictive control model is used to receive the input from the drying model, predict the drying efficiency and energy consumption trend during the drying process, and generate drying prediction results. The graph neural network knowledge graph model is used to receive the input of the quality inspection model, perform defect root cause analysis and quality traceability in the quality inspection process, and generate quality inspection prediction results. The model prediction results are generated by fitting the feeding prediction results, pressing prediction results, molding prediction results, drying prediction results, and quality inspection prediction results.

3. The dynamic adjustment and control system for carton production according to claim 2, characterized in that, The instruction calculation module includes: The adjustment threshold calculation unit dynamically generates adjustment thresholds for each process based on the multimodal production parameters collected by the data acquisition module. The adjustment judgment calculation module compares the feeding prediction results, pressing prediction results, molding prediction results, drying prediction results, and quality inspection prediction results with the adjustment threshold of the corresponding process to determine whether adjustment is needed. The adjustment target value calculation module calculates the corresponding adjustment target value for the process that needs to be adjusted based on the judgment result of the adjustment judgment calculation module, and generates the adjustment instruction based on the generated adjustment target value.

4. The dynamic adjustment and control system for carton production according to claim 3, characterized in that, The cross-process execution module includes: The instruction receiving unit is used to receive the adjustment instruction transmitted by the instruction calculation module and parse the adjustment target value of the corresponding process. A multi-process coupling modeling unit is used to establish a process dynamic model of the dynamic relationship between multiple processes. Based on the adjustment target value, and according to the output of the process dynamic model, the process parameters are fine-tuned in advance to counteract adjustment interference. The program execution module is used to adjust the corresponding process equipment based on the adjustment target value parsed by the instruction receiving unit; The fine-tuning execution module is used to detect the production parameters of the adjusted process in real time, and perform fine-tuning based on the comparison between the production parameters of the process and the trigger target threshold.

5. The cardboard box production dynamic adjustment and control system according to claim 4, characterized in that, The trigger target thresholds include: prediction result deviation value, process coupling interference trigger value, process safety limit value, and quality feedback trigger value.

6. The dynamic adjustment and control system for carton production according to claim 1, characterized in that, The data acquisition module includes: Data sensing units are installed in each process step to collect multimodal data; The data alignment unit is used to bind and mark the multimodal data collected by the data sensing unit with the time series for initial time synchronization; The data transmission unit is used to transmit the multimodal data processed by the data alignment unit.

7. The dynamic adjustment and control system for carton production according to claim 6, characterized in that, The multimodal production parameters include: mechanical parameters, optical parameters, environmental parameters, and material parameters.

8. The dynamic adjustment and control system for carton production according to claim 1, characterized in that, The edge computing module includes: The data receiving unit is used to receive multimodal data collected by the data acquisition module; The data preprocessing unit is used to perform data cleaning and preprocessing on the multimodal data to remove noise and invalid data, and to perform time synchronization alignment, thereby achieving secondary time synchronization. The feature extraction unit is used to extract the input quantities of the feeding model, pressing model, molding model, drying model, and quality inspection model from the multimodal data using a feature extraction algorithm. The feature input unit is used to input the model input data extracted by the feature extraction unit into the hybrid model module.

9. The dynamic adjustment and control system for carton production according to claim 1, characterized in that, It also includes a self-healing security module, which integrates a blockchain instruction storage unit and a three-level fault degradation control mechanism to respond to and handle sudden incidents.

10. A method for dynamic adjustment of carton production executed by a dynamic adjustment and control system for carton production according to any one of claims 1-9, characterized in that, include: Step A, Multimodal Data Acquisition: Collect multimodal production parameters during normal production based on data acquisition modules set in the feeding process, pressing process, molding process, drying process, and quality inspection process; Step B, Model Input Extraction: After preprocessing the multimodal production parameters through the edge computing module, the input quantities of the feeding model, pressing model, molding model, drying model, and quality inspection model are extracted based on the feature extraction algorithm. Step C, Obtaining Process Prediction Results: Based on the input quantities of the feeding model, pressing model, molding model, drying model, and quality inspection model, input them into the corresponding hybrid model module to obtain the model prediction results; Step D, Obtaining the Adjustment Target Value: The instruction calculation module compares the model prediction results corresponding to each process output in Step C with the dynamically generated adjustment thresholds for each process to determine whether adjustment is needed. Then, it calculates and obtains the adjustment target value for the process that needs adjustment and generates an adjustment instruction based on the corresponding generated adjustment target value. Step E, Adjustment Instruction Execution: Based on the output results of the process dynamic model, the cross-process execution module adjusts the production parameters of the corresponding process equipment to overcome process coupling interference, and then executes the adjustment instructions generated by the instruction calculation module to adjust the production equipment of each process. Step F, Target Refinement and Fine-tuning: The cross-process execution module monitors the adjusted production equipment. After reaching the target threshold, the production equipment is fine-tuned a second time to improve the adjustment effect.

Images