Quality control method and system based on single-component bonding silica gel

Abstract

The invention relates to the technical field of silica gel quality control, and discloses a quality control method and system based on single-component bonding silica gel. The system comprises a multi-source data acquisition module, a state estimation module, a dynamic model construction module, a quality calculation module, an abnormity positioning module and an early warning generation module. The multi-source data acquisition module is used for acquiring parameter data such as temperature, pressure, viscosity and curing time when the silica gel is applied; the state estimation module carries out fusion processing on the multi-source data and determines a silica gel quality state estimation value; the dynamic model building module dynamically generates a silica gel bonding physical model according to the operation condition data; the quality calculation module combines the model with multi-source data, and calculates quality parameters of an unmonitored area through a compensation algorithm; the abnormity positioning module positions a quality defect source based on the calculation parameters; the early warning generation module generates graded early warning signals when detecting quality abnormity. The system can solve the traditional management and control problem and realize comprehensive management and control of silica gel quality.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of silica gel quality control, in particular to a quality control method and system based on single-component adhesive silica gel. BACKGROUND

[0002] In the application scenario of single-component adhesive silica gel, quality stability is always an important challenge in the industry throughout the whole process from production and processing to actual bonding operation. The performance of single-component adhesive silica gel is extremely susceptible to the combined influence of external environment and process parameters, and its quality control involves multiple key links. Fluctuations in any of these links can cause quality problems in the final product. Currently, the industry mainly uses single-parameter monitoring methods to monitor the quality of single-component adhesive silica gel, i.e., independently monitoring individual parameters such as temperature, pressure, viscosity, or curing time. This monitoring mode has obvious limitations and cannot effectively analyze the correlation between multiple parameters. For example, when only viscosity anomalies are detected, it is difficult to determine whether the anomalies are caused by temperature fluctuations, pressure changes, or both, making it difficult to trace the quality problems. The existing technology lacks effective fusion processing mechanisms for multi-source quality parameter data. In the application process of silica gel, data such as temperature, pressure, viscosity, and curing time are often scattered in different monitoring devices or systems, and the data are independent of each other, which cannot form a unified quality state evaluation basis. This makes it difficult for operators to accurately grasp the overall quality state of the silica gel and rely on experience for judgment, which can easily lead to misjudgment or missed judgment. For unmonitored areas in the application process of silica gel, the existing technology cannot effectively calculate the quality parameters. Due to the limitations of the arrangement of monitoring equipment in actual production scenarios, real-time monitoring is difficult to achieve in some areas, and these unmonitored areas become blind spots for quality control. Once quality problems occur in these areas, they are often not detected in time and are only exposed when the final product is tested, which causes a lot of waste of cost and delay in production. The current quality control system lacks dynamic adjustment capability. The operating conditions in the silica gel bonding process change with production batches, raw material batches, environmental conditions, and other factors, while existing silica gel bonding models are fixed models that cannot be dynamically adjusted according to changes in actual operating conditions. The accuracy of fixed models will decrease significantly when the operating conditions change greatly, which cannot provide reliable support for quality control and thus affects the stability of product quality. The quality abnormality early warning mechanism in the prior art is not perfect. When detecting a quality abnormality, only a single early warning signal can usually be sent, and hierarchical early warning cannot be performed according to the severity of the abnormality. This makes it difficult for an operator to quickly judge the emergency degree of the abnormality, and targeted countermeasures cannot be taken, which may cause small quality problems to be magnified, affecting production progress and product quality. SUMMARY

[0003] The purpose of the present application is to provide a quality control method and system based on single-component adhesive silicone to solve the problems raised in the background art.

[0004] To achieve the above-mentioned purpose, the present application provides a quality control system based on single-component adhesive silicone, which comprises: A multi-source data acquisition module for acquiring multi-source quality parameter data in the silicone application process, wherein the multi-source quality parameter data comprises temperature data, pressure data, viscosity data and curing time data; A state estimation module for performing fusion processing on the multi-source quality parameter data to determine a silicone quality state estimation value; A dynamic model construction module for dynamically generating a silicone adhesive physical model according to operating condition data; A quality calculation module for combining the silicone adhesive physical model and the multi-source quality parameter data and using a compensation algorithm to calculate quality parameters of an unmonitored area; An abnormality positioning module for positioning a quality defect source based on the calculated quality parameters of the unmonitored area; An early warning generation module for generating a hierarchical early warning signal when a quality abnormality is detected.

[0005] Preferably, the multi-source data acquisition module comprises: A sensor array is arranged at a preset position on a silicone coated surface, wherein the sensor array comprises a temperature sensor subset, a pressure sensor subset and a viscosity sensor subset; The temperature sensor subset is symmetrically distributed in a high-temperature variable region of the silicone coating area, the pressure sensor subset is arranged in a stress concentration region of the silicone coating edge, and the viscosity sensor subset is uniformly distributed in a central region of the silicone coating; The temperature data, the pressure data and the viscosity data are collected in real time by the multi-source data acquisition module.

[0006] Preferably, the state estimation module comprises: Noise characteristic analysis is performed on the multi-source quality parameter data to obtain multi-source quality data noise characteristics; According to the multi-source quality data noise characteristics, a multi-source quality data denoising threshold is set; According to the multi-source quality data denoising threshold, noise identification and filtering preprocessing are performed on the multi-source quality parameter data, and standard multi-source quality parameter data is obtained. Feature extraction fusion and quality state estimation are performed on the standard multi-source quality parameter data, and a silicon gel quality state estimation value is determined.

[0007] Preferably, the dynamic model construction module comprises: The operating condition data is obtained from an external system, and the operating condition data includes environmental temperature data, humidity data, and coating speed data; An equivalent thermal load is generated according to the environmental temperature data, and a flow fluctuation value is calculated according to the coating speed data; The equivalent thermal load and the flow fluctuation value are combined to generate a dynamic load matrix; The dynamic load matrix is input into a reference model established based on silicon gel material parameters to generate the silicon gel bonding physical model.

[0008] Preferably, the quality estimation module comprises: The quality parameter difference of adjacent monitoring points is extracted from the standard multi-source quality parameter data, and the change gradient characteristic quantity of the local area is calculated; The change gradient characteristic quantity is compared with the theoretical gradient of the silicon gel bonding physical model to calculate a quality deviation coefficient; The material parameters of the unmonitored area are iteratively adjusted according to the quality deviation coefficient until the error between the predicted quality parameter of the silicon gel bonding physical model and the measured quality parameter converges to a set range, and a predicted quality parameter is output to construct a quality distribution map.

[0009] Preferably, the abnormal positioning module comprises: The performance curve and historical quality data of the silicon gel material are obtained to establish a mapping relationship between the quality parameter and the defect degree; The quality distribution map is subjected to cycle counting processing to extract an equivalent change amplitude sequence; The equivalent change amplitude sequence is associated with the humidity data in the operating condition data to generate a condition-coupled defect accumulation factor; The quality defect source is located based on the defect accumulation factor.

[0010] Preferably, the warning generation module comprises: A warning rule is set, and the warning rule includes a parameter fluctuation range, a sensitivity adjustment, and a threshold division; Simulated data is generated according to historical normal quality data; The boundary features of normal and abnormal states are learned through a discriminant algorithm; The multi-level warning threshold changing with the application state of silica gel is generated by multi-level division of the discrimination result based on the boundary feature; The real-time quality parameter is compared with the multi-level warning threshold, and the warning level is dynamically adjusted.

[0011] Preferably, the system further comprises: An adaptive strategy module for adjusting the data acquisition frequency based on the real-time data quality condition; The adaptive strategy module sets data acquisition rules, and the data acquisition rules include sampling rate allocation, sensor switching and retransmission mechanism; The data acquisition strategy space is constructed by historical data mining and strategy optimization; The real-time data quality condition is used as a constraint to perform strategy matching in the data acquisition strategy space to obtain an optimized data acquisition strategy.

[0012] Preferably, the system further comprises: An encryption communication module for creating a secure data transmission channel; The encryption communication module configures a key distribution channel and an algorithm encryption channel; The key distribution channel and the algorithm encryption channel are set in parallel to realize the secure data transmission channel.

[0013] Preferably, the present application further comprises a quality control method based on single-component adhesive silica gel, which comprises all modules and method processes of the above-mentioned quality control system based on single-component adhesive silica gel.

[0014] Compared with the prior art, the present application has the following advantages: The multi-source data acquisition module obtains temperature, pressure, viscosity and curing time and other multi-source quality parameter data in the silica gel application process, breaking the limitation of traditional single parameter monitoring and realizing comprehensive coverage of key quality parameters in the silica gel application process. The collection of multi-source data provides rich basic information for subsequent quality analysis and evaluation, so that the judgment of the quality state of silica gel is no longer dependent on a single parameter, but can consider multiple key factors, which is more in line with the actual process of silica gel quality formation. The state estimation module performs fusion processing on the multi-source quality parameter data, which can mine the internal correlation between parameters and form a unified silica gel quality state estimation value. This fusion processing method avoids the problem of data isolation and inability to analyze correlation in single parameter monitoring, allowing operators to accurately grasp the quality state of silica gel as a whole, reducing misjudgment or omission due to scattered data, and improving the accuracy and reliability of quality state evaluation. The dynamic model construction module dynamically generates a silica gel bonding physical model according to the operating condition data, so that the model can adapt to different operating condition changes. In actual production, fluctuations in operating conditions are common phenomena, and the dynamic model can respond to these changes in real time, adjust the model parameters, and ensure that the model always remains consistent with the actual production situation. Compared with traditional fixed models, the dynamic model can more accurately reflect the actual situation of the silica gel bonding process, providing more reliable model support for subsequent quality calculation and analysis.

[0015] The quality calculation module combines the silica gel bonding physical model and multi-source quality parameter data, and uses a compensation algorithm to calculate the quality parameters of the unmonitored area, effectively solving the problem of quality control blind spots in the unmonitored area. Through scientific algorithms and reliable models, the quality parameters of the unmonitored area can be reasonably calculated, enabling operators to understand the quality status of these areas and promptly identify potential quality problems, avoiding cost waste and production delays due to quality problems in unmonitored areas, and achieving quality control of the entire area of the silica gel application process. The abnormal positioning module locates the quality defect source based on the calculated quality parameters of the unmonitored area, making it possible to quickly trace the source of quality problems. In traditional technology, locating the source of quality defects often requires a lot of time and effort, and it is difficult to accurately find the root cause. However, through the fusion analysis of multi-source data and the calculation of unmonitored area parameters, the system can clearly sort out the path and source of quality defects, helping operators quickly locate the problem and reduce the time cost of quality problem tracing, improving the efficiency of problem solving. The early warning generation module generates a hierarchical early warning signal when detecting a quality anomaly, and issues a corresponding level of warning for different severity of quality anomalies. Operators can quickly determine the urgency of the anomaly based on the warning level, and then take different measures. For minor anomalies, adjustments can be made without interrupting production; for serious anomalies, the relevant process can be stopped in time to avoid the problem from becoming larger. This hierarchical warning mechanism makes the handling of quality anomalies more targeted and timely, helping to maintain stable production and ensure product quality. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 The timing diagram of the quality control system based on single-component bonded silica gel described in the present application; Figure 2 The flowchart of data processing for the state estimation module; Figure 3 The flowchart of modeling for the dynamic model construction module; Figure 4 The flowchart of parameter calculation for the quality calculation module; Figure 5 The flowchart of early warning generation for the early warning generation module. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present application will be apparently and completely described with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all the other embodiments obtained by a person of ordinary skill in the art without any creative work fall within the protection scope of the present application.

[0018] Please refer to Figure 1 The present application provides a quality control system based on single-component adhesive silicone, which comprises: The multi-source data acquisition module is responsible for acquiring key quality parameters in the silicone application process, specifically including temperature data, pressure data, viscosity data, and curing time data. These data are derived from a sensor network deployed in the silicone coating area. The state estimation module receives raw data from the multi-source data acquisition module and performs fusion processing on it. This processing process involves data preprocessing and feature extraction, and finally outputs a silicone quality state estimation value that comprehensively reflects the overall quality status of the current silicone. The dynamic model construction module dynamically constructs a silicone bonding physical model based on externally input operating condition data. This model is not static, but is adjusted in real time with the change of operating conditions, to more accurately reflect the actual physical process. The quality calculation module combines the multi-source quality parameter data with the silicone bonding physical model generated by the dynamic model construction module, uses algorithms to calculate the silicone quality parameters in the areas not directly covered by the sensors, and thus constructs more complete quality distribution information. The abnormal positioning module analyzes the quality distribution map based on the quality parameters calculated by the quality calculation module for the unmonitored areas, combines historical data and material characteristics, and identifies and locates the specific source of quality defects. The early warning generation module continuously monitors the system state, and when it detects quality abnormalities, generates different levels of graded early warning signals according to preset rules, prompting the operator to take appropriate measures.

[0019] Example 1: Please refer to Figure 2In the silica coating process, the embodiments are embodied as a highly synergistic data sensing and processing flow. The system deploys a sensor array on the substrate surface of the silica coating, which is based on a deep understanding of the flow characteristics, heat conduction characteristics and stress distribution characteristics of the silica. The deployment of a subset of temperature sensors focuses on the high-temperature variable regions, which are usually located near the heat source or have a significant temperature gradient due to their own chemical reaction heat. For example, in a continuous coating production line, the edges of the silica strip are closer to the heating elements, and the heat dissipation conditions are different from the center area, so multiple temperature sensors are symmetrically installed at these positions to capture possible asymmetric heating or local overheating phenomena. A subset of pressure sensors is installed at the edges of the silica coating area, especially at the start and end points of the coating track and areas with large radii of curvature. These locations are prone to stress concentration due to material shrinkage and substrate constraints during the curing process, and are key points for monitoring interface bonding and internal stress. A subset of viscosity sensors is uniformly distributed in a grid pattern in the center area of the silica coating. The flow field in this area is relatively stable, and the viscosity data can more purely reflect the changes in the rheological properties of the material itself, avoiding the interference of edge effects. All sensors are connected to the data acquisition unit through an industrial bus network to synchronously collect the original readings of temperature, pressure and viscosity signals at a fixed initial frequency, forming an initial set of multi-source quality parameter data.

[0020] After the raw data is transmitted to the state estimation module, it undergoes a series of rigorous signal conditioning processes. The module first analyzes the noise characteristics of the continuously acquired sensor data. This analysis process identifies various interference components in the data, such as periodic fluctuations in the pressure signal caused by high-frequency electromagnetic interference from the production line motor, transient spikes in the thermocouple readings caused by sudden changes in ambient temperature, and transient outliers in the viscosity sensor caused by the passage of small air bubbles. By statistically analyzing long-time series data, the module calculates the amplitude range, main frequency components and their spatial and temporal distribution of each type of noise, thereby obtaining the noise characteristics of multi-source quality data. Based on these characteristics, the module sets differentiated multi-source quality data denoising thresholds for each type of sensor data. For signals dominated by high-frequency noise, the threshold focuses on the frequency domain filter threshold; for occasional spike noise, the threshold is determined based on the statistical outliers of the amplitude.

[0021] The module performs noise identification and filtering preprocessing according to these thresholds. A specific process is as follows: for the data stream from the temperature sensor, a median filter based on a time window is first applied to eliminate occasional glitches caused by power fluctuations; then, a low-pass filter with a cutoff frequency set according to the main noise frequency analyzed previously is used to smooth out high-frequency oscillations while preserving the true temperature trend. For pressure data, since the noise shows strong periodicity, the module enables an adaptive notch filter that can automatically identify and suppress power frequency interference at specific frequencies. The processing of viscosity data is more delicate because the value change is directly related to the material state. After using conventional filtering, the module also adds an outlier detection step based on moving standard deviation. If the standard deviation of a data point at a certain time relative to its neighboring points exceeds the set threshold, the point is determined to be noise and is removed or replaced. After these processes, the raw data is transformed into a set of clean and stable time series, i.e., standard multi-source quality parameter data.

[0022] The preprocessed standard data is sent to the feature extraction and fusion stage. Instead of simply using all data directly, the module extracts key features that represent the system state. For temperature data, the extracted features include but are not limited to: the average temperature of the current coating area, the difference between the maximum and minimum temperatures (temperature uniformity index), and the rate of temperature rise or fall within a certain period of time (curing reaction rate index). The features of pressure data include the average contact pressure, the standard deviation of pressure distribution (reflecting the uniformity of pressure stress), and the curve shape of the pressure change over time at specific points. Viscosity data extracts instantaneous viscosity value, viscosity change rate within a certain period of time, and parameters related to thixotropy characteristics. These feature vectors extracted from different physical quantities are sent to a data fusion unit. The unit normalizes and weights the combination of each feature parameter according to its weight in affecting the final quality state. The weight coefficient may come from the experience value in the historical process database, or be obtained by training a large number of samples through a machine learning model. The fused comprehensive feature vector is finally input into a state estimation algorithm. This algorithm may be a Kalman filter, whose system state variable is the silicone quality state estimate; or it may be a trained neural network model, whose output is a scalar between 0 and 1, and the higher the value, the better the quality state. Through this series of operations, the state estimation module converts the complex multi-sensor readings into a silicone quality state estimate value with clear physical meaning and engineering value.

[0023] Example 2: see Figure 3In the practical application of the silicone bonding process, the execution of the dynamic model construction module is a continuous process that tightly couples the external environment with the internal material response. The module continuously acquires real-time environmental temperature and humidity data from the line's environmental monitoring unit through a specific data interface. These data are typically provided by high-precision temperature and humidity sensors placed above or near the coating workstation and refreshed at a frequency of several times per minute or tens of times per minute. At the same time, the module acquires the operating parameters of the coating equipment from the motion control system, among which the coating speed data are the key input, which can be a set value or an actual measured value fed back by an encoder, reflecting the linear speed at which the silicone is extruded and applied to the substrate.

[0024] After obtaining these operating condition data, the module starts its core calculation process. For the environmental temperature data, the module does not directly use its readings, but converts them into thermal input that affects the silicone material. The system internally stores or has access to a database of thermal-physical properties of the silicone material of this model, including its specific heat capacity, thermal conductivity, and solidification reaction heat, etc. The module uses these parameters, combined with the current environmental temperature and the geometric dimensions of the silicone coating (such as thickness, area), to estimate the heat exchange rate and total amount between the environment and the silicone through built-in heat transfer calculation rules. This process generates an equivalent thermal load, which is a comprehensive indicator that quantifies the heat gained or lost by the silicone per unit time and per unit volume under the current environmental conditions, thereby affecting its temperature change and solidification reaction kinetics.

[0025] For the coating speed data, the module's analysis focuses on its disturbance to the rheological state of the silicone. The stability of the coating speed is directly related to the leveling process and the formation of the internal structure of the silicone after it is extruded. The module continuously monitors the timing data of the coating speed, calculates its average value, standard deviation, and whether there is a trend drift within a time window. A sudden speed increase will cause the silicone to shear thin, and the apparent viscosity will instantaneously decrease, while fluctuations in speed will introduce flow instability, which may leave small air bubbles in the gel or cause uneven thickness. The module converts the amplitude and frequency of speed fluctuations into a dimensionless or physically meaningful flow fluctuation value through a simplified model based on the principles of fluid mechanics. This value represents the intensity of kinetic energy disturbance introduced into the silicone system by mechanical motion.

[0026] The module needs to integrate these two factors from different physical domains: thermal load and flow fluctuation. This is not a simple numerical superposition, but to construct a dynamic load matrix that can reflect the coupling effect of the two. The rows and columns of the matrix may correspond to different influence dimensions and time steps, respectively. For example, one dimension of the matrix represents the components of thermal load at different positions in space, and the other dimension represents the shear stress changes caused by flow fluctuations, and the values of the matrix elements represent the size of the specific type of load at a specific time. The synthesis operation may involve looking up a coupling relationship table previously calibrated through experiments or high-fidelity simulations, or it may be based on a series of empirical rules to weight and combine the effects of the two loads, thereby generating a dynamic load matrix that comprehensively describes the current external conditions on the silicone system.

[0027] This dynamic load matrix is input into the reference model of the system, which is established offline based on the basic material parameters of the single-component silicone, such as its viscosity curve with temperature and shear rate, solidification reaction kinetics equation (describing the relationship between solidification degree and time, temperature), and volume shrinkage characteristics during solidification, etc. The reference model can be a set of parameterized physical equations or a simplified but computationally efficient numerical model framework. The role of the dynamic load matrix is to inject "real-time" driving conditions into this static reference model. The model adjusts its internal temperature field and reaction rate calculation according to the input thermal load, and adjusts its rheological property parameters and stress field calculation according to the flow fluctuation value. After the model is solved, a real-time updated model instance is output that can reflect the physical behavior prediction of the silicone bonding process under the current specific working conditions, i.e. the silicone bonding physical model. This model is no longer universal, but is tailored for this silicone coating line at this moment, under this environment, and its predicted temperature distribution, viscosity change, stress development, and final solidification state will be closer to the actual situation, providing a high-precision theoretical basis for subsequent quality calculation. The whole process is closed-loop and adaptive, as the working condition data continues to flow in, the dynamic load matrix is constantly updated, and the silicone bonding physical model is refreshed accordingly, forming a dynamic modeling capability that can track changes in production conditions.

[0028] Example 3: see Figure 4, focusing on two closely linked processes: quality estimation and anomaly localization. The core task of the quality estimation module is to infer the complete and continuous quality state distribution of the entire silicone coating area using limited sensor data by combining physical models. The module first receives standard multi-source quality parameter data from the state estimation module, which has been pre-processed to remove noise interference and represents reliable temperature, pressure, and viscosity readings at monitoring points. The module extracts the quality parameter difference between adjacent monitoring points from these data. For example, for temperature data, calculate the temperature difference between the positions of sensor A and sensor B; for viscosity data, calculate the viscosity change between the positions of sensor C and sensor D. These differences directly reflect the rate of change of quality parameters in space. Based on these differences between adjacent point pairs, the module further calculates the local area change gradient feature. This feature not only contains the size of the change rate, but also may contain directional information, which describes the spatial change trend of the quality parameter within the sensor network coverage.

[0029] The module compares the actual change gradient feature calculated with the theoretical gradient predicted by the silicone bonding physical model generated by the dynamic model construction module in real time. The theoretical gradient is calculated by the model based on the current input operating condition data and material constitutive relationship, and is the spatial distribution form that the silicone parameters should present under ideal conditions. By comparing the actual observation and theoretical prediction, the module calculates a quality deviation coefficient, which quantifies the consistency of the model prediction and the measured data in the local area. This coefficient becomes the key indicator driving the model correction. The module iteratively adjusts the material parameters or boundary conditions in the silicone bonding physical model corresponding to the unmonitored area according to the quality deviation coefficient. These parameters may include local thermal conductivity, solidification reaction rate constant, or rheological parameters. The adjustment process uses an optimization algorithm to minimize the overall error between the model prediction and the measured values of all monitoring points as the objective function. Iteration continues until the error between the model's predicted output and the measured data of known points converges to a pre-set, acceptable tolerance range. When the iteration converges, the model at this moment is considered to have been well calibrated and can truly reflect the current actual process. The module then runs the calibrated model to output predicted quality parameters for the entire silicone coating area, such as predicted temperature field, predicted viscosity field, and predicted solidification degree field, and constructs a complete and high-resolution spatial quality distribution map based on these parameters.

[0030] The anomaly localization module then performs a deep analysis on this quality distribution map to identify and localize potential defect sources. The module first accesses the database to obtain the intrinsic performance curves of the silicone material, such as its curing kinetics curve (the relationship between the degree of cure and the time-temperature history), the viscosity-temperature curve, and its mechanical property development curve. Meanwhile, the module calls historical quality data, which contains the normal product parameter range recorded in past production and the characteristic data of known defect cases. Based on these information, the module establishes a mapping relationship between the quality parameters and the defect degree. This relationship can be expressed through a function:

[0031] wherein: represents the defect degree, which is a comprehensive index, and the higher the value, the higher the possibility or severity of defects; represents the mapping function relationship obtained through historical data analysis and machine learning training; represents the local temperature; represents the local viscosity; represents the local degree of cure; represents the local shear rate; represents the local humidity history influence factor. This function combines multiple quality parameters to evaluate the degree of deviation of their combined state from the normal range.

[0032] The module performs a cycle count process on the quality distribution map, which aims to quantify the fluctuations of the quality parameters (such as the initial local defect degree calculated using the above function) in the spatial domain. The cycle count analysis can identify the peak, valley, and fluctuation frequency in the parameter distribution, thereby extracting an equivalent variation amplitude sequence that describes the fluctuation strength and distribution characteristics of the quality anomaly in space. Then, the module performs a correlation operation on the extracted equivalent variation amplitude sequence and the humidity data in the operating condition data. Humidity is a key environmental factor that affects the curing process and final performance of silicone, and the correlation operation aims to analyze the statistical correlation or physical causal relationship between humidity changes and quality parameter fluctuations, thereby generating a defect accumulation factor coupled with operating conditions. This factor integrates the influence of environmental humidity and the response of the material itself, more accurately representing the driving force for defect generation. Finally, the module performs a scanning analysis on the quality distribution map based on the calculated defect accumulation factor and the previously established defect mapping relationship, localizing the spatial regions where the defect accumulation factor is significantly higher than the background value. These regions are identified as quality defect sources. The localization results can be specific to a line segment or a region on the coating track, providing clear targets for subsequent process adjustment and quality repair. The entire process from data inference to defect identification forms a complete analysis chain from macroscopic distribution to microscopic localization.

[0033] Example 4: refer toFigure 5 First, a framework for early warning rules is defined, which includes three dimensions: parameter fluctuation range, which specifies the upper and lower limits of each quality monitoring indicator. For example, the temperature monitoring point is allowed to fluctuate within ±5℃ of the set value, the upper limit of pressure fluctuation in the edge area is 15% of the set value, and the viscosity change rate in the center area must not exceed 3% per minute; sensitivity adjustment allows for different detection sensitivities to be set for different process stages or key areas, such as relaxing the viscosity monitoring threshold in the initial coating stage of silicone, while increasing the sensitivity of temperature monitoring when approaching the curing critical point; threshold division establishes a three-level early warning system, classifying abnormal states into three levels: attention (level 1), warning (level 2), and serious (level 3). For example, when the temperature in a certain area exceeds the limit by 5℃ but does not reach 8℃ for 3 consecutive minutes, a level 1 warning is triggered; exceeding the limit by 8℃ triggers level 2; exceeding the limit by 10℃ or the presence of multiple linked abnormalities triggers level 3.

[0034] To accurately define the boundary between normal and abnormal states, the module accesses normal production records from the historical database. These records contain thousands of time-series data points on parameters such as temperature, pressure, and viscosity collected during the production of qualified products. Based on this real historical data, the module employs data augmentation techniques to generate an expanded dataset. This process preserves the statistical distribution characteristics of the original data (such as mean, variance, and autocorrelation) and generates new sequences that conform to the process characteristics through a time-series synthesis algorithm. See Table 1, which shows the simulated temperature data fragments generated by the module.

[0035] Table 1: Temperature data fragments of the simulated silicone coating area.

[0036] Time stamp (min) Zone A temperature (°C) Zone B temperature (°C) Zone C temperature (°C) Zone D temperature (°C) T+0.0 75.2 76.8 74.5 75.9 T+0.5 75.6 77.1 74.8 76.2 T+1.0 76.0 77.3 75.1 76.5 T+1.5 76.3 77.6 75.4 76.8 T+2.0 76.5 77.8 75.6 77.0 ... ... ... ... ... T+10.0 82.1 83.5 81.0 82.7 These simulated data, together with real historical normal data, constitute the training sample set, which is input into the state discrimination algorithm. The algorithm employs a multi-dimensional feature space analysis method to learn the correlation patterns and variation laws among various parameters under normal production conditions. For example, the algorithm can identify a specific covariant relationship between the viscosity rise rate in the central region and the temperature gradient in the edge region during normal curing; or what shape shift the temperature rise curves at each monitoring point should exhibit when the ambient humidity increases. By analyzing thousands of normal samples, the algorithm constructs a multi-dimensional feature boundary characterizing the "healthy" state.

[0037] Based on the learned state boundary features, the early warning generation module performs multi-level threshold division. The division process is based on the degree of abnormal deviation: the first-level threshold corresponds to the buffer zone in the feature space that is close to the normal boundary, where parameter fluctuations are within an acceptable range but require attention; the second-level threshold corresponds to the moderate deviation zone, indicating an abnormal tendency that requires intervention; and the third-level threshold corresponds to the severe abnormal zone that is far from the normal zone. These thresholds are not fixed values, but rather dynamic surfaces that adjust with the process progress. For example, in the early stage of the curing reaction, the temperature control threshold range is relatively wide; when the system detects that the degree of curing has reached 50%, it automatically narrows the allowable temperature fluctuation range and increases the sensitivity of pressure monitoring; if the ambient humidity sensor detects a sudden increase in humidity, the viscosity change rate threshold is simultaneously widened to compensate for the influence of humidity.

[0038] During real-time operation, the early warning generation module continuously receives silicone state parameters from the quality estimation module. Every second, the system spatially matches and compares the latest acquired temperature distribution map, viscosity change rate matrix, and pressure field data of the entire coating area with the multi-level early warning threshold surface currently in effect. The matching process employs parallel computing, with the parameter combinations of each monitoring unit (including actual sensor points and estimated virtual points) located in the feature space. When a parameter combination in a certain area falls into the first-level early warning zone, the system marks the area as yellow and records the offset; if the parameter exceeds the second-level threshold, it is marked as orange and a speed-down command is triggered; when multiple points exceed the third-level threshold or cross-parameter coupling anomalies occur (such as high temperature points accompanied by abnormally low viscosity), it is marked as red and an emergency shutdown protocol is executed. The early warning signal is displayed in real time through the human-machine interface, with different colors covering the coating area schematic diagram to intuitively show the location and level of the anomaly, while simultaneously sending graded alarm codes to the control terminal. The entire process forms a closed loop: new quality data is continuously input, the threshold surface is dynamically updated according to changes in process status and environment, and the early warning level is adjusted in real time according to the real-time matching results, achieving accurate identification and graded response to silicone bonding quality risks.

[0039] Example 5: During the operation of the silicone bonding quality control system, the adaptive strategy module continuously monitors the efficiency of the entire data acquisition process. This module has a built-in data acquisition rule framework, which includes three core operational dimensions: sampling rate allocation rules define the data acquisition frequency adjustment strategy for different sensor types or different monitoring areas; sensor switching rules specify the logic for enabling backup sensors or disabling some sensors under specific conditions; and the retransmission mechanism specifies the automatic retransmission count and priority strategy when data transmission fails. These rules together constitute the basic constraints on data acquisition behavior.

[0040] To build a foundation for decision-making, the adaptive strategy module performs in-depth analysis of historical operational data. The analysis includes sensor readings, communication logs, and corresponding operating condition records from the past few months or years. The analysis process identifies several effective data acquisition patterns: for example, reducing the sampling frequency of pressure sensors in peripheral areas during periods of stable ambient temperature; automatically doubling the sampling rate of sensors in the central area when increased viscosity data fluctuations are detected; and prioritizing temperature data transmission while temporarily delaying the transmission of some historical data during periods of communication channel congestion. By quantitatively evaluating the effectiveness of these historical strategies (e.g., changes in data integrity indicators, bandwidth utilization, and state estimation accuracy), the module selects strategy combinations that perform well in different scenarios, constructing a data acquisition strategy space containing dozens of feasible solutions.

[0041] During system operation, the adaptive strategy module receives real-time data quality assessment feedback from each stage. This real-time data quality status includes: packet loss rate of key sensors, transmission delay statistics, signal-to-noise ratio analysis of data collected from each region, and data validity scores from the state estimation module. These indicators comprehensively reflect the current health status and information value of the data acquisition system. The module uses these real-time conditions as hard constraints to perform rapid search and matching within a pre-built data acquisition strategy space. The matching process employs multi-objective optimization logic: under the premise of meeting minimum data quality requirements (e.g., packet loss rate of key area temperature data <1%, latency of central viscosity data <100ms), the strategy with the lowest system resource consumption is prioritized; when a sharp decline in data quality is detected in a specific area (e.g., noise exceeding the standard of a pressure sensor), a highly robust strategy is switched (e.g., enabling redundant sensors and initiating data verification and retransmission). The optimization process outputs a specific set of optimized data acquisition strategy instructions, such as: "Increase the sampling rate of the temperature sensor in the high-temperature volatile area from 10Hz to 20Hz; disable pressure sensor No. 7 in the stress concentration area and enable its backup sensor; enable a three-time retransmission mechanism for a subset of viscosity sensor data." These instructions are sent to the multi-source data acquisition module in real time for execution, enabling dynamic allocation and fault-tolerant processing of acquisition resources.

[0042] The encrypted communication module operates independently at the data transmission layer, providing security for data exchange between internal modules and with an external monitoring center. This module establishes two parallel core channels: a key distribution channel and an algorithm encryption channel. The key distribution channel is responsible for the lifecycle management of the keys required for encryption. This channel employs a two-way authentication mechanism, generating a session key through secure negotiation between the communicating parties using an asymmetric encryption algorithm during system startup or periodic updates. Key materials are stored in a secure hardware area with strict access control policies. This channel also defines rules for periodic key rotation and an emergency invalidation process to prevent risks associated with the long-term use of a single key.

[0043] The algorithm encryption channel is responsible for protecting the confidentiality and integrity of the actual transmitted data. This channel uses a symmetric encryption algorithm to process all transmitted data streams, including raw sensor readings, state estimates, and warning signals. The specific process is as follows: the sending end obtains the currently valid session key, performs encryption operations on the plaintext data to generate ciphertext, and simultaneously calculates the message authentication code for the data and appends it to the ciphertext; the receiving end, after obtaining the ciphertext, first verifies the validity of the message authentication code to confirm that the data has not been tampered with, and after successful verification, decrypts it using the same key to recover the original data. Two channels operate in parallel: the key distribution channel silently manages the key in the background, only performing low-frequency secure interactions when updates are needed; the algorithm encryption channel continuously processes high-frequency business data streams, and the two work together through a secure interface. This architecture ensures that even if a transmission from the encryption channel is intercepted, an attacker cannot crack the content without a valid key; at the same time, the transmission of the key itself is highly protected. The entire communication process is transparent to all application layer modules; data is automatically encrypted when sent and automatically decrypted and verified when received, forming an end-to-end secure data transmission channel, ensuring the secure flow of silicone bonding quality data throughout the entire processing chain.

[0044] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0045] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A quality control system based on single-component adhesive silicone, characterized in that, The system includes: The multi-source data acquisition module is used to acquire multi-source quality parameter data during the silicone application process. The multi-source quality parameter data includes temperature data, pressure data, viscosity data, and curing time data. The state estimation module is used to fuse the multi-source quality parameter data to determine the estimated value of the silicone quality state. The dynamic model building module is used to dynamically generate a physical model of silicone bonding based on operating condition data. The quality estimation module is used to combine the silicone bonding physical model and the multi-source quality parameter data, and use a compensation algorithm to estimate the quality parameters of the unmonitored area. Anomaly localization module is used to locate the source of quality defects based on the calculated quality parameters of the unmonitored area; The early warning generation module is used to generate graded early warning signals when a quality anomaly is detected.

2. The quality control system based on single-component adhesive silicone according to claim 1, characterized in that, The multi-source data acquisition module includes: A sensor array is arranged at a predetermined position on the silicone-coated surface. The sensor array includes a subset of temperature sensors, a subset of pressure sensors, and a subset of viscosity sensors. The temperature sensor subset is symmetrically distributed in the high-temperature and variable area of ​​the silicone coating region, the pressure sensor subset is arranged in the stress concentration area at the edge of the silicone coating, and the viscosity sensor subset is uniformly distributed in the central area of ​​the silicone coating. The temperature data, pressure data, and viscosity data are collected in real time by the multi-source data acquisition module.

3. A quality control system based on single-component adhesive silicone according to claim 2, characterized in that, The state estimation module includes: Noise characteristics analysis is performed on the multi-source quality parameter data to obtain the noise characteristics of the multi-source quality data; Based on the noise characteristics of the multi-source quality data, a denoising threshold for the multi-source quality data is set; The multi-source quality parameter data is subjected to noise identification and filtering preprocessing according to the multi-source quality data denoising threshold to obtain standard multi-source quality parameter data. Feature extraction and fusion, and quality status estimation are performed on the standard multi-source quality parameter data to determine the estimated value of the silicone quality status.

4. A quality control system based on single-component adhesive silicone according to claim 3, characterized in that, The dynamic model construction module includes: The operating condition data is obtained from an external system, including ambient temperature data, humidity data, and coating speed data. An equivalent thermal load is generated based on the ambient temperature data, and a flow fluctuation value is calculated based on the coating speed data. The equivalent thermal load and the flow fluctuation value are combined to generate a dynamic load matrix; The dynamic load matrix is ​​input into a benchmark model established based on silicone material parameters to generate the silicone bonding physical model.

5. A quality control system based on single-component adhesive silicone according to claim 4, characterized in that, The quality estimation module includes: Extract the quality parameter difference between adjacent monitoring points from the standard multi-source quality parameter data, and calculate the change gradient characteristic of the local area; The quality deviation coefficient is calculated by comparing the changing gradient feature with the theoretical gradient of the silicone bonding physical model. The material parameters of the unmonitored area are iteratively adjusted according to the mass deviation coefficient until the error between the predicted mass parameters and the measured mass parameters of the silicone bonding physical model converges to the set range, and the predicted mass parameters are output to construct a mass distribution map.

6. A quality control system based on single-component adhesive silicone according to claim 5, characterized in that, The anomaly location module includes: Obtain the performance curves and historical quality data of silicone materials, and establish a mapping relationship between quality parameters and defect rate; The mass distribution spectrum is subjected to cyclic counting processing to extract the equivalent change amplitude sequence; The equivalent change amplitude sequence is correlated with the humidity data in the operating condition data to generate a condition-coupled defect accumulation factor. The source of the quality defect is located based on the defect accumulation factor.

7. A quality control system based on single-component adhesive silicone according to claim 6, characterized in that, The early warning generation module includes: Set early warning rules, which include parameter fluctuation range, sensitivity adjustment, and threshold division; Simulation data is generated based on historical normal quality data; The boundary features between normal and abnormal states are learned through a discriminative algorithm; Based on the boundary features, the discrimination results are divided into multiple levels to generate multi-level early warning thresholds that change with the application status of silicone. The real-time quality parameters are compared with the multi-level early warning thresholds to dynamically adjust the early warning level.

8. A quality control system based on single-component adhesive silicone according to claim 7, characterized in that, The system also includes: The adaptive strategy module is used to adjust the data acquisition frequency based on real-time data quality. The adaptive strategy module sets data acquisition rules, which include sampling rate allocation, sensor switching, and retransmission mechanisms. A data acquisition strategy space is constructed through historical data mining and strategy optimization. Using the real-time data quality status as a constraint, policy matching is performed within the data acquisition policy space to obtain an optimized data acquisition policy.

9. A quality control system based on single-component adhesive silicone according to claim 8, characterized in that, The system also includes: An encrypted communication module is used to create a secure data transmission channel; The encrypted communication module is configured with a key distribution channel and an algorithm encryption channel; The key distribution channel and the algorithm encryption channel are set up in parallel to realize the secure data transmission channel.

10. A quality control method based on single-component adhesive silicone, characterized in that, It includes all modules and method flows of a quality control system based on single-component adhesive silicone as described in any one of claims 1 to 9.

Images