A temperature control method and system for foaming processes

By constructing a dynamic coupling model of temperature-viscosity-foam density and a decision tree classification model, the problems of temperature response lag and lack of physical mechanism for control decision in the saddle polyurethane foaming process were solved, achieving precise temperature control and improved product quality stability.

CN123114150APending Publication Date: 2026-11-17TIANJIN UNIVERSITY OF TECHNOLOGY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611200215.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-10
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In the existing technology, the temperature response lag caused by the thermal inertia of the mold in the saddle polyurethane foaming process, the inability to quantify the impact path of temperature fluctuations on product quality, and the lack of physical mechanism basis for temperature control decisions lead to problems such as low temperature control accuracy, uneven product density, and poor batch stability.

Method used

By constructing a dynamic coupling model of temperature, viscosity, and foam density, and utilizing the correlation ratio between stirring rate and material viscosity, combined with a decision tree classification model, feedforward control commands are generated to directly pre-adjust the mold heating power and cooling water channel valves, thereby achieving precise temperature control.

Benefits of technology

It achieves rapid temperature response adaptation, reduces temperature overshoot, improves product quality consistency and yield, reduces energy consumption, and extends equipment life.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123114150A_ABST
    Figure CN123114150A_ABST
Patent Text Reader

Abstract

This invention discloses a temperature control method and system for foaming processes. The method includes: S1, collecting multi-source heterogeneous data and preprocessing the data; S2, constructing a dynamic coupling model of temperature-viscosity-foam density and calculating the indirect influence coefficient of temperature on foam density; S3, generating feedforward control commands in advance based on the changing trend of the indirect influence coefficient of temperature on foam density; S4, calculating the extreme value difference and variance based on periodically collected mold temperature data, constructing a decision tree classification model, and directly generating temperature control strategy commands for each layer; S5, superimposing the feedforward control commands as basic adjustment quantities and the temperature control strategy commands as dynamic adjustment quantities, and outputting control commands to adjust the power of heating rods and the opening degree of cooling valves in each layer. This invention, based on decision tree classification, has clear physical semantics and can directly drive the actuator for temperature control, achieving precise temperature control in the saddle foaming process and effectively improving the stability of foaming.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of foaming process technology, and more specifically to a temperature control method and system for foaming processes. Background Technology

[0002] The saddle is a crucial component of bicycles and electric vehicles, and its comfort, durability, and product consistency directly impact the riding experience and overall product quality. The core manufacturing process for the saddle is polyurethane foaming. In this process, polyurethane raw materials are mixed in a mixer and then injected into a saddle foaming mold. Under specific temperature conditions, a chemical reaction occurs to generate a foam structure, forming the saddle's elastic cushioning layer.

[0003] In the saddle foaming process, temperature is a key process parameter affecting foaming quality. Temperature changes directly affect the viscosity characteristics of polyurethane materials, and viscosity changes further affect the uniformity of foam density distribution. However, in the traditional foaming process, the stirring rate and temperature control rely on manual experience for adjustment, which has the following prominent problems: (1) Temperature response lag: The saddle foaming mold has a large heat capacity and obvious thermal inertia. The existing temperature control strategy parameters are fixed and cannot adapt to the thermal dynamic changes of the mold, resulting in a serious lag in the heating / cooling process, large overshoot, long stabilization time, and large temperature steady-state deviation. (2) Uneven foaming density: The existing control only focuses on the single variable of temperature and ignores the coupling effect of temperature on the viscosity of the material, as well as the indirect effect of viscosity changes on the generation and expansion of bubbles, resulting in uneven foam density distribution and poor consistency of finished product size. (3) Poor batch stability: The control logic with fixed parameters cannot adapt to the changes of different formulations and target temperatures, and lacks online adaptive capability, resulting in significant density fluctuations between batches and a low first-pass yield of products.

[0004] In the prior art, the following solutions are commonly used for temperature control of polyurethane foam: (1) Threshold temperature control: By setting upper and lower temperature thresholds, heating or cooling operations are triggered when the detected temperature exceeds the threshold range. This solution has low control accuracy and slow response, and cannot achieve precise temperature regulation; (2) Fixed parameter PID control: A fixed parameter PID controller is used for closed-loop temperature regulation, but the fixed parameters cannot adapt to the changes in working conditions caused by the thermal inertia of the mold, resulting in large steady-state temperature error and serious overshoot; (3) Fuzzy adaptive PID control: Based on the fuzzy rule table, the temperature deviation and the rate of change of deviation are mapped to the PID parameter adjustment amount, but the fuzzy rules rely on manual experience to set, and the rule base construction is difficult. Difficult, and continuous function mapping lacks precise physical semantic correspondence; (4) Neural network PID control: The PID parameters are adjusted after extracting temperature time-series features by using a deep learning model (such as a time encoder relying on one-dimensional convolution), but the neural network is a black box model, and the control decision is unexplainable and untraceable. In industrial real-time control scenarios, the inference delay is high and the reliability is insufficient; (5) Multivariate temperature coupling analysis: The foaming optimization control is carried out by considering the spatial correlation between multiple temperature measurement points, but this scheme focuses on the spatial coupling relationship between multiple temperature points, rather than the causal transmission chain between temperature, viscosity and density, and cannot reveal the physical mechanism transmission path of temperature fluctuation on the final product quality.

[0005] Therefore, there is an urgent need for a temperature control method for needle foaming to solve problems such as temperature response lag caused by mold thermal inertia in the saddle polyurethane foaming process, unquantifiable path of temperature fluctuations affecting product quality, and lack of physical mechanism basis for temperature control decisions. Summary of the Invention

[0006] This invention addresses at least one technical problem in the prior art by disclosing a temperature control method and system for foaming processes. Based on decision tree classification, with clear physical semantics and the ability to directly drive the actuator for temperature control, this invention achieves precise temperature control during the saddle foaming process, effectively improving foaming stability. It solves problems such as temperature response lag caused by mold thermal inertia in polyurethane foaming processes, unquantifiable impact paths of temperature fluctuations on product quality, and lack of physical mechanism basis for temperature control decisions.

[0007] This invention is achieved through the following technical solution:

[0008] This invention first provides a temperature control method for a foaming process, comprising the following steps:

[0009] S1. Collect multi-source heterogeneous data and preprocess the data;

[0010] S2. Based on the correlation ratio between stirring speed and material viscosity, a dynamic coupling model of temperature-viscosity-foam density is constructed to calculate the indirect influence coefficient of temperature on foam density.

[0011] S3. Based on the changing trend of the indirect influence coefficient of temperature on foam density, feedforward control instructions are generated in advance to directly pre-adjust the heating power and cooling water channel valve opening of each layer of the mold.

[0012] S4. Calculate the extreme value difference and variance based on the periodically collected mold temperature data. Construct a decision tree classification model based on the positive fusion of fluctuation contribution and neighborhood fluctuation coefficient. The decision tree classification model classifies the temperature distribution characteristics by describing the diffusion degree of decision tree nodes and directly generates temperature control strategy instructions for each layer.

[0013] S5. The feedforward adjustment command is used as the basic adjustment quantity, and the temperature control strategy command is used as the dynamic adjustment quantity. The output control command adjusts the power of each layer of heating rods and the opening of the cooling valve to complete the precise temperature control operation.

[0014] This invention introduces the stirring rate-viscosity correlation ratio as a bridging variable and combines it with the indirect influence coefficient of temperature on foam density, thus establishing for the first time a calculable causal transmission model among temperature, viscosity, and density. Compared to existing technologies that rely solely on empirical rules or single parameter thresholds, this invention transforms the impact of temperature fluctuations on product quality into a quantifiable and traceable numerical indicator. This overcomes the technical shortcomings of traditional methods, such as fuzzy causal relationships and the inability to quantitatively attribute causes, providing a quantitative control basis with physical mechanism support for intelligent temperature control systems.

[0015] As a further solution, the method for collecting multi-source heterogeneous data includes: simultaneously collecting four types of physical quantity data—temperature, stirring speed, material viscosity, and foam density—at key parts of the polyurethane foam mixer; arranging multiple thermocouple temperature measurement points in the upper, middle, and lower layers of the saddle foam mold to cover the irregular curved surface area of ​​the saddle; and performing a full-field temperature scan of the outer surface of the mold using an infrared thermal imager, and cross-validating the data collected by the infrared thermal imager with the data from the thermocouple temperature measurement points.

[0016] As a further embodiment, S2 includes:

[0017] S21. Set the ratio α0 between the preset standard stirring rate and the material viscosity, α0 = Rs / Vs, where Rs is the standard stirring rate and Vs is the standard material viscosity.

[0018] S22. Calculate the actual correlation ratio α_t at the current moment in real time, α_t = R_t / V_t, where R_t is the current stirring rate and V_t is the current material viscosity;

[0019] S23. Calculate the indirect influence coefficient of temperature on foam density λ_ind, λ_ind = α_t / α0.

[0020] As a further embodiment, S4 includes:

[0021] S41. Perform feature analysis on the collected mold temperature data. In each control cycle, collect temperature data from all temperature measurement points of the mold and calculate the extreme value difference and variance as basic quantitative indicators of temperature fluctuation.

[0022] S42. Construct a decision tree classification model;

[0023] S43. The decision tree classification model classifies temperature distribution characteristics by describing the diffusion degree of decision tree nodes, and directly generates temperature control strategy instructions for each layer.

[0024] This invention employs a decision tree classification model as the direct decision layer for temperature control. The decision tree branches directly map to physical condition categories and directly output executable commands such as heating power and cooling opening, without requiring intermediate execution units like PID controllers. Compared to the continuous mapping of fuzzy PID and the nonlinear black-box mapping of neural network PID in existing technologies, the decision path of this invention is discretized, hierarchical, and fully interpretable. Each control action can be uniquely traced back to its corresponding physical condition semantics.

[0025] As a further embodiment, S42 includes:

[0026] S421. The fluctuation contribution degree is used as the first classification feature of the decision tree. The fluctuation contribution degree reflects the contribution weight of the temperature fluctuation of a single temperature measurement point to the overall temperature fluctuation.

[0027] S342. The neighborhood fluctuation coefficient is used as the second classification feature of the decision tree. The neighborhood fluctuation coefficient reflects the degree of consistency of temperature fluctuation between the temperature measurement point and its spatially adjacent temperature measurement points.

[0028] S423. The fluctuation contribution degree and the neighborhood fluctuation coefficient are positively integrated to generate a comprehensive classification index;

[0029] S424. The decision tree classification model splits layer by layer based on a comprehensive classification index. Each split node corresponds to a temperature condition category, and the degree of diffusion of the node reflects the degree of dispersion of temperature distribution under that category.

[0030] As a further embodiment, S424 includes:

[0031] The first level of splitting uses the mean of the comprehensive classification index as the splitting threshold to divide the data into two categories: "low comprehensive anomaly" and "high comprehensive anomaly".

[0032] The second layer of splitting uses the extreme value difference as the splitting threshold for the "high comprehensive anomaly" category, further dividing it into two categories: "locally high" and "violent fluctuation".

[0033] The third layer of splitting uses variance as the splitting threshold for the "low-comprehensive anomaly" class, further dividing it into two categories: "uniform and stable" and "slightly fluctuating".

[0034] The diffusion degree of each split node is defined as the standard deviation of all CI values ​​under that node. The greater the diffusion degree, the higher the dispersion of the temperature distribution under that category.

[0035] As a further solution, the mapping rules for the temperature control strategy commands include:

[0036] When the decision tree classification model classifies the temperature as "uniform and stable", the diffusion value is relatively small, generating a maintenance-type temperature control strategy.

[0037] When the decision tree classification model classifies the result as "slight temperature fluctuation", the diffusion degree value fluctuates slightly, generating a mild correction-type temperature control strategy.

[0038] When the decision tree classification model classifies the temperature as "locally high", the diffusion level is moderate, and a corrective temperature control strategy is generated.

[0039] When the decision tree classification model classifies the result as "severe temperature fluctuations", the diffusion level is relatively large, generating a strong suppression-type temperature control strategy.

[0040] As a further embodiment, the control commands in S5 include heating rod power adjustment commands and cooling water channel valve opening adjustment commands. The heating rod power adjustment command adjusts the power supply of the heating rods corresponding to each layer of the mold from 0-100% based on the heating power percentage of each layer, thereby increasing the mold temperature. The cooling water channel valve opening adjustment command adjusts the opening of the cooling water channel valves corresponding to each layer of the mold from 0-100% based on the opening percentage of the cooling valves of each layer, thereby decreasing the mold temperature.

[0041] This invention utilizes a feedforward pre-regulation mechanism based on the indirect influence coefficient of temperature on foam density. It generates control commands in advance based on changes in operating conditions before temperature deviations accumulate, directly adjusting heating power and cooling capacity. Compared to the passive regulation mode in existing technologies that relies solely on feedback control and only acts after a deviation occurs, this invention's feedforward mechanism effectively reduces temperature overshoot, shortens adjustment time, and significantly improves the system's dynamic response performance.

[0042] The present invention also provides a temperature control system for a foaming process, comprising:

[0043] The data acquisition unit is used to simultaneously collect four types of physical quantity data, namely temperature, stirring speed, material viscosity and foam density, at key parts of the polyurethane foam mixer, and to collect layer temperature data of the mold by arranging multiple temperature measuring points in the upper, middle and lower layers of the saddle foam mold.

[0044] The data preprocessing unit denoises and removes outliers from the raw data, and then transmits the processed data to the dynamic coupling unit and the decision tree temperature control unit.

[0045] The dynamic coupling unit constructs a dynamic coupling model of temperature-viscosity-foam density based on the correlation ratio between stirring speed and material viscosity, calculates the indirect influence coefficient of temperature on foam density, and quantifies the indirect influence of temperature on foam density.

[0046] The feedforward pre-adjustment unit generates feedforward control commands in advance based on the changing trend of the indirect influence coefficient of temperature on foam density, and directly pre-adjusts the heating power and cooling water channel valve opening of each layer of the mold.

[0047] The decision tree temperature control unit calculates the extreme value difference and variance based on the periodically collected mold temperature data, constructs a classification index based on the positive fusion of fluctuation contribution and neighborhood fluctuation coefficient, classifies the temperature distribution characteristics by the diffusion degree of decision tree nodes, and directly generates temperature control strategy instructions for each layer.

[0048] The control unit receives feedforward adjustment commands and temperature control strategy commands. The feedforward commands serve as the basic adjustment quantities, and the decision tree strategy commands serve as the dynamic adjustment quantities. The sum of the two serves as the final output control command.

[0049] The execution unit, including the heating power of each layer and the valves of the cooling water channel, is used to receive control commands from the control unit and control and adjust the power of each layer heating rod and the opening of the cooling valves.

[0050] The human-machine interface unit is used to display temperature data, temperature distribution category, temperature strategy type, heating power of each layer of the mold and opening of cooling water channel valves in real time, and provides parameter setting and alarm functions.

[0051] As a further embodiment, the data acquisition unit includes a sensor array and an infrared thermal imager.

[0052] The sensor array is located on the saddle foam mold, with 12 thermocouple temperature measuring points arranged in a three-layer (upper, middle, and lower) configuration, each with four points. The distribution is as follows: the upper layer temperature measuring points are located on the front arc surface area of ​​the saddle at the top of the mold, with a spacing of approximately 30 mm, to monitor the foaming temperature at the front of the saddle; the middle layer temperature measuring points are located on the flat surface area of ​​the saddle seat pad in the middle of the mold, with a spacing of approximately 35 mm, to monitor the foaming temperature in the middle of the saddle; and the lower layer temperature measuring points are located on the rear arc surface area of ​​the saddle at the bottom of the mold, with a spacing of approximately 30 mm, to monitor the foaming temperature at the rear of the saddle.

[0053] An infrared thermal imager is installed on the side of the mold to perform a full-field temperature scan on the outer surface of the mold, with a temperature measurement range of 0-150°C, and is cross-validated with 12 thermocouple data.

[0054] To address the issue of temperature monitoring blind spots in irregular curved surface structures of saddles, this invention employs a strategy of simultaneous acquisition of four types of sensors (temperature, rotational speed, viscosity, and density) at the mixer end and a 12-channel layered data collection strategy at the mold end. Compared to existing technologies where a single sensor or sparse data collection method is insufficient to cover the temperature gradient of irregular curved surfaces, this invention achieves measurable and traceable temperature distribution across the entire curved surface, ensuring the integrity and consistency of temperature gradient data, while suppressing random errors from a single sensing channel.

[0055] The features and beneficial effects of this invention are as follows:

[0056] (1) This invention introduces the stirring rate-viscosity correlation ratio as a bridging variable and combines it with the indirect influence coefficient of temperature on foam density, thus establishing for the first time a calculable causal transmission model among temperature, viscosity, and density. Compared with the existing technology that relies solely on empirical rules or single parameter thresholds for judgment, this invention transforms the impact of temperature fluctuations on product quality into a quantifiable and traceable numerical indicator, overcoming the technical defects of fuzzy causality and inability to quantitatively attribute causes in traditional methods, and providing a quantitative control basis with physical mechanism support for intelligent temperature control systems.

[0057] (2) This invention periodically collects mold temperature data, comprehensively quantifies temperature dispersion and peak deviation through extreme value difference and variance, and distinguishes between single-point fluctuations and interlayer coupling disturbances by integrating fluctuation contribution and neighborhood fluctuation coefficient. Based on the fusion features, a decision tree classification model is constructed, and the degree of tree node diffusion is used to characterize the layered temperature distribution characteristics. Data-driven automatic classification of temperature conditions is achieved without the need to preset fixed temperature control thresholds. After model classification, independent temperature control strategy instructions for each layer can be directly generated, shortening the control response link, realizing layered decoupled regulation, and effectively suppressing temperature oscillations and overshoot. The model rules are visualized, which facilitates debugging and traceability, can accurately identify periodic temperature disturbances, smooth the actions of heating and cooling actuators, stabilize the mold temperature field, and continuously improve the product molding quality and equipment service life.

[0058] (3) This invention uses a fixed feedforward adjustment command as the basic output of layered temperature control. It relies on the feedforward to compensate for constant thermal disturbances such as environmental and material conditions in advance, eliminates static temperature drift, and stabilizes the basic cold and heat load of each layer. Then, the dynamic adjustment amount of the temperature control strategy is superimposed to correct the instantaneous temperature deviation in real time. After the two are superimposed, the power of the heating rod and the opening of the cooling valve are controlled synchronously. On the one hand, the closed-loop adjustment amplitude can be greatly reduced, the temperature overshoot and oscillation can be suppressed, and the high-precision constant temperature of the multi-layer cavity can be achieved without temperature coupling interference between layers. On the other hand, the action of the actuator is smoothed, the life of heating and cooling components is extended, and energy consumption is reduced. At the same time, parameter decoupling facilitates debugging, adapts to the entire process stage of heating, heat preservation and cooling, resists various cold and heat disturbances, stabilizes the product processing temperature environment, and improves the consistency of finished product batches. Attached Figure Description

[0059] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0060] Figure 1 This is a flowchart of a temperature control method for a foaming process according to an embodiment of the present invention;

[0061] Figure 2 This is a schematic diagram of the temperature control system for the foaming process described in an embodiment of the present invention.

[0062] Figure 3 This is a schematic diagram of the multi-source heterogeneous data acquisition and sensor layout scheme described in an embodiment of the present invention;

[0063] Figure 4 This is a schematic diagram of the dynamic coupling analysis model of temperature-viscosity-foam density described in an embodiment of the present invention;

[0064] Figure 5 This is a structural diagram of the saddle foaming mold temperature control system according to an embodiment of the present invention;

[0065] Figure 6 This is a schematic diagram illustrating the generation of temperature control strategy instructions for each layer using the decision tree classification model described in this embodiment of the invention. Detailed Implementation

[0066] To facilitate understanding of the present invention, a more comprehensive description of the present invention will be given below, and embodiments of the present invention will be provided, but this does not limit the scope of the present invention.

[0067] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0068] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0069] like Figures 1 to 6 As shown, a temperature control method for a foaming process includes the following steps:

[0070] S1. Collect multi-source heterogeneous data and preprocess the data, specifically including:

[0071] S11. The method for collecting multi-source heterogeneous data includes: simultaneously collecting four types of physical quantity data, namely temperature, stirring speed, material viscosity and foam density, at key parts of the polyurethane foam mixer; and simultaneously arranging multiple temperature measuring points at the upper, middle and lower layers of the saddle foam mold to collect the layer temperature data of the mold.

[0072] In some embodiments, a K-type thermocouple temperature sensor, a Hall effect speed sensor, a vibration-type online viscosity sensor, and a differential pressure density sensor are deployed on the mixer to simultaneously collect four types of physical quantity data: temperature, stirring speed, material viscosity, and foam density.

[0073] Multiple thermocouple temperature measurement points are arranged in the upper, middle and lower layers of the saddle foam mold to cover the irregular curved surface area of ​​the saddle. The temperature of the outer surface of the mold is scanned in the whole field by an infrared thermal imager. The data collected by the infrared thermal imager is cross-validated with the data of the thermocouple temperature measurement points to improve the accuracy of data collection.

[0074] S12. Methods for preprocessing the collected data include:

[0075] S121 uses a 16-bit high-precision ADC with a sampling frequency set to 100Hz.

[0076] S122. Perform low-pass filtering on the raw data, setting the cutoff frequency to 10Hz to remove high-frequency noise;

[0077] In some embodiments, the cutoff frequency is set to 10Hz, and a fourth-order filter is used to remove high-frequency noise above 10Hz.

[0078] S123. Perform mean smoothing using a sliding window;

[0079] In some embodiments, the window width N=50, that is, 50 sampling points within a 500ms time window, and the mean value within the window is used to replace the current value.

[0080] S124. Use Z-score to remove outliers. Specifically, calculate the Z-score z = (x-μ) / σ for each sampling point, where μ is the mean of the sliding window and σ is the standard deviation of the sliding window. Data points with |z|>3 are identified as outliers and removed, and replaced with the most recent valid value to ensure data quality.

[0081] S2. Based on the correlation ratio between stirring speed and material viscosity, a dynamic coupling model of temperature-viscosity-foam density is constructed to calculate the indirect influence coefficient of temperature on foam density, quantifying the indirect influence of temperature on foam density, specifically including:

[0082] S21. In the polyurethane foaming process, there is an inherent correlation between the stirring rate and the material viscosity; that is, the higher the stirring rate, the greater the shear force on the material, and the lower its apparent viscosity. Under standard operating conditions, this correlation can be expressed as the preset standard correlation ratio α0 = Rs / Vs, where Rs is the standard stirring rate and Vs is the standard material viscosity.

[0083] S22. When the temperature changes, the actual viscosity of the material changes accordingly, and the stirring rate may also be adjusted accordingly. At this time, the actual correlation ratio at the current moment is α_t = R_t / V_t, where R_t is the current stirring rate and V_t is the current material viscosity.

[0084] S23. The indirect influence coefficient of temperature on foam density, λ_ind, reflects the degree to which the current working condition deviates from the standard working condition. Here, λ_ind = α_t / α0.

[0085] λ_ind > 1 indicates that the stirring-viscosity correlation deviates from the standard state under the current operating conditions, and temperature fluctuations have indirectly affected the foam density through viscosity transmission;

[0086] λ_ind = 1 indicates normal operating conditions;

[0087] λ_ind < 1 indicates a deviation in the opposite direction.

[0088] S3. The present invention also includes a feedforward pre-adjustment mechanism: when λ_ind deviates from the preset threshold range, before the temperature deviation has accumulated, a feedforward control command is generated in advance based on the change trend of λ_ind, and the heating power and cooling water channel valve opening of each layer of the mold are pre-adjusted directly to reduce temperature overshoot and shorten the adjustment time.

[0089] S4. Based on the periodically collected mold temperature data, calculate the temperature data characteristic indicators, construct a decision tree classification model to classify and identify temperature conditions, and directly generate mold layered temperature control strategy instructions based on the decision tree classification results. Specifically, this includes:

[0090] S41. Perform feature analysis on the collected mold temperature data. In each control cycle, collect temperature data from all temperature measurement points on the mold, and calculate the extreme value difference and variance as basic quantitative indicators of temperature fluctuation.

[0091] The extreme value difference ΔT is:

[0092] ΔT =T max - T min , among which, T max T represents the highest temperature among all road temperature measurement points. min The lowest temperature among all road temperature measurement points;

[0093] The variance σ² is:

[0094] σ² = Σ(T i - T mean )² / N, where N is the number of temperature measurement points on the mold, and T i Let T be the temperature of the i-th temperature measuring point. mean This is the average value of N temperature measurement points.

[0095] S42. Construct a decision tree classification model, the specific method is as follows:

[0096] S421. The fluctuation contribution degree is used as the first classification feature of the decision tree. The fluctuation contribution degree reflects the contribution weight of the temperature fluctuation of a single temperature measurement point to the overall temperature fluctuation.

[0097] Specifically, the contribution of fluctuations (WC) i (Temperature measurement point i): WC i= σ i ² / Σσ j ², where σ i ² represents the temperature variance of the i-th temperature measurement point over the most recent N control cycles, and Σσ represents the variance of the temperature within the most recent N control cycles. j ² represents the sum of the temperature variances at all road temperature measurement points. WC i The weight of the contribution of the i-th temperature measurement point to the overall temperature fluctuation.

[0098] S422. The neighborhood fluctuation coefficient is used as the second classification feature of the decision tree. The neighborhood fluctuation coefficient reflects the degree of consistency of temperature fluctuation between the temperature measurement point and its spatially adjacent temperature measurement points.

[0099] Specifically, the neighborhood fluctuation coefficient NC i That is, the i-th temperature measurement point, denoted as: NC i = σ i ² / σ neighbors-i ², where σ neighbors-i ² represents the mean temperature variance of the spatially adjacent temperature measurement points of the i-th temperature measurement point, i.e., the average temperature variance of the two adjacent points in the upper layer or the four adjacent points in the middle / lower layer. NC i It reflects the degree of consistency between the temperature fluctuations of the i-th temperature measurement point and its neighborhood.

[0100] S423. The fluctuation contribution degree and the neighborhood fluctuation coefficient are positively integrated to generate a comprehensive classification index;

[0101] Specifically, the comprehensive classification index CI i CI i = WC i × NC i CI i The larger the value, the greater the contribution of the i-th temperature measurement point to the overall fluctuation, and the more inconsistent it is with its neighborhood, meaning the higher the degree of temperature anomaly at that point.

[0102] S424. The decision tree splits layer by layer based on the comprehensive classification index. Each split node corresponds to a temperature condition category. The degree of diffusion of the node reflects the degree of dispersion of the temperature distribution under that category.

[0103] Specifically, the decision tree splits layer by layer based on the CI values ​​of all road temperature measurement points:

[0104] The first layer splits with ΣCI i The mean of the comprehensive classification index was used as the splitting threshold to divide the data into two categories: "low comprehensive anomaly" and "high comprehensive anomaly".

[0105] The second layer of splitting further divides the "high comprehensive anomaly" category into two categories: "locally high" and "drastic fluctuation" using ΔT as the splitting threshold.

[0106] The third level of splitting uses σ² as the splitting threshold for the "low-synthesis anomaly" class.

[0107] It is further divided into two categories: "uniform and stable" and "slightly fluctuating". The diffusion degree of each split node is defined as the standard deviation of all CI values ​​under that node. The greater the diffusion degree, the higher the dispersion of the temperature distribution under that category.

[0108] S43. Based on the diffusion degree numerical distribution corresponding to the temperature distribution category obtained from the decision tree classification model, directly generate the temperature control strategy instructions for each layer (heating power adjustment instructions and cooling water channel valve opening adjustment instructions), as shown in Table 1.

[0109] Table 1

[0110] Temperature distribution categories Diffusion range Heating power adjustment Cooling valve opening Temperature control strategy type Physical semantics Uniform and stable D < 0.1 0% 0% Maintenance Temperature is uniform across all points, maintaining stability as the primary goal. slight fluctuations 0.1 ≤ D < 0.3 10% 10%-20% Mild correction type Temperature fluctuates slightly, adjust accordingly. Locally higher 0.3 ≤ D < 0.6 20%-30% 30%-50% Corrective type Localized high temperatures necessitate rapid stratification correction. violent fluctuations D ≥ 0.6 40%-50% 60%-80% Strongly inhibitory Drastic temperature fluctuations, strong suppression

[0111] The specific mapping rules for generating temperature control strategies are as follows: When the decision tree classification result is "uniform and stable temperature", the diffusion degree value is small, and a maintenance-type temperature control strategy is generated, that is, maintaining the current heating power and closing the cooling water channel to maintain temperature stability; when the decision tree classification model classification result is "slight temperature fluctuation", the diffusion degree value fluctuates slightly, and a mild correction-type temperature control strategy is generated; when the decision tree classification result is "locally high temperature", the diffusion degree value is moderate, and a correction-type temperature control strategy is generated, that is, increasing the opening of the cooling water channel valves and reducing the heating power of the corresponding layer in the deviation area to quickly correct the local temperature deviation; when the decision tree classification result is "drastic temperature fluctuation", the diffusion degree value is large, and a strong suppression-type temperature control strategy is generated, that is, fully opening the cooling water channel valves and significantly reducing the heating power to quickly suppress drastic temperature fluctuations.

[0112] This invention does not use the traditional architecture with PID controllers as the underlying execution unit. The decision tree classification results are no longer mapped to PID parameters, but directly generate executable temperature control strategy instructions, such as heating power percentage and cooling water channel valve opening percentage, realizing direct decision-making from physical coupling analysis to control execution.

[0113] S5. When the current feedforward pre-adjustment mechanism is triggered, the feedforward pre-adjustment command and the temperature control strategy command generated by the decision tree are superimposed: the feedforward command serves as the basic adjustment quantity, the decision tree strategy command serves as the dynamic adjustment quantity, and the sum of the two serves as the final output control command, which drives the heating rods and cooling water channel valves of each layer of the saddle foaming mold to complete the precise temperature control operation and ensure that the temperature overshoot is minimized.

[0114] Specifically, the PLC controller receives feedforward pre-adjustment instructions and temperature control strategy instructions generated by the decision tree, and converts them into the following execution instructions: The heating rod power adjustment instruction adjusts the power supply of the heating rods corresponding to each layer of the mold from 0-100% based on the heating power percentage instructions for each layer, increasing the mold temperature; the cooling water channel valve opening adjustment instruction adjusts the opening of the cooling water channel valves corresponding to each layer of the mold from 0-100% based on the cooling valve opening percentage instructions for each layer, decreasing the mold temperature. The heating rods and cooling water channel valves are deployed at corresponding positions on the upper, middle, and lower layers of the mold to ensure precise temperature control for each layer.

[0115] A temperature control system for a foaming process, comprising:

[0116] The data acquisition unit is used to simultaneously collect four types of physical quantity data, namely temperature, stirring speed, material viscosity and foam density, at key parts of the polyurethane foam mixer, and to collect layer temperature data of the mold by arranging multiple temperature measuring points in the upper, middle and lower layers of the saddle foam mold.

[0117] The data preprocessing unit uses a 16-bit high-precision ADC to acquire raw data, removes high-frequency noise through low-pass filtering, and performs mean smoothing through a sliding window; it establishes an outlier detection mechanism based on Z-score to ensure the accuracy of the raw data, and transmits the processed data to the data processing unit and the decision tree temperature control unit via industrial Ethernet;

[0118] The dynamic coupling unit constructs a dynamic coupling model of temperature-viscosity-foam density based on the correlation ratio between stirring speed and material viscosity, calculates the indirect influence coefficient of temperature on foam density, and quantifies the indirect influence of temperature on foam density.

[0119] The feedforward pre-adjustment unit generates feedforward control commands in advance based on the changing trend of the indirect influence coefficient of temperature on foam density, and directly pre-adjusts the heating power and cooling water channel valve opening of each layer of the mold, thereby reducing temperature overshoot and shortening the adjustment time.

[0120] The decision tree temperature control unit calculates the extreme value difference and variance based on the periodically collected mold temperature data, constructs a classification index based on the positive fusion of fluctuation contribution and neighborhood fluctuation coefficient, classifies the temperature distribution characteristics by the diffusion degree of decision tree nodes, and directly generates temperature control strategy instructions for each layer.

[0121] The control unit receives feedforward adjustment commands and temperature control strategy commands. The feedforward commands serve as the basic adjustment quantities, and the decision tree strategy commands serve as the dynamic adjustment quantities. The sum of the two serves as the final output control command.

[0122] The execution unit, including the heating power of each layer and the valves of the cooling water channel, is used to receive control commands from the control unit and control and adjust the power of each layer heating rod and the opening of the cooling valves.

[0123] The human-machine interface unit is used to display temperature data, temperature distribution category, temperature strategy type, heating power of each layer of the mold and opening of cooling water channel valves in real time, and provides parameter setting and alarm functions.

[0124] The units are interconnected via industrial Ethernet, forming a closed-loop control circuit from data acquisition, data processing, coupled analysis, decision tree classification, temperature control strategy execution and temperature feedback.

[0125] Example 1

[0126] This embodiment uses an electric vehicle saddle polyurethane foam production line as an application scenario to describe in detail the implementation process of the method of the present invention.

[0127] A1. Multi-source heterogeneous data acquisition;

[0128] like Figure 3 The following sensor group is deployed in key parts of the mixer of the automatic foaming machine:

[0129] (1) K-type thermocouple temperature sensor: measuring range 0-200°C, accuracy ±0.5°C, installed in the stirring area of ​​the inner wall of the mixer, used to collect the temperature of the material in the mixer T_mix;

[0130] (2) Hall effect speed sensor: measurement range 0-3000rpm, accuracy ±1rpm, installed on the end of the stirring motor shaft, used to collect stirring speed R_t;

[0131] (3) Vibration-type online viscosity sensor: measurement range 0-10000mPa·s, accuracy ±2%, installed in the mixer outlet pipe, used to collect material viscosity V_t;

[0132] (4) Differential pressure density sensor: measurement range 0.5-2.0 g / cm³, accuracy ±0.5%, installed in the mixer outlet pipe, used to collect foam density ρ_t.

[0133] On the saddle foaming mold, 12 K-type thermocouple temperature measuring points are arranged in a three-layer (upper / middle / lower) configuration with 4 points each: the upper layer temperature measuring points (T1-T4) are arranged on the front arc surface area of ​​the saddle at the top of the mold, with a spacing of about 30mm, to monitor the foaming temperature at the front of the saddle; the middle layer temperature measuring points (T5-T8) are arranged on the flat surface area of ​​the saddle seat pad in the middle of the mold, with a spacing of about 35mm, to monitor the foaming temperature in the middle of the saddle; and the lower layer temperature measuring points (T9-T12) are arranged on the rear arc surface area of ​​the saddle at the bottom of the mold, with a spacing of about 30mm, to monitor the foaming temperature at the rear of the saddle.

[0134] An infrared thermal imager is installed on the side of the mold to perform a full-field temperature scan on the outer surface of the mold, with a temperature measurement range of 0-150°C, and is cross-validated with 12 thermocouple data.

[0135] The data preprocessing methods include: using a 16-bit high-precision ADC with a sampling frequency set to 100Hz. The original data preprocessing flow is as follows: low-pass filtering (cutoff frequency 10Hz, using a 4th-order filter to remove high-frequency noise above 10Hz); sliding window mean smoothing (window width N=50, i.e., 50 sampling points within a 500ms time window, calculating the mean within the window to replace the current value); Z-score outlier removal (calculating the Z-score value z = (x - μ) / σ for each sampling point, where μ is the sliding window mean and σ is the sliding window standard deviation, data points |z|>3 are identified as outliers and removed, replaced by the most recent valid value).

[0136] A2, such as Figure 4 Based on the correlation ratio between stirring speed and material viscosity, a dynamic coupling analysis model of temperature-viscosity-foam density is constructed.

[0137] A21. Determination of the preset standard correlation ratio α0: Under standard operating conditions (target temperature 55°C, standard formula, standard stirring rate), the standard stirring rate Rs and standard material viscosity Vs were collected. Based on experimental measurements, the proposed α0 values ​​for the three formulas are shown in Table 2.

[0138] Table 2

[0139] Formula type Rs (rpm) Vs (mPa·s) <![CDATA[α0]]> High-density formula 1200 450 2.67 medium density formula 1000 380 2.63 Low-density formula 800 310 2.58

[0140] The α0 value is stored in the system database as a benchmark parameter for subsequent real-time comparison.

[0141] A22. Calculation of real-time correlation ratio α_t: In each control cycle, the current stirring rate R_t and the current material viscosity V_t are collected in real time, and α_t = R_t / V_t is calculated.

[0142] A23. Calculation of the indirect influence coefficient λ_ind: λ_ind = α_t / α0.

[0143] λ_ind = -1.0 indicates that the stirring-viscosity correlation is in a standard state under the current operating conditions, and the indirect effect of temperature on foam density is zero;

[0144] λ_ind > 1.0 indicates that the current operating conditions deviate from the standard state, the stirring rate is relatively high compared to the viscosity, and temperature fluctuations have had a positive indirect impact on foam density through viscosity conduction;

[0145] λ_ind < 1.0 indicates that the current operating conditions deviate from the standard conditions, the stirring rate is relatively low compared to the viscosity, and temperature fluctuations have a negative indirect impact on foam density.

[0146] A24. Set the threshold range of λ_ind to [0.90, 1.10]. When λ_ind exceeds this range, trigger feedforward pre-adjustment: when λ_ind > 1.10, it is predicted that the temperature will rise and the viscosity will decrease, so the opening of the valves of each layer of cooling water channel is increased by 5%-15% in advance, linearly mapped according to the deviation of λ_ind, and the heating power is reduced by 2%-10%; when λ_ind < 0.90, it is predicted that the temperature will fall and the viscosity will rise, so the heating power of each layer is increased by 5%-10% in advance, linearly mapped according to the deviation of λ_ind, and the cooling water channel is closed.

[0147] A3, such as Figure 6 A three-step technical approach is adopted to achieve intelligent temperature control decision-making using a decision tree.

[0148] A31. Temperature data feature analysis: In each control cycle, temperature data from 12 temperature measurement points of the mold are collected, and extreme value difference and variance are calculated.

[0149] Extreme value difference: ΔT = T max - T min , among which, T max T represents the highest temperature among the 12 temperature measurement points. min This is the lowest temperature among the 12 temperature measurement points;

[0150] Variance: σ² = Σ(T) i - T mean )² / 12, where Ti is the temperature of the i-th temperature measuring point, and T mean This is the average value of 12 temperature measurement points.

[0151] A32. Construct a decision tree classification model, with the classification features defined as follows:

[0152] A321, Fluctuation Contribution WC i (Temperature measurement point i): WC i = σ i ² / Σσ j ², where σ i ² represents the temperature variance of the i-th temperature measurement point over the most recent N control cycles, and Σσ represents the variance of the temperature within the most recent N control cycles. j ² represents the sum of the temperature variances of all 12 temperature measurement points. WC i The weight of the contribution of the i-th temperature measurement point to the overall temperature fluctuation.

[0153] A322, Neighborhood Fluctuation Coefficient NC i That is, the i-th temperature measurement point: NCi = σ i ² / σ neighbors-i ², where σ neighbors-i ² represents the mean temperature variance of the spatially adjacent temperature measurement points of the i-th temperature measurement point, i.e., the average temperature variance of the two adjacent points in the upper layer or the four adjacent points in the middle / lower layer. NC i It reflects the degree of consistency between the temperature fluctuations of the i-th temperature measurement point and its neighborhood.

[0154] A323, Comprehensive Classification Index (CI) i CI i = WC i × NC i CI i The larger the value, the greater the contribution of the i-th temperature measurement point to the overall fluctuation, and the more inconsistent it is with its neighborhood, meaning the higher the degree of temperature anomaly at that point.

[0155] A324. The decision tree is split layer by layer based on the CI values ​​of all 12 temperature measurement points:

[0156] The first layer splits with ΣCI i The mean value was used as the splitting threshold to divide the data into two categories: "low-comprehension anomalies" and "high-comprehension anomalies".

[0157] The second layer of splitting further divides the "high comprehensive anomaly" category into two categories: "locally high" and "drastic fluctuation" using ΔT as the splitting threshold.

[0158] The third layer of splitting further divides the "low integrated anomaly" category into two subcategories: "uniform and stable" and "slightly fluctuating," using σ² as the splitting threshold. The diffusion degree of each splitting node is defined as the standard deviation of all CI values ​​under that node; a greater diffusion degree indicates a higher degree of dispersion in the temperature distribution within that category.

[0159] Step 3: Based on the decision tree classification results, directly generate temperature control strategy instructions for each layer, as shown in Table 1.

[0160] The temperature control strategy instructions generated by the decision tree are independent heating power percentages and cooling water channel valve opening percentages for each layer, which are directly output to the PLC for execution.

[0161] In addition, when the current feedforward pre-adjustment mechanism is triggered, the feedforward pre-adjustment instruction is superimposed with the temperature control strategy instruction generated by the decision tree: the feedforward instruction serves as the basic adjustment quantity, the decision tree strategy instruction serves as the dynamic adjustment quantity, and the sum of the two serves as the final control instruction output to the PLC, ensuring that temperature overshoot is minimized.

[0162] The A34 PLC controller receives feedforward instructions and decision tree strategy instructions, converting them into the following execution instructions: The heating rod power adjustment instruction adjusts the power supply of the corresponding heating rod in the mold layer from 0-100% based on the heating power percentage instruction for each layer, increasing the mold temperature; the cooling water channel valve opening adjustment instruction adjusts the opening of the corresponding cooling water channel valve in the mold layer from 0-100% based on the cooling valve opening percentage instruction for each layer, decreasing the mold temperature. The heating rods and cooling water channel valves are deployed at corresponding positions on the upper / middle / lower layers of the mold to ensure precise temperature control for each layer.

[0163] Example 2

[0164] The data acquisition unit includes a mixer-side sensor group (K-type thermocouples, Hall effect speed sensor, vibration-type online viscosity sensor, differential pressure density sensor), a mold-side temperature measurement array (12-channel thermocouples layered), and an infrared thermal imager.

[0165] The data preprocessing unit uses a 16-bit high-precision ADC to acquire raw data, continuously acquiring four types of physical quantity data and 12 channels of mold temperature data at a frequency of 100Hz. After preprocessing, the data is transmitted to the data processing unit and the decision tree temperature control unit via industrial Ethernet.

[0166] The dynamic coupling unit, deployed on an industrial computer (IPC), constructs a dynamic coupling model of temperature-viscosity-foam density based on the correlation ratio between stirring speed and material viscosity, calculates the indirect influence coefficient of temperature on foam density, and quantifies the indirect influence of temperature on foam density.

[0167] The feedforward pre-adjustment unit generates feedforward control commands in advance based on the changing trend of the indirect influence coefficient of temperature on foam density, and directly pre-adjusts the heating power and cooling water channel valve opening of each layer of the mold, thereby reducing temperature overshoot and shortening the adjustment time.

[0168] The decision tree temperature control unit calculates the extreme value difference and variance based on the periodically collected mold temperature data, constructs a classification index based on the positive fusion of fluctuation contribution and neighborhood fluctuation coefficient, classifies the temperature distribution characteristics by the diffusion degree of decision tree nodes, and directly generates temperature control strategy instructions for each layer.

[0169] The control unit is a PLC controller that receives feedforward adjustment instructions and temperature control strategy instructions. The feedforward instructions serve as the basic adjustment quantity, and the decision tree strategy instructions serve as the dynamic adjustment quantity. The sum of the two serves as the final output control instruction.

[0170] The execution unit includes mold layer heating rods (2 rods for each of the upper, middle, and lower layers, totaling 6 rods) and cooling water channel valves (21 valves for each of the upper, middle, and lower layers, totaling 6 valves).

[0171] The human-machine interface unit, including the host computer monitoring interface, displays the following information in real time: real-time curves and heat distribution maps of 12 channels of mold temperature data; real-time change curves of the λ_ind indirect influence coefficient; decision tree classification results and the current status of temperature control strategies for each layer; real-time status of heating rod power and cooling water channel valve opening; temperature over-limit alarms and system fault diagnosis information.

[0172] This invention introduces the stirring rate-viscosity correlation ratio as a bridging variable and combines it with the indirect influence coefficient of temperature on foam density, establishing for the first time a calculable causal transmission model among temperature, viscosity, and density. Compared to existing technologies that rely solely on empirical rules or single parameter thresholds, this invention transforms the impact of temperature fluctuations on product quality into a quantifiable and traceable numerical indicator. This overcomes the technical shortcomings of traditional methods, such as fuzzy causal relationships and the inability to quantitatively attribute causes, providing a quantitative control basis with physical mechanism support for intelligent temperature control systems. It is applicable to precise temperature control and foaming in the production process of polyurethane foam for bicycle and electric vehicle saddles.

[0173] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A temperature control method for a foaming process, characterized in that: Includes the following steps: S1. Collect multi-source heterogeneous data and preprocess the data; S2. Based on the correlation ratio between stirring speed and material viscosity, a dynamic coupling model of temperature-viscosity-foam density is constructed to calculate the indirect influence coefficient of temperature on foam density. S3. Based on the changing trend of the indirect influence coefficient of temperature on foam density, feedforward control instructions are generated in advance to directly pre-adjust the heating power and cooling water channel valve opening of each layer of the mold. S4. Calculate the extreme value difference and variance based on the periodically collected mold temperature data, and construct a decision tree classification model. The decision tree classification model directly generates temperature control strategy instructions for each layer. S5. The feedforward adjustment command is used as the basic adjustment quantity, and the temperature control strategy command is used as the dynamic adjustment quantity. The output control command adjusts the power of each layer of heating rods and the opening of the cooling valve to complete the precise temperature control operation.

2. The temperature control method for a foaming process according to claim 1, characterized in that: The method for collecting multi-source heterogeneous data includes: simultaneously collecting four types of physical quantity data—temperature, stirring speed, material viscosity, and foam density—at key parts of the polyurethane foam mixer; arranging multiple thermocouple temperature measurement points in the upper, middle, and lower layers of the saddle foam mold to cover the irregular curved surface area of ​​the saddle; and performing a full-field temperature scan of the outer surface of the mold using an infrared thermal imager, and cross-validating the data collected by the infrared thermal imager with the data from the thermocouple temperature measurement points.

3. The temperature control method for a foaming process according to claim 1, characterized in that: S2 includes: S21. Set the ratio α0 between the preset standard stirring rate and the material viscosity, α0 = Rs / Vs, where Rs is the standard stirring rate and Vs is the standard material viscosity. S22. Calculate the actual correlation ratio α_t at the current moment in real time, α_t = R_t / V_t, where R_t is the current stirring rate and V_t is the current material viscosity; S23. Calculate the indirect influence coefficient of temperature on foam density λ_ind, λ_ind = α_t / α0.

4. The temperature control method for a foaming process according to claim 1, characterized in that: S4 includes: S41. Perform feature analysis on the collected mold temperature data. In each control cycle, collect temperature data from all temperature measurement points of the mold and calculate the extreme value difference and variance as basic quantitative indicators of temperature fluctuation. S42. Construct a decision tree classification model; S43. The decision tree classification model classifies temperature distribution characteristics by describing the diffusion degree of decision tree nodes, and directly generates temperature control strategy instructions for each layer.

5. The temperature control method for a foaming process according to claim 4, characterized in that: S42 includes: S421. The fluctuation contribution degree is used as the first classification feature of the decision tree classification model. The fluctuation contribution degree reflects the contribution weight of the temperature fluctuation of a single temperature measurement point to the overall temperature fluctuation. S342. The neighborhood fluctuation coefficient is used as the second classification feature of the decision tree classification model. The neighborhood fluctuation coefficient reflects the degree of consistency of temperature fluctuation between the temperature measurement point and its spatially adjacent temperature measurement points. S423. The fluctuation contribution degree and the neighborhood fluctuation coefficient are positively integrated to generate a comprehensive classification index; S424. Based on the comprehensive classification index, perform layer-by-layer splitting. Each splitting node in the layer-by-layer splitting process corresponds to a temperature condition category. The diffusion degree of the splitting node reflects the dispersion of the temperature distribution under that category.

6. The temperature control method for a foaming process according to claim 5, characterized in that: S424 includes: The first level of splitting uses the mean of the comprehensive classification index as the splitting threshold to divide the data into two categories: "low comprehensive anomaly" and "high comprehensive anomaly". The second layer of splitting uses the extreme value difference as the splitting threshold for the "high comprehensive anomaly" category, further dividing it into two categories: "locally high" and "drastic fluctuation". The third layer of splitting uses variance as the splitting threshold for the "low-comprehensive anomaly" class, further dividing it into two categories: "uniform and stable" and "slightly fluctuating". The diffusion degree of each split node is defined as the standard deviation of all CI values ​​under that node. The greater the diffusion degree, the higher the dispersion of the temperature distribution under that category.

7. The temperature control method for a foaming process according to claim 4, characterized in that: The mapping rules for the temperature control strategy commands include: When the decision tree classification model classifies the temperature as "uniform and stable", the diffusion value is relatively small, generating a maintenance-type temperature control strategy. When the decision tree classification model classifies the result as "slight temperature fluctuation", the diffusion degree value fluctuates slightly, generating a mild correction-type temperature control strategy. When the decision tree classification model classifies the temperature as "locally high", the diffusion level is moderate, and a corrective temperature control strategy is generated. When the decision tree classification model classifies the result as "severe temperature fluctuations", the diffusion level is relatively large, generating a strong suppression-type temperature control strategy.

8. The temperature control method for a foaming process according to claim 1, characterized in that: The control commands in S5 include heating rod power adjustment commands and cooling water channel valve opening adjustment commands. The heating rod power adjustment command adjusts the power supply of the heating rods corresponding to each layer of the mold from 0-100% according to the heating power percentage of each layer, thereby increasing the mold temperature. The cooling water channel valve opening adjustment command adjusts the opening of the cooling water channel valves corresponding to each layer of the mold from 0-100% according to the opening percentage of the cooling valves of each layer, thereby decreasing the mold temperature.

9. A temperature control system for a foaming process, characterized in that: include: The data acquisition unit is used to simultaneously collect four types of physical quantity data, namely temperature, stirring speed, material viscosity and foam density, at key parts of the polyurethane foam mixer, and to collect layer temperature data of the mold by arranging multiple temperature measuring points in the upper, middle and lower layers of the saddle foam mold. The data preprocessing unit denoises and removes outliers from the raw data, and then transmits the processed data to the dynamic coupling unit and the decision tree temperature control unit. The dynamic coupling unit constructs a dynamic coupling model of temperature-viscosity-foam density based on the correlation ratio between stirring speed and material viscosity, calculates the indirect influence coefficient of temperature on foam density, and quantifies the indirect influence of temperature on foam density. The feedforward pre-adjustment unit generates feedforward control commands in advance based on the changing trend of the indirect influence coefficient of temperature on foam density, and directly pre-adjusts the heating power and cooling water channel valve opening of each layer of the mold. The decision tree temperature control unit calculates the extreme value difference and variance based on the periodically collected mold temperature data, constructs a classification index based on the positive fusion of fluctuation contribution and neighborhood fluctuation coefficient, classifies the temperature distribution characteristics by the diffusion degree of decision tree nodes, and directly generates temperature control strategy instructions for each layer. The control unit receives feedforward adjustment commands and temperature control strategy commands. The feedforward commands serve as the basic adjustment quantities, and the decision tree strategy commands serve as the dynamic adjustment quantities. The sum of the two serves as the final output control command. The execution unit, including the heating power of each layer and the valves of the cooling water channel, is used to receive control commands from the control unit and control and adjust the power of each layer heating rod and the opening of the cooling valves. The human-machine interface unit is used to display temperature data, temperature distribution category, temperature strategy type, heating power of each layer of the mold and opening of cooling water channel valves in real time, and provides parameter setting and alarm functions.

10. A temperature control system for a foaming process according to claim 9, characterized in that: The data acquisition unit includes a sensor array and an infrared thermal imager. The sensor array is located on the saddle foam mold, with 12 thermocouple temperature measuring points arranged in a three-layer (upper, middle, and lower) configuration, each with four points. The distribution is as follows: the upper layer temperature measuring points are located on the front arc surface area of ​​the saddle at the top of the mold, with a spacing of approximately 30 mm, to monitor the foaming temperature at the front of the saddle; the middle layer temperature measuring points are located on the flat surface area of ​​the saddle seat pad in the middle of the mold, with a spacing of approximately 35 mm, to monitor the foaming temperature in the middle of the saddle; and the lower layer temperature measuring points are located on the rear arc surface area of ​​the saddle at the bottom of the mold, with a spacing of approximately 30 mm, to monitor the foaming temperature at the rear of the saddle. An infrared thermal imager is installed on the side of the mold to perform a full-field temperature scan on the outer surface of the mold, with a temperature measurement range of 0-150°C, and is cross-validated with 12 thermocouple data.