A dynamic proportioning control method and system for two-component ms structural adhesive

Abstract

The application provides a dynamic proportioning control method and system for a two-component Ms structural adhesive, relates to the technical field of mixing control, and comprises the following steps: according to the mechanical properties of a mixed product, deploying a proportioning optimizer to perform proportioning decision and dimension mutual verification optimization under low-dimensional projection; according to target proportioning data, performing proportioning control; under the driving of a glue mixer, collecting near-infrared transmission spectrum according to an infrared spectrum probe, and performing mixed uniformity judgment and participation control adjustment decision by using Euclidean distance; and according to dynamic control parameters, performing channel dynamic parameter adjustment control of a main agent pump and a curing agent pump. The application solves the technical problem that, in the prior art, the mixed uniformity of a two-component MS structural adhesive cannot be directly detected online, and the proportioning control relies on indirect feedback, so that the problems of mixed unevenness and proportioning drift cannot be found and corrected in time, and the stability of the two-component MS structural adhesive is affected, and the mixed quality and stability of the MS structural adhesive are improved.

Description

Technical Field

[0001] This application relates to the field of mixing control technology, specifically to a dynamic proportioning control method and system for a two-component Ms structural adhesive. Background Technology

[0002] In the coating process of two-component MS structural adhesives, most existing proportioning control methods rely on indirect feedback based on flow rate and pressure. Mixing is typically accomplished using a volumetric metering pump based on the theoretical displacement ratio in conjunction with a static mixing tube. Due to the lack of direct means of sensing the chemical composition of the mixed colloid, the accuracy of the proportioning can usually only be indirectly assessed through periodic weighing or mechanical property testing of the final product. This makes it difficult to monitor the true state of mixing uniformity in real time during production, leading to the failure to detect mixing inconsistencies and proportioning drift in a timely manner, resulting in batch scrap.

[0003] In summary, the existing technology has technical problems such as the inability to directly detect the mixing uniformity of the two-component MS structural adhesive online, and the reliance on indirect feedback for ratio control, which makes it impossible to detect and correct mixing unevenness and ratio drift in a timely manner, further affecting the stability of the component MS structural adhesive. Summary of the Invention

[0004] The purpose of this application is to provide a dynamic proportioning control method and system for two-component Ms structural adhesives, in order to solve the technical problems in the prior art where the mixing uniformity of two-component MS structural adhesives cannot be directly detected online, and the proportioning control relies on indirect feedback, which leads to the inability to detect and correct mixing unevenness and proportioning drift in a timely manner, further affecting the stability of the component Ms structural adhesives.

[0005] To achieve the above objectives, this application provides a method and system for dynamic proportion control of a two-component Ms structural adhesive.

[0006] Firstly, this application provides a dynamic proportioning control method for a two-component Ms structural adhesive. This method is implemented through a dynamic proportioning control system for the two-component Ms structural adhesive. The dynamic proportioning control method includes: deploying a proportioning optimizer based on a mapping architecture from a high-dimensional target space to a low-dimensional projection space according to the mechanical properties of the mixed product; performing proportioning decisions and dimensional cross-validation optimization under the low-dimensional projection to obtain target proportioning data; wherein the main component and the curing agent component are projected into at least two low-dimensional projection spaces. The mixing ratio is optimized using an iterative feedback guided by low-dimensional projection space cross-validation. Based on the target mixing ratio data, the parameters of the main agent pump channel and the curing agent pump channel of the mixer are configured to control the mixing ratio of the main agent component and the curing agent component. As the mixer is driven, near-infrared transmission spectra are collected by an infrared spectral probe and transmitted back to the automatic control module. By identifying the spectral vectors based on the characteristic absorption peaks of the two components, Euclidean distance is used to judge the mixing uniformity and make control adjustment decisions, generating dynamic control parameters. Based on the dynamic control parameters, the channels of the main agent pump and the curing agent pump are dynamically adjusted and controlled.

[0007] Optionally, the mechanical properties of the hybrid product are obtained, and a high-dimensional target space based on the performance matrix is ​​constructed. Based on the mechanical properties, multiple low-dimensional perspectives are determined, and the high-dimensional target space is projected to construct a low-dimensional projection space. For the short-term adhesion dimension, a normalized first dimensionless index is defined, and for the long-term curing dimension, a normalized second dimensionless index is defined. By mapping the high-dimensional target space and the low-dimensional projection space, a ratio optimizer is constructed.

[0008] Optionally, for the first projection space, a first weight is determined by neighborhood distance, a second weight is determined by the corresponding high-dimensional performance importance, and a first relative center is located by the first weight and the second weight; the first relative center is used as the low-dimensional performance target of the short-term adhesion dimension, and the ratio optimization of the main component and the curing agent component is performed to determine the first ratio data.

[0009] Optionally, for the second projection space, a second proportion data is determined; the second projection space is evaluated using the first proportion data, and an adjustment constraint is applied with a preset step size to determine a first corrected proportion; the first projection space is evaluated using the second proportion data, and an adjustment constraint is applied with a preset step size to determine a second corrected proportion; multiple rounds of cross-validation optimization are performed using the first corrected proportion and the second corrected proportion until the difference between the two data of the Nth corrected proportion updated iteratively is less than a preset value, which is considered convergence, and any one of the Nth corrected proportions is selected as the target proportion data.

[0010] Optionally, a near-infrared spectral probe is installed at the outlet of the mixer to collect near-infrared transmission spectra as the mixer is driven; the characteristic absorption peaks of the functional groups of the main component and the curing agent component in the near-infrared region are determined; an automatic control module is deployed based on the mixing uniformity index of the characteristic absorption peaks; the continuously collected near-infrared transmission spectra are input into the automatic control module to generate dynamic control parameters.

[0011] Optionally, the first frame spectrum and the second frame spectrum are acquired, and the first spectral vector and the second spectral vector are determined, wherein the spectral vector is determined by the spectral band where the characteristic absorption peak is located; the Euclidean distance between the first spectral vector and the second spectral vector is calculated; the mixing uniformity is judged based on the Euclidean distance, and a judgment result is generated, wherein if the Euclidean distance is less than a preset value, it is judged as mixing uniformly.

[0012] Optionally, a spectral fluctuation sequence is generated based on the judgment result of a preset time period; if the fluctuation trend of the spectral fluctuation sequence is periodic, it is judged as a type of anomaly, wherein the periodicity standard is taken as the frequency of the pump stroke, and the type of anomaly is uneven mixing caused by pump pulsation; if the fluctuation trend is a pulse-like spike, it is judged as a type of anomaly, wherein the type of anomaly is a mixer malfunction.

[0013] Optionally, if it is the first type of anomaly, a dynamic control command is generated; based on the dynamic control command, a parameter control control decision is made using Euclidean distance as the adjustment requirement, and the dynamic control parameters are generated, wherein the dynamic control parameters include the speed correction amount of the main agent pump and the curing agent pump; if it is the second type of anomaly, a shutdown and maintenance command is generated.

[0014] Optionally, a mixing uniformity trend curve is plotted with sampling time as the horizontal axis and the quantized value of Euclidean distance as the vertical axis; the pump stroke signal and alarm threshold line are superimposed on the mixing uniformity trend curve to perform time series monitoring and automatic anomaly marking, and the results are displayed on the terminal interface.

[0015] Secondly, this application also provides a dynamic proportioning control system for a two-component Ms structural adhesive, used to execute a dynamic proportioning control method for a two-component Ms structural adhesive as described in the first aspect. The dynamic proportioning control system for the two-component Ms structural adhesive includes: a dimension mapping module, used to deploy a proportioning optimizer based on the mechanical properties of the mixed product using a mapping architecture from a high-dimensional target space to a low-dimensional projection space, performing proportioning decisions and dimension cross-validation optimization under low-dimensional projection to obtain target proportioning data. This is achieved by optimizing the proportioning of the main component and the curing agent component through projection in at least two low-dimensional projection spaces, using a low-dimensional projection space cross-validation... The verification method uses iterative feedback guidance; the parameter configuration module is used to configure the parameters of the main agent pump channel and the curing agent pump channel of the mixer according to the target ratio data, and to execute the ratio control of the main agent component and the curing agent component; the mixing uniformity judgment module is used to collect near-infrared transmission spectrum according to the infrared spectral probe as the mixer is driven, and transmit it back to the automatic control module. By identifying the spectral vector based on the characteristic absorption peak of the two components, the module makes a mixing uniformity judgment and parameter control adjustment decision based on Euclidean distance, and generates dynamic control parameters; the dynamic control module is used to execute dynamic parameter adjustment and control of the main agent pump and the curing agent pump channel according to the dynamic control parameters.

[0016] One or more technical solutions provided in this application have at least the following technical effects or advantages: By deploying a proportioning optimizer based on a mapping architecture from a high-dimensional target space to a low-dimensional projection space, based on the mechanical properties of the mixed product, proportioning decisions and dimensional cross-validation optimization under the low-dimensional projection are performed to obtain target proportioning data. Based on this target proportioning data, the parameters of the main agent pump channel and the curing agent pump channel of the mixer are configured to control the proportioning of the main agent component and the curing agent component. As the mixer drives, near-infrared transmission spectra are acquired using an infrared spectral probe and transmitted back to the automatic control module. By identifying the spectral vectors based on the characteristic absorption peaks of the two components, Euclidean distance is used to determine mixing uniformity and make parameter control decisions, generating dynamic control parameters. Based on these dynamic control parameters, dynamic parameter adjustment and control of the main agent pump and the curing agent pump channels are performed. In other words, by installing a near-infrared spectral probe at the mixer outlet, the spectral vectors of the characteristic absorption peaks of the two components are acquired in real time. Mixing uniformity is quantitatively determined using Euclidean distance, and the Euclidean distance determination result is used as feedback to connect to the automatic control module, generating dynamic control parameters and dynamically adjusting the channel parameters of the main agent pump and the curing agent pump, thereby improving the mixing quality and stability of the MS structural adhesive.

[0017] The above description is merely an overview of the technical solution of this application. To better understand the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating the dynamic proportioning control method for a two-component Ms structural adhesive according to this application.

[0020] Figure 2 This is a schematic diagram of the dynamic proportioning control system for a two-component Ms structural adhesive according to this application.

[0021] Figure labeling: Dimension mapping module 11, parameter configuration module 12, mixing uniformity judgment module 13, dynamic control module 14. Detailed Implementation

[0022] This application provides a dynamic proportioning control method and system for two-component Ms structural adhesives, solving the technical problems in existing technologies where the mixing uniformity of two-component MS structural adhesives cannot be directly detected online, and proportioning control relies on indirect feedback, leading to the inability to promptly detect and correct mixing unevenness and proportioning drift, further affecting the stability of the component Ms structural adhesives. By installing a near-infrared spectral probe at the mixer outlet, the spectral vectors of the characteristic absorption peaks of the two components are acquired in real time. Mixing uniformity is quantitatively determined using Euclidean distance, and the Euclidean distance determination result is used as feedback to connect to an automatic control module, generating dynamic control parameters and dynamically adjusting the channel parameters of the main agent pump and curing agent pump, thereby improving the mixing quality and stability of the MS structural adhesive.

[0023] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0024] Example 1, please refer to the appendix. Figure 1 This application provides a dynamic proportioning control method for a two-component Ms structural adhesive. The method is applied to a dynamic proportioning control system for the two-component Ms structural adhesive, and specifically includes the following steps: Based on the mechanical properties of the hybrid product, a proportion optimizer is deployed using a mapping architecture from a high-dimensional target space to a low-dimensional projection space. This optimizer performs proportion decisions and dimensional cross-validation optimization under low-dimensional projection to obtain target proportion data. Specifically, the proportion optimization of the main component and the curing agent component is performed in at least two low-dimensional projection spaces, and iterative feedback is guided by a low-dimensional projection space cross-validation method.

[0025] Furthermore, this application also includes the following steps: obtaining the mechanical properties of the hybrid product and constructing a high-dimensional target space based on the performance matrix; determining multiple low-dimensional perspectives based on the mechanical properties, projecting the high-dimensional target space onto it to construct a low-dimensional projection space, wherein, for the short-term adhesion dimension, a normalized first dimensionless index is defined, and for the long-term curing dimension, a normalized second dimensionless index is defined; and constructing a proportioning optimizer by mapping the high-dimensional target space and the low-dimensional projection space.

[0026] Specifically, prepare a batch of two-component MS structural adhesive samples with different mixing ratios. The mixing ratio should be set to cover the expected process window, such as the range from relative under-curing to relative over-curing. For each selected mixing ratio, according to the standard sample preparation method, thoroughly mix the main agent and curing agent under vacuum conditions, inject into a standard mold, and cure for a sufficient time under specified temperature and relative humidity conditions, such as 23°C and 50.0% humidity, to obtain the mixed product specimen.

[0027] Following industry-recognized testing methods, the mechanical properties of each blended product at each mixing ratio were determined. For example, tensile strength was measured using a dumbbell-shaped specimen stretched at a constant rate on a universal testing machine until fracture, and the maximum tensile stress was recorded. Elongation at break was measured simultaneously with tensile strength, and the percentage of gauge length elongation at fracture was recorded. Peel strength was measured using a 180-degree peel specimen, determining the peel force per unit width. Curing speed was typically characterized by finger-dry time or stringing time, i.e., the time required from the start of mixing until the colloid surface was no longer tacky. Depth of curing was measured by injecting the colloid into a cylindrical mold of a certain depth and measuring the thickness of the cured layer after a specific time. Elastic modulus was calculated from the slope of the initial linear segment of the tensile stress-strain curve. At least five parallel specimens were prepared for each mixing ratio, and the arithmetic mean was taken as the representative value of the mechanical properties for that ratio.

[0028] The representative values ​​of various mechanical properties corresponding to all mixing ratios are organized into a two-dimensional data array, namely the performance matrix. The row index of the matrix corresponds to different mixing ratios, and the column index corresponds to different mechanical property indicators. The value of each cell in the matrix is ​​the measured average value of that performance at that ratio.

[0029] A high-dimensional target space is constructed based on this performance matrix. Each row of the matrix is ​​considered a data point in the high-dimensional space, and the coordinates of this data point are composed of the mechanical property values ​​of all columns in that row. For example, if there are six mechanical property indicators, then each data point is a six-dimensional vector, with its six dimensions representing tensile strength, elongation at break, peel strength, curing speed, deep curing depth, and elastic modulus, respectively. The data points corresponding to all rows are distributed in this six-dimensional space, forming the high-dimensional target space. The boundary of the high-dimensional target space is determined by the minimum and maximum values ​​of each property, and its internal structure reflects the trade-off relationship of mechanical properties under different mixing ratios.

[0030] Based on mechanical properties, several low-dimensional perspectives are identified: a short-term bonding dimension focusing on ease of application and a long-term curing dimension focusing on structural load-bearing capacity. One low-dimensional perspective focuses on short-term bonding behavior, specifically the performance of the blended product from application to initial curing, with a focus on time-sensitive indicators such as curing speed and initial tack development rate. The other low-dimensional perspective focuses on long-term curing behavior, specifically the intrinsic mechanical properties and durability of the blended product after full curing, with a focus on indicators such as tensile strength, elongation at break, elastic modulus, and uniformity of deep curing.

[0031] Each high-dimensional data point in the high-dimensional target space is projected onto these two low-dimensional perspectives according to a predefined mapping rule. For the short-term adhesion dimension, the original data related to curing speed, initial tack, etc., in the high-dimensional space are normalized and then weighted to form a first dimensionless index with a value between zero and one. For the long-term curing dimension, data such as tensile strength, elongation at break, and elastic modulus are normalized and then weighted to form a second dimensionless index. The original high-dimensional space is thus compressed into a two-dimensional low-dimensional projection space, with the first dimensionless index on the horizontal axis and the second dimensionless index on the vertical axis.

[0032] A mathematical mapping relationship is established between the original mechanical performance data in the high-dimensional target space and two dimensionless indices in the low-dimensional projection space, and a proportion optimizer is constructed. Multiple response surface models are established with the mixing ratio as the independent variable and the various mechanical properties in the high-dimensional space as the dependent variable; simultaneously, a response surface model is established with the mixing ratio as the independent variable and the first and second dimensionless indices in the low-dimensional projection space as the dependent variables.

[0033] Given the desired first and second dimensionless index values, solve for the corresponding mixing ratio. Since the forward mapping is usually nonlinear and not one-to-one, within a reasonable range of the mixing ratio, the optimization objective is to minimize the Euclidean distance between the low-dimensional coordinates output by the forward mapping and the desired low-dimensional coordinates. The optimal mixing ratio is found using gradient descent. An initial guess for the mixing ratio is set, typically choosing the midpoint of a reasonable range as the starting point. If the range is 8.0 to 12.0, the initial value is set to 10.0. The hyperparameters of the gradient descent method are set as follows: learning rate 0.05, convergence condition is that the change in the mixing ratio between two iterations is less than 0.001. Enter the iteration loop. In each iteration, calculate the forward mapping output at the current mixing ratio to obtain the current low-dimensional coordinates; calculate the Euclidean distance between the current low-dimensional coordinates and the desired low-dimensional coordinates; calculate the first derivative of this Euclidean distance with respect to the mixing ratio, giving the mixing ratio a small increment; calculate the changed Euclidean distance; the difference between the two, divided by the increment, is the approximate gradient. The direction of the blending ratio adjustment is determined by the sign of the gradient: a positive gradient indicates that increasing the blending ratio will increase the Euclidean distance, therefore the blending ratio should be decreased; a negative gradient indicates that the blending ratio should be increased. The blending ratio is updated one step in the opposite direction of the gradient according to the set learning rate. After updating, the blending ratio is checked to see if it exceeds a reasonable range; if so, it is restricted to the range boundary. This iteration is repeated until the convergence condition is met, and the final blending ratio is the optimal target mix.

[0034] The constructed proportion optimizer was validated by taking several mixed proportion samples that were not involved in the modeling, measuring their actual mechanical properties, calculating the actual low-dimensional index, and setting the expected low-dimensional index as the actual value. The optimizer then outputs a recommended proportion, and the deviation between the recommended proportion and the actual proportion is compared. If the deviation is within the allowable range, such as ±0.2%, the optimizer is considered effective; otherwise, the number of experimental samples needs to be increased or the fitting model needs to be changed.

[0035] The proportion optimizer provides the ability to reverse map from desired performance indicators to target proportion data, enabling operators to directly obtain the required mixing ratio without repeated trial and error, significantly shortening the process parameter debugging cycle.

[0036] Furthermore, this application also includes the following steps: for the first projection space, a first weight is determined by neighborhood distance, a second weight is determined by the corresponding high-dimensional performance importance, and a first relative center is located by the first weight and the second weight; taking the first relative center as the low-dimensional performance target of the short-term adhesion dimension, the ratio optimization of the main agent component and the curing agent component is performed to determine the first ratio data.

[0037] Furthermore, this application also includes the following steps: determining second proportion data for the second projection space; evaluating the second projection space using the first proportion data and adjusting the constraints with a preset step size to determine a first corrected proportion; evaluating the first projection space using the second proportion data and adjusting the constraints with a preset step size to determine a second corrected proportion; performing multiple rounds of cross-validation optimization using the first corrected proportion and the second corrected proportion until the difference between the two data of the Nth corrected proportion updated iteratively is less than a preset value, which is judged as convergence, and selecting any one of the Nth corrected proportions as the target proportion data.

[0038] Specifically, for the first projection space, for each existing experimental data point, the Euclidean distance to all other data points is calculated, and the average of the smallest distances is taken as the neighborhood distance of that point. The smaller the neighborhood distance, the denser the data around that point, and the stronger its representativeness, thus assigning it a larger first weight. Based on pre-set high-dimensional performance importance, such as tensile strength importance of 0.3, curing speed importance of 0.25, deep curing depth importance of 0.2, elongation at break importance of 0.15, peel strength importance of 0.05, and elastic modulus importance of 0.05, these importance values ​​are normalized and used as the second weight. The coordinate value of each data point is multiplied by the product of the corresponding first weight and the second weight, then summed and divided by the total weight to obtain the first relative center, which is a virtual ideal point representing the most desired low-dimensional coordinates for short-term adhesion considering both data distribution density and mechanical performance importance.

[0039] Optimization of the mixing ratio is performed with the first relative center as the target. Using the established forward mapping model (the aforementioned mapping from mixing ratio to the first dimensionless index), the gradient descent method is employed to find a ratio within a reasonable range of mixing ratios that makes the first dimensionless index output by the forward mapping as close as possible to the coordinate value of the first relative center. This optimization process does not consider the second dimensionless index, and the result is called the first mixing ratio data. The reasonable range of the mixing ratio is defined. For example, if the experiment covers a range of hardener parts from 8.0 to 12.0, the reasonable range can be set to 7.5 to 12.5. An initial mixing ratio is set, usually the midpoint of the reasonable range. An iterative loop is entered. In each iteration, the current mixing ratio is input into the forward mapping model to calculate the corresponding first dimensionless index, and the deviation between this index and the first relative center is calculated. To determine the direction of the next mixing ratio adjustment, the first derivative of the deviation with respect to the mixing ratio needs to be calculated. Since the forward mapping model lacks an explicit analytical derivative, a small increment, such as 0.001 parts of curing agent, is applied to the current mixing ratio. The forward mapping model is then called again to obtain a new dimensionless index. The difference between the two output values ​​is calculated and divided by the increment to approximate the gradient. The sign of the gradient determines the adjustment direction: a positive gradient indicates that increasing the mixing ratio will increase the dimensionless index, potentially increasing the deviation, thus requiring a decrease; a negative gradient indicates an increase. The gradient is multiplied by a pre-set learning rate, such as 0.05, to obtain the update step size for the mixing ratio. This step size is then subtracted from the current ratio to obtain the new ratio. After each update, the new ratio is checked to ensure it remains within a reasonable range; if it exceeds this range, it is constrained to the boundary. This iteration is repeated until the change in the mixing ratio between two consecutive iterations is less than a preset threshold, such as 0.001, or the maximum number of iterations is reached. The resulting mixing ratio is the first proportion data.

[0040] Similarly, using the same method, for the second projection space, the second relative center is first calculated based on neighborhood distance and high-dimensional performance importance. Then, through forward mapping and reverse optimization, the second proportion data is obtained. The first proportion data is taken and input into the forward mapping model of the second projection space. The actual dimensionless index of this proportion in the long-term curing dimension is calculated and compared with the second relative center to obtain the deviation. If the actual dimensionless index is lower than the second relative center, it indicates insufficient long-term performance, and adjustment is needed to increase the amount of hardener. If the actual dimensionless index is higher than the second relative center, it indicates excessive long-term performance, and adjustment is needed to decrease it. During adjustment, single-step correction is performed according to a preset step size, such as 0.1 parts of hardener, and adjustment constraints are applied, such as a maximum of 5 steps per round, and the adjusted proportion cannot exceed the range of 8.0 to 12.0. After the maximum preset number of adjustments, the first corrected proportion is obtained.

[0041] Take the second ratio data and input it into the forward mapping model of the first projection space. Calculate the actual dimensionless index of the ratio in the short-term adhesion dimension and compare it with the first relative center. Similarly, perform reverse correction according to the preset step size and adjustment constraints to obtain the second corrected ratio.

[0042] The values ​​of the first and second corrected proportions are compared, and their absolute difference is calculated. If the difference is less than a preset convergence threshold (e.g., 0.05 parts of curing agent), convergence is considered achieved, and either proportion is randomly selected as the target proportion data. If the difference is still greater than the preset convergence threshold, the first corrected proportion is used as the new first proportion data, and the second corrected proportion is used as the new second proportion data. The above cross-evaluation and correction process is repeated, that is, the latest first corrected proportion is used to evaluate and adjust again in the second projection space, and the latest second corrected proportion is used to evaluate and adjust again in the first projection space. This process is repeated multiple times until the two corrected proportions are sufficiently close. The target proportion data is the final optimization result obtained by randomly selecting one of the two corrected proportions after convergence. Since this process only alternates between one-dimensional short-term and one-dimensional long-term, with different dimensions mutually verifying and fine-tuning, this proportion optimization process is lightweight and efficient compared to direct optimization analysis of high-dimensional multi-index factors.

[0043] For example, taking a certain type of MS structural adhesive as an example, five mixing ratios were obtained in the preliminary test, with the curing agent mass ratio of the main agent being 100:8.0, 100:9.0, 100:10.0, 100:11.0, and 100:12.0, respectively, corresponding to the short-term adhesion dimensionless index S and the long-term curing dimensionless index L. In the first projection space, after neighborhood distance weighting and high-dimensional performance importance weighting, the first relative center was calculated to be 0.66; the first ratio was optimized using the golden section method to obtain 9.2. Similarly, the second relative center in the second projection space was 0.70, and the second ratio was optimized to obtain 10.7. The preset step size was 0.1, and the convergence threshold was 0.05. First round of mutual verification: Substituting the first ratio 9.2 into the second projection space, L=0.70 was obtained, which deviated from the second relative center by 0, so the first corrected ratio remained at 9.2. Substituting the second ratio 10.7 into the first projection space, we get S=0.67, which deviates from the first relative center by +0.01. After adjusting the step size, we get the second corrected ratio 10.6. The difference between the two is 0.4 > 0.05, so we continue iterating. Second round: Using 9.2 to evaluate the second projection space, we get L=0.70, which remains unchanged; using 10.6 to evaluate the first projection space, we get S=0.68, which deviates from 0.66 by +0.02, so we adjust it to 10.5. The difference is 0.3 > 0.05. Third round: Using 10.5 to evaluate the first projection space, we get S=0.69, so we adjust it to 10.4. The difference is 0.2. Fourth round: Using 10.4 to evaluate, we get S=0.70, so we adjust it to 10.3. The difference is 0.1. Fifth round: Using 10.3 to evaluate, we get S=0.71, so we adjust it to 10.2. At this point, the difference between the first corrected ratio 9.2 and the second corrected ratio 10.2 is 0.05, reaching the convergence threshold. 9.2 is then taken as the target ratio data.

[0044] By decomposing the high-dimensional multi-objective optimization problem into two independent optimization and cross-validation processes in two low-dimensional spaces, the computational cost is significantly reduced. When the matching requirements proposed by the two dimensions are inconsistent, repeated cross-calibration eventually converges to a compromise solution that can simultaneously satisfy the basic requirements of both dimensions, avoiding severe deviations in the performance of the other dimension caused by optimization in a single dimension.

[0045] Based on the target ratio data, the parameters of the main agent pump channel and the curing agent pump channel of the mixer are configured to control the ratio of the main agent component and the curing agent component.

[0046] Specifically, the target mixing ratio data is input into the mixer's controller. The controller converts this ratio into the speed ratio of the two pump drive motors based on parameters such as the displacement characteristics of the two pumps, the glue density, and pipeline resistance. The volume or mass of the main agent pump discharged per revolution, and the volume or mass of the hardener pump discharged per revolution, are then determined. Let the displacement per revolution of the main agent pump be Qm, and the displacement per revolution of the hardener pump be Qc. To achieve the mass ratio R = main agent mass / hardener mass, the speed ratio Nm / Nc = (R × ρc × Qc) / (ρm × Qm), where ρm and ρc are the densities of the main agent and hardener, respectively. If the two pumps have identical structures, the speed ratio is directly equal to the reciprocal of the target mixing ratio data multiplied by the density ratio.

[0047] The calculated target speed ratio is used as the main setpoint and written into the servo drives or frequency converters of the main agent pump and the curing agent pump. Auxiliary parameters such as pump start-up acceleration, maximum speed limit, and pressure alarm threshold are set according to the current liquid level in the glue tank and the viscosity of the glue.

[0048] After the mixer is started, the controller simultaneously drives the main agent pump and the hardener pump according to the set speed ratio. Both pumps deliver the adhesive to the inlet of the static mixing tube or dynamic mixing valve through their respective pipelines. During operation, the controller continuously reads the feedback signals from the flow sensors installed at the pump outlets or on the pipelines, calculating the real-time flow rate and the actual mixing ratio. If the deviation between the actual ratio and the target ratio exceeds the allowable range, such as ±0.5%, the controller automatically adjusts the speed of one of the pumps to correct the ratio. The main agent pump speed is maintained at the set value, and only the hardener pump speed is adjusted to match the ratio requirement. The corrected speed command is output to the pump driver via analog signal or bus communication.

[0049] The controller executes a pre-filling and venting procedure, first running at low speed for a few seconds to expel air and unmixed adhesive from the pipeline, and then proceeding to the normal dispensing stage once the flow rate stabilizes. The entire mixing control process continues until the dispensing task is completed or a stop command is received.

[0050] Driven by the mixer, near-infrared transmission spectra are acquired by an infrared spectral probe and transmitted back to the automatic control module. By identifying the spectral vectors based on the characteristic absorption peaks of the two components, the module makes judgments on mixing uniformity and makes control adjustment decisions based on Euclidean distance, generating dynamic control parameters.

[0051] Furthermore, this application also includes the following steps: installing a near-infrared spectral probe at the outlet position of the mixer, and collecting near-infrared transmission spectra as the mixer is driven; determining the characteristic absorption peaks of the functional groups of the main agent component and the curing agent component in the near-infrared region; deploying an automatic control module based on the mixing uniformity index of the characteristic absorption peaks; and inputting the continuously collected near-infrared transmission spectra into the automatic control module to generate dynamic control parameters.

[0052] Specifically, a near-infrared spectral probe is fixedly installed at the outlet of the mixer, ensuring that its optical window is flush with the flow path of the mixed adhesive and that the adhesive flows through the probe's detection area at a stable thickness. A transmission-type probe is selected, with the light source and detector located on opposite sides of the adhesive flow direction. After installation, optical path calibration and baseline correction are required to ensure a flat reference spectrum when there is no adhesive.

[0053] The mixer drive is activated to begin mixing the main agent and curing agent according to the target ratio and allow them to flow to the outlet. Simultaneously, the near-infrared spectroscopy probe is activated to continuously acquire transmission spectra at a predetermined sampling frequency, such as 10 times per second. Each acquired frame of spectrum is digitized and stored in a buffer. Through literature review, the characteristic absorption peak positions of key functional groups in the main agent and curing agent components in the near-infrared region are determined. For example, the silicon-oxygen bonds in the main agent absorb at specific wavelengths, while the amino or epoxy groups in the curing agent absorb at other wavelengths. The presence and intensity variations of characteristic absorption peaks reflect whether the two components are present in the correct proportion and whether they are uniformly distributed. The wavelength range is recorded as the band of interest for subsequent spectral analysis.

[0054] Near-infrared transmission spectroscopy is performed when near-infrared light passes through a layer of adhesive of a certain thickness. Molecules in the adhesive absorb certain wavelengths of light, while the unabsorbed light passes through and is received by a detector, resulting in a spectral curve. The horizontal axis represents wavelength, and the vertical axis represents absorbance. Functional groups are atoms or groups of atoms in a molecule that determine the characteristic chemical properties of the compound. Characteristic absorption peaks are the selective absorption of specific wavelengths of light by specific functional groups in the near-infrared spectrum, appearing as a peak (bulge or depression) on the spectral curve. The positions of characteristic absorption peaks differ for different functional groups; for example, the first-order overtone absorption peak of a carbon-hydrogen bond is approximately between 1700 nm and 1800 nm, while the absorption peak of an oxygen-hydrogen bond is approximately between 1400 nm and 1500 nm.

[0055] The mixing uniformity index measures how uniformly the main agent and curing agent are distributed at a microscopic level after mixing. It is typically calculated by determining the Euclidean distance between multiple continuously acquired near-infrared spectra to obtain a similarity score. When the mixture is completely uniform, two consecutive spectra should be almost identical, and the mixing uniformity index approaches 0.

[0056] An automatic control module is deployed to receive near-infrared spectral data, calculate the mixing uniformity index, determine the current mixing quality, and generate dynamic control parameters for adjusting pump parameters based on preset judgment rules. One or more narrow bands containing characteristic absorption peaks are selected, such as extending 5nm to the left and right of each peak, and the spectral intensity vector within each band is extracted. For two consecutively acquired spectra, the Euclidean distance between their spectral vectors within the band is calculated. This distance value is normalized and used as the mixing uniformity index. Specifically, the Euclidean distance is calculated using the spectral vectors of the current frame and the previous frame. Let the vector dimension be n, and the two frame vectors be A and B, then the Euclidean distance d = sqrt(Σ(A...). i -B i ) 2 ), A i Let B be the i-th vector A. i For the i-th B vector, normalize d to the interval between 0 and 1, and divide by a pre-determined maximum possible distance, such as the distance between the spectra of the pure main agent and the pure curing agent. The normalized value is the mixing uniformity index; the closer the index is to 0, the more uniform the mixing.

[0057] A preset threshold, such as 0.05, is used. When the index is below the preset threshold, the mixing is considered uniform; when it is above the threshold, the mixing is considered uneven. The automatic control module also includes a trend analysis submodule to identify the change pattern of Euclidean distance over time, such as periodic fluctuations or sudden spikes, thereby distinguishing between uniformity fluctuations caused by pump pulsation and hardware failures of the mixer.

[0058] Continuously acquired real-time near-infrared transmission spectra are input frame by frame into the automatic control module. For each frame, a mixing uniformity index is calculated against a reference frame. If the mixing uniformity index is within the allowable range, no control command is output. If the mixing uniformity index exceeds the threshold, the required pump parameter correction amount, i.e., the dynamic control parameters, is calculated based on preset decision logic. These typically include the speed increment of the curing agent pump and the adjustment of the soft-start time. The dynamic control parameters are output to the drivers of the main agent pump and the curing agent pump via fieldbus or analog signals, changing the pump speed in real time to correct the mixing state in a closed loop. The drivers smoothly change the motor speed according to the soft-start time, thereby changing the actual flow rates of the two components and correcting the mixing uniformity. The entire closed-loop process is repeated at the spectral sampling frequency to ensure that the mixing quality remains under control.

[0059] Furthermore, this application also includes the following steps: acquiring a first frame spectrum and a second frame spectrum, determining a first spectral vector and a second spectral vector, wherein the spectral vector is determined based on the spectral band where the characteristic absorption peak is located; calculating the Euclidean distance between the first spectral vector and the second spectral vector; judging the mixing uniformity based on the Euclidean distance, and generating a judgment result, wherein if the Euclidean distance is less than a preset value, it is judged to be mixed uniformly.

[0060] Furthermore, this application also includes the following steps: generating a spectral fluctuation sequence based on the judgment result of a preset time period; if the fluctuation trend of the spectral fluctuation sequence is periodic, it is determined to be a type of anomaly, wherein the periodicity standard is taken as the frequency of the pump stroke, and the type of anomaly is uneven mixing caused by pump pulsation; if the fluctuation trend is a pulse-like spike, it is determined to be a type of anomaly, wherein the type of anomaly is a mixer malfunction.

[0061] Furthermore, this application also includes the following steps: if it is the first type of anomaly, generate a dynamic control instruction; according to the dynamic control instruction, make a parameter control adjustment decision using Euclidean distance as the adjustment requirement, and generate the dynamic control parameters, wherein the dynamic control parameters include the speed correction amount of the main agent pump and the curing agent pump; if it is the second type of anomaly, generate a shutdown and maintenance instruction.

[0062] Specifically, after the mixer has stabilized, the first frame of near-infrared transmission spectrum is acquired, and the spectral bands containing one or more pre-determined characteristic absorption peaks are extracted. For each selected characteristic absorption peak, its center wavelength is used as a reference, and a fixed width is extended to the left and right, such as 5 nm each, forming a continuous narrow band. Within this band, according to the spectrometer's resolution, such as one data point every 5 nm, the absorbance values ​​corresponding to all wavelength points are extracted, and these values ​​are arranged in ascending order of wavelength to form the first spectral vector. After a sampling interval, the second frame of spectrum is acquired. Following the exact same band selection rules as the first frame, the absorbance values ​​of the same wavelength points are extracted and arranged in the same order to form the second spectral vector.

[0063] Calculate the Euclidean distance between the first and second spectral vectors. Subtract the corresponding elements of the two vectors to obtain the difference; square each difference; sum all the squared values; take the square root of the sum to obtain the numerical value of the Euclidean distance.

[0064] The calculated Euclidean distance is compared with a preset value. The preset value is a critical value determined based on the statistical results of a large number of uniformly and non-uniformly mixed samples. If the Euclidean distance is less than the preset value, it indicates that the two spectra are very similar, meaning that the chemical composition of the mixed gel hardly changes at adjacent time points, indicating uniform mixing. If the Euclidean distance is greater than or equal to the preset value, it indicates that there is a significant difference between the two spectra, indicating non-uniform mixing. The judgment result is output, indicating whether the mixing is uniform or non-uniform.

[0065] Instead of analyzing and identifying uniformity based on features or real-time mixing conditions, it judges the consistency of spectral vectors from two separate measurements. This lightweight approach requires less computational power and is well-suited for real-time online adjustments. When the mixture is uniform, the spectra at adjacent time points tend to be consistent, with small distance values; when the mixture is poor, the spectra fluctuate greatly, resulting in large distance values.

[0066] During continuous operation, the mixing uniformity judgment result calculated at each sampling time is stored in a fixed-length circular buffer, the size of which corresponds to a preset time period. When the buffer is full, the oldest value is removed each time a new distance value is obtained, thereby maintaining a temporally continuous spectral fluctuation sequence.

[0067] Trend analysis is performed on the spectral fluctuation sequence. The dominant frequency of the sequence is extracted by fast Fourier transform or by statistically analyzing the time interval between adjacent peaks. Simultaneously, the current stroke frequency of the pump is read from the mixer. The two frequencies are compared. If the difference between the dominant frequency of the spectral fluctuation sequence and the pump stroke frequency is within the allowable error range, such as ±0.1Hz, the fluctuation is considered periodic and synchronous with the pump stroke frequency. This is identified as a type of anomaly, namely, mixing inhomogeneity caused by pump pulsation. Its characteristic is that the Euclidean distance fluctuates periodically at a fixed frequency, and the fluctuation amplitude is usually within a certain range. This can be suppressed by adjusting the pump speed or increasing the damping.

[0068] If there are no obvious periodic fluctuations in the sequence, but there are peaks with amplitudes far exceeding the normal fluctuation range, such as the distance value suddenly increasing to more than 3 times the normal value and falling back within 1 to 2 sampling periods, it is judged as a type II anomaly, that is, a hardware failure of the mixer. This may be caused by channel effect or flow around in the mixer, that is, the colloid is not fully mixed and flows around through gaps or wear, or the internal components of the mixer are deformed or blocked, and the fluid is forced to change its path, etc.

[0069] Different decision paths are executed for different types of anomalies. For one type of anomaly, a dynamic control command is generated, entering the parameter control adjustment decision stage. The difference between the current Euclidean distance and the preset threshold is taken as the adjustment requirement, i.e., the deviation that needs to be reduced. Based on the proportional-integral-derivative (PID) control algorithm, the speed correction amounts for the main agent pump and the curing agent pump are calculated. The speed of the main agent pump is kept constant, and only the speed of the curing agent pump is adjusted because the curing agent flow rate is smaller, resulting in a faster adjustment response. The formula for calculating the speed correction amount is: ΔN = Kp × (current Euclidean distance - threshold) + Ki × integral term + Kd × derivative term, where Kp, Ki, and Kd are pre-tuned coefficients. The correction amount can be positive (increasing the curing agent pump speed) or negative (decreasing the speed), and the specific sign depends on the correspondence between the deviation and the proportioning deviation in the adhesive system. The final generated dynamic control parameters include the curing agent pump speed correction amount and the soft start time.

[0070] For Class II anomalies, no speed adjustment attempts are made; a shutdown and maintenance command is generated directly. A safety relay on the mixer is sent via fieldbus to immediately cut off the power to the drive motors of the main agent pump and the hardener pump, and triggers the audible and visual alarms. Simultaneously, the specific fault code and prompts are displayed on the interactive interface, requiring operator intervention for equipment maintenance. For example, if the Class II anomaly is a mixer malfunction, check the mixing components or seals.

[0071] For example, a preset time period of 10.0s corresponds to 100 frames of spectrum, with a sampling frequency of 10Hz; the pump stroke frequency of the curing agent pump is 2.0Hz; and the Euclidean distance threshold is 0.050. During normal operation, the average Euclidean distance is approximately 0.025. Between 30s and 40s, simulated hardener pump seal wear causes the hardener flow rate to exhibit sinusoidal fluctuations with a frequency of 2.0Hz and an amplitude of ±15%. The Euclidean distance sequence within this 10s is recorded. The sequence exhibits a distinct sinusoidal waveform, with a peak distance value of approximately 0.085, a trough of approximately 0.030, and an average value of 0.058, exceeding the threshold. A fast Fourier transform of the Euclidean distance sequence reveals a significant amplitude at 2.0Hz, consistent with the pump stroke frequency. The automatic control module identifies this as a type of anomaly. Proportional control Kp=100.0, Ki=0, and Kd=0 are invoked. The current Euclidean distance is the average of the last 5 frames, which is 0.072, exceeding the tolerance by 0.022. The speed correction was 2.2 r / min. The current speed of the hardener pump was 17.8 r / min, the target speed was 20.0 r / min, and the slow start time was 0.3 s. A command was sent via the bus, and the hardener pump accelerated to 20.0 r / min within 0.3 s. After the adjustment, the average Euclidean distance decreased to 0.035 over the next 10 s, the periodic fluctuation amplitude decreased, and the mixing uniformity was restored. In another scenario, at 60 s, the internal seal of the simulated static mixer wore out, forming a bypass channel. Some adhesive flowed out directly through the bypass without passing through the mixing blades. Starting from the first frame of 60 s, the Euclidean distance jumped sharply from the normal 0.025 to 0.210. It dropped to 0.195 in the second frame, 0.205 in the third frame, and then fluctuated between 0.190 and 0.220 without periodicity. Five consecutive frames showed a spike that did not subside. The trend analysis submodule detected a spike amplitude exceeding 8 times the normal value without periodicity, classifying it as a Class II anomaly. A shutdown and maintenance command was immediately generated. The digital output port went high, the enable signals for the main agent pump and hardener pump drivers were disconnected, and the motors stopped. Simultaneously, an alarm window popped up on the HMI: "Mixer malfunction, possibly due to internal flow obstruction or blockage. Please check the mixing elements and seals." An audible and visual alarm was activated. Upon arrival, the operator disassembled the mixer and found a worn seal ring; after replacement, the system returned to normal.

[0072] Different control strategies are adopted for different types of anomalies. For pump pulsation that can be adjusted online, dynamic control parameters are automatically generated for closed-loop compensation to avoid production efficiency loss caused by downtime. For mixer hardware failures that cannot be automatically repaired, shutdown and maintenance instructions are decisively generated to prevent the continued production of waste products and protect the equipment from further damage.

[0073] Based on the aforementioned dynamic control parameters, dynamic parameter adjustment and control of the channels of the main agent pump and the curing agent pump are performed.

[0074] Furthermore, this application also includes the following steps: plotting a mixing uniformity trend curve with sampling time as the horizontal axis and the quantized value of Euclidean distance as the vertical axis; superimposing the pump stroke signal and alarm threshold line on the mixing uniformity trend curve, performing time series monitoring and automatic anomaly marking, and displaying it on the terminal interface.

[0075] Specifically, based on the dynamic control parameters, the corrected speed commands are sent to the servo drives of the main agent pump and the curing agent pump via fieldbus. The drives linearly change the motor speed according to the soft start time, adjusting the actual flow rates of the two components towards the target ratio. The speed of the main agent pump is kept constant, and only the speed of the curing agent pump is adjusted because the curing agent has a small flow rate, a fast response, and high adjustment efficiency. After parameter adjustment, the spectrum is collected and the Euclidean distance is calculated to form a closed-loop feedback.

[0076] Simultaneously, a real-time monitoring thread is initiated, dedicated to acquiring and displaying the mixing uniformity trend curve. At fixed time intervals, such as every 0.1 seconds, synchronized with spectral sampling, the quantized Euclidean distance value at the current moment is acquired. This value is used as the ordinate, and the cumulative time since system startup or the start of the current batch is used as the abscissa to plot a point in the chart control on the terminal interface. As time progresses, all points are connected to form a continuous broken line, i.e., the mixing uniformity trend curve.

[0077] The pump's encoder acquires pulse signals per revolution / stroke, converting them into a series of discrete trigger points, such as a short vertical line or a triangular marker at the start of each stroke, with a height equal to the maximum range of the Euclidean distance. These markers are plotted against the background of the trend curve or below it on the same layer, allowing operators to visually observe whether fluctuations in the Euclidean distance are synchronized with the pump stroke.

[0078] A horizontal straight line is set along the vertical axis as the alarm threshold line, equal to the pre-calibrated Euclidean distance alarm threshold, such as 0.05. When a newly drawn Euclidean distance point exceeds the threshold line, a conspicuous mark is added at that point, such as a solid red dot, a flashing border, or a pop-up prompt box. If multiple consecutive points exceed the threshold line, the start and end positions of the abnormal interval will be marked. For periodic fluctuations or peaks identified by the trend analysis submodule, text labels are added to the corresponding positions on the curve, such as abnormal pump pulsation or mixer malfunction.

[0079] All real-time and historical data are stored in a local database. The terminal interface supports timeline zooming, panning, and quick retrieval of abnormal records. Operators can easily grasp the evolution of mixing uniformity, the current state, and the timing and cause of anomalies without additional analysis.

[0080] In summary, the dynamic proportioning control method for a two-component Ms structural adhesive provided in this application has the following technical effects: A proportioning optimizer is deployed based on the mechanical properties of the mixed product using a mapping architecture from a high-dimensional target space to a low-dimensional projection space. Proportioning decisions and dimensional cross-validation optimizations under low-dimensional projection are performed to obtain target proportioning data. Based on the target proportioning data, parameters are configured for the main agent pump channel and the curing agent pump channel of the mixer to control the proportioning of the main agent component and the curing agent component. As the mixer is driven, near-infrared transmission spectra are acquired using an infrared spectral probe and transmitted back to the automatic control module. By identifying the spectral vectors based on the characteristic absorption peaks of the two components, Euclidean distance is used to determine mixing uniformity and make parameter control decisions, generating dynamic control parameters. Based on the dynamic control parameters, dynamic parameter adjustment and management of the main agent pump and the curing agent pump channels are performed. In other words, by installing a near-infrared spectral probe at the mixer outlet, the spectral vectors of the characteristic absorption peaks of the two components are collected in real time. The mixing uniformity is quantitatively judged by Euclidean distance. The Euclidean distance judgment result is used as a feedback quantity and connected to the automatic control module to generate dynamic control parameters. The channel parameters of the main agent pump and the curing agent pump are dynamically adjusted to improve the mixing quality and stability of MS structural adhesive.

[0081] Example 2: Based on the same inventive concept as the dynamic proportioning control method for a two-component Ms structural adhesive in Example 1, this application also provides a dynamic proportioning control system for a two-component Ms structural adhesive. Please refer to the appendix. Figure 2The dynamic proportioning control system for a two-component Ms structural adhesive includes: a dimension mapping module 11, used to deploy a proportioning optimizer based on the mechanical properties of the mixed product using a mapping architecture from a high-dimensional target space to a low-dimensional projection space, performing proportioning decisions and dimensional cross-validation optimization under low-dimensional projection to obtain target proportioning data, wherein the main component and curing agent component are optimized by projection proportioning in at least two low-dimensional projection spaces, and iterative feedback guidance is performed using a low-dimensional projection space cross-validation method; a parameter configuration module 12, used to configure the parameters of the main component pump channel and curing agent pump channel of the mixer according to the target proportioning data, and perform proportioning control of the main component and curing agent component; a mixing uniformity judgment module 13, used to collect near-infrared transmission spectra according to an infrared spectral probe driven by the mixer, and transmit the data back to the automatic control module, and perform mixing uniformity judgment and parameter control adjustment decisions based on Euclidean distance by identifying the spectral vectors based on the characteristic absorption peaks of the two components, generating dynamic control parameters; and a dynamic control module 14, used to perform dynamic parameter adjustment and control of the main component pump and curing agent pump channels according to the dynamic control parameters.

[0082] Furthermore, the dimension mapping module 11 in the dynamic proportioning control system of the two-component Ms structural adhesive is also used for: acquiring the mechanical properties of the mixed product and constructing a high-dimensional target space based on the performance matrix; determining multiple low-dimensional perspectives based on the mechanical properties, projecting the high-dimensional target space to construct a low-dimensional projection space, wherein, for the short-term bonding dimension, a normalized first dimensionless index is defined, and for the long-term curing dimension, a normalized second dimensionless index is defined; and constructing a proportioning optimizer by mapping the high-dimensional target space and the low-dimensional projection space.

[0083] Furthermore, the dimension mapping module 11 in the dynamic proportioning control system of the two-component Ms structural adhesive is also used to: determine a first weight based on neighborhood distance for a first projection space, determine a second weight based on the corresponding high-dimensional performance importance, locate a first relative center based on the first weight and the second weight; and perform proportion optimization of the main component and the curing agent component based on the first relative center as the low-dimensional performance target of the short-term bonding dimension to determine the first proportioning data.

[0084] Furthermore, the dimension mapping module 11 in the dynamic proportioning control system of the two-component Ms structural adhesive is also used for: determining second proportioning data for the second projection space; evaluating the second projection space with the first proportioning data and adjusting the constraints with a preset step size to determine a first corrected proportion; evaluating the first projection space with the second proportioning data and adjusting the constraints with a preset step size to determine a second corrected proportion; performing multiple rounds of cross-validation optimization with the first corrected proportion and the second corrected proportion until the difference between the two data of the Nth corrected proportion updated iteratively is less than a preset value, which is judged as convergence, and selecting any one of the Nth corrected proportions as the target proportioning data.

[0085] Furthermore, the mixing uniformity judgment module 13 in the dynamic proportioning control system of the two-component Ms structural adhesive is also used for: installing a near-infrared spectral probe at the outlet position of the mixer, and collecting near-infrared transmission spectra as the mixer is driven; determining the characteristic absorption peaks of the functional groups of the main component and the curing agent component in the near-infrared region; deploying an automatic control module based on the mixing uniformity index of the characteristic absorption peaks; and inputting the continuously collected near-infrared transmission spectra into the automatic control module to generate dynamic control parameters.

[0086] Furthermore, the mixing uniformity judgment module 13 in the dynamic proportioning control system of the two-component Ms structural adhesive is also used for: acquiring a first frame spectrum and a second frame spectrum, determining a first spectral vector and a second spectral vector, wherein the spectral vector is determined by the spectral band where the characteristic absorption peak is located; calculating the Euclidean distance between the first spectral vector and the second spectral vector; judging the mixing uniformity based on the Euclidean distance, and generating a judgment result, wherein if the Euclidean distance is less than a preset value, it is judged as mixing uniformly.

[0087] Furthermore, the mixing uniformity judgment module 13 in the dynamic proportioning control system of the two-component Ms structural adhesive is also used to: generate a spectral fluctuation sequence based on the judgment result of a preset time period; if the fluctuation trend of the spectral fluctuation sequence is periodic, it is judged as a type of anomaly, wherein the periodicity standard is taken as the frequency of the pump stroke, and the type of anomaly is the mixing unevenness caused by pump pulsation; if the fluctuation trend is a pulse peak, it is judged as a type of anomaly, wherein the type of anomaly is the mixer malfunction.

[0088] Furthermore, the mixing uniformity judgment module 13 in the dynamic proportioning control system of the two-component Ms structural adhesive is also used for: generating a dynamic control command if it is the first type of abnormality; making a parameter control adjustment decision based on the dynamic control command with Euclidean distance as the adjustment requirement, and generating the dynamic control parameters, wherein the dynamic control parameters include the speed correction amount of the main agent pump and the curing agent pump; and generating a shutdown and maintenance command if it is the second type of abnormality.

[0089] Furthermore, the dynamic control module 14 in the dynamic proportioning control system of the two-component Ms structural adhesive is also used to: plot a mixing uniformity trend curve with sampling time as the horizontal axis and the quantized value of Euclidean distance as the vertical axis; superimpose pump stroke signal and alarm threshold line on the mixing uniformity trend curve, perform time series monitoring and automatic abnormal marking, and display it on the terminal interface.

[0090] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The dynamic proportioning control method and specific examples of a two-component Ms structural adhesive in the foregoing embodiment 1 are also applicable to the dynamic proportioning control system of a two-component Ms structural adhesive in this embodiment. Through the foregoing detailed description of the dynamic proportioning control method of a two-component Ms structural adhesive, those skilled in the art can clearly understand the dynamic proportioning control system of a two-component Ms structural adhesive in this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.

[0091] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0092] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for dynamically controlling the mixing ratio of a two-component Ms structural adhesive, characterized in that, include: Based on the mechanical properties of the hybrid product, a proportion optimizer is deployed using a mapping architecture from a high-dimensional target space to a low-dimensional projection space. Proportion decision-making and dimensional cross-validation optimization under low-dimensional projection are performed to obtain target proportion data. Specifically, the proportion optimization of the main component and the curing agent component is carried out by projection proportion optimization in at least two low-dimensional projection spaces, and iterative feedback guidance is carried out by cross-validation of low-dimensional projection spaces. Based on the target ratio data, the parameters of the main agent pump channel and the curing agent pump channel of the mixer are configured to control the ratio of the main agent component and the curing agent component. Driven by the mixer, near-infrared transmission spectrum is acquired by the infrared spectral probe and transmitted back to the automatic control module. By identifying the spectral vector based on the characteristic absorption peaks of the two components, the module makes a judgment on the mixing uniformity and a decision on the parameter control adjustment based on the Euclidean distance, and generates dynamic control parameters. Based on the aforementioned dynamic control parameters, dynamic parameter adjustment and control of the channels of the main agent pump and the curing agent pump are performed; The allocation optimizer is deployed using a mapping architecture from a high-dimensional target space to a low-dimensional projection space, including: Obtain the mechanical properties of hybrid products and construct a high-dimensional target space based on the performance matrix; Based on the mechanical properties, multiple low-dimensional perspectives are determined, and the high-dimensional target space is projected to construct a low-dimensional projection space. For the short-term bonding dimension, a normalized first dimensionless index is defined, and for the long-term curing dimension, a normalized second dimensionless index is defined. A matching optimizer is constructed by mapping the high-dimensional target space with the low-dimensional projection space.

2. The dynamic proportioning control method for a two-component Ms structural adhesive as described in claim 1, characterized in that, Performing allocation decisions under low-dimensional projection includes: For the first projection space, the first weight is determined by the neighborhood distance, the second weight is determined by the corresponding high-dimensional performance importance, and the first relative center is located by the first weight and the second weight. Using the first relative center as the low-dimensional performance target of short-term adhesion dimension, the ratio optimization of the main component and the curing agent component is performed to determine the first ratio data.

3. The dynamic proportioning control method for a two-component Ms structural adhesive as described in claim 2, characterized in that, For the second projection space, determine the second proportion data; The second projection space is evaluated based on the first ratio data, and the first correction ratio is determined by adjusting the constraints with a preset step size. The first projection space is evaluated using the second ratio data, and the second correction ratio is determined by adjusting the constraints with a preset step size. Using the first and second corrected ratios, perform multiple rounds of mutual verification optimization until the difference between the two corrected ratios in the Nth iteration is less than a preset value. If the result is converged, select any one of the corrected ratios in the Nth iteration as the target ratio data.

4. The dynamic proportioning control method for a two-component Ms structural adhesive as described in claim 1, characterized in that, Generate dynamic control parameters, including: A near-infrared spectroscopy probe is installed at the outlet of the mixer to collect near-infrared transmission spectra as the mixer is driven. The characteristic absorption peaks of the functional groups of the identified main component and curing agent component in the near-infrared region; An automatic control module is deployed based on the mixing uniformity index of the aforementioned characteristic absorption peaks; The continuously acquired near-infrared transmission spectra are input into the automatic control module to generate dynamic control parameters.

5. The dynamic proportioning control method for a two-component Ms structural adhesive as described in claim 4, characterized in that, The uniformity of mixing is determined using Euclidean distance, including: Acquire the first frame spectrum and the second frame spectrum, and determine the first spectral vector and the second spectral vector, wherein the spectral vector is determined based on the spectral band where the characteristic absorption peak is located; Calculate the Euclidean distance between the first spectral vector and the second spectral vector; The uniformity of mixing is determined based on the Euclidean distance, and a determination result is generated. If the Euclidean distance is less than a preset value, the mixture is determined to be uniform.

6. The dynamic proportioning control method for a two-component Ms structural adhesive as described in claim 5, characterized in that, Based on the judgment results of the preset time period, a spectral fluctuation sequence is generated; If the fluctuation trend of the spectral fluctuation sequence is periodic, it is judged as a type of anomaly. Among them, the periodicity standard is that it is in sync with the pump stroke. The type of anomaly is the mixing unevenness caused by pump pulsation. If the fluctuation trend is a pulse-like spike, it is judged as a type II anomaly, in which the type II anomaly is a mixer malfunction.

7. The dynamic proportioning control method for a two-component Ms structural adhesive as described in claim 6, characterized in that, If the anomaly is one of the aforementioned types, a dynamic control command will be generated. According to the dynamic control command, the Euclidean distance is used as the adjustment requirement to make the parameter control adjustment decision and generate the dynamic control parameters, wherein the dynamic control parameters include the speed correction amount of the main agent pump and the curing agent pump; If the anomaly is one of the two types mentioned above, a shutdown and maintenance instruction will be generated.

8. The dynamic proportioning control method for a two-component Ms structural adhesive as described in claim 1, characterized in that, After implementing dynamic parameter adjustment and control of the channels for the main agent pump and the curing agent pump, the following is included: Plot the mixing uniformity trend curve with sampling time on the horizontal axis and the quantized value of Euclidean distance on the vertical axis; The pump stroke signal and alarm threshold line are superimposed on the mixing uniformity trend curve to perform time series monitoring and automatic anomaly marking, which are then displayed on the terminal interface.

9. A dynamic proportioning control system for a two-component Ms structural adhesive, characterized in that, The steps for implementing the dynamic proportioning control method for a two-component Ms structural adhesive according to any one of claims 1 to 8, wherein the dynamic proportioning control system for the two-component Ms structural adhesive comprises: The dimension mapping module is used to deploy a proportion optimizer based on the mechanical properties of the mixed product using a mapping architecture from a high-dimensional target space to a low-dimensional projection space. It performs proportion decision-making and dimension cross-validation optimization under low-dimensional projection to obtain target proportion data. Specifically, it optimizes the proportion of the main component and the curing agent component by projection in at least two low-dimensional projection spaces and uses a low-dimensional projection space cross-validation method for iterative feedback guidance. The parameter configuration module is used to configure the parameters of the main agent pump channel and the curing agent pump channel of the mixer according to the target ratio data, and to perform ratio control of the main agent component and the curing agent component. The mixing uniformity judgment module is used to collect near-infrared transmission spectra based on infrared spectral probes while driving the mixer, and transmit the data back to the automatic control module. By identifying the spectral vectors based on the characteristic absorption peaks of the two components, the module makes judgments on mixing uniformity and makes control adjustment decisions based on Euclidean distance, generating dynamic control parameters. The dynamic control module is used to perform dynamic parameter adjustment and control of the channels of the main agent pump and the curing agent pump according to the dynamic control parameters.

Images