A mold partition heating control method for electric energy metering box production

Abstract

The application provides a mold partition heating control method for electric energy metering box production, comprising: obtaining a three-dimensional model of a multi-table position partition intersection area and an adjacent thin-walled area in a mold cavity of electric energy metering box production, calculating a volume ratio of corresponding mold steel materials of the two areas, and identifying a thermal inertia difference caused by a thick-thin transition interface; inputting the thermal inertia difference and a preset initial heating power into a support vector regression machine to predict a transient temperature difference zone formed between the intersection area and the thin-walled area in a heating stage, and extract a maximum temperature difference peak and valley and a time period when the maximum temperature difference peak and valley occur; evaluating resin local overheating and thin-walled solidification risk through the maximum temperature difference peak and valley, determining a first heating power according to the evaluated risk; analyzing an influence of a mismatch between a pushing speed of the intersection area and a pushing speed of the thin-walled area on residual internal stress accumulation at the multi-table position partition intersection, evaluating power redistribution requirements accordingly, and generating an output power distribution scheme of each independent heating circuit.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a method for controlling the partitioned heating of molds in the production of electricity metering boxes. Background Technology

[0002] In the manufacturing of electricity metering boxes, mold zone heating control is a core factor determining product molding quality. Current conventional heating control approaches generally assume that increasing heating power in areas with lower temperatures will accelerate the heating rate in those areas, bringing the overall mold temperature closer to uniformity. However, when dealing with the junction area of ​​multiple metering partitions, the junction area, due to its thick structure, has a much larger steel volume than the surrounding thin-walled areas. This is analogous to simultaneously heating a large pot of water and a cup of water with the same intense heat. If the heating power is blindly increased to quickly reach the target temperature in the thick junction area, the junction area, due to its large volume, absorbs heat slowly, resulting in a slow temperature rise. Meanwhile, the adjacent thin-walled areas, due to their smaller volume, will heat up rapidly under high power. The higher the heating power, the amplified this difference in thermal inertia caused by the sudden increase in volume will be, leading to a significant transient temperature difference between the junction area and the thin-walled areas within a short period. This temperature difference causes the resin in the thin-walled areas to harden rapidly, while the resin in the junction area remains in a flowing state, resulting in a complete disconnect in the curing progress between the two areas. A mismatch in curing speed can cause intense tension within the material, leading to a high concentration of residual internal stress and ultimately causing the box to crack. Therefore, when faced with complex conditions where the volume of steel increases rapidly in different areas of the mold, accurately controlling the transient temperature difference amplified by differences in thermal inertia, avoiding a mismatch in curing progress, and achieving synchronous hardening of the resin and uniform stress distribution throughout the mold cavity becomes a key issue in ensuring the production quality of electricity metering boxes. Summary of the Invention

[0003] This invention provides a method for controlling the zoned heating of molds in the production of electricity metering boxes, mainly including: Obtain a three-dimensional model of the intersection area of ​​the multi-position partition and the adjacent thin-walled area in the mold cavity of the electricity metering box production, calculate the ratio of the corresponding mold steel volume of the two areas of the partition intersection area and the adjacent thin-walled area, and identify the thermal inertia difference caused by the thick and thin transition interface. The thermal inertia difference and the preset initial heating power are input into the support vector regression machine to predict the transient temperature difference band formed between the intersection region and the thin-walled region during the heating stage, and the maximum temperature difference peak and valley are extracted. The risk of localized overheating and premature curing of thin-walled resin is assessed by evaluating the maximum temperature difference peak and valley, and the first heating power is determined based on the assessed risk. The independent heating circuits of each zone of the mold cavity are driven by the first heating power. Based on the surface temperature of the junction area, the surface temperature of the thin-walled area, and the local heat flux density of the junction area, the expansion path and influence range of the transient temperature difference zone along the thick-thin transition interface are identified. Based on the expansion path and influence range of the transient temperature difference band along the thick-thin transition interface, a random forest tree is used to identify the physical state stage of the resin in the junction region and the thin-walled region, and the propagation speed of the curing front in the junction region and the thin-walled region is analyzed. The effect of the mismatch between the advancing speed in the intersection region and the advancing speed in the thin-walled region on the accumulation of residual internal stress at the intersection of multiple surface plate partitions is analyzed. Based on this, the power redistribution requirement is assessed, and the output power distribution scheme of each independent heating circuit is generated.

[0004] Furthermore, the process of obtaining a three-dimensional model of the intersection area of ​​the multi-position partitions and the adjacent thin-walled area in the mold cavity for producing the electricity metering box, calculating the ratio of the corresponding mold steel volumes of the two areas (the partition intersection area and the adjacent thin-walled area), and identifying the thermal inertia difference caused by the thick-thin transition interface includes: Obtain the design drawing of the mold cavity, and separate the multi-level partition plate intersection area and the adjacent thin-walled area from the design drawing. Import them into the triangular mesh modeling tool to obtain the 3D solid model of the multi-level partition plate intersection area and the 3D solid model of the adjacent thin-walled area, and mark the contact surface along the thickness transition interface. Obtain the solid volume of the mold steel corresponding to the two areas of the partition plate intersection area and the adjacent thin-walled area through mesh volume integration. Divide the solid volumes of the mold steel corresponding to the two areas of the partition plate intersection area and the adjacent thin-walled area to obtain the ratio of the steel volume. Project the ratio of the steel volume point by point along the contact surface to the discrete nodes of the thickness transition interface to obtain the thermal inertia difference distribution of the thickness transition interface.

[0005] Furthermore, the step of inputting the thermal inertia difference and the preset initial heating power into a support vector regression machine to predict the transient temperature difference band formed between the intersection region and the thin-walled region during the heating stage, and extracting the maximum temperature difference peak and valley, includes: The discrete node value sequence of thermal inertia difference distribution is concatenated with the preset initial heating power to form an input vector, which is then fed into a support vector regression machine pre-trained with historical heating test data to obtain the temperature difference sequence between the intersection region and the thin-walled region on each discrete node during the heating stage. The temperature difference sequence is concatenated along the thick-thin transition interface to form the transient temperature difference band. Local extreme points are scanned along the heating time axis to obtain the maximum temperature difference peak and valley. The heating time coordinates corresponding to the temperature difference sequence are read, and the heating time intervals where the temperature difference peak and valley are located are extracted to obtain the occurrence time period and pair it with the maximum temperature difference peak and valley for output.

[0006] Furthermore, determining the first heating power based on the assessed risk includes: The overall risk level is determined based on the risk level of thin-walled pre-curing and the risk level of local overheating in the intersection area. When the overall risk level exceeds the preset warning level, the overheating risk is deemed prominent. The preset power reduction level table is then called, and the preset initial heating power is reduced according to the level in the preset power reduction level table that matches the overall risk level to obtain the first heating power. When the overall risk level does not exceed the preset warning level, the overheating risk is deemed under control, and the preset initial heating power is used as the first heating power.

[0007] Furthermore, the method of driving the independent heating circuits of each zone of the mold cavity with the first heating power, and identifying the expansion path and influence range of the transient temperature difference band along the thick-thin transition interface based on the surface temperature of the intersection area, the surface temperature of the thin-walled area, and the local heat flux density of the intersection area, includes: The first heating power drives the independent heating circuits of each zone of the mold cavity to operate synchronously. Contact thermocouples arranged on the surfaces of the intersection area and the thin-walled area collect surface temperature at a fixed sampling period. Thin-film heat flux meters embedded in the steel thickness direction of the intersection area synchronously collect the local heat flux density of the intersection area, obtaining a temperature and heat flux density sampling sequence synchronously aligned along the thick-thin transition interface. The temperature difference gradient is obtained by dividing the temperature difference between the two surfaces of the intersection area and the thin-walled area by the geometric distance between the sampling points along the thick-thin transition interface. The local heat flux density is defined as positive pointing to the thin-walled area and negative pointing to the intersection area. The sign of the local heat flux density is used to determine the direction of heat conduction, obtaining the propagation direction of the transient temperature difference band along the thick-thin transition interface. Discrete nodes with temperature difference gradient values ​​higher than a preset sweep threshold are connected in series to form an extension path. Discrete nodes on both sides of the extension path whose temperature difference gradient falls back to within the preset sweep threshold are defined as boundaries, obtaining the influence range.

[0008] Furthermore, the method of identifying the physical state stages of the resin in the intersection region and the thin-walled region using random forest trees includes: The surface temperature time series and local heat flux density time series of each discrete node in the intersection region and the thin-walled region are collected along the extended path. For each discrete node, the temperature rise slope, the second-order difference temperature sign change point, and the duration of the heat flux density continuous stable segment are extracted as state stage feature vectors. The state stage feature vectors are organized into a state stage feature vector set along the boundary of the influence range and fed into a random forest tree that has been trained in advance with thermosetting resin curing test data. The output category labels of the random forest tree correspond to four types of resin physical states in sequence: viscous flow stage, gel stage, exothermic stage, and curing stage. A corresponding category label is obtained for each discrete node.

[0009] Furthermore, the analysis of the solidification front advancement velocity of the intersection region and the thin-walled region includes: Along the extended path, the boundary line between the discrete nodes of adjacent viscous flow stage and gel stage is determined as the gel interface position. The moment when the temperature rise slope changes from steep to flat in the temperature time sequence corresponding to the discrete node of the exothermic stage is taken as the starting point of the exothermic platform, and the moment when the temperature rise slope changes from flat to steep is taken as the ending point of the exothermic platform. The discrete nodes of adjacent gel stage and curing stage are connected in series to form the curing front profile. For the curing front profile on each side of the intersection region and the thin-walled region, the displacement increment of the curing front profile along the normal of the thick-thin transition interface at two adjacent sampling moments is divided by the corresponding sampling interval to obtain the curing front advancement speed of the intersection region and the thin-walled region.

[0010] Furthermore, the analysis of the mismatch between the propulsion velocity in the intersection region and the propulsion velocity in the thin-walled region and its impact on the accumulation of residual internal stress at the intersection of multiple epitope septa includes: The difference between the solidification front advance speed of the intersection region and the solidification front advance speed of the thin-walled region is calculated along the thick-thin transition interface to form a solidification front advance speed difference sequence. The distribution of the difference sequence on the thick-thin transition interface is mapped point by point to the geometric position of the intersection of the multi-epitope partition to obtain the solidification front advance mismatch zone.

[0011] Furthermore, after obtaining the solidified leading edge propulsion mismatch band, the process includes: The mismatch zone of the curing front is mapped onto the residual internal stress conversion table pre-calibrated with the strain-stress constitutive relationship of thermosetting resin curing. The corresponding residual internal stress value is found point by point. The residual internal stress is formed by connecting the points along the intersection of the multi-position partitions. The continuous segment in the residual internal stress accumulation distribution that exceeds the preset accumulation threshold is determined as the stress accumulation core zone.

[0012] Furthermore, the generation of the output power distribution scheme for each independent heating circuit includes: determining the power adjustment level corresponding to each independent heating circuit based on the position coordinates of the stress accumulation core zone and the difference sequence of the solidification front advancement speed; outputting an enhancement level or a maintenance level for the heating circuit in the intersection area; outputting an suppression level or a delay level for the heating circuit in the thin-walled area; and summarizing the levels to form the output power distribution scheme.

[0013] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention discloses a mold zoning heating control method for the production of electricity metering boxes. Addressing the challenges of asynchronous curing and stress concentration caused by significant differences in wall thickness between the intersection area of ​​multiple metering partitions and adjacent thin-walled areas, the method first obtains three-dimensional models of the two areas and calculates the ratio of steel volumes to quantify the thermal inertia difference at the thickness transition interface. This value, along with the initial heating power, is input into a support vector regression machine to predict the transient temperature difference band and the timing of the maximum temperature difference peak and trough during the heating phase. Based on this, the risk of localized resin overheating and premature curing of the thin-walled area is assessed. If overheating is significant, the power is reduced to a preset level; if under control, the initial power is used to determine the first heating power driving each... Independent circuits operate in zones; surface temperature and local heat flux density are collected to identify the expansion path and influence range of the temperature difference zone along the transition interface. Random forest tree is used to identify the gel interface position, the start and end points of the exothermic platform, and the outline of the curing front. The advancement speed of the curing front in the two regions is obtained. Based on the influence of speed mismatch on the accumulation of residual internal stress, the output power of each circuit is redistributed. The intersection area is moderately enhanced, while the thin-walled area is suppressed and delayed, so that the curing front of the entire cavity advances synchronously. This suppresses temperature difference impact, homogenizes residual stress, and eliminates stress concentration at the intersection, significantly improving the dimensional stability and molding yield of the metering box molded parts. Attached Figure Description

[0014] Figure 1 This is a flowchart of a mold zoning heating control method for the production of electricity metering boxes according to the present invention.

[0015] Figure 2 This is a schematic diagram of a mold zoning heating control method for the production of electricity metering boxes according to the present invention.

[0016] Figure 3 This is another schematic diagram of a mold zoning heating control method for the production of electricity metering boxes according to the present invention. Detailed Implementation

[0017] To further understand the content of this invention, a detailed description of the invention is provided in conjunction with the accompanying drawings and embodiments. The specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention. It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0018] like Figures 1-3 This embodiment of a mold zoning heating control method for the production of electricity metering boxes may specifically include: S101. Obtain the three-dimensional model of the intersection area of ​​the multi-position partition plate and the adjacent thin-walled area in the mold cavity of the power metering box production, calculate the ratio of the corresponding mold steel volume of the two areas, and identify the thermal inertia difference brought about by the thick-thin transition interface.

[0019] The design drawings of the mold cavity for the production of the electricity metering box are obtained. From these drawings, the intersection area of ​​the multi-position partitions and the adjacent thin-walled area are separated. These two areas are imported into a triangular mesh modeling tool to obtain 3D solid models of the intersection area and the thin-walled area, respectively. The contact surfaces along the thickness transition interface are then marked. For the 3D solid models of the intersection area and the thin-walled area, the solid volumes of the corresponding mold steel for each area are obtained through mesh volume integration. The solid volumes of the corresponding mold steel for each area are divided to obtain the ratio of their respective steel volumes. This ratio is then projected point-by-point along the contact surface onto discrete nodes of the thickness transition interface to obtain the thermal inertia difference distribution between the intersection area and the thin-walled area along the thickness transition interface.

[0020] In one embodiment, the design file of the mold cavity for producing the energy metering box is stored in a standard three-dimensional solid format, containing the overall geometry of the multi-position partitions. The multi-position partition intersection area is segmented from the design file according to the geometric features of the partition intersection, and a thin-walled area adjacent to the intersection area along the thickness-thin transition interface is segmented according to the thin-walled feature. The two areas are imported into a triangular mesh modeling tool, and the solid surfaces are discretized using triangular facets to obtain the three-dimensional solid models of the intersection area and the thin-walled area. The contact surface node sequence of the thickness-thin transition interface is marked on the common contact surface of the two models.

[0021] For example, the mesh volume integration operation calculates the volume of each tetrahedral element within the three-dimensional solid model and sums them up to obtain the solid volume of the mold steel in the corresponding region. Let the solid volume of the steel in the intersection region be V1 and the solid volume of the steel in the thin-walled region be V2, then the ratio of the steel volumes is R = V1 / V2, where V1 and V2 are both measured in cubic millimeters, and R is a dimensionless value that reflects the degree of difference in the volume of the steel in the two regions.

[0022] It should be noted that the ratio R of the steel volume does not act uniformly on the thickness transition interface, but is projected point by point along the contact surface node sequence onto the discrete nodes of the thickness transition interface.

[0023] Specifically, each discrete node carries a local volume ratio mapping value, determined by the relative difference between the steel volume of the intersection region and the thin-walled region in the neighboring area of ​​that node. By concatenating the mapping values ​​of all discrete nodes, the thermal inertia difference distribution between the intersection region and the thin-walled region along the thick-thin transition interface is obtained. This thermal inertia difference distribution reflects the significant variation in the heat absorption capacity of the steel at different locations on the thick-thin transition interface, providing a geometric basis for predicting transient temperature differences during the subsequent heating stage. In another embodiment, the design drawing of the mold cavity for the production of the energy metering box is obtained. The intersection region of the multi-position partition and the adjacent thin-walled region are separated from the design drawing. The two regions are imported into a triangular mesh modeling tool to obtain a three-dimensional solid model of the intersection region and a three-dimensional solid model of the thin-walled region, and the contact surface between them along the thick-thin transition interface is marked. For the three-dimensional solid models of the intersection region and the thin-walled region, the solid volumes V1 and V2 of the corresponding mold steel in the two regions are obtained through mesh volume integration calculation, and the ratio of the steel volumes R = V1 / V2 is calculated. The contact surface is discretized into a triangular mesh to obtain N discrete nodes. For each node, a preset distance of 3 mm is extended along the normal direction into the intersection region and the thin-walled region, respectively. The local thickness d1i of the intersection region and the local thickness d2i of the thin-walled region at that node are sampled, and the local thickness ratio at that node is calculated. The normalized volume distribution coefficient at that node is obtained by multiplying the global volume ratio R by the local thickness ratio ri. Normalize the wi values ​​of all nodes until their sum is 1. Multiply the normalized volume distribution coefficient by the product ρc of the steel density ρ and specific heat capacity c to obtain the thermal inertia difference value at that node. Where ΔT0 is a reference temperature difference of 50 degrees Celsius, the thermal inertia difference distribution H1 to HN between the intersection region and the thin-walled region along the thick-thin transition interface is obtained.

[0024] S102. Input the thermal inertia difference and the preset initial heating power into the support vector regression machine to predict the transient temperature difference band formed between the intersection area and the thin-walled area during the heating stage, and extract the maximum temperature difference peak and valley and the time period of their occurrence.

[0025] The thermal inertia difference distribution between the intersection region and the thin-walled region along the thick-thin transition interface is obtained. The discrete node value sequence of the thermal inertia difference distribution is concatenated with a preset initial heating power to form an input vector, which is then fed into a support vector regression machine pre-trained with historical heating test data to obtain the temperature difference sequence between the intersection region and the thin-walled region at each discrete node during the heating stage. The temperature difference sequence is concatenated along the thick-thin transition interface to form a transient temperature difference band. For the temperature difference sequence of each discrete node in the transient temperature difference band, local extreme points are scanned along the heating time axis. The local maximum point is recorded as the temperature difference peak, and the local minimum point is recorded as the temperature difference valley. The highest value is selected from all temperature difference peaks, and the lowest value is selected from all temperature difference valleys to obtain the maximum temperature difference peak-valley. For the maximum temperature difference peak and valley, read the corresponding heating time coordinates in the temperature difference sequence, extract the heating time interval where the temperature difference peak is located and the heating time interval where the temperature difference valley is located, obtain the occurrence time period of the maximum temperature difference peak and valley, and output the maximum temperature difference peak and valley with the occurrence time period.

[0026] In one embodiment, the support vector regression machine is a supervised regression learner that minimizes structural risk. It predicts continuous outputs by finding a fitting function in the feature space that ensures all samples deviate from the regression hyperplane within a preset tolerance band. Considering the characteristics of the temperature difference between the junction region and the thin-walled region during the mold heating stage, which fluctuates over time and exhibits strong nonlinear coupling with the steel volume ratio and heating power, the support vector regression machine can complete the fitting prediction of the transient temperature difference band under small sample experimental data conditions.

[0027] Specifically, the input vector of the support vector regression machine is composed of two parts. The first part is the thermal inertia difference value of discrete nodes on the thick-thin transition interface. The thermal inertia difference values ​​of N discrete nodes spread along the thick-thin transition interface are arranged sequentially and denoted as h1, h2 to hN. The second part is the preset initial heating power P0, which is the power level set by each independent heating circuit of the mold cavity when the temperature rise is started. The concatenated input vector is in the form of h1, h2 to hN connected end to end with P0, forming a feature vector of length N+1, which is fed into the support vector regression machine.

[0028] It should be noted that the training samples for the support vector regression machine are derived from historical heating tests conducted by the mold factory in the early stages. Each set of samples corresponds to a complete heating test record, which includes the sequence of thermal inertia differences calculated from the steel volume ratio of the intersection region and the thin-walled region of the mold cavity used in the test, the heating power level set for the test, and the time series of temperature difference values ​​synchronously collected at fixed time intervals by thermocouples embedded on the surfaces of the intersection region and the thin-walled region during the heating process. The input vectors of the historical samples and the corresponding temperature difference time series are fed into the support vector regression machine in pairs for parameter optimization, resulting in the trained support vector regression machine.

[0029] It is understood that the transient temperature difference band refers to a two-dimensional distribution of temperature differences along the thick-thin transition interface during the heating phase, which evolves continuously with heating time. One dimension is the discrete node position on the thick-thin transition interface, and the other dimension is the heating time. Each cell records the difference between the surface temperature of the intersection region and the surface temperature of the thin-walled region at that position and at that moment. Further, for the temperature difference sequence corresponding to each discrete node in the transient temperature difference band, adjacent values ​​are compared point by point along the heating time axis from the start time to the end time. Values ​​at time points that are larger than both the values ​​at the previous and subsequent time points are marked as local maxima, i.e., temperature peaks; values ​​at time points that are smaller than both the values ​​at the previous and subsequent time points are marked as local minima, i.e., temperature valleys. Multiple temperature peaks and valleys typically appear at a discrete node, corresponding to multiple temperature fluctuations caused by differences in thermal inertia during the heating process.

[0030] For example, the selection process for the maximum temperature difference peak and valley is as follows: First, all temperature difference peaks obtained from discrete nodes are aggregated into a temperature difference peak set, and the highest value in the temperature difference peak set is taken as the maximum temperature difference peak; then, all temperature difference valleys obtained from discrete nodes are aggregated into a temperature difference valley set, and the lowest value in the temperature difference valley set is taken as the maximum temperature difference valley. The maximum temperature difference peak and the maximum temperature difference valley are collectively referred to as the maximum temperature difference peak and valley. For the maximum temperature difference peak and valley, the continuous time period before and after the time point where the temperature difference value of the maximum temperature difference peak is located in the temperature difference value sequence, from rising to leveling off and then starting to fall, is read as the temperature rise time interval of the temperature difference peak; the continuous time period before and after the time point where the temperature difference value of the maximum temperature difference valley is located, from falling to leveling off and then starting to rise, is read as the temperature rise time interval of the temperature difference valley. The above two temperature rise time intervals are combined into the occurrence period of the maximum temperature difference peak and valley, and are paired with the value of the maximum temperature difference peak and valley for output.

[0031] Preferably, for situations where historical heating test data is scarce, a support vector regression machine with a radial basis function kernel can be used. The thermal inertia difference values ​​and the initial heating power in the input vector are normalized before being input, ensuring that the prediction accuracy of the transient temperature difference band during the heating stage meets the requirements of mold zone heating control for identifying the maximum temperature difference peaks and valleys and their occurrence periods. In another embodiment, the thermal inertia difference distribution along the thick-thin transition interface between the intersection region and the thin-walled region is obtained. The value sequence of N discrete nodes is uniformly sampled along the transition interface and concatenated with a preset initial heating power to form an input vector of length N+1. The heating time axis is divided into M discrete time points with a step size of 0.5 seconds. For each discrete time point, N support vector regression machines with scalar output are trained. The i-th model at the k-th time point takes the same N+1-dimensional vector as input and outputs the temperature difference between the intersection region and the thin-walled region at the i-th node at that time point. A radial basis function is selected as the kernel function, with a penalty coefficient of 10 and an insensitive bandwidth of 0.05 degrees Celsius. The training samples are the measured values ​​of the transition interface temperature difference sampled every 0.5 seconds in historical heating experiments. During prediction, the current N+1-dimensional input vector is distributed to all N x M models, and backfilled into an N x M matrix according to the node number vertically and the time number horizontally, thus obtaining the transient temperature difference band. For the row vector corresponding to each discrete node in the transient temperature difference band, local extreme points are scanned along the heating time axis. The local maximum points are recorded as temperature difference peaks, and the local minimum points are recorded as temperature difference valleys. The highest value is selected from all temperature difference peaks, and the lowest value is selected from all temperature difference valleys to obtain the maximum temperature difference peak-valley. For the maximum temperature difference peak and valley, read the corresponding heating time coordinates in the row vector, extract the heating time interval where the temperature difference peak is located and the heating time interval where the temperature difference valley is located, obtain the occurrence time period of the maximum temperature difference peak and valley, and output the maximum temperature difference peak and valley with the occurrence time period.

[0032] S103. Assess the risk of localized overheating and premature curing of thin-walled resin by assessing the maximum temperature difference peak and valley, and determine the first heating power based on the assessed risk.

[0033] The maximum temperature difference peaks and valleys and their occurrence periods are obtained. For the highest value among the maximum temperature difference peaks, a preset overheating threshold is compared to determine whether the resin temperature in the thin-walled region exceeds the resin gelation initiation temperature during the occurrence period, thus obtaining the thin-walled region's pre-curing risk level. For the lowest value among the maximum temperature difference valleys, a preset hysteresis threshold is compared to determine whether the resin temperature in the junction region remains stagnant in the flow dynamics for an extended period during the occurrence period, thus obtaining the junction region's local overheating risk level. The higher of the thin-walled region's pre-curing risk level and the junction region's local overheating risk level is taken as the comprehensive risk level. If the comprehensive risk level exceeds a preset warning level, the overheating risk is deemed prominent; if the comprehensive risk level does not exceed the preset warning level, the overheating risk is deemed controlled. In the case of prominent overheating risk, a preset power reduction level table is called, and the preset initial heating power is reduced according to the level in the table that matches the comprehensive risk level to obtain the first heating power. In the case of controlled overheating risk, the preset initial heating power is used as the first heating power.

[0034] In one embodiment, the determination of the risk level of thin-walled pre-curing revolves around the resin gelation initiation temperature. The resin gelation initiation temperature refers to the critical temperature at which the thermosetting resin filling the mold cavity transitions from a viscous flow state to a gel state during heating. Above this critical temperature, the resin molecular chains begin to cross-link, and the viscosity increases sharply. Before production, the mold manufacturer conducts differential scanning calorimetry (DSC) tests on the resin grade used to obtain the corresponding gelation initiation temperature, which is then stored in the process parameter library for use in this solution.

[0035] Specifically, for the highest value among the maximum temperature difference peaks, the discrete node location of the temperature difference peak is used as a base point. The occurrence time of the temperature difference peak is mapped to the predicted sequence of the surface temperature of the thin-walled region, and the peak value of the surface temperature of the thin-walled region during the occurrence time is read. Then, the peak value is compared with the resin gelation initiation temperature. If the peak value exceeds the preset overheating threshold corresponding to the gelation initiation temperature, it is determined that the resin in the thin-walled region has entered the gelation state in advance, and the risk level of thin-walled region solidification is recorded as high. If the peak value does not exceed the preset overheating threshold but is close to it, the risk level of thin-walled region solidification is recorded as medium. If the peak value is much lower than the preset overheating threshold, the risk level of thin-walled region solidification is recorded as low. Further, for the lowest value among the maximum temperature difference valleys, the determination of the local overheating risk level of the confluence region is based on the resin flow dynamic residence time. The predicted sequence of the surface temperature of the confluence region during the occurrence time of the temperature difference valley is read, and the continuous duration of the temperature in the sequence that is continuously lower than the resin gelation initiation temperature is counted. The continuous duration is compared with a preset lag threshold. The preset hysteresis threshold refers to the upper limit of the relative hysteresis window formed when the resin in the confluence region has entered the exothermic plateau period after the resin in the flow dynamic residence time in the confluence region has been too long. If the continuous duration exceeds the preset hysteresis threshold, the local overheating risk level of the confluence region is recorded as high; if the continuous duration is close to the preset hysteresis threshold, the local overheating risk level of the confluence region is recorded as medium; if the continuous duration is much lower than the preset hysteresis threshold, the local overheating risk level of the confluence region is recorded as low.

[0036] It should be noted that when the risk level of thin-walled pre-curing and the risk level of local overheating in the intersection area are combined into a comprehensive risk level, the higher risk level shall be used.

[0037] For example, if the risk level of thin-walled pre-curing is high and the risk level of local overheating in the intersection area is medium, then the overall risk level is high. This merging method ensures that any prominent risk in any area can be reflected in the overall risk level.

[0038] For example, the preset warning level is set to medium. If the overall risk level reaches high and exceeds the preset warning level, the overheating risk is determined to be prominent. If the overall risk level is low or medium and does not exceed the preset warning level, the overheating risk is determined to be under control.

[0039] In one possible implementation, the preset power reduction level table is pre-stored in the process controller. The level table is indexed by the comprehensive risk level, and each index entry corresponds to a power reduction coefficient. For index entries with a high comprehensive risk level, the corresponding power reduction coefficient is a value between 70% and 80% of the initial heating power; for index entries with an even higher comprehensive risk level, the corresponding power reduction coefficient is a lower value. When an overheating risk is determined to be significant, the power reduction coefficient matching the comprehensive risk level is read from the level table and denoted as k. The preset initial heating power P0 is multiplied by k to obtain the first heating power P1, i.e., P1 = P0 × k.

[0040] It is understandable that, for situations where the risk of overheating is under control, the preset initial heating power P0 is sufficient to meet the process requirement that the temperature difference between the intersection area and the thin-walled area will not cause resin curing mismatch. Therefore, the preset initial heating power is used as the first heating power P1, i.e., P1=P0.

[0041] Preferably, the power reduction coefficient of the preset power reduction level table can be set with multiple sets of values ​​for different resin grades and different mold specifications. When the process controller calls the level table, it selects the corresponding value set according to the grade and specification corresponding to the current mold cavity, so that the first heating power can be adapted to the different steel volumes in the multi-level partition intersection area and the different wall thicknesses in the thin-walled area. In another embodiment, the maximum temperature difference peak and valley and their occurrence time are obtained. For the highest value among the maximum temperature difference peaks, a preset overheating threshold is compared to determine whether the resin temperature in the thin-walled area exceeds the resin gelation initiation temperature during the occurrence time, thus obtaining the risk level of thin-walled premature curing. For the lowest value among the maximum temperature difference valleys, a preset hysteresis threshold is compared to determine whether the resin temperature in the intersection area remains in the flow state for a long time during the occurrence time, thus obtaining the risk level of local overheating in the intersection area. The preset overheating threshold Tth is obtained by offsetting the resin gel initial temperature Tg downwards by a safety margin ΔT, i.e., Tth = Tg - ΔT, where ΔT is between 3 and 8 degrees Celsius, used to provide early warning before the thin-wall temperature exceeds Tg. A peak thin-wall temperature exceeding Tg is classified as high-level, between Tth and Tg as medium-level, and below Tth as low-level. The preset hysteresis threshold is determined by the cumulative duration of temperature retention in the flow dynamics at the junction area: exceeding 90 seconds is high-level, 60 to 90 seconds as medium-level, and below 60 seconds as low-level. The higher of the thin-wall pre-curing risk level and the junction area local overheating risk level is taken as the comprehensive risk level. The preset warning level is set to medium-level. If the comprehensive risk level reaches or exceeds medium-level, the overheating risk is considered prominent; if the comprehensive risk level is below medium-level, the overheating risk is considered controlled. For situations where the overheating risk is significant, a preset power reduction level table is invoked, and the preset initial heating power P0 is reduced according to the level in the table that matches the overall risk level, resulting in a first heating power P1 = P0 * k. The level table is divided into two levels: when the overall risk level is medium, the reduction coefficient k = 0.85 to 0.9 is used, and when the overall risk level is high, k = 0.7 to 0.75 is used. For situations where the overheating risk is under control, the preset initial heating power is used as the first heating power.

[0042] S104. Drive the independent heating circuits of each zone of the mold cavity to operate with the first heating power, collect the surface temperature of the intersection area, the surface temperature of the thin-walled area, and the local heat flux density of the intersection area, and identify the expansion path and influence range of the transient temperature difference zone along the thick-thin transition interface.

[0043] The first heating power drives the independent heating circuits of each zone of the mold cavity to operate synchronously. Multiple contact thermocouples are arranged along the thickness transition interface direction on the surface of the junction area to collect the surface temperature of the junction area at a fixed sampling period. Similarly, contact thermocouples are arranged along the same interface direction on the surface of the thin-walled area to collect the surface temperature of the thin-walled area at the same sampling period. A thin-film heat flux meter embedded in the steel thickness direction of the junction area synchronously collects the local heat flux density of the junction area, resulting in a temperature and heat flux density sampling sequence synchronously aligned along the thickness transition interface. For this temperature and heat flux density sampling sequence, the surface temperature difference between adjacent measuring points is divided by the geometric distance between the two measuring points along the interface direction to obtain the temperature change rate in the interface direction. The local heat flux density of the junction area is defined as positive pointing towards the thin-walled area and negative pointing towards the junction area. The sign of the local heat flux density is used to determine the direction of heat conduction from the junction area to the thin-walled area or from the thin-walled area to the junction area, thus obtaining the propagation direction of the transient temperature difference band along the thickness transition interface. Based on the propagation direction of the transient temperature difference band along the thick-thin transition interface, discrete nodes with a temperature change rate higher than a preset sweep threshold along the interface direction are connected in series to form an extension path. Discrete nodes whose temperature change rate along the interface direction on both sides of the extension path falls back to within the preset sweep threshold are delineated as boundaries, thus obtaining the extension path and influence range of the transient temperature difference band along the thick-thin transition interface.

[0044] In one embodiment, the mold cavity is divided into several zones according to the intersection area of ​​multiple partition plates, adjacent thin-walled areas, and the surrounding conventional area. Each zone is powered by a corresponding independent heating circuit. The independent heating circuit typically consists of an embedded electric heating tube, a power controller, and power feedback wiring. The power controller receives a first heating power command and outputs the corresponding power to the electric heating tube according to the command, so that the steel in each zone enters the heating stage synchronously according to a predetermined heating curve.

[0045] Specifically, contact thermocouples are arranged at several measuring points evenly distributed along the thickness transition interface on the surface of the intersection area of ​​the mold cavity. Each measuring point is equipped with one K-type thermocouple, and the thermocouple probes are attached to the steel surface of the intersection area and coupled via thermally conductive silicone grease. An equal number of contact thermocouples are mirrored along the surface of the thin-walled area, ensuring a one-to-one correspondence between the measuring points in the thin-walled area and the intersection area along the thickness transition interface. A thin-film heat flow meter is embedded in the thickness direction of the steel in the intersection area. The thin-film heat flow meter has a structure of two thermopile layers sandwiching a calibration thermally conductive layer, capable of outputting a millivolt-level electrical signal proportional to the heat flux density passing through its thickness direction. The outputs of each thermocouple and each thin-film heat flow meter are fed into the same data acquisition instrument. The acquisition instrument synchronously samples all channels at a fixed sampling period, typically between 0.5 seconds and 2 seconds, obtaining a temperature and heat flux density sampling sequence synchronously aligned along the thickness transition interface. Furthermore, the temperature gradient calculation is performed on adjacent pairs of measuring points in the intersection region and the thin-walled region along the thick-thin transition interface. Let the surface temperature sampling value of the intersection region at a certain pair of measuring points be Tj, and the surface temperature sampling value of the thin-walled region be Tb. Let the geometric distance between the two measuring points along the thick-thin transition interface be d. Then the temperature gradient at that point is... The temperature gradient sequence is calculated sequentially from all the measuring points arranged along the thick-thin transition interface.

[0046] It should be noted that the sign of the local heat flux density in the convergence region is determined by the potential polarity of the two thermopile layers of the thin-film heat flux meter. When heat passes through the calibrated heat-conducting layer from the interior of the steel in the convergence region towards the steel in the thin-walled region, the potential of the upper thermopile is higher than that of the lower layer, and the output millivolt-level electrical signal is positive; when heat flows back from the steel in the thin-walled region into the steel in the convergence region, the potential polarity reverses, and the output millivolt-level electrical signal is negative.

[0047] In one possible implementation, the propagation direction of the transient temperature difference band is determined as follows: for the same pair of measuring points, if the temperature gradient G at that point is positive and the local heat flux density at that point is positive, then it is determined that heat flows from the convergence region to the thin-walled region, and the propagation direction is marked as outward expansion; if the temperature gradient G at that point is positive and the local heat flux density at that point is negative, then it is determined that heat flows back from the thin-walled region into the convergence region, and the propagation direction is marked as backflow. The propagation directions of all measuring point pairs along the thick-thin transition interface are summarized to obtain the propagation direction sequence of the transient temperature difference band along the thick-thin transition interface.

[0048] For example, the preset sweep threshold is the allowable critical value of the temperature difference gradient, which is preset according to the specification of the resin's lateral temperature difference tolerance during the heating stage in the mold factory's process manual. It is generally taken as a value between 0.3 degrees Celsius per millimeter and 0.5 degrees Celsius per millimeter. Along the thickness transition interface, discrete nodes with a temperature difference gradient G higher than the preset sweep threshold are connected in series according to their positions on the thickness transition interface. The broken line formed by the series connection is the expansion path of the transient temperature difference band along the thickness transition interface.

[0049] It is understood that the boundary delineation on both sides of the expansion path is performed as follows: Starting from each discrete node on the expansion path, extending along the normal direction perpendicular to the thick-thin transition interface towards the inner side of the steel in the intersection region and the inner side of the steel in the thin-walled region respectively, the first discrete node where the temperature difference gradient G falls back to within the preset sweep threshold is found, and this discrete node is marked as the boundary point at that location. All boundary points are connected end to end along the expansion path to form two boundary lines. The strip-shaped region enclosed by the two boundary lines and the expansion path together is the influence range of the transient temperature difference band along the thick-thin transition interface.

[0050] Preferably, for situations where the mold cavity structure is complex and a single sweep threshold is insufficient to cover all measurement point pairs, the preset sweep threshold can be set with multiple values ​​according to different segments of the thick-thin transition interface. The power controller identifies the expansion path and influence range according to the value group corresponding to each segment, so that the identification result of the transient temperature difference band is adapted to the different steel volumes in the intersection area of ​​the multi-level partition and the different wall thicknesses in the thin-walled area.

[0051] S105. Based on the expansion path and influence range of the transient temperature difference zone along the thick-thin transition interface, a random forest tree is used to identify the physical state stage of the resin in the junction region and the thin-walled region, and the curing front advancement speed in the junction region and the thin-walled region is analyzed.

[0052] The expansion path and influence range of the transient temperature difference band along the thick-thin transition interface are obtained. The surface temperature time series and local heat flux density time series of each discrete node in the intersection region and thin-walled region are collected along the expansion path. For each discrete node, the heating slope, the second-order temperature difference sign change point, and the duration of the continuous stable heat flux density period are extracted as state stage feature vectors. These state stage feature vectors are organized along the boundary of the influence range to form a state stage feature vector set. The state stage feature vector set is fed into a random forest tree pre-trained with thermosetting resin curing test data. The output category labels of the random forest tree correspond to four types of resin physical states: viscous flow stage, gel stage, exothermic stage, and curing stage. For each discrete node, a corresponding category label is obtained. The boundary line between adjacent viscous flow stage and gel stage discrete nodes along the expansion path is determined as the gel interface position. Along the extended path, the moment when the temperature rise slope changes from steep to flat in the temperature time sequence corresponding to the discrete node of the exothermic stage is taken as the starting point of the exothermic platform, and the moment when the temperature rise slope changes from flat to steep is taken as the ending point of the exothermic platform. The discrete nodes of the adjacent gel stage and the curing stage are connected in series to form the curing front profile. For the curing front profile on one side of the intersection region and the thin-walled region, the displacement increment of the profile along the normal of the thick-thin transition interface at two adjacent sampling moments is divided by the corresponding sampling interval to obtain the curing front advancement speed of the intersection region and the thin-walled region.

[0053] In one embodiment, the thermosetting resin undergoes four physical states sequentially during the heating stage of the mold cavity: viscous flow, gelation, exothermic reaction, and curing. In the viscous flow stage, the resin molecular chains can still flow freely, and the temperature rises approximately linearly with the heat flow. In the gelation stage, the molecular chains begin to crosslink, the viscosity increases sharply, but exothermic reaction has not yet started. In the exothermic stage, the crosslinking reaction releases heat in a concentrated manner, the temperature shows a distinct step-like plateau, and there is a stable output of heat flux density. In the curing stage, the crosslinked network is basically formed, and the temperature rises approximately linearly again with external heating. Identifying the spatial distribution of these four states along the thickness transition interface is a prerequisite for determining the position of the curing front and its advancement speed.

[0054] Specifically, along the expansion path of the transient temperature difference band, three types of state stage features are extracted for each discrete node on the expansion path. The first type is the temperature rise slope, which is the mean of the first-order difference of the surface temperature time series of the discrete node within a sliding time window, denoted as s. The width of the sliding time window is selected as an integer multiple of the sampling period, generally taking a value between 5 and 10 sampling points. The second type is the sign change point of the second-order difference of the temperature, which is the time coordinate of the second-order difference of three adjacent sampling points in the surface temperature time series of the discrete node changing from positive to negative or from negative to positive. The sign change point reflects the inflection point of the temperature rise rhythm and is marked as the set of inflection moments corresponding to the discrete node. The third type is the duration of the continuous stable heat flux density segment, which is the duration of the continuous segment in the local heat flux density time series corresponding to the discrete node where the fluctuation amplitude between adjacent sampling points does not exceed a preset fluctuation threshold, denoted as f. The ordered array consisting of the values ​​s, the number of elements in the set of turning points, and f is used as the state stage feature vector corresponding to the discrete node. The state stage feature vectors of all discrete nodes are arranged in positional order along the boundary of the influence range to form a state stage feature vector set.

[0055] It should be noted that the random forest tree is a supervised classification learner formed by combining several CART decision trees in parallel. Each decision tree takes the feature vector as input, recursively divides the trees according to the node information gain criterion to obtain the class prediction, and then synthesizes the predictions of all decision trees into the final class label by majority voting.

[0056] In one possible implementation, the training samples for the random forest tree are derived from thermosetting resin curing experiments. Each set of samples consists of a state-stage feature vector, compiled based on the characteristic performance of four physical states—viscous flow stage, gel stage, exothermic stage, and curing stage—under a standard heating curve, along with corresponding manually labeled category labels. The training samples are then fed into the random forest tree for learning, resulting in a trained random forest tree.

[0057] Specifically, each state stage feature vector in the set of state stage feature vectors is sequentially fed into a trained random forest tree to obtain the category label corresponding to the discrete node. The category label value is one of the following: viscous flow stage, gel stage, exothermic stage, or solidification stage. The category labels of all discrete nodes are arranged in positional order along the expansion path to obtain a sequence of physical state stages distributed along the expansion path. Further, the physical state stage sequence is scanned along the expansion path from the intersection region to the thin-walled region. The boundary between two adjacent discrete nodes, where the former is categorized as the viscous flow stage and the latter as the gel stage, is determined as the gel interface location. For each discrete node categorized as the exothermic stage, its surface temperature time sequence is traced back. The moment when the heating slope s drops from above a preset steep rise threshold to below the preset steep rise threshold is taken as the starting point of the exothermic platform. The moment when the heating slope s jumps back from below the preset steep rise threshold to above the preset steep rise threshold is taken as the ending point of the exothermic platform. The preset steep rise threshold is preset based on the typical heating slope of the resin grade before and after entering the exothermic platform under the standard heating curve, and is generally taken as a value between 0.05 degrees Celsius per second and 0.15 degrees Celsius per second.

[0058] It is understood that the concatenation of the curing front profile uses the boundary points between adjacent discrete nodes labeled as the gel stage and the curing stage as primitives. All boundary points are connected end-to-end according to their positional order on the expansion path to obtain a curve extending along the thick-thin transition interface; this curve is the curing front profile. A segment of the curing front profile on one side of the intersection region is denoted as the curing front profile intersection branch, and a segment on the other side of the thin-walled region is denoted as the curing front profile thin-walled branch.

[0059] For example, the process of calculating the curing front advancement velocity is as follows: For the intersection branch of the curing front contour, take the displacement increment Δx1 of the intersection branch along the normal direction of the thick-thin transition interface at two adjacent sampling times, divide it by the sampling interval Δt between the two sampling times, and obtain the curing front advancement velocity v1 = Δx1 ÷ Δt in the intersection region; For the thin-walled branch of the curing front contour, take the displacement increment Δx2 of the thin-walled branch along the normal direction of the thick-thin transition interface at two adjacent sampling times, divide it by the same sampling interval Δt, and obtain the curing front advancement velocity v2 = Δx2 ÷ Δt in the thin-walled region. v1 and v2 are the curing front advancement velocities in the intersection region and the thin-walled region, respectively.

[0060] Preferably, the number of decision trees in the random forest tree is between 100 and 300. When each decision tree is generated, two components from the feature vector of the state stage are randomly selected to participate in node division, so that the recognition results of the gel interface position, the start and end points of the exothermic platform and the contour of the curing front remain stable for the combination of different steel volumes in the intersection area of ​​the multi-epitope partitions and different wall thicknesses in the thin-walled area in the mold cavity.

[0061] S106. Analyze the impact of the mismatch between the advancing speed in the intersection area and the advancing speed in the thin-walled area on the accumulation of residual internal stress at the intersection of multiple surface plate partitions. Based on this, assess the power redistribution requirements and generate the output power distribution scheme for each independent heating circuit.

[0062] The curing front advancement velocities of the intersection region and the thin-walled region are obtained. The difference between these velocities is calculated along the thick-thin transition interface to obtain a curing front advancement velocity difference sequence. The distribution of this difference sequence along the thick-thin transition interface is then mapped point-by-point to the geometric position at the intersection of the multi-epitope partitions, resulting in a curing front advancement mismatch zone. For this curing front advancement mismatch zone, its difference values ​​are mapped to a residual internal stress conversion table pre-calibrated using the thermosetting resin curing strain-stress constitutive relationship. The corresponding residual internal stress values ​​are then looked up point-by-point and connected along the intersection of the multi-epitope partitions to form a residual internal stress accumulation distribution. Continuous segments exceeding a preset accumulation threshold in this residual internal stress accumulation distribution are identified as stress accumulation core zones. The position coordinates of the stress accumulation core zone and the difference sequence of the solidification front advance speed are used as input features and fed into a random forest tree that has been trained in advance with historical power adjustment test data. The output of the random forest tree is the power adjustment level corresponding to each independent heating circuit. For the heating circuit in the intersection area, the output is an enhancement level or a maintenance level. For the heating circuit in the thin-walled area, the output is an suppression level or a delay level. The output power allocation scheme of each independent heating circuit is formed by summarizing the levels.

[0063] In one embodiment, the accumulation of residual internal stress at the junction of the multi-epitope septa originates from the inconsistency in the advancement speed of the curing front between the junction region and the thin-walled region. When the curing front of the thin-walled region advances before that of the junction region, the resin in the thin-walled region forms a cross-linked network first, constraining the resin in the junction region, which is still in the gel or exothermic stage. This constraint accumulates at the junction along the thickness transition interface, forming a local peak of residual internal stress. Conversely, when the curing front of the junction region advances before that of the thin-walled region, stress accumulation in the opposite direction will also occur at the junction. The magnitude of the difference in advancement speed between the two regions has a monotonic correspondence with the magnitude of the accumulated residual internal stress.

[0064] Specifically, the construction method of the solidification front advancement velocity difference sequence is as follows: the solidification front advancement velocity v1 in the intersection region and the solidification front advancement velocity v2 in the thin-walled region are mapped one-to-one according to the discrete node positions along the thick-thin transition interface, and the difference Δv = v1 - v2 is calculated point by point to obtain the difference Δv sequence laid out along the thick-thin transition interface. A positive difference Δv indicates that the solidification front advancement in the intersection region is faster, and a negative difference Δv indicates that the solidification front advancement in the thin-walled region is faster. The distribution of the difference Δv sequence on the thick-thin transition interface is projected onto the actual geometric coordinates of the intersection of the multi-epipod partitions according to their corresponding geometric node positions, presenting a banded spatial distribution along the intersection. This banded spatial distribution is the solidification front advancement mismatch zone. The width of the mismatch zone reflects the lateral expansion of the mismatch at the intersection, and the intensity of the mismatch zone reflects the absolute magnitude of the difference Δv at each node.

[0065] It should be noted that the residual internal stress conversion table is constructed based on the pre-calibrated strain-stress constitutive relationship of the thermosetting resin. The principle is that in the continuous segment from gelation to complete curing, the material modulus monotonically increases with the degree of cross-linking. When there is a difference in the advancement speed of the curing front in adjacent regions, the cured region forms a constraint strain on the adjacent uncured region. The product of the constraint strain and the modulus yields the constraint stress. Based on this constitutive relationship, the mold manufacturer uses the discrete value of the difference Δv as input in the early experimental stage, and the residual internal stress measured by the resistance strain gauge attached to the sample surface as output, establishing a correlation between Δv and the residual internal stress values, and storing this correlation as a residual internal stress conversion table.

[0066] Specifically, for each discrete node on the mismatch zone advancing along the curing front, its difference value Δv is substituted into the residual internal stress conversion table to obtain the corresponding residual internal stress value. The residual internal stress values ​​of each discrete node are then concatenated along the geometric coordinate sequence of the intersection of the multi-position partitions to obtain the residual internal stress concentration distribution. Further, the stress concentration core zone is determined by scanning point-by-point along the residual internal stress concentration distribution, marking discrete nodes with residual internal stress values ​​higher than a preset concentration threshold as concentration points. Consecutive adjacent concentration points are merged into a segment. The preset concentration threshold is pre-set based on the upper limit of safe stress for the resin grade under standard curing processes, generally taking a value between 50% and 70% of the resin's ultimate tensile stress. The resulting continuous segment is the stress concentration core zone, and the coordinates of the two ends and the center of the stress concentration core zone are recorded according to the geometric coordinates of the intersection of the multi-position partitions.

[0067] In one possible implementation, the training samples for the random forest tree originate from power adjustment experiments previously conducted by the mold factory. Each set of samples consists of the location coordinates of the stress concentration core zone recorded in the previous experiments, the sequence of differences in the solidification front advance speed Δv under the corresponding working conditions, and the power adjustment level label that was actually used in the current experiment and achieved uniform solidification. The training samples are fed into the random forest tree for parameter optimization according to the majority voting criterion to obtain the trained random forest tree.

[0068] It is understood that the power adjustment levels consist of four levels. The "Enhancement Level" increases the output power of the heating circuit in the junction area by a preset amplification factor, typically between 1.05 and 1.20. The "Maintainment Level" keeps the output power of the heating circuit in the junction area unchanged. The "Suppression Level" reduces the output power of the heating circuit in the thin-walled area by a preset reduction factor, typically between 0.80 and 0.95. The "Delay Level" postpones the start of the heating stage required to reach the heat release stage of the heating circuit in the thin-walled area by a preset delay window, where the preset delay window is graded from shortest to longest based on the absolute magnitude of the current difference Δv.

[0069] For example, the position coordinates of the stress accumulation core zone and the difference Δv between the solidification front advancement speeds are concatenated into an input feature vector and fed into a trained random forest tree. For each independent heating circuit in the mold cavity, its corresponding power adjustment level label is obtained. For heating circuits in the intersection area, the enhancement or maintenance level is read; for heating circuits in the thin-walled area, the suppression or delay level is read. The level labels of each independent heating circuit are arranged and summarized in circuit number order to obtain the output power allocation scheme for each independent heating circuit.

[0070] Preferably, when the mold cavity structure is complex and several unconnected stress accumulation core zones are distributed at the intersection of multiple surface plate partitions, the random forest tree independently outputs a corresponding power adjustment level for each stress accumulation core zone. The power controller issues power commands in parallel loop by loop according to the output power distribution scheme, so that the solidification front advance speed v1 of the intersection area and the solidification front advance speed v2 of the thin-walled area tend to be similar along the thick-thin transition interface, and the residual internal stress accumulation distribution tends to be uniform along the intersection of multiple surface plate partitions. In another embodiment, the position coordinates of the stress accumulation core zone and the difference sequence of the solidification front advance speed are used as input features and fed into a pre-trained random forest tree. The random forest tree consists of two subtrees. Before training, the historical power adjustment test data are marked according to the location of the heating loop. Samples falling at the intersection of multiple surface plate partitions are assigned to the intersection group, and samples falling in the thin-walled transition area are assigned to the thin-walled group. The convergence group subtree uses velocity differences, core zone coordinates, and partition thickness ratios from historical samples as features, with the label set limited to two levels: enhanced and maintenance. Examples include an enhanced level with a 15% to 30% power increase and a maintenance level that maintains rated power. The thin-walled group subtree is trained using similar features, with the label set limited to two levels: suppressed and delayed. Examples include a suppressed level with a 10% to 25% power decrease and a delayed level with a 30 to 60 second delay. In the prediction phase, the circuit location is first looked up in the circuit number table, and then the features are fed into the corresponding subtree. The subtree output is the final level for that circuit, eliminating the need to add location filtering rules beyond the classifier output. The output power allocation scheme for each independent heating circuit is then formed by summarizing the levels of each circuit.

[0071] The above-disclosed embodiments are merely preferred embodiments of the present invention and should not be construed as limiting the scope of the invention. Those skilled in the art will understand that implementing all or part of the above-described embodiments and making equivalent changes in accordance with the claims of the present invention are still within the scope of the invention.

Claims

1. A method for controlling the zoned heating of molds in the production of electricity metering boxes, characterized in that, The method includes: Obtain a three-dimensional model of the intersection area of ​​the multi-position partition and the adjacent thin-walled area in the mold cavity of the electricity metering box production, calculate the ratio of the corresponding mold steel volume of the two areas of the partition intersection area and the adjacent thin-walled area, and identify the thermal inertia difference caused by the thick and thin transition interface. The thermal inertia difference and the preset initial heating power are input into the support vector regression machine to predict the transient temperature difference band formed between the intersection region and the thin-walled region during the heating stage, and the maximum temperature difference peak and valley are extracted. The risk of localized overheating and premature curing of thin-walled resin is assessed by evaluating the maximum temperature difference peak and valley, and the first heating power is determined based on the assessed risk. The independent heating circuits of each zone of the mold cavity are driven by the first heating power. Based on the surface temperature of the junction area, the surface temperature of the thin-walled area, and the local heat flux density of the junction area, the expansion path and influence range of the transient temperature difference zone along the thick-thin transition interface are identified. Based on the expansion path and influence range of the transient temperature difference band along the thick-thin transition interface, a random forest tree is used to identify the physical state stage of the resin in the junction region and the thin-walled region, and the propagation speed of the curing front in the junction region and the thin-walled region is analyzed. The effect of the mismatch between the advancing speed in the intersection region and the advancing speed in the thin-walled region on the accumulation of residual internal stress at the intersection of multiple surface plate partitions is analyzed. Based on this, the power redistribution requirement is assessed, and the output power distribution scheme of each independent heating circuit is generated.

2. The method for controlling the zoned heating of a mold for the production of electricity metering boxes according to claim 1, characterized in that, The process of obtaining a three-dimensional model of the intersection area of ​​multiple metering partitions and the adjacent thin-walled area in the mold cavity for producing the power metering box, calculating the ratio of the corresponding mold steel volumes of the two areas, and identifying the thermal inertia differences caused by the thick-thin transition interface includes: Obtain the design drawing of the mold cavity, and separate the multi-level partition plate intersection area and the adjacent thin-walled area from the design drawing. Import them into the triangular mesh modeling tool to obtain the 3D solid model of the multi-level partition plate intersection area and the 3D solid model of the adjacent thin-walled area, and mark the contact surface along the thickness transition interface. Obtain the solid volume of the mold steel corresponding to the two areas of the partition plate intersection area and the adjacent thin-walled area through mesh volume integration. Divide the solid volumes of the mold steel corresponding to the two areas of the partition plate intersection area and the adjacent thin-walled area to obtain the ratio of the steel volume. Project the ratio of the steel volume point by point along the contact surface to the discrete nodes of the thickness transition interface to obtain the thermal inertia difference distribution of the thickness transition interface.

3. The method for controlling the zoned heating of a mold for the production of electricity metering boxes according to claim 2, characterized in that, The method of inputting thermal inertia differences and preset initial heating power into a support vector regression machine to predict the transient temperature difference band formed between the intersection region and the thin-walled region during the heating stage, and extracting the maximum temperature difference peak and valley, includes: The discrete node value sequence of thermal inertia difference distribution is concatenated with the preset initial heating power to form an input vector, which is then fed into a support vector regression machine pre-trained with historical heating test data to obtain the temperature difference sequence between the intersection region and the thin-walled region on each discrete node during the heating stage. The temperature difference sequence is concatenated along the thick-thin transition interface to form the transient temperature difference band. Local extreme points are scanned along the heating time axis to obtain the maximum temperature difference peak and valley. The heating time coordinates corresponding to the temperature difference sequence are read, and the heating time intervals where the temperature difference peak and valley are located are extracted to obtain the occurrence time period and pair it with the maximum temperature difference peak and valley for output.

4. The method for controlling the zoned heating of a mold for the production of electricity metering boxes according to claim 3, characterized in that, The determination of the first heating power based on the assessed risk includes: The overall risk level is determined based on the risk level of thin-walled pre-curing and the risk level of local overheating in the intersection area. When the overall risk level exceeds the preset warning level, the overheating risk is deemed prominent. The preset power reduction level table is then called, and the preset initial heating power is reduced according to the level in the preset power reduction level table that matches the overall risk level to obtain the first heating power. When the overall risk level does not exceed the preset warning level, the overheating risk is deemed under control, and the preset initial heating power is used as the first heating power.

5. A method for controlling the zoned heating of a mold for the production of electricity metering boxes according to claim 1, characterized in that, The method of driving independent heating circuits in each zone of the mold cavity with a first heating power, and identifying the expansion path and influence range of the transient temperature difference band along the thick-thin transition interface based on the surface temperature of the intersection area, the surface temperature of the thin-walled area, and the local heat flux density of the intersection area, includes: The first heating power drives the independent heating circuits of each zone of the mold cavity to operate synchronously. Contact thermocouples arranged on the surfaces of the intersection area and the thin-walled area collect surface temperature at a fixed sampling period. Thin-film heat flux meters embedded in the steel thickness direction of the intersection area synchronously collect the local heat flux density of the intersection area, obtaining a temperature and heat flux density sampling sequence synchronously aligned along the thick-thin transition interface. The temperature difference gradient is obtained by dividing the temperature difference between the two surfaces of the intersection area and the thin-walled area by the geometric distance between the sampling points along the thick-thin transition interface. The local heat flux density is defined as positive pointing to the thin-walled area and negative pointing to the intersection area. The sign of the local heat flux density is used to determine the direction of heat conduction, obtaining the propagation direction of the transient temperature difference band along the thick-thin transition interface. Discrete nodes with temperature difference gradient values ​​higher than a preset sweep threshold are connected in series to form an extension path. Discrete nodes on both sides of the extension path whose temperature difference gradient falls back to within the preset sweep threshold are defined as boundaries, obtaining the influence range.

6. A method for controlling the zoned heating of a mold for the production of electricity metering boxes according to claim 5, characterized in that, The method of identifying the physical state stages of the resin in the intersection region and the thin-walled region using random forest trees includes: The surface temperature time series and local heat flux density time series of each discrete node in the intersection region and the thin-walled region are collected along the extended path. For each discrete node, the temperature rise slope, the second-order difference temperature sign change point, and the duration of the heat flux density continuous stable segment are extracted as state stage feature vectors. The state stage feature vectors are organized into a state stage feature vector set along the boundary of the influence range and fed into a random forest tree that has been trained in advance with thermosetting resin curing test data. The output category labels of the random forest tree correspond to four types of resin physical states in sequence: viscous flow stage, gel stage, exothermic stage, and curing stage. A corresponding category label is obtained for each discrete node.

7. A method for controlling the zoned heating of a mold for the production of electricity metering boxes according to claim 6, characterized in that, The analysis of the solidification front advancement velocity in the intersection region and the thin-walled region includes: Along the extended path, the boundary line between the discrete nodes of adjacent viscous flow stage and gel stage is determined as the gel interface position. The moment when the temperature rise slope changes from steep to flat in the temperature time sequence corresponding to the discrete node of the exothermic stage is taken as the starting point of the exothermic platform, and the moment when the temperature rise slope changes from flat to steep is taken as the ending point of the exothermic platform. The discrete nodes of adjacent gel stage and curing stage are connected in series to form the curing front profile. For the curing front profile on each side of the intersection region and the thin-walled region, the displacement increment of the curing front profile along the normal of the thick-thin transition interface at two adjacent sampling moments is divided by the corresponding sampling interval to obtain the curing front advancement speed of the intersection region and the thin-walled region.

8. A method for controlling the zoned heating of a mold for the production of electricity metering boxes according to claim 7, characterized in that, The analysis of the mismatch between the propulsion velocity in the intersection region and the propulsion velocity in the thin-walled region and its impact on the accumulation of residual internal stress at the intersection of multiple epitope septa includes: The difference between the solidification front advance speed of the intersection region and the solidification front advance speed of the thin-walled region is calculated along the thick-thin transition interface to form a solidification front advance speed difference sequence. The distribution of the difference sequence on the thick-thin transition interface is mapped point by point to the geometric position of the intersection of the multi-epitope partition to obtain the solidification front advance mismatch zone.

9. A method for controlling the zoned heating of a mold for the production of electricity metering boxes according to claim 8, characterized in that, After obtaining the solidified leading edge propulsion mismatch band, the process includes: The mismatch zone of the curing front is mapped onto the residual internal stress conversion table pre-calibrated with the strain-stress constitutive relationship of thermosetting resin curing. The corresponding residual internal stress value is found point by point. The residual internal stress is formed by connecting the points along the intersection of the multi-position partitions. The continuous segment in the residual internal stress accumulation distribution that exceeds the preset accumulation threshold is determined as the stress accumulation core zone.

10. A method for controlling the zoned heating of a mold for the production of electricity metering boxes according to claim 9, characterized in that, The output power allocation scheme for generating each independent heating circuit includes: Based on the position coordinates of the stress accumulation core zone and the difference sequence of the solidification front advancement speed, the power adjustment level corresponding to each independent heating circuit is determined. For the heating circuit in the intersection area, the power output is enhanced or maintained. For the heating circuit in the thin-walled area, the power output is suppressed or delayed. The output power distribution scheme is formed by summarizing the power output levels.

Images