Automatic Control Method for Carbon Black Masterbatch Production Equipment Based on PCL

Abstract

The present invention relates to the technical field of carbon black masterbatch production data processing, and specifically discloses an automatic control method for carbon black masterbatch production equipment based on PCL, which is used to solve problems such as poor carbon black dispersion uniformity and low conductive threshold regulation accuracy caused by insufficient multi-source heterogeneous data fusion and lack of cross-process parameter dynamic coupling. This method collects heterogeneous data streams of mixing equipment, extrusion equipment and detection equipment in real time, uses a sliding window mechanism for time-domain alignment processing and extracts frequency-domain features, triggers the iterative update of the knowledge base through real-time quality data feedback, and establishes a buffer material level closed-loop feedback mechanism to ensure the continuity of material transmission. Through the technical path of combining data fusion, virtual simulation and dynamic optimization, the present invention improves the carbon black dispersion uniformity and the regulation accuracy of conductive performance, and at the same time reduces the process commissioning cycle and the risk of abnormal shutdown, and is applicable to the intelligent production control of high-performance conductive carbon black masterbatch.

Description

Technical Field

[0001] The present invention relates to the technical field of carbon black masterbatch production data processing, and particularly to an automatic control method for carbon black masterbatch production equipment based on PCL. Background Art

[0002] Carbon black masterbatch is a pre-dispersed granule formed by coating carbon black as a functional filler with a carrier resin and dispersing additives. It is mainly used to prepare conductive or antistatic polymer composites, and its performance indicators depend on the dispersion degree of carbon black in the resin matrix and the interfacial bonding strength. Existing carbon black masterbatch production equipment usually consists of a three-roll mill, a mixer, a kneader, a twin-screw extruder, and a pelletizing unit. Among them, the three-roll mill grinds by squeezing the surface areas of three horizontally arranged rollers to break up the pigment particle aggregates in the color paste and produce primary particles. The mixer can preliminarily mix the pigments processed by the three-roll mill with the resin carrier. The pigments and carriers preliminarily mixed by the mixer are then mixed and plasticized by the kneader to form a semi-finished product of the color masterbatch. The twin-screw extruder completes melt blending and dispersion strengthening, and the pelletizing unit then performs granulation and forming.

[0003] The production process of carbon black masterbatch involves the dynamic coupling of multi-dimensional process parameters such as kneading temperature, extrusion pressure, feeding rate, and screw speed. Although the existing equipment control system has a data acquisition function, it has the technical defect of insufficient multi-source heterogeneous data fusion. Specifically, the thermodynamic parameters (temperature gradient, torque value) of the kneader and the rheological parameters (melt pressure, shear rate) of the twin-screw extruder belong to independent databases with different sampling frequencies and data structures, resulting in difficulties in constructing a dynamic correlation model between process parameters. Taking the control of carbon black dispersion uniformity as an example, there is no real-time mapping relationship between the particle size distribution data of carbon black aggregates in the kneading stage and the melt rheological curve data in the extrusion stage, resulting in the inability of the kneader rotor speed to be adaptively adjusted according to the feedback of the melt pressure of the extruder. This data processing breakpoint directly leads to an increase in the probability of secondary aggregation of carbon black, and the fluctuation range of the volume resistivity of the material expands to 1-2 orders of magnitude. The existing system lacks a time series alignment and characteristic parameter extraction mechanism for cross-process data, severely restricting the closed-loop control accuracy of the carbon black dispersion process. Summary of the Invention

[0004] Aiming at the deficiencies of the existing technology, the present invention provides an automatic control method for carbon black masterbatch production equipment based on PCL, which is used to solve the technical pain points of poor carbon black dispersion uniformity, low conductive threshold regulation accuracy, and limited process energy efficiency optimization caused by insufficient multi-source heterogeneous data fusion and lack of cross-process parameter dynamic coupling in existing carbon black masterbatch production equipment.

[0005] To solve the above technical problems, the specific technical solutions of the present invention are as follows:

[0006] The present invention provides an automatic control method for a carbon black masterbatch production device based on PCL, including:

[0007] Real-time collect heterogeneous data streams of mixing equipment, extrusion equipment and detection equipment, perform time-domain alignment processing on the heterogeneous data streams to generate a synchronous process parameter sequence, and extract the frequency-domain features and waveform features of equipment operation parameters;

[0008] Based on the equipment operation parameters, construct a multi-dimensional production task simulator including raw material ratio parameters, generate simulation data sets for different ratio scenarios through digital twin technology, and establish a hierarchical model knowledge base according to the hierarchical relationship of raw material ratio, process parameters and quality indicators;

[0009] When receiving a new production task, parse the task parameters and, based on the current equipment state characteristics, match candidate control models from the hierarchical model knowledge base through the dynamic time warping algorithm;

[0010] Perform multi-objective optimization sorting on the candidate control models according to the quality indicator priority, and convert the key parameters of the selected model into cross-device collaborative control instructions;

[0011] During the execution of the control instructions, collect production quality data in real time, and trigger the iterative update of the hierarchical model knowledge base based on the quality deviation detection results.

[0012] Further, for the automatic control method for a carbon black masterbatch production device based on PCL of the present invention, the time-domain alignment processing includes:

[0013] Adopt a sliding window mechanism to perform sampling frequency compensation on the asynchronous data streams of the mixing equipment and the extrusion equipment;

[0014] Perform outlier detection on the compensated data sequence to generate a set of process parameters with continuous time stamp alignment.

[0015] Further, for the automatic control method for a carbon black masterbatch production device based on PCL of the present invention, the construction of the multi-dimensional production task simulator includes:

[0016] Establish a constraint relationship model between raw material ratio parameters and the physical boundary of the extrusion equipment;

[0017] Drive the digital twin system to simulate the material dispersion process through the constraint relationship model, and generate a training data set including the correlation curve between mixing temperature and torque.

[0018] Further, for the automatic control method for a carbon black masterbatch production device based on PCL of the present invention, the establishment of the hierarchical model knowledge base includes:

[0019] Divide the basic scenario models according to the raw material ratio types as the first-level index;

[0020] Store the corresponding optimized combination of process parameters as a secondary index under the basic scenario model;

[0021] Associate the process parameter combination with the quality inspection threshold to form a three - level quality assessment standard.

[0022] Furthermore, for the automatic control method of the carbon black masterbatch production equipment based on PCL of the present invention, the dynamic time warping algorithm includes:

[0023] Perform Fourier transform encoding on the device pre - heating state data collected in real time;

[0024] Calculate the dynamic time warping distance between the encoded feature vector and the feature matrix of the knowledge base model;

[0025] Select the models with bending distance less than the preset threshold as the candidate control model set;

[0026] Verify the device compatibility of the selected candidate control models;

[0027] When it is detected that the current device hardware version does not match the model requirements, automatically trigger the update of the physical boundary parameters of the knowledge base.

[0028] Furthermore, for the automatic control method of the carbon black masterbatch production equipment based on PCL of the present invention, the multi - objective optimization sorting includes:

[0029] Construct a weighted evaluation function including the quality qualification rate, energy consumption coefficient, and equipment load rate;

[0030] Dynamically adjust the weight allocation ratio of the evaluation function according to the production task type;

[0031] Force - load a foreign object detection enhancement model for the high - carbon black ratio scenario.

[0032] The technical necessity of forcibly loading a foreign object detection enhancement model in the scenario of high carbon black ratio stems from the non-linear amplification effect of the high carbon black filling system on the sensitivity to foreign objects. When the mass fraction of carbon black exceeds 25%, its surface can adsorb resin molecular chains to form a dense coating layer, resulting in the melt light transmittance dropping below the signal-to-noise ratio critical value of the conventional visual detection system, and the existing gray threshold segmentation algorithm fails to extract the morphological features of metal chips and gel masses. At this time, it is necessary to introduce the joint feature analysis of melt rheological parameters and dielectric constant: by high-frequency sampling the power spectral density function of the melt pressure fluctuation signal, capturing the characteristic frequency band corresponding to the local flow field distortion caused by foreign objects (usually distributed in the range of 10 - 50 Hz), and combining with the abnormal offset of the dielectric loss factor detected by a broadband dielectric spectrometer, a multi-modal foreign object classifier based on support vector machine is constructed. At the same time, the formation of the carbon black percolation network makes the resistivity response of the system to conductive foreign objects show an exponential change characteristic, and metal particles with a volume fraction of 0.1% can cause a shift in the percolation threshold. It is necessary to adopt dynamic differential resistance monitoring technology to reconstruct the real-time spatial electric field distribution of the four-probe array to locate the local conductive channel variation area caused by foreign objects. The enhancement model further integrates a torque harmonic component analysis module for the internal mixer. When a foreign object causes an instantaneous load mutation of the rotor, the amplitude and phase characteristics of the third harmonic of the torque signal are extracted and dynamically time-warped and matched with the pre-stored foreign object mechanical response map in the knowledge base to trigger the collaborative control logic of the feeding system emergency stop and reverse blockage clearing program, thereby maintaining process stability and product performance consistency.

[0033] Further, in the automatic control method of the carbon black masterbatch production equipment based on PCL of the present invention, the conversion into cross-device collaborative control instructions includes:

[0034] Generating a regulating instruction for the line speed of the dispersing disc according to the torque spectrum characteristics of the mixing equipment;

[0035] Calculating the screw speed compensation amount based on the extrusion pressure balance model and generating a speed correction instruction;

[0036] Dynamically setting the speed gradient parameter of the pelletizing equipment according to the particle size distribution characteristics.

[0037] Further, in the automatic control method of the carbon black masterbatch production equipment based on PCL of the present invention, the real-time acquisition of production quality data includes:

[0038] Constructing a multi-dimensional quality evaluation matrix including specific surface area, DBP oil absorption value, volume resistivity, and unit energy consumption;

[0039] When the continuous batch data in the evaluation matrix exceeds the associated quality threshold, triggering the retraining of the mixing temperature compensation model.

[0040] Further, in the automatic control method of the carbon black masterbatch production equipment based on PCL of the present invention, the iterative update includes:

[0041] When no foreign object features are detected, expand the dimension of the foreign object detection atlas in the knowledge base;

[0042] Initiate the digital twin re-simulation process for the scenario model with a matching failure rate exceeding the threshold;

[0043] Establish a traceable process database including the historical optimal model version;

[0044] The traceable process database includes:

[0045] Store the device firmware information and raw material batch data corresponding to each historical version model;

[0046] Provide a comparative analysis function of production process parameters based on the time dimension.

[0047] Furthermore, the automatic control method for the carbon black masterbatch production equipment based on PCL according to the present invention further includes:

[0048] Establish a buffer material level closed-loop feedback mechanism between the discharge valve of the mixing equipment and the feeding port of the extrusion equipment;

[0049] When abnormal material transfer continuity is detected, adaptively adjust the timing relationship between the feeding rate and the opening degree of the discharge valve.

[0050] The beneficial effects of the present invention:

[0051] Through the sliding window mechanism and frequency domain feature extraction technology, the present invention realizes the time domain alignment and feature fusion of asynchronous data streams of equipment such as internal mixers and extruders, eliminating the problem of missing cross-process parameter coupling caused by data delay in the existing independent PLC control system. By constructing a unified data representation framework, the system can accurately capture the dynamic correlation between the torque fluctuation of the internal mixer and the melt pressure of the extruder, and then generate cross-device collaborative control instructions such as the linear velocity of the dispersion disk and the screw speed. This technical chain effectively inhibits the secondary agglomeration of carbon black, significantly improves the dispersion uniformity of carbon black, and lays a foundation for the stable regulation of electrical conductivity.

[0052] The multi-dimensional production task simulator based on digital twin technology, combined with the dynamic matching mechanism of the hierarchical model knowledge base, breaks through the limitations of the existing process relying on manual experience trial and error. The system pre-judges the safety boundary of process parameters through constraint modeling of raw material ratio and equipment physical boundary; at the same time, it uses the rapid call of historical optimal parameter combinations and multi-objective optimization sorting to achieve the rapid production and accurate parameter adaptation of new formulations. For example, for the high carbon black ratio scenario, the system automatically loads the foreign object detection enhancement model and the melt pressure compensation strategy, significantly shortening the process debugging cycle and improving product consistency.

[0053] Through the closed-loop feedback mechanism of the real-time quality assessment matrix and the traceable process database, the system has the ability of dynamic iteration and self-optimization. When quality deviation or abnormal equipment status is detected, it triggers the retraining of the hybrid temperature compensation model and the re-simulation of the digital twin, so that the process parameters can continuously adapt to the raw material fluctuations and equipment aging. At the same time, the closed-loop feedback mechanism of the buffer material level suppresses the risk of material shortage or overflow through the dynamic adjustment of the material transfer timing, ensuring the continuity of production. This technical solution greatly reduces the probability of abnormal shutdowns, extends the service life of equipment, and provides reliable guarantee for the efficient and stable production of carbon black masterbatch. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, for those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on the drawings.

[0055] Figure 1 It is a flowchart of the automatic control method for the carbon black masterbatch production equipment based on PCL provided by the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0056] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the specific embodiments and corresponding drawings of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention. The technical solutions provided by each embodiment of the present invention will be described in detail below with reference to the drawings.

[0057] To better understand the objectives of the present invention, the present invention will be further described in detail below.

[0058] As Figure 1 shown, the present invention provides an automatic control method for a carbon black masterbatch production equipment based on PCL, including:

[0059] Step S101, collect heterogeneous data streams of mixing equipment, extrusion equipment and detection equipment in real time, perform time-domain alignment processing on the heterogeneous data streams to generate a synchronous process parameter sequence, and extract the frequency-domain characteristics and waveform characteristics of the equipment operation parameters;

[0060] Aiming at the problem of data sampling frequency differences between mixing equipment (such as a mixer) and extrusion equipment (such as a twin-screw extruder), a sliding window mechanism is used to resample and compensate the asynchronous data streams. For example, the torque data collected once per second by the mixer and the melt pressure data of the extruder 10 times per second are aligned on the time axis through window overlapping interpolation.

[0061] Perform outlier detection based on statistical distribution (such as the 3σ principle) on the synchronized data sequence, eliminate outliers caused by sensor noise or equipment transient fluctuations, and generate a continuous and stable process parameter sequence.

[0062] Extract the frequency-domain energy distribution characteristics of equipment parameters through fast Fourier transform (FFT), and at the same time use waveform decomposition technology to capture morphological characteristics such as steep change points and periodicity of parameter curves, providing multi-dimensional feature vectors for subsequent model matching.

[0063] This step solves the problems of time-axis misalignment of multi-source data and missing feature dimensions in existing independent PLC control systems through time-domain alignment and feature extraction. For example, the dynamic correlation between the torque fluctuation of a mixer and the melt pressure of an extruder requires accurately synchronized timing data to establish an effective model, thereby avoiding regulation lag caused by data delay during the carbon black dispersion process.

[0064] Step S102: Construct a multi-dimensional production task simulator including raw material ratio parameters based on the equipment operation parameters, generate simulation data sets for different ratio scenarios through digital twin technology, and establish a hierarchical model knowledge base according to the hierarchical relationship of raw material ratio, process parameters, and quality indicators;

[0065] Establish a physical boundary constraint model for the extrusion equipment based on raw material ratio parameters (such as carbon black content, resin type). For example, in the case of a high carbon black ratio scenario, it is necessary to limit the upper limit of the screw speed to avoid exceeding the melt pressure.

[0066] When simulating the material dispersion process in the digital twin system, associate and model the mixing temperature gradient and torque change curve to generate a training data set including predicted values of carbon black dispersion degree.

[0067] The hierarchical model knowledge base is organized in a three-level structure: the first-level index is classified according to raw material ratio (such as PA6-based / PS-based), the second-level index stores the corresponding process parameter combinations (such as mixing temperature range, screw speed ratio), and the third-level index associates the quality detection threshold (such as volume resistivity ≤ 1×10^3 Ω·cm).

[0068] Through the hierarchical modeling of raw material - process - quality, the limitations of the empirical setting of existing process parameters are broken through. For example, when the production task switches to a new type of conductive masterbatch, the system can automatically match the process parameter combination of historical similar ratio scenarios according to the carbon black content, avoiding repeated trial and error and significantly shortening the process debugging cycle.

[0069] Step S103: When receiving a new production task, parse the task parameters and based on the current equipment state characteristics, match a candidate control model from the hierarchical model knowledge base through the dynamic time warping algorithm;

[0070] Dynamic Time Warping (DTW) matching: Fourier encode the temperature rise curve during the equipment preheating stage, calculate its dynamic bending distance from the feature matrix of the knowledge base model, and screen out candidate models with similar shapes.

[0071] Step S104, perform multi-objective optimization and sorting on the candidate control models according to the quality index priority, and convert the key parameters of the selected model into cross-device collaborative control instructions;

[0072] Multi-objective optimization: When constructing a weighted evaluation function, assign a higher weight to the quality qualification rate for the high-precision conductive masterbatch task, while focus on optimizing the energy consumption coefficient for the cost reduction task.

[0073] Step S105, during the execution of the control instructions, collect production quality data in real time, and trigger the iterative update of the hierarchical model knowledge base based on the quality deviation detection results.

[0074] Instruction conversion and iteration: Convert the optimized screw speed compensation amount into PID control parameters executable by the equipment; when the carbon black dispersion deviation is detected online to exceed ±5%, trigger the re-simulation and update of the corresponding process parameter combination in the knowledge base.

[0075] This closed-loop control mechanism realizes autonomous optimization from data to decision-making. For example, when the feeding rate of the extruder suddenly changes, resulting in melt pressure fluctuations, the system quickly identifies the abnormality through real-time pressure feedback, dynamically adjusts the collaborative relationship between the rotor speed of the internal mixer and the screw speed of the extruder, suppresses the phenomenon of carbon black secondary agglomeration, and ensures the stability of the dispersion uniformity.

[0076] Specifically, for the automatic control method of the carbon black masterbatch production equipment based on PCL described in the present invention, the time domain alignment processing includes:

[0077] Adopt a sliding window mechanism to perform sampling frequency compensation on the asynchronous data streams of the mixing equipment and the extrusion equipment;

[0078] Perform outlier detection on the compensated data sequence to generate a set of process parameters with continuous time stamp alignment.

[0079] The adoption of a sliding window mechanism for sampling frequency compensation includes the following technical contents:

[0080] Asynchronous data stream processing: For the sampling frequency differences between the mixing equipment (such as an internal mixer) and the extrusion equipment (such as a twin-screw extruder) (for example, the internal mixer is once per second and the extruder is ten times per second), adopt a sliding window mechanism to perform interpolation compensation on the low-frequency data stream. For example, through linear interpolation or cubic spline interpolation, expand the torque data of the internal mixer on the time axis to the same sampling density as the melt pressure data of the extruder.

[0081] Window overlap strategy: Set the window overlap ratio (such as 50%) to achieve smooth data transition between adjacent windows and avoid the loss of key features due to data truncation. For example, in the critical stage of torque mutation in the internal mixer, the complete waveform features can be retained by overlapping windows.

[0082] Timestamp alignment: The interpolated data stream is timestamped to associate the process parameters of the mixing equipment and the extrusion equipment under a unified time base. For example, the torque data of the 5th second of the internal mixer is accurately matched with the melt pressure data of the 5th second of the extruder.

[0083] In existing independent PLC control systems, the data of mixing equipment and extrusion equipment are misaligned in time axis due to sampling frequency differences, making it impossible to establish a dynamic correlation model. For example, the coordinated adjustment of the torque change of the internal mixer and the melt pressure of the extruder requires synchronous time series data. The sliding window mechanism eliminates data delays between devices through interpolation compensation and time calibration, providing a unified time benchmark for subsequent process parameter correlation analysis, thereby improving the real-time performance of cross-device control.

[0084] Performing outlier detection to generate an alignment parameter set includes the following technical contents:

[0085] Outlier detection method: The 3σ principle (triple standard deviation method) based on statistical distribution is used to identify abnormal data. For example, the mean and standard deviation of the torque series of the internal mixer are calculated, and data points that are beyond the mean ±3σ range are eliminated.

[0086] Dynamic threshold adjustment: Dynamically adjust the outlier threshold according to the equipment operation stage (such as preheating and steady-state production). For example, a larger fluctuation range is allowed during the equipment startup phase to avoid misjudging transient data as outliers.

[0087] Data repair mechanism: The removed outliers are filled with linear interpolation of the previous and next data or the average value of adjacent windows to improve the continuity of the time series. For example, the torque anomaly of the internal mixer is replaced by the sliding average of the previous and next 1 second data.

[0088] Outlier detection solves the problem of abnormal data interference caused by sensor noise or transient fluctuations of equipment (such as voltage mutations and mechanical vibrations). For example, if the melt pressure sensor of the extruder produces abnormal peaks due to instantaneous blockage, directly using such data will mislead the optimization of process parameters. Through dynamic thresholds and data repair, a continuous and stable set of process parameters is generated, providing high-quality input for subsequent feature extraction and model matching, thereby improving the reliability of control instructions.

[0089] Specifically, the PCL-based automatic control method for carbon black masterbatch production equipment of the present invention, the construction of a multi-dimensional production task simulator includes:

[0090] Establish a constraint relationship model between the raw material ratio parameters and the physical boundaries of the extrusion equipment;

[0091] Drive the digital twin system to simulate the material dispersion process through the constraint relationship model, and generate a training data set including the correlation curve of mixing temperature and torque.

[0092] Raw material parameter mapping: Establish a quantitative correlation between the ratio parameters such as carbon black content and resin melt index and the physical boundaries of the extrusion equipment (such as the maximum screw speed and the upper limit of melt pressure). For example, when the carbon black content is high (>30%), the increase in melt viscosity is derived through a rheological model, and the screw speed is dynamically limited to avoid melt fracture.

[0093] Material property modeling: Based on the rheological characteristic curves of resin types (such as PA6, PS), establish a shear rate-viscosity relationship model at different temperatures, and define the temperature setting range for each section of the extruder. For example, for high-viscosity resins, the temperature of the melt conveying section needs to be reduced to prevent local overheating.

[0094] Equipment capacity matching: According to the screw geometric parameters (length-diameter ratio, screw groove depth) and driving power, construct a physical constraint model of screw speed-torque-production capacity, so that the simulation parameters are within the actual load range of the equipment.

[0095] In the existing process, the matching of raw material ratio and equipment parameters depends on manual experience, which easily leads to parameter overrun (such as melt pressure overpressure alarm). The present invention realizes the automatic prediction of the parameter safety boundary by quantifying the correlation between raw material characteristics and the physical boundaries of the equipment. For example, when the carbon black content in the formula increases, the system automatically calculates the increase in melt viscosity and dynamically adjusts the upper limit of the screw speed of the extruder to avoid the decrease in dispersion degree or equipment overload caused by insufficient shear force.

[0096] Multi-physical field coupling simulation: Couple the thermodynamic model (temperature field distribution of the internal mixer) and the fluid mechanics model (melt flow of the extruder) in the digital twin system to simulate the dispersion process of carbon black particles in the resin. For example, simulate the influence of the rotor speed of the internal mixer on the crushing efficiency of carbon black aggregates, and generate a correlation curve of torque-dispersion degree.

[0097] Dynamic boundary condition loading: Use the equipment parameter limits (such as the maximum screw speed) output by the constraint relationship model as the boundary conditions of the digital twin simulation, so that the simulation scenario is consistent with the actual equipment capacity. For example, when simulating a high carbon black ratio, the screw speed is forcibly limited not to exceed the equipment safety threshold.

[0098] Characteristic data extraction: Extract key parameters such as mixing temperature gradient and torque fluctuation frequency from the simulation results, and construct a correlation data set including process parameters and quality indicators. For example, record the steady-state torque fluctuation range when the internal mixing temperature reaches the set value as a reference feature for the dispersion degree to meet the standard.

[0099] The existing trial-and-error method requires a large amount of materials and time to verify process parameters. However, the present invention pre-acts the production process through digital twin technology, significantly shortening the process debugging cycle. For example, for the new conductive masterbatch formula, the system automatically loads the rheological parameters of the corresponding raw materials, simulates the carbon black dispersion state at different internal mixer temperatures, and outputs the optimal temperature-speed combination. This process not only avoids the risk of parameter overrun in actual production but also can predict in advance whether the dispersion uniformity meets the standard, providing high-confidence training data for the knowledge base.

[0100] Specifically, for the automatic control method of the carbon black masterbatch production equipment based on PCL of the present invention, the establishment of the hierarchical model knowledge base includes:

[0101] Dividing the basic scenario model according to the raw material ratio type as the first-level index;

[0102] Storing the corresponding optimized combination of process parameters under the basic scenario model as the second-level index;

[0103] Associating the process parameter combination with the quality inspection threshold to form the third-level quality evaluation standard.

[0104] Dividing the basic scenario model (first-level index) according to the raw material ratio:

[0105] Raw material classification rule: Divide the basic scenario according to the resin matrix type (such as PA6, PS) and the carbon black content range (such as 5%-10%, 10%-20%). For example, define "PA6-based - 15% carbon black" as an independent scenario model.

[0106] Scene feature extraction: Extract key features for each raw material ratio scenario, including the resin melt index, carbon black specific surface area, additive type, etc., to form the unique identification code of the scenario.

[0107] Index storage structure: Store the scenario model using a hash table or tree structure, and achieve fast retrieval through the ratio feature code. For example, when inputting a PA6-based formula with a carbon black content of 18%, the system automatically matches to the "PA6-based - 15%-20% carbon black" scenario branch.

[0108] In the existing process, the empirical correspondence relationship between the raw material ratio and the process parameters is scattered and unstructured, resulting in a long debugging cycle for new formulas. Through the scientific classification and feature coding of the raw material ratio, the system can quickly locate historical similar scenarios. For example, when the production task switches to "PS-based - 25% carbon black", directly call the optimized parameter combination under the corresponding scenario to avoid repeated trial and error.

[0109] Storing the optimized combination of process parameters (second-level index):

[0110] Parameter combination generation: Based on digital twin simulation data and historical production data, screen out parameter combinations such as the internal mixer temperature range, screw speed ratio, and feeding rate that meet the quality indicators. For example, for the "PA6-based - 15% carbon black" scenario, store the optimized combination of an internal mixer temperature of 180 - 190 °C and a screw speed of 60 - 70 rpm.

[0111] Parameter correlation constraint: Establish logical association rules between parameters, such as the linkage relationship between the upper limit of the internal mixer torque and the melt pressure threshold of the extruder, to avoid storing invalid or conflicting parameter combinations.

[0112] Compressed storage technology: Adopt parameter matrix or eigenvector to compress and store process parameters. For example, normalize parameters such as temperature, speed, and pressure and encode them as multi-dimensional vectors to reduce storage redundancy.

[0113] The existing parameter settings rely on manual experience, making it difficult to ensure the coordination of multi-device parameters. By storing the verified optimized parameter combinations, the system can directly call cross-device collaborative control instructions. For example, for the high carbon black ratio scenario, the system automatically loads the combination of "high internal mixer temperature + low screw speed" to inhibit the secondary agglomeration of carbon black and improve the dispersion uniformity.

[0114] Associated quality detection threshold (three-level evaluation standard):

[0115] Quality index mapping: Dynamically bind the process parameter combination with quality detection thresholds such as specific surface area (NSA ≥ 200 10 3 cm 3 / g) and volume resistivity (≤ 1 × 10³ Ω·cm). For example, when the volume resistivity detection value corresponding to parameter combination A exceeds the threshold, it is automatically marked as "needs optimization".

[0116] Dynamic threshold adjustment: Adaptively update the quality threshold according to raw material batch fluctuations (such as changes in carbon black particle size distribution). For example, when the D50 particle size of carbon black increases by 5%, automatically relax the blackness value threshold to NSA ≥ 180 10 3 cm 3 / g.

[0117] Quality assessment model: Construct a decision tree or logistic regression model to judge whether the current parameter combination meets the quality requirements. For example, when the internal mixer temperature is lower than the set range, trigger a "lack of dispersion" warning and recommend a temperature increase compensation strategy.

[0118] The existing quality control relies on post-event detection and cannot real-time associate process parameters with quality results. Through the three-level quality assessment standard, the system can predict quality risks during the production process. For example, when it is detected that the melt pressure of the extruder exceeds the threshold associated with the current parameter combination, immediately adjust the screw speed to avoid non-compliance of electrical conductivity, and achieve a "parameter-quality" closed-loop control.

[0119] Specifically, for the automatic control method of the carbon black masterbatch production equipment based on PCL of the present invention, the dynamic time warping algorithm includes:

[0120] Perform Fourier transform encoding on the equipment preheating state data collected in real time;

[0121] Calculate the dynamic time warping distance between the encoded feature vector and the feature matrix of the knowledge base model;

[0122] Select the models with bending distance less than the preset threshold as the candidate control model set;

[0123] Verify the equipment compatibility of the selected candidate control models;

[0124] When it is detected that the current equipment hardware version does not match the model requirements, automatically trigger the update of the physical boundary parameters of the knowledge base.

[0125] Fourier transform encoding of equipment preheating data:

[0126] Segmented feature extraction: Divide the time series data in the equipment preheating stage (such as mixer heating, extruder temperature rise) into sub-stages such as initial temperature rise and steady state maintenance according to the temperature change trend, perform Fourier transform on the data of each stage separately, and extract the frequency domain energy distribution characteristics. For example, analyze the proportion of the low-frequency component of the torque fluctuation when the mixer rotor starts to characterize the mechanical load characteristics.

[0127] Feature dimensionality reduction encoding: Sort the spectrum data after Fourier transform according to the energy intensity, retain the first N principal components (such as the frequency bands with energy proportion exceeding 90%), and generate a compact feature vector. For example, the high-frequency noise in the mixer preheating stage is filtered, and only the low-frequency features reflecting the bearing friction characteristics are retained.

[0128] Existing methods directly use time domain data to match models, which are easily affected by the time stretching of the equipment preheating curve (such as slower temperature rise in winter) or instantaneous noise interference. Fourier transform effectively captures the essential state of the equipment preheating process (such as enhanced low-frequency vibration caused by bearing wear) through frequency domain energy feature extraction, improving the robustness of subsequent model matching.

[0129] Calculation of dynamic time warping distance:

[0130] Processing of non-aligned sequences: For the equipment preheating time difference (such as different heating efficiencies of new and old equipment), non-linearly align the feature vectors through the dynamic time warping (DTW) algorithm to eliminate the deviation caused by the time axis stretching. For example, match the 30-second preheating curve of the current mixer with the 25-second historical curve in the knowledge base in terms of morphology.

[0131] Similarity quantification: Calculate the cumulative distance of the aligned sequences, integrating the frequency-domain feature differences (such as the main frequency offset) and the time-series morphological differences (such as the peak position offset) to generate a multi-dimensional similarity score.

[0132] The DTW algorithm solves the dependence of the existing Euclidean distance on the strict alignment of time series. For example, the change in the heating rate caused by the environmental temperature difference during the preheating process of the extruder will not affect the accurate judgment of the "preheating completed state", thus improving the accuracy of candidate model screening.

[0133] Candidate control model screening:

[0134] Dynamic threshold setting: Dynamically adjust the bending distance threshold according to the production task type (such as high-precision conductive masterbatch or general masterbatch). For example, a more stringent threshold is adopted for the conductive masterbatch task to prioritize quality, while the threshold for the general masterbatch task can be appropriately relaxed to improve efficiency.

[0135] Priority ranking: Arrange the candidate models in ascending order of the bending distance, and at the same time superimpose the weights of additional dimensions such as the quality pass rate and the equipment load rate to generate a comprehensive sorted list.

[0136] Existing screening methods only rely on a single distance metric and are prone to ignoring the multi-objective trade-offs in actual production. Through dynamic threshold and multi-dimensional ranking, the system can balance quality, efficiency, and equipment life. For example, when the bending distances of multiple models are similar, the parameter combination with lower historical energy consumption is preferentially selected.

[0137] Equipment compatibility verification:

[0138] Hardware parameter verification: Compare the compatibility of parameters such as the equipment firmware version, sensor type (such as the pressure sensor range), and drive motor power stored in the candidate model with the current equipment. For example, when screening out models that are only compatible with servo motor drives, if the current equipment is an asynchronous motor, it will be automatically excluded.

[0139] Physical boundary constraint: Verify whether the model parameters (such as the upper limit of the screw speed and the torque threshold of the internal mixer) exceed the mechanical load-bearing capacity of the current equipment. For example, when the model requires the extruder speed to be 120 rpm while the current equipment supports a maximum of 100 rpm, an alarm is triggered and the model is excluded.

[0140] Compatibility verification prevents the risk of "optimization on paper". For example, an old model extruder cannot withstand a high-speed model due to the limitation of the screw material, and directly loading such a model may cause equipment damage. Through the double verification of hardware parameters and physical boundaries, the feasibility of control instructions is improved.

[0141] Knowledge base physical boundary update:

[0142] Boundary parameter recalculation: When detecting hardware upgrades of the device (such as replacing a high-power motor), call the digital twin system to re-simulate the physical boundaries (such as the maximum screw speed and the upper limit of the melt pressure), and update the constraint conditions of the associated models in the knowledge base.

[0143] Incremental model iteration: Only retrain the scenario models affected by hardware changes (such as the high-speed related models), retain the original parameter combinations of other scenarios, and reduce the consumption of computing resources.

[0144] The existing system requires manual maintenance of the matching relationship between device parameters and models, resulting in a lag in response. The present invention enables real-time synchronization between the knowledge base and the device status by automatically triggering boundary updates. For example, after the extruder is upgraded, the system automatically removes the original speed limit and loads an optimized model adapted to the new hardware, without the need for production suspension and debugging.

[0145] The dynamic time warping algorithm solves the contradiction between device state drift and model solidification through a technical chain of state encoding → flexible matching → compatibility verification → closed-loop update:

[0146] Anti-interference matching: Frequency domain feature extraction and the DTW algorithm eliminate timing deviations and improve the accuracy of model screening;

[0147] Safety control guarantee: Compatibility verification prevents the device from operating beyond limits;

[0148] Enhanced self-adaptability: The dynamic update of physical boundaries enables the knowledge base to continuously adapt to device upgrades.

[0149] Specifically, for the automatic control method of the carbon black masterbatch production equipment based on PCL described in the present invention, the multi-objective optimization sorting includes:

[0150] Construct a weighted evaluation function including the qualified rate of product quality, the energy consumption coefficient, and the equipment load rate;

[0151] Dynamically adjust the weight distribution ratio of the evaluation function according to the type of production task;

[0152] Forcefully load a foreign object detection enhancement model for high carbon black ratio scenarios.

[0153] Construct a weighted evaluation function:

[0154] Quantification of the qualified rate of product quality: Based on on-line detection data (such as specific surface area, DBP oil absorption value, volume resistivity), calculate the qualified rate weight in real time. For example, map the volume resistivity deviation range (±10%) to a quality score.

[0155] Modeling of the energy consumption coefficient: Correlate the device power data with the energy consumption per unit output to establish a dynamic energy consumption model. For example, when the melt pressure of the extruder rises by 10%, the system automatically calculates the corresponding energy consumption increment and converts it into an energy consumption coefficient.

[0156] Equipment load rate assessment: The load status of the equipment is deduced in real time through parameters such as motor current and bearing temperature. For example, the ratio of screw speed to rated power is used as the load rate index.

[0157] The optimization of existing processes often solely pursues quality or energy consumption indicators, resulting in equipment overload or efficiency loss. Through a weighted evaluation function, the system can dynamically balance quality, energy efficiency, and equipment life. For example, when the quality score is close to the threshold, the system automatically reduces the energy consumption weight to prioritize ensuring the product qualification rate; while when the equipment load rate exceeds the limit, it focuses on reducing the speed to extend the equipment life.

[0158] Dynamic adjustment of weight allocation:

[0159] Task type recognition: Parse the task tags in the production order (such as "high-precision conductive grade" and "general cost reduction type"), and call the preset weight template. For example, the quality weight in the conductive grade task is set to 70%, and the energy consumption weight in the cost reduction type task is increased to 60%.

[0160] Real-time working condition adaptation: Fine-tune the weight ratio according to the equipment status (such as the wear degree of the extruder screw). For example, when it is detected that the screw wear exceeds the limit, reduce the load rate weight to avoid further wear.

[0161] Priority arbitration mechanism: When there are multi-object conflicts (such as improving quality requires increasing energy consumption), select the Pareto optimal solution according to the optimization results of historical data.

[0162] Dynamic weight allocation solves the problem that the fixed weight model cannot adapt to diverse production requirements. For example, when producing during the low electricity price period at night, the system can temporarily increase the energy consumption weight to preferentially select low-energy consumption processes, while when rushing to produce high-value orders, it prioritizes ensuring quality stability to achieve the optimal allocation of resources.

[0163] Forced loading of foreign object detection enhancement model:

[0164] High carbon black scenario recognition: Automatically trigger the enhanced detection mode through raw material ratio parameters (carbon black content ≥ 25%).

[0165] Expansion of detection dimensions: On the basis of conventional visual detection, add abnormal analysis of melt rheology parameters (such as pressure mutation detection) and infrared spectrum foreign object feature matching. For example, when it is detected that the melt pressure curve shows non-periodic fluctuations, the visual system is linked to focus on the suspected contaminated area.

[0166] Dynamic model loading: Call the pre-trained foreign object detection model (such as the metal chip and gel particle feature library) from the knowledge base to improve the detection sensitivity and response speed.

[0167] In the scenario of high carbon black ratio, carbon black particles are likely to cover up foreign objects or form pseudo-agglomerates, and existing detections are prone to missed judgments. Through the multi-dimensional analysis of the enhanced model, the system can distinguish real dispersion defects from foreign object interference. For example, when the rheological property differences between carbon black agglomerates and metal chips are identified, the system automatically adjusts the speed of the internal mixer to break up the agglomerates, and at the same time triggers the foreign object removal mechanism to avoid abnormal electrical conductivity.

[0168] Multi-objective optimization ranking systematically solves the problem of objective conflicts in process optimization through dynamic trade-off and scenario enhancement:

[0169] Comprehensive decision-making ability: The weighted evaluation function realizes the collaborative optimization of quality, energy consumption, and equipment status, avoiding sub-optimal solutions caused by single objectives;

[0170] Flexible adaptation: The weight dynamic adjustment mechanism adapts to diverse production requirements and enhances the process generalization ability;

[0171] Risk pre-control: The enhanced model for foreign object detection in the high carbon black scenario actively intercepts quality hazards to ensure product consistency.

[0172] Specifically, for the automatic control method of carbon black masterbatch production equipment based on PCL described in the present invention, the conversion into cross-device collaborative control instructions includes:

[0173] Generating a regulation instruction for the linear velocity of the dispersion disk according to the torque spectrum characteristics of the mixing equipment;

[0174] Calculating the screw speed compensation amount based on the pressure balance model of the extruder and generating a speed correction instruction;

[0175] Dynamically setting the rotational speed gradient parameter of the pelletizing equipment according to the particle size distribution characteristics.

[0176] Regulation of the linear velocity of the dispersion disk based on the torque spectrum:

[0177] Spectrum feature analysis: Performing a fast Fourier transform (FFT) on the torque time series data of the mixing equipment (such as an internal mixer), and extracting the energy distributions in the low frequency band (0 - 5 Hz) and the high frequency band (5 - 20 Hz). When the proportion of low frequency energy is high, it indicates that the carbon black agglomerates are not fully broken, and the linear velocity of the dispersion disk needs to be increased to enhance the shear strength.

[0178] Dynamic mapping strategy: Establishing the correlation rule between the energy proportion of the torque frequency band and the linear velocity. For example, when the proportion of low frequency energy exceeds 50%, the linear velocity is increased by 5% - 10% in a gradient; if the proportion of high frequency energy continues to rise, a fine adjustment of the linear velocity is triggered to maintain the dispersion stability.

[0179] Closed-loop feedback control: According to the on-line dispersion degree detection result of the material at the inlet of the extruder (such as the fluctuation range of specific surface area), reversely correcting the adjustment amplitude of the linear velocity to suppress the risk of resin degradation caused by over-shearing.

[0180] The existing control of the dispersion disk relies on fixed rotational speed parameters and cannot adapt to the dynamic changes of carbon black aggregates. Through torque spectrum analysis, the system can accurately identify the dispersion state: low-frequency energy reflects large-size aggregates, and high-frequency energy characterizes the degree of dispersion refinement. For example, when the low-frequency torque of the internal mixer suddenly increases, the system automatically increases the linear speed to enhance the shearing effect, synchronously linking the feeding rate of the extruder to avoid abnormal fluctuations in the melt pressure caused by uneven dispersion.

[0181] Extruder pressure balance and speed compensation:

[0182] Pressure dynamic modeling: Based on the data of the melt pressure sensor of the extruder, a differential relationship model between the pressure change rate and the screw speed is constructed. For example, when the pressure rise rate exceeds the threshold, the compensation amount for reducing the screw speed is deduced to balance the feeding rate.

[0183] Compensation amount calculation: According to the pressure fluctuation direction (rise / fall) and historical trend, the proportional-integral (PI) control algorithm is used to generate a speed correction command. For example, when it is detected that the melt pressure exceeds the limit for 3 consecutive seconds, the speed compensation amount is dynamically calculated according to the integral value of the pressure deviation (such as reducing 2-3 rpm per minute).

[0184] Multi-segment coordination: Separate pressure data of different segments (melting segment, mixing segment) of the extruder are modeled respectively, and segmented speed compensation commands are generated to optimize the melt flow uniformity.

[0185] The existing adjustment of the screw speed lags behind the change of the melt pressure, which is likely to cause fluctuations in the carbon black dispersion degree. Through the pressure balance model, the system can predict the pressure trend and make early compensation. For example, when the sudden increase in the feeding rate causes the pressure in the melting segment to rise, the system synchronously reduces the speed of the mixing segment to expand the melt residence time window, avoiding the secondary aggregation of carbon black caused by too high shear rate, thereby stabilizing the volume resistivity index.

[0186] Dynamic setting of the rotational speed gradient of the pelletizing equipment:

[0187] Particle size characteristic mapping: The particle size distribution data of the masterbatch (such as D50, D90) are obtained through an on-line particle size analyzer, and an association model between the particle size concentration and the rotational speed gradient of the pelletizing knife is established. For example, when the D90 value exceeds 800 μm, the rotational speed ratio of the high-speed blade group is increased to refine the particles.

[0188] Gradient parameter generation: According to the target particle size range (such as 300-500 μm), a rotational speed gradient band is set, and the running time ratio of the low-speed, medium-speed, and high-speed knife groups is dynamically allocated. For example, when it is detected that the particle size distribution is skewed to the right, the working time of the high-speed knife group is extended to reduce the proportion of small particle sizes.

[0189] Guarantee of material flow consistency: Link the pressure monitoring data at the die head of the tandem extruder, adjust the matching relationship between the feeding rate and the cutter speed of the pelletizer, and prevent abnormal particle morphology caused by sudden changes in material flow rate.

[0190] Existing pelletizing equipment uses fixed rotational speed parameters and is difficult to adapt to the melt viscosity differences caused by changes in carbon black content. Through particle size distribution feedback, the system can dynamically adjust the rotational speed gradient of the cutter group. For example, when the melt viscosity of the high carbon black ratio is high and the fluidity is poor, the system reduces the proportion of the low-speed cutter group to avoid material accumulation, and at the same time increases the rotational speed of the high-speed cutter group to improve particle uniformity and reduce the energy consumption of the subsequent screening process.

[0191] Cross-device collaborative control instructions systematically solve the process coordination problems in carbon black masterbatch production through the technical chain of state perception → dynamic modeling → instruction linkage:

[0192] Improvement of dispersion uniformity: Torque spectrum analysis realizes precise control of dispersion intensity and reduces the probability of carbon black secondary agglomeration;

[0193] Guarantee of melt stability: The pressure balance model predictively adjusts the screw speed to suppress fluctuations in conductivity;

[0194] Optimization of particle size consistency: The rotational speed gradient of the pelletizer dynamically matches the melt characteristics, reducing the proportion of over-fine or over-coarse particles. The present invention provides a closed-loop optimization framework for the full-process automatic control of carbon black masterbatch production.

[0195] Specifically, for the automatic control method of carbon black masterbatch production equipment based on PCL described in the present invention, the real-time acquisition of production quality data includes:

[0196] Construct a multi-dimensional quality evaluation matrix including blackness value, volume resistivity, and unit energy consumption;

[0197] When the data of consecutive batches in the evaluation matrix exceed the associated quality threshold, trigger the retraining of the mixing temperature compensation model.

[0198] Construct a multi-dimensional quality evaluation matrix:

[0199] Multi-source data fusion: Real-time collect volume resistivity and unit energy consumption data through on-line detection equipment. For example, use a four-probe resistivity tester to dynamically measure the conductivity of the masterbatch, and synchronously collect the energy consumption data (kW·h / kg) of the internal mixer and the extruder.

[0200] Dynamic weight assignment: Adjust the quality index weights according to the production task type. For example, in the task of conductive masterbatch, the weight of volume resistivity is set to 60%.

[0201] Data normalization: Standardize indicators with different dimensions (for example, the resistivity unit is Ω·cm, and the energy consumption unit is kW·h / kg) to eliminate the impact of magnitude differences on the evaluation results.

[0202] Existing quality inspection relies on offline sampling inspection, resulting in data lag. The multi-dimensional matrix dynamically reflects the production quality status through real-time data fusion. For example, when the melt pressure fluctuation of the extruder causes the carbon black dispersion to decrease, the correlation between the increase in blackness value and the decrease in resistivity can be captured by the matrix, quickly locating the root cause of process abnormalities (such as insufficient mixing temperature or improper screw speed).

[0203] Trigger the retraining of the mixing temperature compensation:

[0204] Dynamic threshold setting: Set the quality threshold based on historical optimal batch data (such as the average value of the previous 100 batches ± 3σ), and dynamically adjust according to the characteristics of raw material batches (such as the carbon black particle size distribution). For example, when the D50 particle size of carbon black increases, relax the specific surface area ≤ 180 10 3 g / cm 3 .

[0205] Continuous anomaly determination: When the same quality index (such as volume resistivity) exceeds the limit for 3 consecutive batches, it is determined as a systematic deviation, triggering the retraining of the mixing temperature compensation model.

[0206] Incremental retraining: Combine the current anomaly batch data with historical optimal data, and simulate the temperature compensation strategy (such as increasing the mixing temperature by 2 - 5°C) through the digital twin system to generate an optimized model adapted to the current raw material characteristics.

[0207] In the existing process, temperature compensation relies on manual experience adjustment, with slow response speed and easy introduction of new errors. The present invention realizes precise closed-loop control through a real-time data-driven retraining mechanism. For example, when the replacement of carbon black batches causes the dispersion to decrease, the system automatically identifies the coordinated deterioration trend of specific surface area and resistivity, triggers the update of the mixing temperature compensation model, and dynamically increases the heating power of the mixer to compensate for the change in raw material characteristics.

[0208] The real-time quality data acquisition and retraining trigger mechanism solves the pain points of existing quality control through the technical path of multi-dimensional monitoring → dynamic threshold determination → model self-optimization:

[0209] Quality closed-loop control: The multi-dimensional matrix captures process abnormalities in real time, avoiding batch quality accidents;

[0210] Improved adaptability: Incremental retraining enables the temperature compensation model to continuously adapt to raw material fluctuations;

[0211] Energy efficiency collaborative optimization: The unit energy consumption index is involved in the evaluation to prevent the out-of-control energy consumption caused by quality improvement. The present invention provides a double guarantee for the quality stability and process economy of carbon black masterbatch production.

[0212] Specifically, for the automatic control method of the carbon black masterbatch production equipment based on PCL of the present invention, the iterative update includes:

[0213] When no foreign object feature is detected, expand the dimension of the foreign object detection map in the knowledge base;

[0214] Start the digital twin re-simulation process for the scenario model with a matching failure rate exceeding the threshold;

[0215] Establish a traceable process database including the historical optimal model version;

[0216] The traceable process database includes:

[0217] Store the device firmware information and raw material batch data corresponding to each historical version model;

[0218] Provide a comparative analysis function of production process parameters based on the time dimension.

[0219] Foreign object detection map dimension:

[0220] Multi-modal feature fusion: When the on-line detection system (such as a vision camera, metal detector) captures an unrecorded foreign object signal, extract the morphological features (contour, size), material features (spectral reflectivity) and rheological features (abnormal melt pressure fluctuation pattern) of the foreign object, and construct a composite feature vector.

[0221] Map dynamic update: Compare the similarity between the new foreign object feature and the existing map in the knowledge base. If the matching degree is lower than the threshold, automatically create a new map category and associate a disposal strategy (such as shutdown alarm, pneumatic rejection). For example, when an unknown gel particle is detected, record its infrared spectral feature and update it to the "organic foreign object" map branch.

[0222] Cross-device linkage: The foreign object feature library is linked with the control parameters of the internal mixer and the extruder. When the same type of foreign object appears repeatedly, automatically optimize the feeding rate or screw speed to reduce the probability of foreign object mixing.

[0223] The existing foreign object detection relies on a fixed feature library and cannot cope with new pollutants (such as new additive residues). By dynamically expanding the map dimension, the system can independently identify unrecorded foreign object types. For example, when a small amount of metal chips are mixed into the carbon black, the system quickly locates the pollution source and updates the disposal strategy through the time correlation between the melt pressure mutation and the metal detection signal, reducing the batch quality risk.

[0224] Digital twin re-simulation process trigger:

[0225] Dynamic monitoring of failure rate: Count the number of matching failures of the knowledge base model in continuous production tasks (for example, 3 out of 10 tasks fail to meet the quality threshold), calculate the failure rate and compare it with the preset threshold (such as 25%).

[0226] Re-simulation parameter reconstruction: Resample the raw material ratio and equipment status data of the failure scenario, adjust the boundary conditions of the digital twin model (such as melt viscosity correction coefficient and screw wear compensation parameter), and generate a training data set suitable for the current working conditions.

[0227] Model incremental training: Only the sub-models associated with the failed scenario are retrained, and the optimized parameters of other scenarios are retained to reduce computing resource consumption. For example, for the matching failure of the "PP-based-carbon black 30%" scenario, only the material dispersion process under this ratio is re-simulated.

[0228] Existing model updates require full retraining, which is inefficient. By triggering local re-simulation based on the failure rate, the system can quickly repair model deviations in specific scenarios. For example, when the original speed model fails due to extruder screw wear, the re-simulation process generates a compensation model based on the current screw gap parameters to restore the stability of the melt pressure.

[0229] Construction of traceable process database:

[0230] Versioned storage: Store the historical optimal model version and its associated metadata by timestamp, including device firmware version (such as PLC program version number), sensor calibration records, and raw material batch information (carbon black D50 particle size, resin melt index).

[0231] Index optimization: Establish a multi-dimensional index structure to support rapid retrieval of historical process data by time range, raw material type, equipment model, etc. For example, when searching for the task "PA6-based conductive masterbatch in 2023", the optimal mixing temperature parameters for the corresponding period are automatically associated.

[0232] Comparative analysis tool: Provides a timeline sliding comparison function to visualize the correlation and change trends between process parameters and quality indicators in different periods. For example, compare the volume resistivity distribution under the new and old screw speed models to assist in process optimization decisions.

[0233] Existing process data is stored in a scattered manner, making it difficult to trace the source of abnormalities. With a traceable database, technicians can quickly locate historical problems. For example, when the resistivity of a batch of products is abnormal, tracing back to the raw material batch data reveals fluctuations in the specific surface area of ​​carbon black, thereby optimizing the incoming material inspection standards. In addition, the comparative analysis function supports historical verification of process parameters to avoid repetitive trial and error.

[0234] The iterative update mechanism systematically improves the process adaptability through a technical closed loop of dynamic learning → local optimization → data backtracking:

[0235] Foreign object prevention and control upgrade: The atlas is dynamically expanded to enhance the identification and disposal efficiency of unknown pollutants;

[0236] Model rapid repair: The targeted resimulation of failure scenarios shortens the model iteration cycle by more than 50%;

[0237] Knowledge precipitation and reuse: The traceable database supports the structured accumulation and rapid call of process optimization experience. The present invention provides a closed-loop self-optimization ability for the continuous process improvement of carbon black masterbatch production.

[0238] Specifically, the automatic control method for carbon black masterbatch production equipment based on PCL described in the present invention further includes:

[0239] Establish a buffer material level closed-loop feedback mechanism between the discharge valve of the mixing equipment and the feeding port of the extrusion equipment;

[0240] When the abnormality of material transfer continuity is detected, adaptively adjust the timing relationship between the feeding rate and the opening degree of the discharge valve.

[0241] Establishment of buffer material level closed-loop feedback mechanism:

[0242] Multi-sensor fusion monitoring: Deploy a laser level gauge and a weighing sensor in the buffer bin to collect the material accumulation height and quality data in real time, and comprehensively judge the material level state in combination with the vibration sensor signal (such as amplitude frequency) at the feeding port of the extruder.

[0243] Dynamic threshold setting: Dynamically adjust the safe material level range according to material characteristics (carbon black content, resin fluidity). For example, when the material with a high carbon black ratio has poor fluidity, raise the lower limit of the material level by 10%-15% to prevent material shortage.

[0244] Anti-interference filtering processing: Perform moving average filtering and mutation suppression processing on the material level sensor signal to eliminate instantaneous interference caused by material collapse or mechanical vibration.

[0245] In the existing open-loop control, the discharge valve and the feeding equipment operate independently, and it is easy to cause empty bin or overflow of the buffer bin due to timing deviation. The closed-loop feedback improves the continuity of material transfer through real-time material level monitoring and dynamic threshold adjustment. For example, when it is detected that the material level in the buffer bin continues to be lower than the lower limit, the system automatically delays the closing time of the discharge valve and simultaneously increases the feeding rate of the extruder to maintain the stability of material supply.

[0246] Detection and adaptive adjustment of abnormal material transfer:

[0247] Abnormal mode recognition: Analyze the correlation between the material level change rate and the equipment operation parameters. For example, when the opening degree of the discharge valve reaches 100% but the material level still continues to drop, it is determined that the feeding of the internal mixer is insufficient or the feeding rate is too high.

[0248] Timing Relationship Optimization: Establish a timing correlation model between the actions of the discharge valve and the feeding rate, and dynamically adjust the action interval through the proportional-derivative (PD) control algorithm. For example, when the feeding rate increases by 20%, the opening time of the discharge valve is advanced by 0.5 seconds to match the feeding demand.

[0249] Cross-device Cooperative Control: Link the opening degree of the discharge valve of the internal mixer and the rotational speed of the screw of the extruder. When an abnormal material level in the buffer bin is detected, synchronously reduce the screw rotational speed and decrease the opening degree of the discharge valve to prevent material accumulation or excessive shearing.

[0250] In the existing methods, abnormal material transmission often causes severe fluctuations in the melt pressure of the extruder (such as a sudden drop in pressure caused by material breakage), affecting the uniformity of carbon black dispersion. Through timing self-adaptation adjustment, the system can quickly respond to abnormal working conditions. For example, when the carbon black agglomerates block the discharge valve and cause a feeding interruption, the system immediately reduces the rotational speed of the extruder and triggers a reverse blockage clearing program, and at the same time adjusts the subsequent production rhythm to compensate for the downtime, minimizing the production of defective products to the greatest extent.

[0251] The closed-loop feedback mechanism for the buffer material level solves the problem of the stability of cross-device material transmission through the technical path of real-time perception → dynamic decision-making → cooperative control:

[0252] Continuous Production Assurance: Dynamic material level monitoring and device cooperation prevent material breakage or overflow, improving the operation efficiency of the production line;

[0253] Optimization of Dispersion Uniformity: A stable material supply reduces fluctuations in the melt pressure of the extruder, inhibiting the secondary agglomeration of carbon black;

[0254] Abnormal Quick Response: The timing self-adaptation adjustment mechanism reduces the abnormal downtime of the equipment by more than 30%. The present invention provides key control assurance for the efficient and continuous operation of carbon black masterbatch production.

[0255] The technical terms of the present invention are explained as follows:

[0256] PCL (Programmable Control Logic, programmable control logic):

[0257] Definition: A programmable control architecture based on an industrial automation system, used to realize the logical control and cooperative scheduling of carbon black masterbatch production equipment.

[0258] Application Scenario: In the present invention, PCL serves as the core control platform, responsible for multi-source heterogeneous data fusion, cross-device instruction issuance, and real-time feedback processing, replacing the decentralized control mode of the existing independent PLC units.

[0259] Dynamic Time Warping Algorithm (Dynamic Time Warping, DTW):

[0260] Definition: An algorithm for processing the morphological similarity matching of time series data, which supports the alignment of sequences with different lengths or time axis stretching.

[0261] Application scenario: Used to match the preheating curve of the current device with the historical model in the knowledge base, eliminate the timing deviation caused by equipment aging or environmental differences, and improve the accuracy of model screening.

[0262] Digital Twin technology:

[0263] Definition: A technology that real-time maps the physical device state through a virtual model, supporting multi-physical field simulation and parameter optimization of the production process.

[0264] Application scenario: Build a virtual simulation environment for the carbon black dispersion process, simulate the melt rheological behavior under different raw material ratios, and generate an optimized combination of process parameters.

[0265] Multi-source heterogeneous data fusion:

[0266] Definition: The process of time alignment and feature extraction for asynchronous and multi-structured data from different devices (internal mixer, extruder, testing equipment). Application scenario: Unify heterogeneous data such as the torque of the internal mixer and the melt pressure of the extruder into a synchronous process parameter sequence to support cross-process dynamic coupling analysis.

[0267] Hierarchical model knowledge base:

[0268] Definition: A process knowledge storage system organized in a three-level structure of raw material ratio, process parameters, and quality indicators, supporting scenario-based model retrieval and invocation.

[0269] Application scenario: Quickly match historical optimized parameters according to the production task type (such as PA6-based / PS-based) to shorten the process debugging cycle.

[0270] Buffer level closed-loop feedback mechanism:

[0271] Definition: A closed-loop control method that real-time monitors the material transfer state through a level sensor and dynamically adjusts the device action timing.

[0272] Application scenario: Prevent material accumulation or material shortage between the discharging of the internal mixer and the feeding of the extruder to ensure production continuity.

[0273] Fourier Transform Encoding:

[0274] Definition: A data processing method that converts a time-domain signal into a frequency-domain energy distribution feature, used to extract the core features of the device operating state.

[0275] Application scenario: Conduct frequency-domain analysis on the torque fluctuation data of the internal mixer to identify the key frequency band characteristics of the carbon black dispersion state.

[0276] Melt pressure balance model:

[0277] Definition: A mathematical model that describes the dynamic relationship between the melt pressure of the extruder, screw speed, and feeding rate. Application scenario: Generate screw speed compensation commands based on the real-time pressure change trend to suppress the fluctuation of carbon black dispersion.

[0278] Volume Resistivity:

[0279] Definition: A key indicator characterizing the electrical conductivity of materials, with the unit of Ω·cm. The lower the value, the better the electrical conductivity.

[0280] Application scenario: As the core parameter for evaluating the quality of carbon black masterbatch, it is used for closed-loop optimization of the mixing temperature and screw speed.

[0281] Foreign object detection enhancement model:

[0282] Definition: A model for pollutant identification and disposal strategy based on multi-modal data (vision, spectroscopy, rheology).

[0283] Application scenario: For high carbon black ratio scenarios, enhance the detection sensitivity and rejection efficiency of foreign objects such as metal chips and gel particles.

[0284] Traceable process database:

[0285] Definition: A structured database that stores historical optimal process models, equipment status, and raw material batch data, supporting comparative analysis in the time dimension.

[0286] Application scenario: Quickly trace the production parameters of quality abnormal batches, assisting in root cause analysis and process optimization.

[0287] Quality qualification rate weighted evaluation function:

[0288] Definition: A dynamic scoring model that comprehensively considers quality indicators such as blackness value and volume resistivity, supporting adaptive adjustment of weights.

[0289] Application scenario: Dynamically optimize the priority of process objectives according to different production task types (such as high-precision conductive grade / general grade).

[0290] Explanation of the relevance of technical terms:

[0291] The above technical terms jointly construct a technical chain of "data perception → virtual simulation → dynamic matching → closed-loop control" for the production of carbon black masterbatch:

[0292] The data layer (PCL, multi-source heterogeneous data fusion) realizes the comprehensive perception of equipment status;

[0293] The model layer (digital twin, hierarchical knowledge base) supports the intelligent optimization of process parameters;

[0294] The control layer (dynamic time warping, melt pressure balance model) ensures the cross-device collaboration accuracy;

[0295] The quality layer (volume resistivity, foreign object detection) realizes the consistency of product performance.

[0296] The implementation steps of the specific embodiments of the present invention are as follows:

[0297] Equipment startup and data synchronization:

[0298] Equipment startup: The internal mixer, twin-screw extruder, and in-line inspection equipment are powered on, and the PCL system automatically loads the equipment firmware version information (internal mixer V3.2, extruder V5.1).

[0299] Data acquisition and synchronization:

[0300] Internal mixer: Temperature (unit: °C) and torque (unit: N·m) data are collected once per second;

[0301] Extruder: Melt pressure (unit: MPa) and zone temperatures (unit: °C) data are collected 10 times per second;

[0302] In-line inspection equipment: Specific surface area and volume resistivity (unit: Ω·cm) are collected in real time.

[0303] Time-domain alignment: The sliding window mechanism (window length 1 second, overlap rate 50%) is used to interpolate and compensate the low-frequency data of the internal mixer to generate a synchronized time series parameter sequence.

[0304] Hierarchical model knowledge base matching:

[0305] Scenario matching:

[0306] Input task parameters (resin, carbon black content 20%), and the system matches the "ABS-based - carbon black 15% - 25%" scenario model from the first-level index of the knowledge base;

[0307] The corresponding process parameter combination is retrieved from the second-level index: internal mixing temperature 180 - 190 °C, screw speed 65 - 75 rpm, feeding rate 80 kg / h;

[0308] The quality threshold is loaded from the third-level index: volume resistivity ≤ 1.2 × 10³ Ω·cm (safety redundancy design).

[0309] Dynamic time warping (DTW) verification:

[0310] Fourier encode the temperature rise curve of the internal mixer during the equipment preheating stage (0 - 5 minutes), calculate its dynamic bending distance from the knowledge base model, and screen out 3 candidate models;

[0311] Verify equipment compatibility: Exclude models that require an extruder speed > 80 rpm (the current equipment supports a maximum of 75 rpm).

[0312] Multi-objective optimization control:

[0313] Load the weighted evaluation function:

[0314] Since the task type is "high-precision conductive grade", the weight distribution is: qualified rate of quality 60%, energy consumption coefficient 25%, equipment load rate 15%;

[0315] Screen out the model with the highest comprehensive score: internal mixer temperature 185 °C, screw speed 70 rpm, feeding rate 85 kg / h.

[0316] Send cross-equipment collaborative control instructions:

[0317] Internal mixer: Generate an instruction to increase the linear velocity of the dispersion disc to 25 m / s based on torque spectrum analysis (low-frequency energy accounts for 45%);

[0318] Extruder: Calculate the screw speed compensation amount of -2 rpm (actual operation is 68 rpm) based on the melt pressure balance model;

[0319] Pelletizer: Set the high-speed knife group ratio to 60% and the medium-speed knife group ratio to 30% according to the D50 = 420 μm feedback by the on-line particle size analyzer.

[0320] Real-time quality monitoring and closed-loop update:

[0321] Multi-dimensional quality assessment:

[0322] Continuous 5 batches of detection data: volume resistivity (0.9 - 1.1×10³ Ω·cm), blackness value (L* = 23 - 24), unit energy consumption (0.85 kW·h / kg);

[0323] Determine that the quality meets the standard and the evaluation matrix is normal.

[0324] Foreign object detection trigger:

[0325] In the 6th batch, abnormal fluctuations (±15%) in the melt pressure were detected, and the vision system was linked to identify 0.5 mm metal chips;

[0326] Expand the foreign object atlas in the knowledge base, add the feature of "metal chips - spectral reflectivity > 90%", and trigger the reverse blockage clearing program of the extruder.

[0327] Model iteration and update:

[0328] Trace back the raw material batch number corresponding to the abnormal batch (carbon black Lot#2309) and the equipment status (extruder screw wear degree 2%).

[0329] The digital twin system simulates the scenario of metal chips mixing in and updates the foreign object detection sensitivity parameters of the "ABS-based - 20% carbon black" model.

[0330] Closed-loop feedback of buffer material level:

[0331] Abnormal handling:

[0332] During the production of the 7th batch, the buffer silo material level continuously remained below the lower limit (30%), and the system determined that the opening degree of the internal mixer discharge valve was insufficient.

[0333] Adaptive adjustment: The opening degree of the discharge valve was increased from 80% to 95%, and the feeding rate was decreased from 85 kg / h to 80 kg / h, and the material level balance was restored within 10 seconds.

[0334] Effect of the embodiment:

[0335] Process efficiency: The task switching time was shortened from 4 hours to 20 minutes.

[0336] Quality indicators: The fluctuation range of volume resistivity was reduced by 50%, and the standard deviation of blackness value ≤ 0.5.

[0337] Energy efficiency optimization: The unit energy consumption decreased by 15%, and the equipment load rate was stable at 85% - 90%.

[0338] Summary of the embodiment:

[0339] Through the full-process verification of data synchronization → model matching → multi-objective optimization → closed-loop control in this embodiment, the practical application value of the technical solution of the present invention in the production of carbon black masterbatch is proved, significantly improving the process stability and product consistency, and providing reliable technical support for the large-scale production of high-performance conductive masterbatch.

[0340] The present invention effectively solves the problems of dispersion uniformity and process energy efficiency optimization in the production of carbon black masterbatch through the efficient fusion and dynamic collaborative control of multi-source heterogeneous data. First, aiming at the problems of different data sampling frequencies and structures of multiple devices, a sliding window mechanism is used to perform time-domain alignment processing on the asynchronous data streams of the mixing equipment and the extrusion equipment, and combined with outlier detection to generate synchronous process parameter sequences. By extracting the frequency-domain features and waveform features of the equipment operation parameters, a unified data representation framework is constructed to eliminate the time-axis deviation of cross-process parameters such as internal mixer temperature and extrusion pressure, laying a foundation for the establishment of a dynamic correlation model.

[0341] Secondly, a multi-dimensional production task simulator is constructed through digital twin technology. By combining the raw material ratio parameters with the equipment physical boundary constraint model, a simulation data set under different ratio scenarios is generated. A hierarchical model knowledge base is established based on the hierarchical relationship among the raw material ratio, process parameters, and quality indicators, and the dynamic time warping algorithm is used to match real-time tasks with historical models. Through Fourier transform encoding features and dynamic time warping distance calculation, candidate control models compatible with the current equipment state are screened out. Then, combined with multi-objective optimization sorting such as the quality pass rate and energy consumption coefficient, cross-device collaborative control instructions are generated to achieve precise linkage adjustment of parameters such as the line speed of the dispersing disc of the internal mixer and the screw speed of the extruder.

[0342] Finally, a closed-loop optimization is formed through real-time quality data feedback and knowledge base iterative update. A multi-dimensional quality assessment matrix is constructed to monitor key indicators such as blackness value and volume resistivity. When it is detected that the continuous batch data exceeds the threshold, the retraining of the mixing temperature compensation model is automatically triggered. At the same time, a traceable process database is established to store the historical optimal model version and equipment operation data, supporting the rapid traceability of abnormal scenarios and digital twin re-simulation. Combining the buffer level closed-loop feedback mechanism to dynamically adjust the material transfer timing, ensuring production continuity and process stability, thereby continuously improving the carbon black dispersion uniformity and the regulation accuracy of the conductivity threshold.

Claims

1. The automatic control method of carbon black masterbatch production equipment based on PCL is characterized in that: include: Collect heterogeneous data streams of mixing equipment, extrusion equipment and detection equipment in real time, perform time domain alignment processing on the heterogeneous data streams to generate a synchronous process parameter sequence, and extract frequency domain characteristics and waveform characteristics of equipment operation parameters; A multi-dimensional production task simulator including raw material ratio parameters is constructed based on the equipment operation parameters, simulation data sets of different ratio scenarios are generated through digital twin technology, and a hierarchical model knowledge base is established according to the hierarchical relationship between raw material ratios, process parameters and quality indicators; When receiving a new production task, parsing the task parameters and matching the candidate control model from the hierarchical model knowledge base through a dynamic time warping algorithm based on the current equipment state characteristics; Perform multi-objective optimization sorting on the candidate control models according to the quality indicator priorities, and convert key parameters of the selected models into cross-device collaborative control instructions; During the execution of the control instructions, production quality data is collected in real time, and iterative updating of the hierarchical model knowledge base is triggered based on the quality deviation detection results.

2. The automatic control method for carbon black masterbatch production equipment based on PCL according to claim 1, characterized in that: The time domain alignment process includes: A sliding window mechanism is used to compensate the sampling frequency of the asynchronous data streams of the mixing device and the extrusion device; Outlier detection is performed on the compensated data sequence to generate a set of process parameters with continuously aligned timestamps.

3. The automatic control method for carbon black masterbatch production equipment based on PCL according to claim 1, characterized in that: The construction of a multi-dimensional production task simulator comprises: Establish the constraint relationship model between raw material ratio parameters and physical boundaries of extrusion equipment; The constraint relationship model is used to drive the digital twin system to simulate the material dispersion process and generate a training data set including a mixing temperature and torque correlation curve.

4. The automatic control method for carbon black masterbatch production equipment based on PCL according to claim 1, characterized in that: The establishment of the hierarchical model knowledge base includes: Divide the basic scenario model by raw material ratio type as the first-level index; Storing the corresponding process parameter optimization combination under the basic scenario model as a secondary index; The process parameter combinations are associated with quality detection thresholds to form a three-level quality assessment standard.

5. The automatic control method for carbon black masterbatch production equipment based on PCL according to claim 1, characterized in that: The dynamic time warping algorithm includes: Perform Fourier transform encoding on the real-time collected equipment preheating status data; Calculate the dynamic time warping distance between the encoded feature vector and the feature matrix of the knowledge base model; Screening models whose bending distance is less than a preset threshold as a set of candidate control models; Verify the compatibility of equipment for the selected candidate control models; When it is detected that the current device hardware version does not match the model requirements, the knowledge base physical boundary parameter update is automatically triggered.

6. The automatic control method for PCL-based carbon black masterbatch production equipment according to claim 1, characterized in that: The multi-objective optimization sorting includes: Construct a weighted evaluation function including quality qualification rate, energy consumption coefficient and equipment load rate; Dynamically adjust the weight distribution ratio of the evaluation function according to the production task type; The foreign body detection enhancement model is forced to load for high carbon black ratio scenarios.

7. The automatic control method for PCL-based carbon black masterbatch production equipment according to claim 1, characterized in that: The conversion into a cross-device collaborative control instruction includes: Generate a dispersion disk linear speed adjustment instruction according to the torque spectrum characteristics of the mixing device; Calculate the screw speed compensation amount based on the extruder pressure balance model and generate a speed correction instruction; The speed gradient parameters of the pelletizing equipment are dynamically set according to the particle size distribution characteristics.

8. The automatic control method for PCL-based carbon black masterbatch production equipment according to claim 1, characterized in that: The real-time collection of production quality data includes: A multidimensional quality assessment matrix including specific surface area, DBP oil absorption value, volume resistivity and specific energy consumption was constructed; When consecutive batches of data in the evaluation matrix exceed an associated quality threshold, retraining of the hybrid temperature compensation model is triggered.

9. The automatic control method for PCL-based carbon black masterbatch production equipment according to claim 1, characterized in that: The iterative update includes: When unrecorded foreign body features are detected, the foreign body detection graph dimension in the knowledge base is expanded; Initiate the digital twin re-simulation process for scenario models whose matching failure rate exceeds the threshold; Establish a traceable process database including historical optimal model versions; The traceable process database includes: Store the equipment firmware information and raw material batch data corresponding to each historical version model; Provides comparative analysis of production process parameters based on the time dimension.

10. The automatic control method for PCL-based carbon black masterbatch production equipment according to claim 7, characterized in that: Also includes: Establish a closed-loop feedback mechanism for buffering material level between the discharge valve of the mixing equipment and the feed port of the extrusion equipment; When abnormal material transmission continuity is detected, the timing relationship between the feeding rate and the opening of the discharge valve is adaptively adjusted.

Images