A method and system for analyzing the homogeneity of a plastic melt

Abstract

The application relates to the technical field of plastic processing, and discloses a plastic melting uniformity analysis method and system. The method comprises the following steps: collecting fluctuation of a supply amount and material distribution data, obtaining a temperature deviation fluctuation interval and a material abnormal distribution graph; combining the temperature deviation and a dynamic disturbance frequency to judge a fluctuation type and generate temperature distribution abnormality prediction data; extracting key features from the temperature distribution abnormality prediction data to determine a plastic flow characteristic change trend; integrating dynamic characteristic data to evaluate the influence degree of a melting process on uniformity; quantifying a shear rate mutation response based on the influence degree to adjust a melting time fluctuation interval parameter; generating a uniformity distribution graph through multi-scenario simulation to determine a product quality quantization index; generating an adjustment strategy according to the index and iteratively optimizing the adjustment strategy to finally determine melting process control parameters. The application solves the problems of poor melting uniformity and control lag caused by disturbance coupling, and improves the stability of a plastic melting process and the consistency of product quality.

Description

A method and system for analyzing the uniformity of plastic melt flow Technical Field

[0001] This application relates to the field of plastic processing technology, and in particular to a method and system for analyzing the uniformity of plastic melt. Background Technology

[0002] In plastic molding processes such as injection molding, extrusion, and blow molding, the uniformity of materials during the melting stage is a key factor determining the final product's mechanical properties, dimensional accuracy, surface quality, and long-term stability. Melt uniformity not only affects the product's mechanical strength and appearance but also directly relates to the energy efficiency and material utilization rate of the production process. With the increasing demands for performance consistency in high-end plastic products and the urgent need for refined process control in intelligent manufacturing, real-time monitoring, intelligent analysis, and adaptive regulation of the melting process have become core technological challenges for the transformation and upgrading of the plastics processing industry.

[0003] Current methods for controlling the uniformity of plastic melting processes have significant limitations. Common technical approaches primarily rely on the independent monitoring and adjustment of single-point or local parameters: most systems measure the temperature at specific locations in the melting zone using only a few thermocouples and perform open-loop or simple closed-loop control based on preset heating curves or screw speeds. This method cannot comprehensively perceive real-time changes in the feed rate and abnormal material spatial distribution caused by fluctuations in the feeding system, material agglomeration, equipment wear, or environmental changes during the melting process. For example, in twin-screw extrusion, the dispersion state of the material directly affects heat transfer efficiency, but existing methods lack the ability to quantitatively monitor the real-time distribution of material particles. When disturbances occur during production, traditional methods struggle to distinguish the type of disturbance in a timely manner, let alone assess the specific impact mechanisms of different disturbances on the melt temperature field, shear field, and flow characteristics. Although some improved methods attempt to introduce visual inspection or ultrasonic monitoring, these technologies often only focus on surface phenomena or single physical quantities, failing to achieve multi-dimensional integrated analysis of temperature data, material distribution data, flow data, and equipment operation data. At the control strategy level, while the widely used PID control can achieve temperature stability, its fixed parameters are difficult to adapt to dynamic conditions such as nonlinear changes in material viscosity and sudden changes in shear rate during melting. More importantly, existing methods lack the ability to predictively assess melt uniformity: the system can only passively adjust after uniformity problems have already appeared, and cannot predict uniformity trends based on real-time data and optimize parameters in advance. Under complex production conditions, this lag control often leads to problems such as large batch-to-batch quality fluctuations, decreased yield of high-quality products, and increased energy consumption. In addition, traditional methods generally lack a systematic iterative optimization mechanism, and cannot improve control parameters by continuously learning from production data, resulting in low process optimization efficiency.

[0004] To address the above deficiencies, this application combines real-time multi-source data acquisition, intelligent identification of fluctuation types, and flow characteristic prediction and analysis to achieve early diagnosis and adaptive iterative optimization of process parameters for multiple disturbances such as supply fluctuations and abnormal distribution during the melting process. This solves the problem of uneven temperature and material distribution caused by disturbance coupling, and improves the control accuracy, uniformity stability, and product quality consistency of the plastic melting process. Summary of the Invention

[0005] This application provides a method and system for analyzing the uniformity of plastic melting, which solves the problem of uneven temperature and material distribution caused by disturbance coupling, and improves the control accuracy, uniformity stability and product quality consistency of the plastic melting process.

[0006] In a first aspect, this application provides a method for analyzing the uniformity of plastic melt flow, the method comprising:

[0007] Step S101: Collect supply fluctuation data and material distribution data in real time, obtain the fluctuation range of temperature distribution deviation in the melting area and the abnormal distribution map of material particle dispersion state, and determine the supply fluctuation amplitude and material distribution abnormality index.

[0008] Step S102: Based on the fluctuation range of the supply and the abnormal material distribution index, combined with the temperature distribution deviation and dynamic disturbance frequency distribution, determine the fluctuation type and generate abnormal temperature distribution prediction data.

[0009] Step S103: Extract key features from the temperature distribution anomaly prediction data to determine the changing trend of plastic flow characteristics;

[0010] Step S104: Based on the changing trend of the plastic flow characteristics, integrate the real-time collected dynamic characteristic data and evaluate the degree of influence of the melting process on the uniformity of plastic melting.

[0011] Step S105: Based on the degree of influence, quantify the abrupt response of the shear rate and adjust the melting time fluctuation range parameter;

[0012] Step S106: Based on the adjusted melting time fluctuation range parameters, generate uniformity distribution maps under different disturbance conditions through multi-scenario simulation to determine the quantitative indicators of product quality.

[0013] Step S107: Based on the product quality quantification index, generate an adjustment strategy for uneven dispersion, iteratively optimize the melting uniformity improvement parameters, and determine the final melting process control parameters based on the optimized parameters.

[0014] Secondly, this application provides a plastic melt uniformity analysis system for implementing the aforementioned plastic melt uniformity analysis method, the system comprising:

[0015] The data acquisition module is used to collect supply fluctuation data and material distribution data in real time, obtain the fluctuation range of temperature distribution deviation in the melting area and the abnormal distribution map of material particle dispersion state, and determine the supply fluctuation amplitude and material distribution abnormality indicators.

[0016] The fluctuation analysis module is used to determine the fluctuation type and generate temperature distribution anomaly prediction data based on the fluctuation amplitude of the supply and the abnormal material distribution index, combined with the temperature distribution deviation and dynamic disturbance frequency distribution.

[0017] The feature extraction module is used to extract key features from the temperature distribution anomaly prediction data to determine the changing trend of plastic flow characteristics;

[0018] The impact assessment module is used to integrate real-time collected dynamic characteristic data based on the changing trend of the plastic flow characteristics to assess the degree of impact of the melting process on the uniformity of plastic melting.

[0019] The fluctuation range adjustment module is used to quantify the abrupt response of the shear rate based on the degree of influence and adjust the melting time fluctuation range parameter.

[0020] The simulation quality control module is used to generate uniformity distribution maps under different disturbance conditions through multi-scenario simulation based on the adjusted melting time fluctuation range parameters, and to determine the quantitative indicators of product quality.

[0021] The strategy iteration module is used to generate adjustment strategies for uneven dispersion based on the product quality quantification indicators, iteratively optimize the melting uniformity improvement parameters, and determine the final melting process control parameters based on the optimized parameters.

[0022] This application proposes a method and system for analyzing the uniformity of plastic melting, which solves the problem of uneven temperature and material distribution caused by disturbance coupling, and improves the control accuracy, uniformity stability, and product quality consistency of the plastic melting process. Compared with the prior art, the beneficial effects of the technical solution of this application are at least as follows:

[0023] First, it can collect and integrate multi-source data in real time, including supply fluctuations, material distribution status, and melting zone temperature distribution, to achieve multi-dimensional comprehensive perception of the melting process status. This overcomes the limitations of traditional methods that rely on single-point or local monitoring, and improves the comprehensiveness and accuracy of process status description.

[0024] Secondly, based on fused data, it can intelligently identify fluctuation types, distinguish between gradual and sudden disturbance modes, and generate anomaly prediction data by combining dynamic disturbance frequency and temperature distribution deviation, thereby achieving early warning and cause location of potential uniformity problems, enhancing the predictability and pertinence of process control.

[0025] Third, by extracting key features from the predicted data and analyzing the changing trends of plastic flow characteristics, the impact of the melting process on uniformity can be dynamically assessed, providing a quantitative basis for parameter adjustment and solving the problems of traditional control methods relying on experience and having a delayed response.

[0026] Fourth, the shear rate mutation response is quantified based on the degree of influence, and the melting time fluctuation range parameter is adaptively adjusted, realizing the dynamic matching of process parameters and process state, and improving the system's adaptability and control accuracy under nonlinear and time-varying conditions.

[0027] Fifth, based on the adjusted parameters, multi-scenario simulations are performed to generate uniformity distribution maps under different disturbance conditions and determine product quality quantification indicators, thereby realizing predictive evaluation and quality prediction of melting results and supporting process optimization and risk avoidance before production.

[0028] Sixth, based on quantitative indicators, iteratively generate and optimize adjustment strategies to ultimately determine the control parameters of the melting process, forming a closed-loop optimization mechanism of "monitoring-analysis-prediction-adjustment-verification" to continuously improve melting uniformity, ensure stable and consistent product quality, and reduce energy consumption and scrap rate. Attached Figure Description

[0029] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 is a schematic flowchart of a plastic melt uniformity analysis method according to this application;

[0031] Figure 2 is a temperature distribution anomaly prediction diagram in this application;

[0032] Figure 3 shows the prediction results of flow resistance by the support vector regression algorithm in this application.

[0033] Figure 4 is a diagram of the training process of the energy loss correction evaluation model in this application;

[0034] Figure 5 is a comparison chart of the comprehensive performance indicators in this application;

[0035] Figure 6 is a schematic diagram of the structure of a plastic melt uniformity analysis system according to this application. Detailed Implementation

[0036] This application provides a method and system for analyzing the uniformity of plastic melt flow. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0037] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to Figure 1. An embodiment of a method for analyzing the uniformity of plastic melt flow in this application includes:

[0038] Step S101: Collect supply fluctuation data and material distribution data in real time, obtain the fluctuation range of temperature distribution deviation in the melting area and the abnormal distribution map of material particle dispersion state, and determine the supply fluctuation amplitude and material distribution abnormality index.

[0039] In one specific embodiment, step S101 includes the following steps:

[0040] The flow sensor and image sensor in the plastic processing equipment collect real-time data on supply fluctuations and material distribution, and extract data on temperature distribution deviation and material particle dispersion state from them.

[0041] The temperature distribution deviation data is decomposed into a time series, and the maximum and minimum values ​​are extracted from the decomposed time series to form the fluctuation range of the temperature distribution deviation.

[0042] Perform regional density analysis on the particle dispersion data of the material, mark the areas with density below the average value as abnormal areas, and generate an abnormal distribution map of the particle dispersion state of the material.

[0043] Calculate the width of the temperature distribution deviation fluctuation range and use the width as the fluctuation range of the supply.

[0044] The proportion and intensity of abnormal areas in the statistical anomaly distribution map are analyzed, and the proportion and intensity are weighted and fused to obtain the material distribution anomaly index.

[0045] Specifically, in the plastic melting and processing scenario, flow sensors and image sensors installed in the plastic processing equipment work together to collect data. The flow sensor captures the change value of the supply once per second, while the image sensor synchronously collects the material position coordinate information. From the raw data collected by the two types of sensors, temperature distribution deviation data and material particle dispersion state data are directly filtered and separated, providing a direct data source for subsequent data processing.

[0046] For temperature distribution deviation data, wavelet transform is used for time series decomposition. Daubechies wavelet is selected as the mother wavelet function, and the decomposition scale is set to 5 levels. Multi-scale decomposition is performed on the time series signal corresponding to the temperature distribution deviation data to separate the trend component and the fluctuation component. The trend component reflects the slow change pattern of temperature, while the fluctuation component captures the rapid disturbance of temperature. By traversing all data points of the decomposed time series, the maximum and minimum values ​​are extracted to construct the fluctuation range of temperature distribution deviation. For example, if the maximum temperature value is 152℃ and the minimum temperature value is 138℃ after decomposition, then the fluctuation range of temperature distribution deviation is [138℃, 152℃].

[0047] For the particle dispersion data, a gridding method is used to divide the molten area into 10×10 equal-area sub-regions. The number of particles per unit volume in each sub-region is calculated to obtain the regional density. The density of each sub-region is compared with the average density of all sub-regions. Sub-regions with a density 30% or more below the average density are marked as abnormal regions. The distribution of abnormal regions is presented in the form of a heatmap, where the color depth is positively correlated with the degree of abnormality. Red indicates high abnormal regions with a density more than 50% below the average density, and orange indicates medium abnormal regions with a density 30% to 50% below the average density. At the same time, Gaussian filtering is applied to the heatmap, with a Gaussian kernel size of 3×3 and a standard deviation of 1.2. The pixel values ​​are weighted and averaged using the Gaussian kernel to reduce noise interference, generating an abnormal distribution map of the particle dispersion state.

[0048] Calculate the width of the temperature distribution deviation fluctuation range. The width value is the supply fluctuation amplitude. If the fluctuation range is [138℃, 152℃], then the supply fluctuation amplitude is 152℃ - 138℃ = 14℃. Statistically calculate the proportion of the abnormal area to the entire melting area in the abnormal distribution map. Simultaneously, calculate the ratio of the abnormal area density to the average density to determine the abnormal intensity. Set the weight of the proportion to 0.4 and the weight of the intensity to 0.6. Use the formula P = 0.4Q + 0.6R (where P is the material distribution anomaly index, Q is the abnormal area proportion, and R is the abnormal intensity) to weight and fuse the proportion and intensity to obtain the material distribution anomaly index. For example, if the abnormal area proportion is 0.35 and the abnormal intensity is 0.7, then the material distribution anomaly index is 0.4 × 0.35 + 0.6 × 0.7 = 0.56.

[0049] During data processing, the supply change values ​​collected by the flow sensor directly correspond to the temperature distribution deviation data, and the material position coordinates collected by the image sensor directly correspond to the material particle dispersion state data. The raw and processed data form a clear source correlation. When decomposing the temperature distribution deviation data using wavelet transform, multi-scale decomposition separates different characteristics of temperature changes. The decomposition scale ensures accurate capture of temperature fluctuations at different frequencies, providing data support for precise acquisition of fluctuation ranges. The maximum and minimum values ​​of the decomposed data directly determine the boundaries of the fluctuation range. When processing the material particle dispersion state data using gridded analysis and Gaussian filtering, the number of grid divisions ensures the refinement of region density calculations, while the kernel size and standard deviation parameters of the Gaussian filter ensure the smoothness and accuracy of the anomaly distribution map, reducing noise interference in anomaly region identification. The supply fluctuation amplitude is directly derived from the fluctuation range width, and the material distribution anomaly index is obtained through weighted fusion of the proportion and intensity of anomaly regions. A quantitative correspondence is formed between the processed data and the final index.

[0050] Step S102: Based on the fluctuation range of supply and the abnormal indicators of material distribution, combined with the temperature distribution deviation and the dynamic disturbance frequency distribution, determine the fluctuation type and generate temperature distribution anomaly prediction data.

[0051] In one specific embodiment, step S102 includes the following steps:

[0052] Based on the fluctuation range of supply and abnormal material distribution indicators, combined with temperature distribution deviation and dynamic disturbance frequency distribution, a data comparison and analysis method is used to determine the fluctuation type.

[0053] If the fluctuation range of the supply exceeds the preset range threshold and the material distribution anomaly index is higher than the preset distribution threshold, the fluctuation type is determined to be a gradual disturbance mode.

[0054] The degree of uneven heat conduction and the time delay of dynamic characteristics are extracted from the gradual disturbance mode, and the material flow influence value and the cumulative value of distribution anomaly are obtained through multi-dimensional feature mapping.

[0055] Based on the influence value of material flow and the cumulative value of distribution anomalies, predictive data for temperature distribution anomalies are generated.

[0056] Specifically, based on the determined fluctuation range of the supply and abnormal material distribution indicators, combined with real-time collected temperature distribution deviation data and dynamic disturbance frequency distribution data, a data comparison and analysis method is used to determine the fluctuation type. The preset amplitude threshold is set according to the rated operating parameters of the plastic processing equipment and the material melting process requirements. For example, for the PP plastic extrusion process, the preset amplitude threshold is set to 12℃, and the preset distribution threshold is set to 0.5. The dynamic disturbance frequency distribution is obtained by performing Fourier transform on the vibration signal and motor current fluctuation signal during the operation of the processing equipment. The Fourier transform uses the Hanning window function, and the sampling frequency is set to 100Hz. After transformation, frequency distribution data in the range of 0~50Hz is obtained, which directly reflects the frequency characteristics of various disturbances during equipment operation. The fluctuation range of the supply is numerically compared with a preset range threshold, and the abnormal material distribution index is quantitatively compared with a preset distribution threshold. If the calculated fluctuation range of the supply is 14°C, which exceeds the preset range threshold of 12°C, and the abnormal material distribution index is 0.56, which is higher than the preset distribution threshold of 0.5, then the fluctuation type is determined to be a gradual disturbance mode. This judgment logic directly addresses the technical problem that traditional methods cannot distinguish the type of disturbance, and achieves accurate identification of gradual disturbances.

[0057] Under the gradual perturbation mode, the degree of heat conduction non-uniformity is calculated using a heat conduction model. This non-uniformity is calculated based on a heat conduction model constructed using Fourier's law, with the core expression of Fourier's law: heat flux density vector = -material thermal conductivity × temperature gradient vector. The material thermal conductivity is pre-set to 0.23 W / (m·K) based on the plastic material. A three-dimensional steady-state heat conduction control equation is constructed using a 1:1 geometric model of the molten region. The actual boundary temperature of the molten region is set as the boundary condition, and room temperature (25℃) as the initial condition. The model is discretized into finite mesh elements and solved using the finite volume method. During the calculation, the temperature values ​​of all spatially discrete points are first extracted from the temperature distribution deviation data, and then calculated using the finite volume method. The three-dimensional temperature gradient at each point is calculated using a fractional method. The heat flux density is obtained by substituting it into Fourier's law. Then, the heat conduction rate of each region is calculated by combining the heat transfer area of ​​each grid cell. Subsequently, the average heat conduction rate of all regions is calculated, and the absolute difference between the heat conduction rate of each region and the average value is calculated. The ratio of this difference value to the average heat conduction rate is used as the quantification result of the degree of heat conduction non-uniformity. The dynamic characteristic time delay is calculated by a cross-correlation analysis algorithm. Temperature distribution deviation data and dynamic disturbance frequency distribution data are selected as input sequences. The time lag range is set to 0~10 seconds with a step size of 0.1 seconds. The correlation coefficient under different lag times is calculated. The lag time corresponding to the maximum value of the correlation coefficient is the dynamic characteristic time delay.

[0058] A multidimensional feature mapping algorithm is used to process the degree of uneven heat conduction and the time delay of dynamic characteristics. The multidimensional feature mapping algorithm adopts a radial basis function neural network. The input layer of the network has 2 neurons, which correspond to the degree of uneven heat conduction and the time delay of dynamic characteristics, respectively. The hidden layer has 15 neurons, the activation function is a Gaussian function, and the function variance parameter is set to 0.8. The output layer has 2 neurons, which output the material flow influence value and the cumulative value of distribution anomaly, respectively. During network training, 1000 sets of sample data on the degree of uneven heat conduction and dynamic characteristic time delay under different working conditions were selected. The corresponding material flow influence value and distribution anomaly cumulative value were obtained through experimental measurement. The sample data were divided into training set and test set in a 7:3 ratio. The number of training iterations was set to 500, and the learning rate was set to 0.01. The mean square error between the network prediction value and the actual measurement value was minimized by the gradient descent method. After training, the network model can achieve accurate mapping from input features to output values. Among them, the material flow influence value quantifies the degree of interference of uneven heat conduction on material flow, and the distribution anomaly cumulative value reflects the superposition effect of material distribution anomaly caused by dynamic characteristic time delay. This mapping process solves the technical defect of traditional methods that cannot quantify the specific impact of disturbance on the melting process and establishes a direct correlation between disturbance features and changes in melting state.

[0059] Based on the obtained material flow impact values ​​and cumulative distribution anomaly values, a temperature distribution anomaly prediction model is constructed. This model employs a Long Short-Term Memory (LSTM) network. The input layer consists of two feature vectors: the material flow impact value and the cumulative distribution anomaly value. The hidden layer contains three LSTM units, with 64 units per layer. The dropout rate is set to 0.2 to prevent overfitting. The output layer outputs the predicted temperature distribution deviation values ​​for the next five time steps, with a time step interval of 1 second. During model training, three months of historical production data were selected, containing 50,000 sets of material flow impact values, cumulative distribution anomaly values, and corresponding subsequent temperature distribution deviation data. These data were divided into training and validation sets in an 8:2 ratio. The Adam optimizer was used during training, with the mean squared error loss function. The number of iterations was 300, and the initial learning rate was 0.001, decreasing by 50% every 100 iterations. After training, the currently calculated material flow impact values ​​and cumulative distribution anomaly values ​​are input into the LSTM model, which outputs the predicted temperature distribution deviation data for each location in the melting area within the next five seconds—the temperature distribution anomaly prediction data. This process, through a well-defined model structure, training parameters, and input / output settings, enables predictive analysis of temperature distribution anomalies. It overcomes the technical bottleneck of traditional methods that can only passively respond to uniformity issues, providing lead time for subsequent parameter adjustments and ensuring the stability of the melting process.

[0060] Step S103: Extract key features from the temperature distribution anomaly prediction data to determine the changing trend of plastic flow characteristics.

[0061] In one specific embodiment, step S103 includes the following steps:

[0062] The local temperature difference peak and temperature gradient change rate of the plastic melting zone are extracted from the temperature distribution anomaly prediction data as key features;

[0063] A correlation model between local temperature difference peak and plastic viscosity change was established, and a mapping relationship between temperature gradient change rate and plastic flow resistance was constructed.

[0064] The plastic viscosity fluctuation value corresponding to the local temperature difference peak is calculated by the correlation model, and the plastic flow resistance increment corresponding to the temperature gradient change rate is obtained by the mapping relationship.

[0065] By combining the fluctuation values ​​of plastic viscosity and the increment of plastic flow resistance, the change law of plastic fluidity in the molten state is analyzed, and the trend of plastic flow characteristics is determined.

[0066] Specifically, the temperature distribution anomaly prediction data includes predicted temperature values ​​at different time steps and spatial locations within the melting region. Based on this data, the local temperature difference peak and temperature gradient change rate of the plastic melting region are extracted as key features. The local temperature difference peak is obtained by traversing the temperature data of each spatial location within the prediction time range. For each spatial coordinate point, the difference between its predicted maximum and minimum temperature values ​​is calculated, and the maximum value among all differences is the local temperature difference peak. For example, in a certain set of prediction data, the local temperature difference peak obtained by traversing is 12℃. The temperature gradient change rate is obtained by calculating the spatial gradient of the temperature distribution data. The finite difference method is used, with a spatial step size of 0.01m. For the three-dimensional temperature data of each spatial point, the partial derivatives in the x, y, and z directions are calculated respectively. The square root of the sum of the squares of the partial derivatives in the three directions is used to obtain the temperature gradient of that point. Then, the ratio of the difference in temperature gradient between adjacent time steps to the time interval is calculated to obtain the temperature gradient change rate. The time interval is set to 1 second. For example, the calculated temperature gradient change rate is 0.8℃ / (m·s). This feature extraction method directly solves the technical problem that traditional methods only focus on a single physical quantity and lack the ability to capture key features of dynamic changes in temperature distribution, thus providing accurate data support for subsequent flow characteristic analysis.

[0067] A correlation model between local temperature difference peaks and plastic viscosity changes was established. The model was constructed using a backpropagation (BP) neural network, with the local temperature difference peak as the input layer and the plastic viscosity fluctuation as the output layer. The neural network consists of an input layer, hidden layers, and an output layer. The input layer has one neuron, the hidden layers have two layers with 20 neurons each, and the ReLU activation function is used. The output layer has one neuron and uses a linear activation function. During model training, experimental data of different plastic materials were selected as training samples, with 800 sets of samples. Each set of samples includes the measured value of the local temperature difference peak and the corresponding measured value of the plastic viscosity fluctuation. The viscosity was measured using a rotational rheometer at 180℃ and a shear rate of 100s. -1 Measurements were obtained under the specified conditions. The samples were divided into training and test sets in a 7:3 ratio. The number of training iterations was set to 1000, and the learning rate was 0.005. An adaptive moment estimation optimization algorithm was used to adjust the network weights and biases, with the mean squared error used as the loss function. When the loss function value was below 1×10⁻⁶, the network weights and biases were adjusted. -4 Training is stopped when the time is right. Once the training is complete, the model can map local temperature difference peaks to viscosity fluctuation values.

[0068] Simultaneously, a mapping relationship between the temperature gradient change rate and the flow resistance of plastics was constructed. This mapping was implemented using a support vector regression algorithm, with a radial basis function kernel set to 0.5 and a penalty coefficient of 10. The training samples consisted of 600 sets of measured data on the temperature gradient change rate and corresponding flow resistance under different operating conditions. The flow resistance was calculated by measuring the pressure difference between the inlet and outlet of the molten zone using a pressure sensor with a measurement accuracy of 0.01 MPa. After standardizing the sample data, it was input into the support vector regression model. Five-fold cross-validation was used to optimize the model parameters, determining the optimal kernel function parameters and penalty coefficient. This ensured that the correlation coefficient between the model's predicted values ​​and the measured values ​​reached above 0.95, guaranteeing the accuracy of the mapping relationship.

[0069] The extracted local temperature difference peak is input into the trained BP neural network model to calculate the corresponding plastic viscosity fluctuation value. For example, inputting a local temperature difference peak of 12℃ will result in a viscosity fluctuation value of 200 Pa·s. The temperature gradient change rate is input into the support vector regression model to obtain the corresponding plastic flow resistance increment. For example, inputting a temperature gradient change rate of 0.8℃ / (m·s) will result in a flow resistance increment of 5 MPa. The plastic viscosity fluctuation value directly reflects the viscosity difference within the material, while the flow resistance increment reflects the additional resistance encountered by the material during flow. Together, they constitute the core parameters describing the flowability of the molten state, overcoming the technical deficiency of traditional methods that cannot quantify the specific impact of temperature changes on flow characteristics.

[0070] By combining the viscosity fluctuation value and the increase in flow resistance of plastics, the fluidity variation law of plastics in the molten state is analyzed. The viscosity fluctuation percentage is obtained by calculating the ratio of the viscosity fluctuation value to the initial viscosity. The initial viscosity is preset according to the plastic material; for example, the initial viscosity of PE plastic is 800 Pa·s, and the viscosity fluctuation percentage is 25%. Simultaneously, the resistance increment percentage is obtained by calculating the ratio of the increase in flow resistance to the initial flow resistance. The initial flow resistance is calculated to be 20 MPa based on the equipment's rated parameters, and the resistance increment percentage is 25%. Based on the changing trends of the viscosity fluctuation percentage and the resistance increment percentage, and considering their synchronicity over time, the changing trend of plastic flow characteristics is determined. If both the viscosity fluctuation percentage and the resistance increment percentage show an upward trend, and the change exceeds a preset threshold, it is judged that the flow characteristics are developing in the direction of deteriorating fluidity. This analysis process establishes a direct correlation between temperature anomalies and changes in flow characteristics, providing a clear basis for subsequent assessment of the impact on the melting process and achieving accurate prediction of changes in flow characteristics.

[0071] Step S104: Based on the changing trend of plastic flow characteristics, integrate the dynamic characteristic data collected in real time and evaluate the degree of influence of the melting process on the uniformity of plastic melting.

[0072] In one specific embodiment, performing step S104 includes the following steps:

[0073] Based on the changing trend of plastic flow characteristics, the range of plastic viscosity variation and data noise intensity are extracted;

[0074] The dynamic characteristic data collected in real time is integrated through multi-source data fusion technology. The dynamic characteristic data includes supply fluctuation data and material distribution data within the processing equipment.

[0075] If the viscosity of the plastic changes beyond the preset range and the data noise intensity is higher than the preset limit, the energy loss of the melting process will be corrected and evaluated based on the dynamic characteristic data after fusion.

[0076] Based on the revised evaluation results, the extent to which the melting process affects the uniformity of plastic melting is determined.

[0077] Specifically, based on the changing trends of plastic flow characteristics, the range of plastic viscosity variation and data noise intensity are further extracted from the plastic viscosity fluctuation values ​​and plastic flow resistance increments obtained by feature extraction and model calculation from the temperature distribution anomaly prediction data. The range of plastic viscosity variation is determined by statistically analyzing the difference between the maximum and minimum values ​​of all viscosity fluctuation values ​​within a certain time window (e.g., 10 seconds). For example, if the maximum viscosity fluctuation value is 350 Pa·s and the minimum is 150 Pa·s within this time window, then the range of plastic viscosity variation is 200 Pa·s. The data noise intensity is calculated using the root mean square error method. Based on the time series of viscosity fluctuation values, a sliding window size of 5 data points is set. The sum of squares of the deviations between the data in each window and the window mean is calculated, and the square root of the average value is taken to obtain the data noise intensity. If the calculated result is 25 Pa·s, then the data noise intensity for that period is 25 Pa·s.

[0078] Multi-source data fusion technology integrates real-time acquired dynamic characteristic data, including supply fluctuation data collected by flow sensors within the processing equipment and material distribution data collected by image sensors. The multi-source data fusion technology employs a Kalman filter algorithm, treating supply fluctuation data and material distribution data as two independent observation sources to establish state equations and observation equations. The state equations are centered on the dynamic equilibrium state of the melting process, with the state vector containing two state variables: the supply stability coefficient and the material distribution uniformity. The state transition matrix is ​​set to [[0.98,0.02],[0.03,0.97]], and the process noise covariance matrix is ​​set to diag([0.01,0.01]). In the observation equations, the observation matrix for supply fluctuation data is set to [1,0], and the observation matrix for material distribution data is set to [0,1]. The observation noise covariance matrix is ​​set according to the sensor accuracy; the observation noise covariance for the flow sensor is 0.008, and the observation noise covariance for the image sensor is 0.012. Through the prediction-update iterative process of Kalman filtering, for each set of real-time data received (sampling interval of 1 second), the current state vector is first predicted based on the state equation, and then the predicted value is corrected by combining the observation equation to obtain the fused dynamic characteristic data. This data contains comprehensive dynamic information on supply and material distribution, which solves the technical problem that traditional methods have failed to achieve multi-dimensional data fusion.

[0079] The preset range is determined based on the melting process requirements of the plastic material. For example, for ABS plastic injection molding, the preset viscosity variation range is 100~180 Pa·s, and the preset limit is set to 20 Pa·s based on sensor performance and the noise range allowed by the process. When the plastic viscosity variation range (200 Pa·s) exceeds the preset range (100~180 Pa·s), and the data noise intensity (25 Pa·s) is higher than the preset limit (20 Pa·s), the energy loss in the melting process is corrected and evaluated based on the fused dynamic characteristic data. The energy loss correction evaluation uses a BP neural network model. The model input layer consists of four input features: the supply stability coefficient, material distribution uniformity, plastic viscosity variation range, and data noise intensity from the fused dynamic characteristic data. The hidden layer has three layers, with 30, 20, and 15 neurons in each layer, respectively. The activation function for all layers is the Sigmoid function. The output layer is the energy loss correction coefficient for the melting process. During model training, 5000 sets of experimental data were selected as training samples. Each set of samples included measured values ​​of supply stability coefficient, material distribution uniformity, plastic viscosity variation range, data noise intensity, and the corresponding measured value of energy loss correction coefficient (calculated by measuring the input-output energy difference during the melting process using a calorimeter). The samples were divided into training and test sets in an 8:2 ratio. The number of training iterations was set to 800, with an initial learning rate of 0.003. The learning rate decreased by 40% every 200 iterations. The gradient descent method was used to minimize the mean absolute error between the predicted and measured values. Training was stopped when the mean absolute error of the training set was below 0.02. After training, the fused dynamic characteristic data, extracted plastic viscosity variation range, and data noise intensity were input into the BP neural network model, which outputs the energy loss correction coefficient. For example, if the output result is 1.3, the initial energy loss calculation value (obtained by multiplying the equipment's rated power by the operating time) is corrected based on this coefficient. The corrected energy loss = initial energy loss × energy loss correction coefficient.

[0080] Based on the revised energy loss results, the degree of influence of the melting process on the uniformity of plastic melting is determined. A quantitative model of the influence degree is established, using the revised energy loss as the core input parameter, and combining the deviation rate of the plastic viscosity variation range from the preset interval and the deviation rate of the data noise intensity from the preset limit value, to construct a multiple linear regression equation: Where Y is the quantified value of the impact, E is the corrected energy loss, Vr is the deviation rate of the plastic viscosity variation range ((actual viscosity variation range - preset upper limit) / preset upper limit), Nr is the deviation rate of the data noise intensity ((actual data noise intensity - preset limit) / preset limit), and a, b, c, and d are regression coefficients. These coefficients were obtained by fitting 1000 sets of sample data, resulting in a=0.005, b=0.3, c=0.25, and d=0.1. For example, if the corrected energy loss is 800J, the deviation rate of the plastic viscosity variation range is (200-180) / 180≈0.111, and the deviation rate of the data noise intensity is (25-20) / 20=0.25, then the quantified value of the impact is Y=0.005×800+0.3×0.111+0.25×0.25+0.1=4.1958. This quantitative value directly corresponds to the level of influence. The larger the value, the more significant the negative impact of the melting process on uniformity. This process, through multi-parameter fusion and quantitative modeling, solves the technical deficiency of traditional methods that cannot accurately assess the degree of influence of the melting process on uniformity. It provides a quantitative basis for subsequent parameter adjustments, makes up for the shortcomings of traditional control that relies on experience judgment, and makes the assessment of the degree of influence more objective and accurate, meeting the needs of intelligent manufacturing for refined process control.

[0081] Step S105: Based on the degree of influence, quantify the abrupt response of the shear rate and adjust the melting time fluctuation range parameter.

[0082] In one specific embodiment, step S105 includes the following steps:

[0083] Collect shear rate data and dynamic signal data of the plastic melting zone;

[0084] Signal abrupt change features are extracted from dynamic signal data, and these features are correlated with shear rate data to establish a correspondence between shear rate and signal abrupt change.

[0085] A quantitative analysis of the abrupt response of shear rate in the corresponding relationship is performed to generate quantitative analysis results;

[0086] When the impact exceeds the preset impact threshold, the parameters of the melting time fluctuation range are adjusted based on the quantitative analysis results.

[0087] Specifically, when collecting shear rate data and dynamic signal data in the plastic melting zone, data acquisition is accomplished using shear rate sensors and dynamic signal sensors installed in the melting chamber of the plastic processing equipment. The shear rate sensor adopts the torque measurement principle and collects the shear rate data of the molten plastic under the shearing action of the screw at a sampling frequency of 50Hz. The dynamic signal sensor synchronously collects the motor current signal and cavity vibration signal during equipment operation, with a sampling frequency set to 100Hz. The data collected by the two types of sensors are transmitted to the data processing unit in real time through the equipment data bus. The shear rate data directly reflects the shear stress state of the material during the melting process, while the dynamic signal data contains signal change information caused by equipment operation disturbances, providing raw data support for subsequent abrupt feature extraction.

[0088] When extracting signal abrupt change features from dynamic signal data, the motor current signal and the cavity vibration signal are preprocessed separately. A moving average filtering algorithm is then used to denoise both types of signals, with a moving average window size of 20 data points. Abnormal fluctuation points in the original data are replaced by the mean of the data within the window, eliminating the interference of sensor noise on signal abrupt change identification. The first-order difference method is then used to calculate the signal change rate of the denoised dynamic signal data. The difference formula is as follows: ,in Let k be the rate of change of the signal at time k. Let k be the dynamic signal value at time k. For the dynamic signal value at time k-1, the threshold values ​​for the signal rate of change are set to 0.05V / ms (for current signals) and 0.2mm / s. 2 (For vibration signals) When the rate of change of the signal at a certain moment exceeds the corresponding threshold, that moment is marked as a signal mutation point. The signal amplitude, mutation duration, and mutation slope corresponding to the mutation point are extracted as signal mutation features. Each mutation point corresponds to a set of three-dimensional feature vectors.

[0089] When correlating signal mutation features with shear rate data, a one-to-one correspondence between the signal mutation feature vector and the shear rate data at the same time point is established using the timestamp as the matching benchmark, constructing a three-dimensional data matrix containing time, signal mutation feature, and shear rate dimensions. The Pearson correlation coefficient algorithm is used to quantify the correlation strength between the two, calculating the correlation coefficient between each feature component of the signal mutation features and the shear rate data. The correlation coefficient R is calculated as follows:

[0090]

[0091] in For the i-th group of signal abrupt change characteristic component values, This is the mean of the characteristic component. The data represents the shear rate at time i. Let n be the mean of the shear rate data and n be the number of data groups. Set the correlation coefficient threshold to 0.7, filter out the feature components with correlation coefficients higher than the threshold, and construct a model of the correspondence between shear rate and signal mutation. This model clarifies which signal mutation features will cause significant changes in shear rate, and solves the technical problem that traditional methods cannot establish a correlation between perturbation signals and shear states.

[0092] When quantifying the abrupt response of shear rate in the corresponding relationship, a support vector machine (SVM) algorithm is used to construct a quantization model. The model input is a filtered signal abrupt response feature vector, and the output is the quantized value of the shear rate abrupt response. During model training, 1200 sets of historical data are selected as training samples. Each set of samples contains a signal abrupt response feature vector and the corresponding measured value of the shear rate abrupt response (obtained by the difference between the data before and after the abrupt response collected by a shear rate sensor). The samples are divided into training and test sets in an 8:2 ratio. The training set is used for model parameter optimization, and the test set is used for model performance verification. A multinomial kernel function is selected for the model kernel function, with a kernel function parameter of 3 and a penalty coefficient of 15. The SMO algorithm is used to solve for the optimal classification hyperplane, and the number of training iterations is set to 500. When the mean absolute error of the test set is less than 0.03s... -1 Training is stopped when the signal mutation feature vector extracted in real time is input into the model, and the output shear rate mutation response quantification value is obtained. This value directly reflects the intensity and range of influence of the shear rate mutation, realizing a quantitative description of the shear rate mutation response and making up for the technical deficiency of traditional methods that cannot quantify the impact of shear mutation.

[0093] The preset influence threshold is determined based on the uniformity requirements of the plastic processing technology. For example, for the PC plastic extrusion process, the preset influence threshold is set to 0.65. When the quantified value of the influence obtained in step S104 is greater than this preset influence threshold, it indicates that the negative impact of the melting process on uniformity has exceeded the allowable range, and the melting time fluctuation range parameter needs to be adjusted. The parameter is adjusted according to the mapping relationship between the quantified value of the shear rate mutation response and the melting time adjustment amount, and a mapping function is constructed. ,in This is an adjustment amount for melting time. This is the quantization value of the shear rate abrupt change response. This is the proportionality coefficient. The baseline adjustment was obtained by fitting data from 500 sets of experimental data. If the quantization value of the shear rate abrupt change response is 0.5 s... -1 Then the melting time adjustment amount The initial melt time fluctuation range parameter is based on the equipment's rated process settings, for example, an initial range of [25s, 35s]. The upper and lower limits of this range are adjusted synchronously according to the calculated adjustment amount. The adjusted melt time fluctuation range is [25+0.7, 35+0.7] = [25.7s, 35.7s]. If the quantized response value of the shear rate mutation is negative, it indicates that the shear rate mutation has a positive impact on uniformity. In this case, the melt time fluctuation range is adjusted negatively according to the same logic. This adjustment process directly guides the optimization of melt time parameters through the quantified shear rate mutation response data, establishing a dynamic matching mechanism between process disturbances and process parameter adjustments. This solves the technical problem that traditional PID control with fixed parameters cannot adapt to dynamic conditions such as shear rate mutations, achieving adaptive adjustment of melt time parameters, improving the response speed and control accuracy of the melting process to dynamic disturbances, and ensuring the uniformity of plastic melting.

[0094] Step S106: Based on the adjusted melting time fluctuation range parameters, generate uniformity distribution maps under different disturbance conditions through multi-scenario simulation to determine the quantitative indicators of product quality.

[0095] In one specific embodiment, step S106 includes the following steps:

[0096] Based on the adjusted melting time fluctuation range parameters, select the actual working condition range corresponding to the dynamic disturbance frequency distribution, supply fluctuation amplitude and material distribution anomaly index, and construct multiple sets of disturbance conditions;

[0097] For each set of disturbance conditions, a simulation model of the melting process matching the plastic processing equipment is built, and the adjusted melting time fluctuation range parameters are input into the simulation model;

[0098] The melting process is simulated using a simulation model, and the temperature distribution data and material particle dispersion data of the melting region are output under each set of disturbance conditions.

[0099] Temperature uniformity is calculated based on temperature distribution data, and dispersion uniformity coefficient is obtained by combining material particle dispersion data.

[0100] Based on temperature uniformity and dispersion uniformity coefficient, a uniformity distribution map is generated for each set of perturbation conditions.

[0101] Extract the overall uniformity score and the proportion of abnormal areas from the uniformity distribution map, and determine the overall uniformity score and the proportion of abnormal areas as quantitative indicators of product quality.

[0102] Specifically, based on the adjusted melting time fluctuation range parameters and actual operating data of plastic processing equipment, the actual operating conditions corresponding to the dynamic disturbance frequency distribution, supply fluctuation amplitude, and material distribution anomaly index are selected. The dynamic disturbance frequency distribution selects typical frequency points in the 0~50Hz range, and divides them into 11 frequency values ​​at 5Hz intervals. The supply fluctuation amplitude is based on the statistical distribution of historical production data, and five gradient values ​​of 8℃, 10℃, 12℃, 14℃, and 16℃ are selected. The material distribution anomaly index selects five quantification levels of 0.4, 0.5, 0.6, 0.7, and 0.8. Multiple sets of disturbance conditions are constructed through a three-factor full combination method. Each set of disturbance conditions contains a unique combination of dynamic disturbance frequency, supply fluctuation amplitude, and material distribution anomaly index, forming a total of 11×5×5=275 sets of disturbance conditions, comprehensively covering the disturbance scenarios that may occur during the production process.

[0103] For each set of disturbance conditions, a simulation model of the melting process matching the plastic processing equipment was built. This model was constructed based on the finite element method. The geometric model was modeled at a 1:1 scale according to the actual melting cavity size of the equipment, with a cavity length of 1.2m and an inner diameter of 0.15m. The mesh was generated using tetrahedral elements, with 800,000 elements and a mesh quality factor controlled above 0.8. The physical field settings included a heat conduction field, a fluid flow field, and a material diffusion field. The thermal conductivity of the heat conduction field was set to 0.23 W / (m·K) based on the plastic material, and the specific heat capacity was set to 1800 J / (kg·K). The fluid flow field was described using the Navier-Stokes equations, and the material diffusion field was modeled using Fick's second law. The adjusted melting time fluctuation range parameters were used as the input boundary conditions of the simulation model. The parameters included the upper limit value, the lower limit value, and the fluctuation period, which was set to 5 seconds to ensure that the model input was consistent with the actual process parameters. The melting process was simulated using a simulation model with a simulation time step set to 0.01 seconds. The total simulation duration was 1.5 times the upper limit of the adjusted melting time fluctuation range to ensure the melting process reached a stable state. During the simulation, the model solved the governing equations of each physical field through numerical iteration. After each time step, it output the temperature data and particle concentration data of each spatial node in the melting region at that moment, i.e., temperature distribution data and particle dispersion data. The output accuracy of the temperature data was 0.1℃, and the output accuracy of the particle concentration data was 0.01 kg / m³. 3 The simulation results for each set of perturbation conditions include complete time series data and spatial distribution data.

[0104] Temperature uniformity is calculated based on temperature distribution data using the standard deviation coefficient method. First, the mean and standard deviation of temperature data at all spatial nodes within the melting region are calculated. Temperature uniformity is equal to 1 minus the ratio of the standard deviation to the mean. This calculation method quantifies the dispersion of temperature distribution through the ratio of the standard deviation to the mean. The temperature uniformity value ranges from 0 to 1; the closer to 1, the more uniform the temperature distribution. Then, a dispersion uniformity coefficient is obtained by combining material particle dispersion data using the coefficient of variation method. The coefficient of variation of material particle concentration at each spatial node is calculated. The dispersion uniformity coefficient is equal to 1 minus the coefficient of variation, where the coefficient of variation is equal to the ratio of the standard deviation of particle concentration to the mean of particle concentration. The dispersion uniformity coefficient also ranges from 0 to 1; a larger value indicates more uniform particle dispersion. Based on temperature uniformity and dispersion uniformity coefficient, a uniformity distribution map is generated for each set of disturbance conditions and presented in the form of a two-color overlay thermogram. The horizontal axis represents the axial length of the melting cavity, and the vertical axis represents the radial position. Temperature uniformity is mapped through a red color spectrum, with the red depth increasing as temperature uniformity increases. Dispersion uniformity coefficient is mapped through a green color spectrum, with the green depth increasing as dispersion uniformity coefficient increases. In the overlay image, the yellow area represents the area with both high temperature uniformity and dispersion uniformity coefficient, and the purple area represents the area with both low uniformity, intuitively reflecting the uniformity distribution state within the melting area.

[0105] The uniformity score and the proportion of abnormal areas are extracted from the uniformity distribution map. The uniformity score is calculated using a weighted summation method: Uniformity Score = 0.6 × Temperature Uniformity + 0.4 × Dispersion Uniformity Coefficient. The weight of temperature uniformity is set to 0.6, and the weight of the dispersion uniformity coefficient is set to 0.4, reflecting the weighted relationship between temperature distribution and material dispersion on melt uniformity. The score ranges from 0 to 1. The proportion of abnormal areas is calculated using an image recognition algorithm. Areas with a temperature uniformity below 0.7 or a dispersion uniformity coefficient below 0.7 are defined as abnormal areas. The proportion of abnormal areas is obtained by statistically analyzing the ratio of the volume of abnormal areas to the total volume of the melt cavity. The uniformity score and the proportion of abnormal areas are established as quantitative indicators of product quality, directly linking melt uniformity to product quality. This solves the technical problem of traditional methods lacking quantitative assessment of melt uniformity, providing clear data support for the generation of subsequent adjustment strategies. Furthermore, by using multi-scenario simulations to predict product quality under different disturbance conditions, it overcomes the limitation of traditional methods that can only passively adjust after quality problems appear, achieving proactive control of the melting process.

[0106] Step S107: Based on the product quality quantification indicators, generate adjustment strategies for uneven dispersion, iteratively optimize the melting uniformity improvement parameters, and determine the final melting process control parameters based on the optimized parameters.

[0107] In one specific embodiment, step S107 includes the following steps:

[0108] By comparing the quantitative indicators of product quality with the preset qualified standards, the target quantitative indicators that have not been met are identified.

[0109] Based on the target quantitative indicators, combined with the quantitative results of the influence of the melting process on the uniformity of plastic melting and the abrupt response of shear rate, the core causes of uneven dispersion were identified.

[0110] Based on the core causes, an initial adjustment strategy is generated that includes fine-tuning parameters for melting time and temperature compensation parameters;

[0111] The initial adjustment strategy is applied to the melting process of plastic processing equipment, and the supply fluctuation data, material distribution data and temperature distribution data are collected in real time by sensors.

[0112] Based on the collected data, the product quality quantitative indicators are recalculated to determine whether the recalculated product quality quantitative indicators all meet the preset qualification standards.

[0113] If not, optimize the initial adjustment strategy and apply the optimized initial adjustment strategy to the melting process of the plastic processing equipment again until the product quality quantitative indicators reach the preset qualified standards. Then, determine the last optimized initial adjustment strategy as the final melting process control parameter.

[0114] If so, the initial adjustment strategy should be directly determined as the final melting process control parameter.

[0115] Specifically, the product quality quantitative indicators are compared with the preset qualification standards. The preset qualification standards set the uniformity comprehensive score threshold at 0.85 and the abnormal area proportion threshold at 0.15. The uniformity comprehensive score and abnormal area proportion obtained in step S106 are compared with their corresponding thresholds. If, under a certain disturbance condition, the uniformity comprehensive score is 0.78, lower than the threshold of 0.85, and the abnormal area proportion is 0.22, higher than the threshold of 0.15, then these two indicators are selected as non-compliant target quantitative indicators. This is combined with the impact degree quantitative value obtained in step S104 (e.g., 4.1958) and the shear rate mutation response quantitative value obtained in step S105 (e.g., 0.5s). -1The core causes of uneven dispersion were identified. A causal relationship model was constructed, with the target quantitative index deviation value (uniformity comprehensive score deviation = 0.85-0.78 = 0.07, abnormal area proportion deviation = 0.22-0.15 = 0.07), the influence degree quantitative value, and the shear rate mutation response quantitative value as inputs. A decision tree algorithm was used to classify the causes. The root node of the decision tree is the influence degree quantitative value, the branch nodes are the shear rate mutation response quantitative value and the target quantitative index deviation value, respectively, and the leaf nodes are the core cause types. The algorithm determined that the core causes were localized low temperatures due to uneven heat conduction and insufficient material dispersion caused by shear rate mutations.

[0116] Targeting the core causes, an initial adjustment strategy is generated, including melt time fine-tuning parameters and temperature compensation parameters. The melt time fine-tuning parameters are calculated based on the quantized value of the shear rate abrupt response, using a linear mapping formula: ,in M represents the fine-tuning amount of the melting time, and M is the quantized value of the shear rate abrupt change response. Substituting M=0.5s... -1 ,get If the original adjusted melting time fluctuation range was [25.7s, 35.7s], then the fine-tuned range is [25.7 + 0.5, 35.7 + 0.5] = [26.2s, 36.2s]. The temperature compensation parameter is determined based on the quantified value of the degree of influence, using a proportional-integral algorithm. ,in Y represents the temperature compensation amount, and Y represents the quantified value of the degree of influence. The difference between the width of the temperature distribution deviation fluctuation range and the preset width (e.g., the preset width is 10℃, the actual width is 14℃, ΔV=4℃) is substituted into Y=4.1958. ,get For the low-temperature areas marked in the abnormal distribution map, the temperature compensation parameter is set to 1.66℃.

[0117] The initial adjustment strategy was applied to the melting process of the plastic processing equipment. Flow sensors, image sensors, and temperature sensors were used to collect real-time data on supply fluctuations, material distribution, and temperature distribution at sampling frequencies of 1Hz, 1Hz, and 50Hz, respectively. The sampling duration was twice the upper limit of the adjusted melting time fluctuation range (i.e., 36.2 × 2 = 72.4 s) to ensure coverage of the complete melting cycle. Based on the collected data, the product quality quantification indicators were recalculated. The recalculation process followed the processing logic of steps S101 to S106: time-series decomposition of the newly collected temperature distribution data was performed to extract the fluctuation range; regional density analysis was performed on the material distribution data to generate an anomaly distribution map; the fluctuation type was determined by combining the dynamic disturbance frequency distribution; key features were extracted to determine the trend of flow characteristic changes; the degree of impact was assessed; melting time parameters were adjusted; and a uniformity distribution map was generated through a simulation model. Finally, a new comprehensive uniformity score and anomaly area percentage were obtained. Assuming the recalculated comprehensive uniformity score was 0.82, still below the threshold of 0.85, and the anomaly area percentage was 0.18, above the threshold of 0.15, the initial adjustment strategy needed to be optimized. The optimization process employs a gradient descent algorithm, with the objective functions being maximizing the overall uniformity score and minimizing the proportion of abnormal regions. The optimization objective value is calculated as 0.7 × overall uniformity score - 0.3 × proportion of abnormal regions. Melting time fine-tuning parameters and temperature compensation parameters are used as optimization variables, with a learning rate of 0.05 and 100 iterations. In each iteration, the parameter values ​​are adjusted according to the gradient direction of the objective function. After the first optimization, the melting time fine-tuning is increased by 0.3 s, resulting in an adjusted range of [26.5 s, 36.5 s], and the temperature compensation is increased by 0.4℃, adjusting to 2.06℃. The optimized adjustment strategy is then applied again to the melting process, repeating the data acquisition and quality quantification index calculation steps. If, after the third optimization, the overall uniformity score is 0.86 and the proportion of abnormal regions is 0.14, both meeting the preset qualification standards, then the adjustment strategy after the third optimization is determined as the final melting process control parameters.

[0118] This process accurately identifies the core causes by comparing target quantitative indicators with qualified standards, generates targeted adjustment strategies based on quantitative data, and continuously corrects parameters through an iterative optimization mechanism. This solves the technical problems of traditional methods lacking a systematic iterative optimization mechanism and having low process optimization efficiency. At the same time, through real-time data acquisition and feedback, the adjustment strategy is dynamically optimized, overcoming the shortcomings of traditional control methods such as slow response and inability to adapt to dynamic working conditions. This ensures that the product quality quantitative indicators meet the qualified standards and guarantees the uniformity of plastic melting and the consistency of product quality.

[0119] Please refer to Figure 2, which shows the temperature distribution anomaly prediction graph. It displays the trends of three curves: historical temperature, measured temperature, and predicted temperature. The three curves are closely aligned, and the difference between the predicted and measured temperatures is kept within a small range. Furthermore, the predicted curve extends 5 seconds into the future based on historical data, showing a slight upward trend. The LSTM model used in this method has high prediction accuracy and can accurately predict temperature change trends, overcoming the shortcomings of traditional methods that passively respond to temperature anomalies. This provides advance time for subsequent extraction of key features and adjustment of melting time parameters, ensuring the stability of the melting process.

[0120] Please refer to Figure 3, which shows the prediction results of flow resistance using the support vector regression algorithm. The scatter plots of the actual and predicted flow resistance values ​​are distributed along the ideal prediction line. This demonstrates that the temperature gradient change rate and flow resistance mapping relationship constructed by this method has extremely high accuracy. The goodness of fit between the predicted and actual values ​​is close to 1, accurately quantifying the impact of temperature gradient changes on flow resistance. This solves the technical deficiency of traditional methods in being unable to quantitatively correlate temperature and flow characteristics, providing a scientific basis for judging the changing trend of plastic flow characteristics through the increment of flow resistance.

[0121] Please refer to Figure 4, which shows the training process of the energy loss correction evaluation model, illustrating the differences between the training and test sets. The score trends with the number of training epochs show that both curves continuously increase with the number of iterations, then plateau after 800 epochs, and finally... The score is higher than the "excellent" benchmark. This indicates that the BP neural network model of this method is well-trained and has reliable accuracy. The model's input layer includes four features: the stability coefficient of the supply, the uniformity of material distribution, the range of plastic viscosity variation, and the data noise intensity. The hidden layer has three layers. After 800 iterations and optimization with a gradually decreasing learning rate, it has accurate predictive capabilities on both the training and test sets. This demonstrates that the model can accurately correct and evaluate the energy loss in the melting process based on multi-source fusion data, overcoming the problem of large energy loss evaluation errors in traditional methods. It provides an accurate tool for quantifying the impact of the melting process on uniformity through multiple linear regression equations.

[0122] Please refer to Figure 5, which is a comparison chart of comprehensive performance indicators. It shows a quantitative comparison between the proposed method and the traditional method in six indicators: compliance rate, uniformity, stability, response speed, adaptability, and comprehensive score. The values ​​of all indicators of the proposed method are significantly higher than those of the traditional method. This indicates that the proposed method solves the defects of the traditional method, which relies on single-point monitoring, has control lag, and weak adaptability, through a closed-loop mechanism of "monitoring-analysis-prediction-adjustment-verification". Under dynamic working conditions, the proposed method can quickly respond to disturbances such as sudden changes in shear rate and fluctuations in supply, and ensure melt uniformity through adaptive parameter adjustment, which significantly improves product quality consistency and production stability.

[0123] Please refer to Figure 6. The following describes a plastic melt uniformity analysis system according to an embodiment of this application. The plastic melt uniformity analysis system includes:

[0124] The data acquisition module is used to collect supply fluctuation data and material distribution data in real time, obtain the fluctuation range of temperature distribution deviation in the melting area and the abnormal distribution map of material particle dispersion state, and determine the supply fluctuation amplitude and material distribution abnormality indicators.

[0125] The fluctuation analysis module is used to determine the fluctuation type and generate temperature distribution anomaly prediction data based on the fluctuation amplitude of supply and material distribution anomaly indicators, combined with temperature distribution deviation and dynamic disturbance frequency distribution.

[0126] The feature extraction module is used to extract key features from temperature distribution anomaly prediction data to determine the changing trend of plastic flow properties.

[0127] The impact assessment module is used to integrate real-time dynamic characteristic data based on the changing trends of plastic flow characteristics to assess the degree of impact of the melting process on the uniformity of plastic melting.

[0128] The fluctuation range adjustment module is used to adjust the melting time fluctuation range parameter by quantifying the abrupt response of the shear rate based on the degree of impact.

[0129] The simulation quality control module is used to generate uniformity distribution maps under different disturbance conditions through multi-scenario simulation based on the adjusted melting time fluctuation range parameters, and to determine the quantitative indicators of product quality.

[0130] The strategy iteration module is used to generate adjustment strategies for uneven dispersion based on product quality quantification indicators, iteratively optimize the parameters for improving melt uniformity, and determine the final melt process control parameters based on the optimized parameters.

[0131] Through the collaborative efforts of the aforementioned components, the system constructs an integrated melt uniformity control system encompassing "data perception, intelligent analysis, predictive evaluation, and closed-loop optimization." This system enables real-time monitoring, cause diagnosis, trend prediction, and adaptive iterative optimization of process parameters for multi-source disturbances during plastic melting.

[0132] The data acquisition module, serving as the system's sensing foundation, collects multi-dimensional data in real time from multiple sources, including supply fluctuations, material distribution, and temperature distribution. This enables comprehensive perception of the melting process, overcoming the limitations of traditional single-point monitoring and providing high-quality, timely data input for subsequent analysis. The fluctuation analysis module performs intelligent analysis based on the collected data. By integrating multi-dimensional information such as temperature deviation and dynamic disturbance frequency, it achieves accurate identification of fluctuation types and predictive assessment of temperature distribution anomalies. This provides data support for early warning and disturbance cause localization, enhancing the system's predictive capabilities. The feature extraction module extracts key features from the predicted data, establishing a correlation model between temperature anomalies and plastic flow characteristics. This enables quantitative analysis of flow state change trends, providing a scientific basis for subsequent impact assessment and parameter adjustment. The impact assessment module combines flow characteristic change trends and dynamic characteristic data. Through multi-source data fusion and modeling calculations, it quantitatively assesses the degree of impact of the melting process on uniformity, resolving the ambiguity of traditional experience-based judgments and improving the objectivity and accuracy of the assessment. The fluctuation range adjustment module, based on the quantification results of the impact degree and combined with the shear rate mutation response analysis, adaptively adjusts the parameters of the melting time fluctuation range, enabling process parameters to dynamically match changes in process state and enhancing the system's adaptability under nonlinear and time-varying conditions. The simulation quality control module, by constructing multi-scenario simulation models, predicts the melting uniformity distribution under different disturbance conditions based on parameter adjustments, generates a visualized uniformity distribution map, and determines quantitative indicators of product quality, achieving pre-assessment of melting results and prediction of quality risks. The strategy iteration module generates and optimizes targeted adjustment strategies based on the comparison between quantitative quality indicators and preset standards. Through a closed-loop iterative mechanism of "collection-analysis-adjustment-verification," it continuously optimizes the melting process control parameters, ultimately achieving a stable improvement in melting uniformity and ensuring consistent product quality. The collaborative operation of this system forms a complete intelligent control closed loop, not only solving the problems of poor uniformity and control lag caused by disturbance coupling in traditional melting control, but also significantly improving the stability, control accuracy, and product quality consistency of the plastic melting process through predictive analysis and adaptive optimization, providing reliable technical support for the intelligent upgrading of the plastics processing industry.

[0133] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the methods and systems described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0134] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for analyzing the uniformity of plastic melt flow, characterized in that, Includes the following steps: Step S101: Real-time acquisition of supply fluctuation data and material distribution data; obtaining the fluctuation range of temperature distribution deviation and the abnormal distribution map of material particle dispersion state within the melting area; determining the supply fluctuation amplitude and material distribution anomaly index; the supply fluctuation amplitude is obtained based on the width of the fluctuation range, and the material distribution anomaly index is obtained based on the weighted fusion of the proportion and intensity of the abnormal area in the abnormal distribution map; Step S102: Based on the supply fluctuation amplitude and the material distribution anomaly index, combined with the temperature distribution deviation and dynamic disturbance frequency distribution, determining whether the fluctuation type is a gradual disturbance mode through comparison with the corresponding preset threshold; extracting the degree of uneven heat conduction and dynamic characteristic time delay from the gradual disturbance mode; obtaining temperature distribution anomaly prediction data through multi-dimensional feature mapping; Step S103. Extract key features from the temperature distribution anomaly prediction data to determine the trend of plastic flow characteristics; S104. Based on the trend of plastic flow characteristics, integrate the real-time collected dynamic characteristic data to evaluate the degree of influence of the melting process on the uniformity of plastic melting; S105. Quantify the abrupt response of shear rate based on the degree of influence and adjust the melting time fluctuation range parameter; S106. Based on the adjusted melting time fluctuation range parameter, generate uniformity distribution maps under different disturbance conditions through multi-scenario simulation to determine the quantitative indicators of product quality; S107. Based on the quantitative indicators of product quality, generate adjustment strategies for uneven dispersion, iteratively optimize the melting uniformity improvement parameters, and determine the final melting process control parameters based on the optimized parameters.

2. The method according to claim 1, characterized in that, Step S101 includes: real-time acquisition of supply fluctuation data and material distribution data through flow sensors and image sensors in the plastic processing equipment, and extraction of temperature distribution deviation data and material particle dispersion state data from them; time series decomposition processing of the temperature distribution deviation data, and extraction of maximum and minimum values ​​from the decomposed time series to form the fluctuation range of temperature distribution deviation; regional density analysis of the material particle dispersion state data, marking areas with density below the average value as abnormal areas, and generating an abnormal distribution map of material particle dispersion state; calculating the width of the temperature distribution deviation fluctuation range, and using the width as the supply fluctuation amplitude; statistically analyzing the proportion and intensity of abnormal areas in the abnormal distribution map, and weightedly fusing the proportion and intensity to obtain material distribution abnormality indicators.

3. The method according to claim 2, characterized in that, Step S102 includes: determining the fluctuation type based on the supply fluctuation amplitude and the material distribution anomaly index, combined with the temperature distribution deviation and dynamic disturbance frequency distribution, using a data comparison analysis method; if the supply fluctuation amplitude exceeds a preset amplitude threshold and the material distribution anomaly index is higher than a preset distribution threshold, determining the fluctuation type as a gradual disturbance mode; extracting the degree of uneven heat conduction and dynamic characteristic time delay from the gradual disturbance mode, obtaining the material flow influence value and the cumulative value of distribution anomaly through multi-dimensional feature mapping; and generating temperature distribution anomaly prediction data based on the material flow influence value and the cumulative value of distribution anomaly.

4. The method according to claim 1, characterized in that, Step S103 includes: extracting the local temperature difference peak and temperature gradient change rate of the plastic melting region from the temperature distribution anomaly prediction data as key features; establishing a correlation model between the local temperature difference peak and the plastic viscosity change, and simultaneously constructing a mapping relationship between the temperature gradient change rate and the plastic flow resistance; calculating the plastic viscosity fluctuation value corresponding to the local temperature difference peak through the correlation model, and obtaining the plastic flow resistance increment corresponding to the temperature gradient change rate through the mapping relationship; combining the plastic viscosity fluctuation value and the plastic flow resistance increment to analyze the fluidity change law in the plastic melting state and determine the trend of plastic flow characteristics.

5. The method according to claim 1, characterized in that, Step S104 includes: extracting the plastic viscosity variation range and data noise intensity based on the trend of the plastic flow characteristics; integrating real-time collected dynamic characteristic data through multi-source data fusion technology, wherein the dynamic characteristic data includes supply fluctuation data and material distribution data within the processing equipment; if the plastic viscosity variation range exceeds a preset range and the data noise intensity is higher than a preset limit value, then performing a correction assessment on the energy loss of the melting process based on the fused dynamic characteristic data; and determining the degree of influence of the melting process on the uniformity of plastic melting based on the correction assessment results.

6. The method according to claim 1, characterized in that, Step S105 includes: collecting shear rate data and dynamic signal data of the plastic melting zone; extracting signal mutation features from the dynamic signal data and associating the signal mutation features with the shear rate data to establish a correspondence between shear rate and signal mutation; performing quantitative analysis on the shear rate mutation response in the correspondence to generate quantitative analysis results; and adjusting the parameters of the melting time fluctuation range according to the quantitative analysis results when the degree of influence is greater than a preset influence threshold.

7. The method according to claim 1, characterized in that, Step S106 includes: selecting the actual operating conditions corresponding to the dynamic disturbance frequency distribution, supply fluctuation amplitude, and material distribution anomaly index based on the adjusted melting time fluctuation range parameters, and constructing multiple sets of disturbance conditions; for each set of disturbance conditions, building a melting process simulation model matched with the plastic processing equipment, and inputting the adjusted melting time fluctuation range parameters into the simulation model; simulating the melting process through the simulation model, and outputting the temperature distribution data and material particle dispersion data of the melting area under each set of disturbance conditions; calculating the temperature uniformity based on the temperature distribution data, and obtaining the dispersion uniformity coefficient by combining the material particle dispersion data; generating a uniformity distribution map for each set of disturbance conditions based on the temperature uniformity and the dispersion uniformity coefficient; extracting the uniformity comprehensive score and the proportion of abnormal areas from the uniformity distribution map, and determining the uniformity comprehensive score and the proportion of abnormal areas as product quality quantification indicators.

8. The method according to claim 1, characterized in that, Step S107 includes: comparing the product quality quantification indicators with the preset qualification standards to screen out the target quantification indicators that have not met the standards; based on the target quantification indicators, combined with the degree of influence of the melting process on the uniformity of plastic melting and the quantification results of the sudden change response of shear rate, locating the core cause of uneven dispersion; generating an initial adjustment strategy including melting time fine-tuning parameters and temperature compensation parameters for the core cause; applying the initial adjustment strategy to the melting process of the plastic processing equipment, and collecting supply fluctuation data, material distribution data and temperature distribution data in real time through sensors; recalculating the product quality quantification indicators based on the collected data, and determining whether the recalculated product quality quantification indicators all meet the preset qualification standards; if not, optimizing the initial adjustment strategy, and applying the optimized initial adjustment strategy to the melting process of the plastic processing equipment again until the product quality quantification indicators meet the preset qualification standards, and determining the last optimized initial adjustment strategy as the final melting process control parameter; if yes, directly determining the initial adjustment strategy as the final melting process control parameter.

9. A plastic melt uniformity analysis system, used to implement the plastic melt uniformity analysis method as described in any one of claims 1 to 8, characterized in that, The plastic melt uniformity analysis system includes: a data acquisition module for real-time acquisition of supply fluctuation data and material distribution data, obtaining the fluctuation range of temperature distribution deviation and the abnormal distribution map of material particle dispersion state within the melting area, and determining the supply fluctuation amplitude and material distribution anomaly index; a fluctuation analysis module for determining the fluctuation type and generating temperature distribution anomaly prediction data based on the supply fluctuation amplitude and the material distribution anomaly index, combined with temperature distribution deviation and dynamic disturbance frequency distribution; a feature extraction module for extracting key features from the temperature distribution anomaly prediction data to determine the changing trend of plastic flow characteristics; and an impact assessment module for determining the... The system describes the changing trends of plastic flow characteristics, integrates real-time collected dynamic characteristic data, and assesses the impact of the melting process on the uniformity of plastic melting. A fluctuation range adjustment module is used to quantify the abrupt response of shear rate based on the stated impact level and adjust the melting time fluctuation range parameters. A simulation quality control module is used to generate uniformity distribution maps under different disturbance conditions through multi-scenario simulation based on the adjusted melting time fluctuation range parameters, determining quantitative indicators of product quality. A strategy iteration module is used to generate adjustment strategies for uneven dispersion based on the quantitative indicators of product quality, iteratively optimize the melting uniformity improvement parameters, and determine the final melting process control parameters based on the optimized parameters.

Images