Method for improving enhancement performance of polypropylene mineral filler for flow battery pole frame

Abstract

The invention discloses a method for improving the enhancement performance of a polypropylene mineral filler for a flow battery pole frame, and particularly relates to the technical field of material processing quality control, and the method comprises the following steps: constructing a processing dynamic characteristic data set by collecting processing heat flow parameters, shear rate fluctuation values and interface energy change data in real time; under the condition that the heat flow stability and the shear stress consistency meet the condition, abnormal behavior segments are identified to form candidate analysis areas; further extracting two indexes of dynamic distribution stability and interface activation coherence, and inputting the two indexes into the prediction model to generate an enhanced fitness coefficient; according to the coefficient matching optimization path, target-oriented behavior reconstruction of the machining scheme is achieved; according to the method, the machining dynamic data are collected, the double-index representation model is constructed, and the prediction mechanism is introduced, so that the precise recognition and quantitative evaluation of the machining behavior are realized, the predefined machining scheme is matched in combination with the adaptation degree coefficient, a path optimization closed loop under data driving is formed, and the machining stability and the control precision are improved.

Description

Technical Field

[0001] This invention relates to the field of material processing quality control technology, and more specifically, to a method for improving the reinforcing performance of polypropylene mineral fillers used in flow battery frames. Background Technology

[0002] Polypropylene mineral fillers used in flow battery frames refer to a polymer composite material technology that incorporates mineral fillers into the injection-molded material of the flow battery casing to enhance its performance. In flow battery systems, the battery casing not only provides structural support and cell stacking positioning but also must possess good mechanical strength, chemical stability, electrical insulation, and dimensional stability. Therefore, using pure polypropylene (PP) alone is insufficient to meet the high requirements of chemical corrosion resistance, structural rigidity, and heat distortion temperature in the operating environment of flow batteries. Thus, by introducing mineral fillers, such as talc, wollastonite, glass microspheres, kaolin, or mica, into the polypropylene resin matrix, the overall material properties can be effectively improved, forming a "polypropylene mineral filler" composite system.

[0003] The main advantages of this composite material system are as follows: First, mineral fillers typically have good dimensional stability and thermal conductivity, which can improve the dimensional retention of injection molded products and reduce warping and deformation caused by thermal expansion and contraction, especially important when the battery cell frame is in a long-term temperature difference environment; Second, some mineral materials (such as talc and wollastonite) have excellent chemical inertness, which can enhance the acid and alkali resistance and corrosion resistance of PP materials, thereby maintaining the stability of the material under long-term contact with electrolyte and dual loads of temperature and chemical reaction; Third, mineral fillers can also improve the rigidity and pressure resistance of polypropylene materials, enhance the structural load-bearing capacity of the shell under high load operation of battery stack pressing and fluid circulation, and prevent the shell from deforming or cracking due to stress concentration; Fourth, the distribution characteristics of mineral fillers themselves help to improve the internal stress distribution of the material, thereby suppressing common quality defects such as shrinkage, air threads, whitening, and silver threads during the injection molding process, and improving the appearance quality and consistency of the product.

[0004] It is worth noting that in the application scenarios of injection-molded parts such as flow battery electrode frames, in order to balance laser welding performance and optical transmittance requirements, the type, particle size, and content of mineral fillers must be strictly screened and optimized to avoid affecting the welding effect due to uneven filler particles or interference from optical properties. Furthermore, the compound design of polypropylene mineral fillers must also consider the control of material leaching under long-term immersion in electrolytes (such as sulfuric acid or fluoride solutions) to meet the high purity requirements of battery systems for metal ion (such as iron, calcium, aluminum, etc.) leaching concentrations of <5ppm or 100ppm. Therefore, "polypropylene mineral fillers" are not merely a simple physical mixture, but a customized composite material technology path guided by structural performance, chemical stability, molding quality, and functional adaptability. They are a key material foundation for achieving reliable encapsulation, long-life operation, and low-cost mass production of flow battery electrode frames.

[0005] As a type of chemical power source suitable for large-scale energy storage systems, flow batteries face complex environmental conditions involving electrolyte corrosion, high-frequency charge-discharge stress, and temperature gradient shocks. These factors impose multi-dimensional performance requirements on the mechanical strength, thermal stability, and interfacial sealing of the casing material. Currently, the industry commonly uses polypropylene as the matrix material for the flow battery electrode frame, supplemented with inorganic mineral fillers such as talc, wollastonite, and mica powder to enhance its overall performance.

[0006] However, existing technologies mainly focus on three aspects: filler type selection, filler ratio setting, and interface modifier addition, aiming to enhance and control these aspects through the intrinsic properties of the material. These methods have significant limitations when dealing with complex nonlinear processing behaviors: on the one hand, the filler dispersion is easily affected by instantaneous heat flow fluctuations and shear stress disturbances, exhibiting unstable spatial arrangement characteristics; on the other hand, interface activation reactions often experience local instability due to fluctuations in process parameters such as mixing rhythm and thermal field structure, making it difficult to achieve continuous optimization of macroscopic performance.

[0007] Especially in the application of flow battery electrode frames, the shell must meet the requirements of large-area and high-consistency molding. Microscopic processing anomalies will be directly amplified into interface defects or structural vulnerabilities. Therefore, simply adjusting the parameters of the material itself is insufficient to meet the demand for the coordinated development of big data-driven processes and structural controllability. Therefore, this invention proposes a method for improving the reinforcement performance of polypropylene mineral fillers for flow battery electrode frames, aiming to solve the above problems. Summary of the Invention

[0008] To achieve the above objectives, the present invention provides the following technical solution: A method for improving the reinforcing performance of polypropylene mineral fillers used in flow battery frames includes the following steps: Step 1: During the mixing and processing of polypropylene and mineral fillers, real-time data on processing heat flux parameters, shear rate fluctuations, and interfacial energy changes are collected to form a dataset characterizing the dynamic features of the processing. Step 2: After data acquisition, an initial analysis is performed based on the thermal flow stability index and shear stress consistency index during the processing. Under the premise that the processing state meets the stability criteria for entering the identification stage, behavioral segments with abnormal distribution or discontinuous interface bonding in the data set are identified, and these behavioral segments are extracted to form candidate analysis regions. Step 3: For the candidate analysis region, calculate based on the gradient characteristics, transient response characteristics and stress energy distribution characteristics formed in the dataset, and extract two processing behavior indices. The first index reflects the dynamic distribution stability of the filler in the mixing zone, and the second index reflects the time continuity of the interface activation reaction. Step 4: Input the first index and the second index as input variables into the pre-trained prediction model, output the enhancement fitness coefficient, and use the enhancement fitness coefficient as a decision benchmark to quantify the overall performance strength of the current processing path. Step 5: Based on the category to which the enhancement fit coefficient belongs in the corresponding response interval defined in the prediction model, match a predefined processing scheme, i.e., a target path in the optimization path set, and perform target-oriented behavior reconstruction based on the target path.

[0009] In a preferred embodiment, during the mixing and processing of polypropylene and mineral fillers, the methods for real-time acquisition of processing heat flux parameters, shear rate fluctuation values, and interfacial energy change data include: By deploying temperature sensors at multiple fixed locations along the processing path, temperature data at each location at different times is collected. The rate of temperature change per unit length is calculated by dividing the temperature difference between two adjacent measuring points by the distance between the two points. This rate of temperature change is then multiplied by the known thermal conductivity constant of polypropylene to obtain the change in heat flux density per unit area. This change in heat flux density is the basis for expressing the processing heat flux parameters. By monitoring the torque change curve of the mixing processing device per second at a set speed, and combining the speed change amplitude, the instantaneous shear rate within the same time period is calculated, and the shear rate fluctuation value is represented by the difference between the maximum and minimum shear rates within the same time period. Interfacial energy change data is obtained by analyzing the energy transfer characteristics of the interface between the filler and polypropylene during processing. Specifically, based on continuous temperature monitoring data, the interfacial tension change range in the heating and cooling sections of the process is identified, the dynamic change range of the tension value between adjacent filler particles per unit volume is recorded, and the interfacial energy change data is calculated by multiplying the total tension change within the range with the change area of ​​the interface. The above three types of data are all based on real-time temperature information, shear force change information and interface contact behavior records during the processing. They form a data set of processing dynamic characteristics through physical quantity measurement and mathematical conversion.

[0010] In a preferred embodiment, after data acquisition is completed, the initial analysis based on the thermal flow stability index and shear stress consistency index during the processing can be performed in the following ways: The processing heat flux parameters are arranged in time sequence. By calculating the average difference of the change in heat flux density in any two adjacent time periods, it is determined whether the processing heat flux parameters are in a stable range. When the change in heat flux density in multiple time periods fluctuates within a preset error range, it is defined as the heat flux stability reaching the evaluation standard. The shear stress consistency index is determined by segmenting the shear rate fluctuation value within a set time window and calculating the variance level of each fluctuation value to determine whether it meets the stability threshold. When the variance of the shear rate fluctuation value in multiple time periods is less than a given limit value, it is determined that the shear stress change trend is consistent. The thermal flow stability index and the shear stress consistency index are used together as the judgment criteria. Only when both of them simultaneously meet the steady state criterion can the initial analysis be deemed qualified and the subsequent identification stage be entered.

[0011] In a preferred embodiment, after data acquisition is completed, identifying behavioral segments in the dataset that exhibit abnormal distribution or discontinuous interface bonding, and extracting these behavioral segments to form candidate analysis regions, includes the following methods: A synchronous analysis matrix was constructed by combining processing heat flux parameters, shear rate fluctuation values, and interface energy change data along the time axis. The minimum analysis interval for each time period was set, and a sliding window method was used to compare the trends of the change curves for every five consecutive time periods. If the average change of the processing heat flux parameter between two consecutive windows exceeds 20%, it is defined as a sudden change in heat flux behavior. If the variance growth rate of the shear rate fluctuation value between two windows exceeds 30%, it is defined as shear-enhancing behavior. If the difference between the first derivative values ​​of the interface energy change data curves in adjacent time periods exceeds the preset slope deviation threshold, it is defined as interface bonding discontinuity behavior. Interface bonding discontinuity behavior is limited to the slope discontinuity that appears in the interface tension change curve, rather than physical interface decoupling. If any parameter meets any of the above abnormal conditions within a set continuous time period, it is marked as an abnormal behavior segment, and the abnormal behavior segment, along with the two adjacent segments before and after it, is included in the analysis scope; all time periods included in the analysis scope are logically integrated to form a candidate analysis region.

[0012] In a preferred embodiment, the first index is calculated as follows: Within the time window corresponding to the candidate analysis region, a continuous time period is divided in units of one second. Image processing methods are used to extract the two-dimensional spatial distribution image of the filler particles in each time period, and the coordinate set of all particles is extracted. The average Euclidean distance of the position change of the same particle in two consecutive time periods is calculated to obtain the average position offset distance within the time window. Then, the local density of all particles in each time period is calculated. Taking the number of particles per unit area in the same time period as the density benchmark, the difference between the maximum density and the average density in the local density sequence is selected and its relative proportion is calculated, which is defined as the local dispersion fluctuation factor. After normalizing the average position offset distance and the local dispersion fluctuation factor, the average position offset distance is used as the independent variable and the local dispersion fluctuation factor is used as the dependent variable. A continuous three-second time period is used as a sliding window. Within each sliding window, three parameter pairs consisting of average position offset distance and local dispersion fluctuation factor are formed. Linear regression fitting is performed on all parameter pairs in each sliding window, and the fitting slope of the linear function is extracted. The fitting slope parameter sequence generated by each segment of the linear function is arranged by time, and the standard deviation of the sequence is calculated as the slope variability index. At the same time, the fitting residual of each segment of the linear function is calculated as the sum of squares of the differences between the fitted value and the original data, and a residual energy sequence is constructed. The coefficient of variation of the residual energy sequence is taken as the residual volatility index. The slope variability index and the residual fluctuation index reflect the stability of the trend changes of filler particles in spatial movement and density fluctuation, and constitute a numerical expression of the dynamic distribution stability of the filler in the mixing zone. The weighted average of the slope variability index and the residual fluctuation index is defined as the first index.

[0013] In a preferred embodiment, the second index is calculated as follows: For each anomalous behavior segment within the candidate analysis region, the interface energy change data sequence within that anomalous behavior segment is extracted and sampled at a frequency of once per second to construct a tension time series. For the tension time series within each segment of abnormal behavior, the first-order difference method is used to calculate the rate of tension change between adjacent time points, and the standard deviation of all tension change rates within the segment is calculated and defined as the slope change factor of the segment. Simultaneously, identify the peak point in the tension sequence, calculate the ratio of the difference between the peak point and the average tension value in the segment, and define it as the tension jump factor of the segment; The slope change factor and tension jump factor of each abnormal behavior segment are averaged separately, and the arithmetic mean of the two is calculated with equal weights to form the fusion factor. The total duration of all abnormal behavior segments within the candidate analysis area is calculated and then compared with the total duration of the entire candidate analysis area. This ratio is defined as the abnormal coverage rate. Finally, the product of the fusion factor and the anomaly coverage rate is used as the basis for the numerical expression of the second index. The fusion factor reflects the internal consistency of each anomalous behavior, and the anomaly coverage rate reflects the time proportion of the behavior in the candidate region. The product of the two constitutes the comprehensive stability expression of the interface activation behavior in the entire candidate region, that is, the second index.

[0014] In a preferred embodiment, the prediction model is any one of the following three structures: I. A model based on a one-dimensional convolutional neural network structure; II. A model based on a dual-hidden-layer feedforward neural network structure; III. Model based on gradient boosting tree structure.

[0015] In a preferred embodiment, matching a predefined processing scheme based on the category to which the enhancement fit coefficient belongs in the response interval defined by the prediction model refers to: The range of the enhanced fit coefficient is divided into three intervals: below 0.6, between 0.6 and 0.8, and above 0.8. When the enhancement adaptability coefficient is less than 0.6, the processing scheme is defined as unstable, and the structural disturbance behavior reconstruction in the matching processing scheme optimization path is carried out, including three operations: adjusting the packing addition order, resetting the particle size distribution, and extending the blending time. When the enhancement adaptability coefficient is between 0.6 and 0.8, the processing scheme is defined as an intermediate adjustable level, and the interface control behavior reconstruction in the matching processing scheme optimization path is carried out, including the adjustment of filler surface treatment method, calibration of interface bonding agent ratio, and correction of mixing temperature step configuration. When the enhancement adaptability coefficient is higher than 0.8, the processing scheme is defined as a high adaptability level, and the rhythm fine-tuning behavior reconstruction in the matching processing scheme optimization path is carried out, including the fine-tuning of the cooling process time delay, the revision of the pressure curve in the holding stage, and the adjustment of the thermal field uniformity balance in the final molding process.

[0016] The technical effects and advantages of this invention are as follows: This invention achieves full-process data recording by real-time acquisition of processing heat flux parameters, shear rate fluctuations, and interfacial energy changes during the mixing and processing of polypropylene and mineral fillers, and constructing a dataset of processing dynamic characteristics. This dataset provides a continuous, objective, and quantifiable basis for subsequent analysis and judgment at each stage, helping to identify potential problems through data fluctuation trends before obvious defects appear in the processing behavior. This avoids the lag and uncertainty inherent in traditional manual inspection methods and enhances the process perception capability of the processing behavior.

[0017] This invention identifies abnormal behavior segments during processing through initial analysis based on thermal flow stability and shear stress consistency indices, extracting candidate analysis regions. Based on this, a first and a second index are calculated to reflect the dynamic distribution stability of the packing material and the time-dependent continuity of the interfacial activation reaction, respectively, achieving a two-dimensional characterization of the internal state of processing behavior. This dual-index structure can more precisely capture the dynamic patterns manifested in the complex interactions between the packing material and the interface, providing a clear and quantifiable basis for subsequent behavioral decisions.

[0018] This invention imports a first and a second index as input variables into a prediction model, outputs an enhanced fit coefficient, and matches a predefined processing scheme to the category to which this coefficient belongs in the response interval, thus realizing a closed-loop control logic between the processing path and the processing scheme. This method effectively realizes a linkage mechanism between the evaluation of processing behavior and the reconstruction of the scheme, so that processing optimization no longer relies on experience-based judgment, but is based on data-driven decision-making, ensuring that the processing scheme is always adjusted around the goal orientation, and has stronger adaptive and fine-tuning capabilities. Attached Figure Description

[0019] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings; Figure 1 This is a schematic diagram of the method for improving the reinforcing performance of polypropylene mineral fillers used in flow battery frames according to the present invention.

[0020] Figure 2 This is a partial structural diagram of the injection-molded electrode frame of the flow battery in this invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0022] Reference Figure 1 The following examples were obtained: Example 1: A method for improving the reinforcing performance of polypropylene mineral filler in flow battery frames, comprising the following steps: Step 1: During the mixing and processing of polypropylene and mineral filler, real-time data on processing heat flux parameters, shear rate fluctuations, and interfacial energy changes are collected to form a dataset characterizing the dynamic features of the processing. This step is used to construct a multi-dimensional quantitative characterization basis for the processing process. The collected data not only covers the thermodynamic changes, mechanical shear fluctuations, and interfacial energy transfer characteristics in the processing physical state, but also constitutes a dynamic response information sequence through time synchronization, which facilitates the identification and pattern modeling of subsequent behavioral segments, thereby providing original data sources and evaluation benchmarks for subsequent analysis, judgment, and decision-making.

[0023] Step Two: After data acquisition, an initial analysis is performed based on the thermal flux stability index and shear stress consistency index during the processing. Provided the processing state meets the stability criteria for entering the identification stage, segments with abnormal distribution or discontinuous interface bonding are identified in the data set, and these segments are extracted to form candidate analysis regions. This step is used to achieve preliminary judgment and screening of processing behavior, ensuring that behavior segment identification is only performed when processing conditions show stable thermal flux parameters and controlled shear fluctuations, thus avoiding the risk of misidentification under unstable conditions. The stability judgment mechanism in this stage helps to build a hierarchical classification basis for processing behavior, ensuring that the subsequently extracted candidate analysis regions have good contextual continuity and behavioral consistency.

[0024] Step 3: For the candidate analysis region, based on the gradient characteristics, transient response characteristics, and stress energy distribution characteristics formed in the dataset, two processing behavior indices are extracted. The first index reflects the dynamic distribution stability of the packing material in the mixing zone, and the second index reflects the time-dependent continuity of the interfacial activation reaction. The significance of this step is that by extracting representative statistical indicators, the spatiotemporal evolution behavior in the candidate analysis region is characterized and modeled, transforming the original processing data into structured parameters with discriminative significance. The first index focuses on the uniformity and positional stability of the packing material during the micro-mixing process, while the second index reflects the continuity and fluctuation trend of the interfacial reaction behavior over time. Together, they construct a two-dimensional core criterion for evaluating processing behavior.

[0025] Step four involves inputting the first and second indices as input variables into the pre-trained prediction model, outputting an enhancement fit coefficient. This coefficient serves as the decision benchmark, quantifying the overall performance strength of the current processing path. This step aims to introduce the inference capabilities of the offline-trained model, enabling the representation and transformation of complex processing behaviors at the abstract index level. By inputting the two behavioral indices into a model with non-linear mapping capabilities, the implicit behavioral patterns are comprehensively evaluated, and the overall processing performance of the current path is expressed as a single numerical output: the enhancement fit coefficient. This coefficient becomes the core decision-making factor for determining the optimal path in the subsequent behavior reconstruction process.

[0026] Step 5: Based on the category of the enhancement fit coefficient within the corresponding response interval defined in the prediction model, match a target path from the predefined processing scheme (i.e., the set of optimization paths), and perform goal-oriented behavior reconstruction based on the target path. This step establishes a mapping rule between the enhancement fit coefficient and the predefined response interval to achieve the classification and selection of optimization paths under different processing states. Each response interval corresponds to a reconstruction strategy under different dimensions, including paths such as structural perturbation, interface adjustment, or rhythm fine-tuning. The behavior reconstruction operation under each path clearly points to the correction method of specific processing conditions, ultimately achieving the goal-oriented optimization behavior execution that specifically adjusts the processing structure and control variables.

[0027] In the process of blending polypropylene with mineral fillers, polypropylene is the polymer matrix material and mineral fillers are inorganic particulate reinforcing materials. The two are blended in the processing equipment under heating. In order to obtain the behavioral response characteristics under the processing state, it is necessary to collect key parameters reflecting heat transfer, shear and interface effects in real time, including processing heat flux parameters, shear rate fluctuation values ​​and interface energy change data.

[0028] Temperature sensors are deployed at multiple fixed locations along the processing path to collect temperature data at different times. These sensors, thermocouples or infrared temperature sensing units with microsecond-level response times, are installed in the high-temperature melting section, shear conversion section, and pressure stabilization section of the twin-screw mixer's processing chamber. The temperature data collected by each sensor is recorded with a timestamp, forming a corresponding two-dimensional time-space temperature distribution matrix. The rate of temperature change per unit length is calculated by dividing the temperature difference between two adjacent measuring points by the distance between them. It is an indicator describing the rate of heat conduction along the axial or radial direction of the screw, expressed in degrees Celsius per millimeter. Its calculation formula is: ; , This represents the temperature readings of two adjacent temperature sensors at the same time point, where i is the temperature sensor number. The distance between two adjacent temperature sensors is used to calculate the change in heat flux density per unit area. This change in temperature rate is multiplied by the known thermal conductivity constant of polypropylene. The thermal conductivity constant of polypropylene is an intrinsic parameter of the material, typically ranging from 0.1 to 0.25 W / m·K. The change in heat flux density (q) is expressed in W / m², and the calculation formula is: ; The thermal conductivity of polypropylene is the known thermal conductivity constant of polypropylene. This change in heat flux density is the basis for expressing the processing heat flux parameters. The processing heat flux parameters characterize the local heat conduction intensity during the processing and are key physical quantities for measuring heating uniformity and energy input response effects.

[0029] By monitoring the torque change curve per second of the mixing and processing device at a set speed, and combining the speed change amplitude, the instantaneous shear rate within the same time period is calculated. The mixing and processing device is specifically a twin-screw extruder. The torque sensor is installed at the screw shaft end to record the load change value per unit time in real time, forming a curve with time as the horizontal axis and torque as the vertical axis. The instantaneous shear rate refers to the relative movement speed between material layers per unit time, which is calculated based on the coupling relationship between the equipment speed and load resistance. The difference between the maximum and minimum shear rates within the same time period is used to represent the shear rate fluctuation value. The shear rate fluctuation value quantifies the rheological fluctuation amplitude within this time period, reflecting the stability and volatility of the stress transmission process during processing.

[0030] Interfacial energy change data is obtained by analyzing the energy transfer characteristics of the interface between the filler and polypropylene during processing. The interfacial energy change data reflects the energy interaction characteristics caused by the difference in surface tension and the inconsistency in thermal response during the filler-matrix contact process, and is often used to measure the degree of interfacial adhesion and the state of reaction activation. Specifically, this includes identifying the variation in interfacial tension between the heating and cooling zones based on continuous temperature monitoring data. The heating and cooling zones are automatically identified by the positive and negative slope changes in the temperature-time curve. Interfacial tension is sampled using a miniature stress sensor embedded in the tension testing unit and expressed as mN / m. The dynamic variation range of tension values ​​between adjacent filler particles per unit volume is recorded. Continuous image acquisition of local areas of the filler and polypropylene under mixed processing conditions is performed using a high-resolution dynamic optical microscopy imaging device to obtain the positional distribution image of filler particles within the field of view per unit time. Image recognition algorithms are used to extract particle contours from each frame of the image, determine the center coordinates of each particle, and calculate the spatial distance between any two adjacent particles. Using the spatial distance between filler particles as the basis for variables, and combining it with the interface temperature value corresponding to the current moment of mixed processing, the following empirical function relationship is substituted to estimate the interfacial tension: the interfacial tension value equals a constant coefficient multiplied by the negative first power of the distance between particles, and then multiplied by a temperature correction term, i.e.: ; This represents the estimated interfacial tension of the k-th and j-th particle pairs at time t. This represents the spatial distance between the center points of the k-th and j-th pairs of particles. This represents the temperature difference between time t and the processing reference temperature, such as 220℃, expressed in degrees Celsius. This represents the distance-tension constant coefficient obtained through experimental fitting, with a value ranging from 0.1 to 0.5 N·μm / m². The temperature correction factor ranges from 0.002 to 0.01 °C. All parameter values ​​are determined based on a standard test platform under calibrated processing equipment conditions, and the particle tension response range is normalized and modeled through univariate control experiments. This method maps the distance and temperature parameters of all particle pairs per unit volume to corresponding interfacial tension values, forming a tension sequence for subsequent statistical processing. Then, the interfacial energy change data is calculated by multiplying the total tension change within this range with the corresponding instantaneous contact interface change area. All three types of data are based on real-time temperature information, shear force change information, and interfacial contact behavior records during processing, forming a dataset of processing dynamic characteristics through physical quantity measurement and mathematical conversion. This dataset is organized with a unified timestamp and stored in a structured manner in the analysis platform, serving as the basic data carrier for subsequent judgment of processing stability, identification of behavioral segments, and extraction of processing indices.

[0031] After data acquisition, the initial analysis based on the thermal flux stability index and shear stress consistency index during processing involves: sequentially arranging the processing heat flux parameters, which refer to the change in heat flux density measured per unit area per unit time within the processing cavity. This parameter is obtained by acquiring temperature gradient information at multiple time points along the processing path and converting it using the thermal conductivity constant of polypropylene material. To facilitate trend identification, these parameters are arranged into an ordered sequence in chronological order. By calculating the average difference in the magnitude of heat flux density change between any two adjacent time periods, it is determined whether the processing heat flux parameter is within a stable range. To eliminate the interference of individual instantaneous fluctuations on the overall trend judgment, a sliding window method is used during the analysis to select two adjacent time periods as comparison units. The heat flux density values ​​of all measuring points in the two time periods are averaged separately, and the difference between these two average values ​​is calculated. Subsequently, similar differences within multiple time periods are statistically analyzed, and their arithmetic mean is taken as a reference standard for the overall magnitude of heat flux density change. When the change in heat flux density over multiple time periods fluctuates within a preset error range, it is defined as meeting the thermal flux stability evaluation standard. This error range is set based on the equipment calibration accuracy and material response characteristics, and is generally between 5% and 10% of the unit heat flux density change. If, in any five consecutive time periods, the change amplitude is less than the upper limit of this error range in at least four of them, it is considered that the heat flux conduction behavior during that processing period did not undergo abrupt changes and meets the thermal flux stability evaluation standard.

[0032] The shear stress consistency index is established by segmenting shear rate fluctuations within a set time window. Shear rate fluctuation refers to the difference between the maximum and minimum shear rate values ​​within a unit time period, characterizing the degree of local rheological fluctuation in the hybrid processing unit at a specific operating speed. The set time window length is typically five seconds. The shear rate fluctuation data sequence is divided into several segments using an equal-length sliding window method, forming multiple relatively independent statistical intervals. The variance level of the fluctuation values ​​in each segment is calculated to determine whether it meets the stability threshold. The variance of the shear rate fluctuation data within each time period is calculated to obtain a statistical index of its degree of change. The stability threshold is set based on historical process databases and actual processing tolerances, generally requiring the variance of the shear rate fluctuation values ​​to be below one percent. If the variance value of a certain time period is less than this threshold, it indicates that the shear stress is operating within a reasonable fluctuation range during that period.

[0033] When the variance of shear rate fluctuation values ​​across multiple time periods is less than a given limit, the shear stress change trend is considered consistent. The mean variance of shear rate fluctuation values ​​across all time periods is statistically analyzed; if at least 80% of the time periods meet the stability threshold condition, the overall shear behavior is considered to have a consistent trend, and this is marked as shear stress consistency meeting the standard. The heat flux stability index and the shear stress consistency index are used together as the judgment criteria. That is, within any time window, only when the heat flux density change amplitude meets the stability requirement and the variance level of the shear rate fluctuation value simultaneously meets the consistency requirement is the current processing state considered to have sufficient stability in the overall physical parameter dimension. Only when both simultaneously meet the stability criteria is the initial analysis deemed acceptable, and the subsequent identification stage begins. If only heat flux stability or only shear stress consistency meets the standard, it does not constitute a trigger condition for subsequent identification. Only when the stability criteria for both heat flux and shear are simultaneously met can the stage of identifying, extracting, and constructing candidate analysis regions for abnormal behavior segments in the dataset proceed. This approach ensures that the basic conditions for identification are based on sufficient process stability, avoids introducing occasional disturbances that could lead to misidentification of behavioral segments, and guarantees the accuracy of subsequent analysis.

[0034] After data acquisition, the method for identifying and extracting candidate analysis regions from segments of data that exhibit abnormal distribution or discontinuous interface bonding is as follows: A synchronous analysis matrix is ​​constructed by arranging processing heat flux parameters, shear rate fluctuation values, and interface energy change data along a time axis. These three types of parameters are extracted from the temperature sensor, torque-speed coupling monitoring module, and interface tension conversion module of the processing equipment, respectively. Each type of parameter is recorded once at the same timestamp, forming a unified three-dimensional data recording unit. With a sampling period of one second, the three types of data corresponding to each time point are assembled into a single row, arranged with time as the horizontal axis, ultimately forming a matrix data structure composed of multiple rows of records. This matrix is ​​the synchronous analysis matrix.

[0035] A minimum analysis interval is set for each time period, and a sliding window method is used to compare the trends of the change curves for every five consecutive time periods. The minimum analysis interval is set to one second, based on a sampling frequency of once per second. The sliding window is defined as performing trend statistics on the change curves of the three types of parameters within any consecutive five-second time interval. For each window, the mean, volatility, derivative changes, etc., of the current segment are compared with the previous segment (i.e., the previous window) to generate trend comparison indicators for identifying abnormal changes. If the average change of the processing heat flux parameter between two consecutive windows exceeds 20%, it is defined as a heat flux mutation. The average value of the processing heat flux parameter within each window is defined as the heat flux center value of that window. The difference between the average values ​​of two windows divided by the mean of the previous window is the percentage change. When this percentage exceeds 20%, the behavior is judged as a heat flux mutation. For example, if the average heat flux in the previous window is 50 units and in the next window is 65 units, the change is 30 percentage points, exceeding 20%, constituting a mutation.

[0036] If the variance growth rate of the shear rate fluctuation value between two windows exceeds 30%, it is defined as shear enhancement behavior. Within each window, the variance of the shear rate fluctuation value is calculated, reflecting the severity of the shear fluctuation within that window. The variance growth rate is calculated by subtracting the variance of the previous window from the variance of the subsequent window, then dividing by the variance of the previous window, and multiplying the result by 100 to obtain the percentage. If this value exceeds 30%, the shear behavior is considered significantly enhanced in the current time period. For example, if the variance of the previous window is four and the variance of the subsequent window is six, the growth rate is 50%, which indicates enhanced behavior.

[0037] If the difference in the first derivative of the interface energy change data curve between adjacent time periods exceeds a preset slope deviation threshold, it is defined as an interfacial discontinuity behavior. The first derivative value, i.e., the tension change slope, is obtained by dividing the difference in values ​​between two adjacent time points by the time interval. If the difference in the first derivative between two adjacent time points exceeds a preset threshold, usually set to 50% of the average derivative of the processing state, it is considered a slope jump behavior. This behavior refers to a sudden change in the rate of interface energy release or absorption, reflecting the discontinuity of the interface chemical or physical activation behavior.

[0038] Interface discontinuity behavior is defined as the discontinuous slope of the interface tension change curve, rather than physical interface decoupling. In data analysis, this behavior is not equivalent to macroscopic deboundage such as material structure delamination or filler debonding; rather, it is judged by the discontinuity of the derivative in the tension-time relationship, representing a behavioral level rather than a structural level. When any parameter meets any of the above abnormal conditions within a defined continuous time period, it is marked as an abnormal behavior segment. During the sliding window iteration process, each second is used as an independent positioning unit. If a second belongs to the center of a window containing abrupt behavior, it is marked as an abnormal behavior point. Two or more consecutive abnormal behavior points constitute an abnormal behavior segment and are assigned an abnormality type label. This abnormal behavior segment, along with the two adjacent segments before and after it, is included in the analysis scope. To eliminate boundary ambiguity and fluctuation interference, the time period within two seconds before and after the start time of the abnormal segment is included to construct a contextual behavior extension band, avoiding fragmented misidentification. All time periods included in the analysis scope are logically integrated to form a candidate analysis region. If multiple independent abnormal behavior segments overlap after adjacent supplementation, they are merged into one large segment. All merged segments are spliced ​​together in chronological order to form a complete set of time intervals without overlap or omissions, which is defined as the candidate analysis region for subsequent processing behavior index calculation.

[0039] The calculation method of the first index includes: dividing the continuous time period in units of one second within the time window corresponding to the candidate analysis area. The start and end time of the candidate analysis area is formed by merging the abnormal behavior segment identified in the previous step with the additional time periods before and after it. Within this area, the continuous data record is divided into multiple time periods of one second in length according to the sampling frequency of once per second, so as to provide a unified time reference for the consistent processing of subsequent image and location data.

[0040] Two-dimensional spatial distribution images of filler particles within each time period are extracted using image processing methods, and the coordinate sets of all particles are extracted. Images acquired by optical imaging equipment every second are segmented, and edge recognition and center localization algorithms are used to obtain the projected positions of particles in the images. The center position of each particle is marked using horizontal and vertical pixels in the image coordinate system. Each time period is used as an analysis unit, and the center coordinates of all filler particles within that period are recorded to form a two-dimensional coordinate set for spatial variation calculation. The average Euclidean distance of the position change of the same particle in two consecutive time periods is calculated to obtain the average position offset distance within the time window. A particle labeling algorithm is used to track the position coordinate changes of filler particles with the same number in two adjacent time periods, and the displacement value is calculated according to the Euclidean formula in two-dimensional space, i.e., the distance between the coordinate points of each particle at time t and t+1. The displacement values ​​of all particles are averaged to obtain the average position offset distance within the candidate analysis area, which serves as a measure of the macroscopic flow performance of particles during the mixing process.

[0041] Next, the local density of all particles in each time period is statistically analyzed. Taking the number of particles per unit area in the same time period as the density benchmark, the difference between the maximum density and the average density in the local density sequence is selected, and their relative ratio is calculated, which is defined as the local dispersion fluctuation factor. Each frame of the image is divided into several analysis grids of equal area. The number of filler particles in each grid area is counted, and the instantaneous density of the local area is defined by dividing the number of particles by the grid area. The grid area with the largest local density in the whole image is found to obtain the maximum density value. At the same time, the density values ​​of all grid areas are averaged to obtain the overall average density value. The ratio of the difference between the two and the average density value represents the degree of density imbalance in the current time period. This ratio is the local dispersion fluctuation factor, which is used to measure the local fluctuation of the mixing uniformity.

[0042] After normalizing the average position offset distance and the local dispersion fluctuation factor, since there are essential differences in their dimensions and numerical ranges, they need to be converted into relative ratios between zero and one by means of maximum value normalization or standard deviation normalization, so that they can be used as comparable continuous variables in regression analysis.

[0043] Using the average positional offset distance as the independent variable and the local dispersion fluctuation factor as the dependent variable, a trend fitting relationship between the two variables was established to analyze their coupling performance in spatial movement and local concentration changes. A continuous three-second time period was used as a sliding window, forming three sets of parameter pairs consisting of the average positional offset distance and the local dispersion fluctuation factor within each window. In the candidate analysis region, a sliding window was defined with a three-second step size, containing three consecutive time periods corresponding to three sets of normalized parameter pairs.

[0044] Linear regression fitting is performed on all parameter pairs for each sliding window, and the slope of the fitted linear function is extracted. A univariate linear function is fitted using the least squares method for the three parameter pairs, and its slope is recorded as an indicator of local trend. Each sliding window is fitted once, yielding multiple fitted slopes, which are arranged in sliding order to form a slope sequence. The fitted slope parameter sequence generated by each linear function is arranged over time, and the standard deviation of this sequence is calculated as an indicator of slope variability. All slope values ​​obtained from the sliding fit are used as a data sequence to reflect the trend and volatility of the packing behavior over time. The standard deviation of this slope sequence is calculated, which is the slope variability index used to measure the stability of particle behavior.

[0045] Simultaneously, the fitting residuals of each linear function segment are calculated as the sum of squares of the differences between the fitted value and the original data, constructing a residual energy sequence. For the linear fitting results within each sliding window, the sum of squares of the deviations between the predicted value and the actual observed value is calculated as the fitting residual energy for that segment. The residual energies of all sliding windows are arranged in chronological order to form a residual energy sequence, reflecting the overall stability of the particle behavior model fitting. The coefficient of variation of the residual energy sequence is taken as the residual volatility index; the ratio of its standard deviation to its mean is calculated for the residual energy sequence to obtain the coefficient of variation, representing the proportional relationship between the degree of data volatility and the average level. The smaller this value, the more uniform the overall residuals.

[0046] Slope variability and residual volatility indices reflect the stability of the trend changes in the spatial movement and density fluctuations of filler particles. The former focuses on the consistency of the trend slope changes, while the latter reflects the stability of the fitting error over time. Both evaluate the controllability of the mixing process from different perspectives. These indices constitute a numerical expression of the dynamic distribution stability of the filler within the mixing zone. A weighted average of the two indices is used to summarize a unified quantitative index between spatial distribution and density consistency, serving as the basis for describing the dynamic behavior of the filler in the processing mixing zone. The weighted average of the slope variability and residual volatility indices is defined as the first index. The weighting ratio can be set empirically, typically using equal weights or weighting based on the performance of stable operating conditions in historical processing cases. The final average value is the first index, used to describe the strength of the temporal stability of the filler distribution state within the candidate analysis area.

[0047] The calculation method of the second index includes: for each abnormal behavior segment in the candidate analysis area, extracting the interface energy change data sequence within that abnormal behavior segment, and sampling it once per second to construct a tension time series; the candidate analysis area consists of multiple abnormal behavior segments, each of which refers to a continuous time period in which a significant trend of change appears in the processing heat flux parameters, shear rate fluctuation values, or interface energy change data identified in step three; the interface energy change data comes from the total energy quantification of the change in interface tension between filler particles and polypropylene per unit volume, and extracting the tension value change once per second from the acquired interface energy change data to form a tension time series, representing the continuous change trajectory of interface tension within that time period.

[0048] For each segment of abnormal behavior, the tension time series is calculated using a first-order differencing method to determine the rate of tension change between adjacent time points. The standard deviation of all tension change rates within the segment is then calculated and defined as the slope change factor for that segment. The first-order differencing method involves subtracting the tension values ​​at consecutive time points t and t+1 to obtain the rate of tension change, expressed in tension values ​​per second. All change rates are combined to form a slope sequence, reflecting the instantaneous intensity change of the tension fluctuation process within that segment. The standard deviation of this slope sequence is calculated, representing the fluctuation level of the tension change within that segment, and serves as the slope change factor for that segment, reflecting the stability of the interfacial tension within the abnormal segment. Simultaneously, the peak point in the tension sequence is identified, and the ratio of the difference between the peak point and the average tension value in the segment is calculated, which is defined as the tension jump factor of the segment. The peak point is the highest value in the tension time series, representing the strong instantaneous behavior of interfacial energy transfer. The average tension value of the segment is the arithmetic mean of all tension values ​​in the sequence. The ratio obtained by dividing the difference between the peak point and the average value by the average value is the tension jump factor, which represents the ratio of the maximum jump amplitude in the interfacial reaction of the segment to the overall fluctuation level, and is used to measure the intensity of local extreme behavior.

[0049] The slope change factor and tension jump factor of each anomalous behavior segment are averaged separately, and the arithmetic mean of the two is calculated with equal weights to form a fusion factor. The slope change factor and tension jump factor of all anomalous behavior segments in the candidate analysis area are calculated separately, and the arithmetic mean of all slope change factors and all tension jump factors are calculated separately. The two are added together and divided by two to obtain the fusion factor, which represents the unified characterization of the interface activation volatility and extreme value jump degree of all anomalous behavior segments in the candidate area.

[0050] The total duration of all anomalous behavior segments within the candidate analysis region is calculated and its ratio to the total duration of the entire candidate analysis region is defined as the anomalous coverage rate. The durations of all anomalous behavior segments within the candidate analysis region are summed to obtain the total duration of anomalous behavior. This total duration is then divided by the total duration of the candidate analysis region to obtain the percentage value, which is the anomalous coverage rate, reflecting the proportion of time spent by anomalous behavior within that region. Finally, the product of the fusion factor and the anomalous coverage rate is used as the basis for the numerical expression of the second index. The fusion factor reflects the internal consistency of each anomalous behavior segment, and the anomalous coverage rate reflects the proportion of time spent by that behavior within the candidate region. Their product constitutes the comprehensive stability expression of the interface activation behavior across the entire candidate region, i.e., the second index. This index value, along with the first index, is used as an input variable in the prediction model in subsequent steps, representing the quantitative expression of the interface response between the filler and polypropylene in the time and intensity dimensions.

[0051] Step four involves inputting the first and second indices as input variables into the pre-trained prediction model, outputting the enhancement fitness coefficient, which serves as the decision criterion to quantify the overall performance strength of the current processing path; the prediction model can be any one of the following three structures: I. A model based on a one-dimensional convolutional neural network structure. This model constructs a multi-layer neural network framework including an input layer, a one-dimensional convolutional layer, a pooling layer, a fully connected layer, and an output layer. The two-dimensional input sequence consisting of a first exponent and a second exponent is fed into the input layer, where feature extraction and representation compression are performed sequentially. The input layer receives first and second exponent data vectors generated at fixed sampling intervals. These vectors are processed by a one-dimensional convolutional layer to extract their changing trends and local combination features. The convolutional kernel length and stride are set empirically. The pooling layer performs maximum or average downsampling on the convolutional output to reduce feature dimensionality and suppress local fluctuations. The output then enters the fully connected layer to complete the nonlinear combination of multi-dimensional feature mappings. Finally, the output layer outputs an enhancement fit coefficient, a real number between zero and one, representing the degree of fit of the current processing path in the pre-training sample space. This structure is suitable for archived processing behavior data from a large number of historical samples. Supervised learning is used during training, with label data representing the judgment rating level of the corresponding processing behavior. After optimization by minimizing the loss function, stable and convergent prediction model parameters are obtained.

[0052] II. A model based on a dual-hidden-layer feedforward neural network structure, a typical multilayer perceptron model, structurally includes an input layer, two hidden layers, and an output layer. The input layer accepts two floating-point values, a first exponent and a second exponent, and simultaneously feeds them into the first hidden layer. The first hidden layer uses an activation function to map the input to a high-dimensional space, extracting nonlinear features; the second hidden layer further reconstructs and compresses these features to form a feature fusion vector, which then enters the output layer to generate the enhancement fitness coefficient. The connection weights of each neuron in the model are updated during the training phase using a backpropagation algorithm, and the optimization objective function is the sum of squared errors between the enhancement fitness coefficient and the historical processing evaluation. This structure features high computational efficiency and network transparency, making it suitable for training with small-sample or moderately complex data structures, especially applicable to scenarios with fewer predictor variables and sufficiently standardized input indicators.

[0053] III. The Gradient Boosting Tree-Based Model: This model uses a gradient boosting decision tree as its basic architecture. Its core is constructed iteratively from multiple weak classifiers (regression trees) following a residual reduction mechanism. The first and second indices of the input variables are considered as node partitioning references in the two-dimensional feature space. Each tree is refitted based on the residuals from the previous model's output, and the results are accumulated to form the current total prediction output, which is ultimately the augmented fit coefficient. This structure does not require the assumption that the input data distribution is linearly separable and has a high adaptability to the interactions between features, making it particularly suitable for complex data patterns with obvious nonlinear boundaries. During the training phase, the model uses the label results of historical processed data as the learning objective, constructing a tree model along the negative gradient direction for each prediction error, iteratively improving the overall regression accuracy of the model. All three models described above optimize the model parameters through pre-training. The training samples come from historical sample datasets with known processing evaluation results, and the training objective is to construct a quantitative functional relationship mapping from "first index + second index" to "augmented fit coefficient".

[0054] Step five involves matching the enhancement fit coefficient to a predefined processing scheme (i.e., a target path from the optimized path set) within the corresponding response interval defined in the prediction model, and then reconstructing the target-oriented behavior based on that path. Specifically, this includes dividing the enhancement fit coefficient into three intervals: below 0.6, between 0.6 and 0.8, and above 0.8. This value originates from the regression result output by the prediction model in step four and is a continuous variable between zero and one, reflecting the degree of fit of the current processing behavior within the historical training sample space.

[0055] I. When the enhancement adaptability coefficient is below 0.6, the processing scheme is defined as unstable. This level indicates that the current processing path deviates significantly in the dynamic processing parameter space, making it difficult to maintain the continuous coordination of filler distribution and interfacial bonding behavior. The optimized path at this time is a structural disturbance-type behavior reconstruction, including the following three operations: 1. Filler addition sequence adjustment: This refers to resetting the entry time sequence of mineral fillers of different particle sizes, types, or surface states into the polypropylene melt in the original processing path to interrupt the original unfavorable distribution trend. For example, injecting easily agglomerated particles later can prevent premature aggregation at high temperatures. 2. Particle size distribution resetting: This refers to adjusting the filler particle size ratio range based on particle size analysis results before processing. By introducing a wider range of particle gradations, the filling and arrangement structure between particles is optimized, thereby reducing local voids or high-pressure points. 3. Extended blending time: This refers to appropriately extending the blending time during the melt blending stage to ensure sufficient dispersion of the filler in the melt and increase the opportunity for shear action to reshape interfacial bonding.

[0056] For example, if the first index is consistently low and the second index fluctuates drastically during actual processing, and the enhancement adaptation coefficient is 0.53, it is considered a typical level of instability. In this case, a combined approach can be adopted, such as post-injection of fine particles, changing the particle size distribution from a 70% concentrated type to a wide distribution type, and extending the blending time from 30 seconds to 40 seconds, to trigger structural disturbance adjustment.

[0057] II. When the compatibility coefficient is between 0.6 and 0.8, the processing scheme is defined as an intermediate adjustable level. This level indicates that the main parameters in the processing process have a certain consistency, but there are fluctuations in interfacial activity control or material compatibility. The matching optimization path at this time is an interface-controlled behavior reconstruction, including the following three operations: 1. Filler surface treatment adjustment: This refers to using different surface coupling agents, plasma treatment, or silanization techniques to reconstruct the polarity and affinity characteristics of the filler surface to improve its affinity with the polypropylene interface. 2. Interfacial bonding agent ratio calibration: This refers to adjusting the addition ratio of interfacial activators or compatibilizers in the processing formulation, such as controlling the dosage of maleic anhydride-grafted polypropylene (MAH-PP) or other functionalized polymers. 3. Mixing temperature gradient configuration correction: This refers to redistributing the temperature curves of the heating, shearing, and cooling sections in the processing process to form a more reasonable thermal gradient, which is beneficial for the filler to achieve gradual integration at different stages.

[0058] For example, if the first index during processing is higher than 0.5 but fluctuates greatly, the second index is between 0.7 and 0.8, and the enhancement adaptability coefficient is 0.71, then the interface-controlled path reconstruction strategy can be implemented by increasing the surface modification temperature by 10 degrees Celsius, adjusting the proportion of binder from 2% to 3%, and fine-tuning the melting section temperature from 185 degrees to 190 degrees.

[0059] III. When the enhancement adaptability coefficient is higher than 0.8, the processing scheme is defined as a high adaptability level. This level indicates that the current processing path has shown a good synergistic mechanism in terms of filler distribution and interface response behavior. The optimization path no longer targets structural or formulation adjustments, but rather a matching rhythm fine-tuning behavior reconstruction, including the following three operations: Cooling process time delay fine-tuning: refers to the precise control of the cooling rate in the cooling section time setting after molding and discharge, avoiding stress concentration or interface re-aggregation due to excessively rapid cooling. Pressure curve revision in the holding pressure stage: refers to adjusting the rise slope, holding time, and fall rhythm of the pressure during the holding pressure process in injection molding or extrusion molding to better match it with the material rheological state. Final molding process thermal field uniformity balance adjustment: refers to adding thermal field feedback control points inside the mold or cooling cavity to adjust the heating or heat dissipation capacity of local areas to reduce the temperature difference inside the molding zone.

[0060] For example, when the prediction model outputs an enhancement fit coefficient of 0.86, it indicates that the filler state and interface behavior are highly coordinated. In this case, the cooling period can be extended from 20 seconds to 22 seconds, the pressure holding time increased by 5%, and thermal monitoring feedback devices can be added to both sides of the mold cavity to perform rhythm fine-tuning adjustments. In summary, the three fit levels and nine target paths correspond to different performance intensity regions, forming a responsive optimization system for processing behavior. Each solution uses the enhancement fit coefficient as a quantitative input to achieve predictable reconfiguration of behavior.

[0061] like Figure 2 As shown, this diagram is a partial structural view of the injection-molded electrode frame of a flow battery, primarily illustrating the injection molding details of the housing sealing area, flow channel design, and hole reinforcement structure. Different colors in the diagram represent different functional areas, including: Yellow area: indicates the outer structure of the shell, including mechanical fixing areas such as mounting frame, screw holes, and positioning holes; Pink area: indicates the closed channel area of ​​the flow channel, which is closely related to the liquid flow path and may include the laser-welded edge sealing area; The blue thin line area represents a refined microstructure flow channel, glue injection tank, or flow restriction device; Circular magnified area: used for high-precision inspection of key details, such as reinforcing ribs around screw holes, structural transition design of fine hole edges, or flow guide protrusions.

[0062] The image above is a magnified view of a portion of the flow battery's frame structure, showing the injection molding details of the casing's flow channel sealing area and the hole reinforcement structure. Different colors represent different functional blocks: the yellow area represents the overall structural frame and connecting holes, serving for positioning and mechanical fixation; the pink area represents the fluid sealing path, ensuring sealing performance and mechanical stability after multiple stacked units are assembled by precisely controlling its surface contour and weld seam position; the blue lines represent the injection groove or flow channel microstructure, controlling the electrolyte flow rate and distribution uniformity between the plates and frames. The magnified view reveals the reinforcement ribs and boundary transition design at the hole edges, which helps improve the pressure resistance and crack resistance around the screw holes. The overall design embodies the design philosophy of high-precision injection molding, fine fluid control, and reliable stacking sealing, ensuring the safe and reliable operation of the flow battery system under high pressure and high flow environment. This structure also requires precise injection molding through material selection (such as PP+GF composite material) and mold hot runner control technology.

[0063] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0064] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0065] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0066] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0067] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for improving the reinforcing performance of polypropylene mineral fillers used in flow battery electrode frames, characterized in that, Includes the following steps: Step 1: During the mixing and processing of polypropylene and mineral fillers, real-time data on processing heat flux parameters, shear rate fluctuations, and interfacial energy changes are collected to form a dataset characterizing the dynamic features of the processing. Step 2: After data acquisition, an initial analysis is performed based on the thermal flow stability index and shear stress consistency index during the processing. Under the premise that the processing state meets the stability criteria for entering the identification stage, behavioral segments with abnormal distribution or discontinuous interface bonding in the data set are identified, and these behavioral segments are extracted to form candidate analysis regions. Step 3: For the candidate analysis region, calculate based on the gradient characteristics, transient response characteristics and stress energy distribution characteristics formed in the dataset, and extract two processing behavior indices. The first index reflects the dynamic distribution stability of the filler in the mixing zone, and the second index reflects the time continuity of the interface activation reaction. Step 4: Input the first index and the second index as input variables into the pre-trained prediction model, output the enhancement fitness coefficient, and use the enhancement fitness coefficient as a decision benchmark to quantify the overall performance strength of the current processing path. Step 5: Based on the category to which the enhancement fit coefficient belongs in the corresponding response interval defined in the prediction model, match a predefined processing scheme, i.e., a target path in the optimization path set, and perform target-oriented behavior reconstruction based on the target path.

2. The method for improving the reinforcing performance of polypropylene mineral filler for flow battery frames according to claim 1, characterized in that, In the process of mixing polypropylene with mineral fillers, methods for real-time acquisition of processing heat flux parameters, shear rate fluctuations, and interfacial energy changes include: By deploying temperature sensors at multiple fixed locations along the processing path, temperature data at each location at different times is collected. The rate of temperature change per unit length is calculated by dividing the temperature difference between two adjacent measuring points by the distance between the two points. This rate of temperature change is then multiplied by the known thermal conductivity constant of polypropylene to obtain the change in heat flux density per unit area. This change in heat flux density is the basis for expressing the processing heat flux parameters. By monitoring the torque change curve of the mixing processing device per second at a set speed, and combining the speed change amplitude, the instantaneous shear rate within the same time period is calculated, and the shear rate fluctuation value is represented by the difference between the maximum and minimum shear rates within the same time period. Interfacial energy change data is obtained by analyzing the energy transfer characteristics of the interface between the filler and polypropylene during processing. Specifically, this includes identifying the range of interfacial tension changes in the heating and cooling zones based on continuous temperature monitoring data, recording the dynamic range of tension values ​​between adjacent filler particles per unit volume, and calculating the interfacial energy change data by multiplying the total tension change within this range with the change area of ​​the interface.

3. The method for improving the reinforcing performance of polypropylene mineral filler for flow battery frames according to claim 2, characterized in that, After data acquisition is completed, the initial analysis based on the thermal flow stability index and shear stress consistency index during the processing can be performed in the following ways: The processing heat flux parameters are arranged in time sequence. By calculating the average difference of the change in heat flux density in any two adjacent time periods, it is determined whether the processing heat flux parameters are in a stable range. When the change in heat flux density in multiple time periods fluctuates within a preset error range, it is defined as the heat flux stability reaching the evaluation standard. The shear stress consistency index is determined by segmenting the shear rate fluctuation value within a set time window and calculating the variance level of each fluctuation value to determine whether it meets the stability threshold. When the variance of the shear rate fluctuation value in multiple time periods is less than a given limit value, it is determined that the shear stress change trend is consistent. The thermal flow stability index and the shear stress consistency index are used together as the judgment criteria. Only when both of them simultaneously meet the steady state criterion can the initial analysis be deemed qualified and the subsequent identification stage be entered.

4. The method for improving the reinforcing performance of polypropylene mineral filler for flow battery electrode frames according to claim 3, characterized in that, After data acquisition is completed, methods for identifying behavioral segments with abnormal distributions or discontinuous interface connections in the dataset and extracting these segments to form candidate analysis regions include: A synchronous analysis matrix was constructed by combining processing heat flux parameters, shear rate fluctuation values, and interface energy change data along the time axis. The minimum analysis interval for each time period was set, and a sliding window method was used to compare the trends of the change curves for every five consecutive time periods. If the average change of the processing heat flux parameter between two consecutive windows exceeds 20%, it is defined as a sudden change in heat flux behavior. If the variance growth rate of the shear rate fluctuation value between two windows exceeds 30%, it is defined as shear-enhancing behavior. If the difference in the first derivative of the interface energy change data in adjacent time periods exceeds the preset slope deviation threshold, it is defined as interface bonding discontinuity behavior. If any parameter meets any of the above abnormal conditions within a set continuous time period, it is marked as an abnormal behavior segment, and the abnormal behavior segment, along with the two adjacent segments before and after it, is included in the analysis scope; all time periods included in the analysis scope are logically integrated to form a candidate analysis region.

5. The method for improving the reinforcing performance of polypropylene mineral filler for flow battery electrode frames according to claim 4, characterized in that, The calculation methods for the first index include: Within the time window corresponding to the candidate analysis region, a continuous time period is divided in units of one second. Image processing methods are used to extract the two-dimensional spatial distribution image of the filler particles in each time period, and the coordinate set of all particles is extracted. The average Euclidean distance of the position change of the same particle in two consecutive time periods is calculated to obtain the average position offset distance within the time window. Then, the local density of all particles in each time period is calculated. Taking the number of particles per unit area in the same time period as the density benchmark, the difference between the maximum density and the average density in the local density sequence is selected and its relative proportion is calculated, which is defined as the local dispersion fluctuation factor. After normalizing the average position offset distance and the local dispersion fluctuation factor, the average position offset distance is used as the independent variable and the local dispersion fluctuation factor is used as the dependent variable. A continuous three-second time period is used as a sliding window. Within each sliding window, three parameter pairs consisting of average position offset distance and local dispersion fluctuation factor are formed. Linear regression fitting is performed on all parameter pairs in each sliding window, and the fitting slope of the linear function is extracted. The fitting slope parameter sequence generated by each segment of the linear function is arranged by time, and the standard deviation of the sequence is calculated as the slope variability index. At the same time, the fitting residual of each segment of the linear function is calculated as the sum of squares of the differences between the fitted value and the original data, and a residual energy sequence is constructed. The coefficient of variation of the residual energy sequence is taken as the residual volatility index. The slope variability index and the residual fluctuation index reflect the stability of the trend changes of filler particles in spatial movement and density fluctuation, and constitute a numerical expression of the dynamic distribution stability of the filler in the mixing zone. The weighted average of the slope variability index and the residual fluctuation index is defined as the first index.

6. The method for improving the reinforcing performance of polypropylene mineral filler for flow battery frames according to claim 5, characterized in that, The calculation methods for the second index include: For each anomalous behavior segment within the candidate analysis region, the interface energy change data sequence within that anomalous behavior segment is extracted and sampled at a frequency of once per second to construct a tension time series. For the tension time series within each segment of abnormal behavior, the first-order difference method is used to calculate the rate of tension change between adjacent time points, and the standard deviation of all tension change rates within the segment is calculated and defined as the slope change factor of the segment. Simultaneously, identify the peak point in the tension sequence, calculate the ratio of the difference between the peak point and the average tension value in the segment, and define it as the tension jump factor of the segment; The slope change factor and tension jump factor of each abnormal behavior segment are averaged separately, and the arithmetic mean of the two is calculated with equal weights to form the fusion factor. The total duration of all abnormal behavior segments within the candidate analysis area is calculated and then compared with the total duration of the entire candidate analysis area. This ratio is defined as the abnormal coverage rate. Ultimately, the product of the fusion factor and the anomaly coverage rate is used as the basis for the numerical expression of the second index.

7. The method for improving the reinforcing performance of polypropylene mineral filler for flow battery electrode frames according to claim 6, characterized in that, The prediction model can be any one of the following three structures: I. A model based on a one-dimensional convolutional neural network structure; II. A model based on a dual-hidden-layer feedforward neural network structure; III. Model based on gradient boosting tree structure.

8. The method for improving the reinforcing performance of polypropylene mineral filler for flow battery electrode frames according to claim 7, characterized in that, Based on the category to which the enhancement fit coefficient belongs in the corresponding response interval defined in the prediction model, matching the predefined processing scheme refers to: The range of the enhanced fit coefficient is divided into three intervals: below 0.6, between 0.6 and 0.8, and above 0.

8. When the enhancement fit coefficient is less than 0.6, the processing scheme is defined as the instability level, and the structural perturbation behavior in the optimization path of the processing scheme is reconstructed. When the enhancement adaptability coefficient is between 0.6 and 0.8, the processing scheme is defined as an intermediate adjustable level, and the interface control behavior in the processing scheme optimization path is refactored. When the enhancement adaptability coefficient is higher than 0.8, the processing scheme is defined as a high adaptability level, and the rhythm fine-tuning behavior reconstruction in the matching processing scheme optimization path is performed.

Images