A method for real-time analysis of deformation amount in a forming stage of a ceramic product

Abstract

The application discloses a real-time analysis method for deformation amount of ceramic product forming stage, and relates to the technical field of data analysis. The method comprises collecting deformation data of a green body in a forming and drying stage, including interlayer shrinkage, shear stress, edge displacement and environmental humidity. By analyzing the correlation between porosity distribution and shrinkage and shear stress, the water migration blockage level is evaluated, and the interlayer shear stress growth trend is predicted. Then, combined with the edge displacement and humidity data, the layered crack risk position is identified. Finally, the data is integrated to predict the interlayer bonding quality attenuation path, and a warning signal is generated. The method solves the problems of hidden interlayer debonding and layered cracking caused by porosity difference, uneven water migration and shear stress accumulation during the forming and drying process of multi-layer composite green body, realizes real-time perception, risk dynamic quantification and early warning of the whole link from micro-block mechanism to macro-bonding quality.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, and in particular to a method for real-time analysis of deformation during the forming stage of ceramic products. Background Technology

[0002] The forming stage of ceramic products directly affects the structural stability of the green body and the service life of the final product. In the production of multi-layer composite green bodies, the bonding quality between layers is a key factor affecting the overall strength.

[0003] Existing technologies typically improve interlayer bonding by adjusting pressure or optimizing clay formulation, but these methods overlook the subtle differences in material properties between different layers. When the porosity of each clay layer differs significantly, simply increasing pressure or changing the formulation is insufficient to balance moisture migration and shrinkage coordination.

[0004] There is an inherent contradiction between interlayer compression force and moisture migration. While increasing the compression force can make the interlayer bond tighter, it will reduce the local porosity of the green body, making it difficult for moisture to migrate outward through the less porosity layers. Taking a three-layer composite green body as an example, when the porosity of the middle layer is low and the porosity of the upper and lower layers is high, the middle layer acts as a barrier to hinder moisture flow. The upper and lower layers shrink faster due to moisture evaporation, while the middle layer shrinks slower. This difference in shrinkage rate between the layers generates shear stress. This stress first accumulates at the edges of the green body, forming micro-cracks that gradually propagate inward, eventually leading to interlayer separation.

[0005] The root cause of the above contradiction lies in the lack of real-time monitoring and analysis capabilities for the deformation of the billet during the forming and drying process. Because it is impossible to accurately grasp the shrinkage state of each layer at different times, adjustments to process parameters can only rely on experience, making it difficult to achieve a dynamic balance between interlayer pressing force and moisture migration. Consequently, it is impossible to effectively prevent interlayer defects caused by inconsistent shrinkage. Summary of the Invention

[0006] To address the technical problems mentioned above, this invention discloses a method for real-time analysis of deformation during the forming stage of ceramic products, comprising the following steps: Step S1: Obtain deformation monitoring data of the multi-layer composite preform during the forming and drying stages. The deformation monitoring data includes real-time shrinkage rate readings between adjacent layers, interlayer interface shear stress readings, edge micro-displacement readings, and ambient humidity readings. Step S2: Identify the porosity distribution characteristics of each layer of mud based on the deformation monitoring data, and evaluate the blocking strength of the porosity distribution on the interlayer moisture migration channel by combining the correlation between the real-time shrinkage rate reading and the interlayer interface shear stress reading, and determine the moisture migration blocking level under the action of high pressure. Step S3: Correlation analysis is performed on the moisture migration blocking level, the real-time shrinkage rate reading, and the interlayer interface shear stress reading to evaluate the cumulative increase of interlayer shear stress caused by the uncoordinated shrinkage of each layer during the drying stage, and to obtain the growth trend of interlayer shear stress. Step S4: Based on the interlaminar shear stress growth trend, combined with the edge micro-displacement reading and the ambient humidity reading, identify the initial risk location of the delamination crack in the edge region and generate a list of potential debonding locations. Step S5: Integrate the deformation monitoring data according to the potential debonding location list, predict the interlayer bonding quality decay path of the multilayer composite preform, generate an early warning signal and output it to the molding monitoring system.

[0007] Furthermore, in step S2, identifying the porosity distribution characteristics of each layer of clay includes: Extract the real-time shrinkage rate readings of each layer from the deformation monitoring data, calculate the shrinkage rate difference between adjacent layers, and determine the porosity layer type based on the comparison result of the shrinkage rate difference with the preset shrinkage threshold. The porosity layer type includes high porosity layer and low porosity layer. Calculate the porosity gradient value between adjacent layers. When the porosity gradient value exceeds a preset gradient threshold, it is determined that there is a risk of water migration channel blockage at the corresponding interlayer interface. Based on the distribution range of the porosity gradient values ​​and the fluctuation range of the interlayer interface shear stress readings, the porosity distribution characteristics of each interlayer interface are determined.

[0008] Furthermore, the evaluation of the cumulative increase in interlaminar shear stress in step S3 includes: Obtain interlayer interface shear stress readings at multiple time points during the drying process, and calculate the shear stress increment between adjacent time points; The shear stress increment value is matched with the real-time shrinkage rate reading at the corresponding time point, and the correlation coefficient between the shrinkage rate change and the shear stress increment is analyzed. When the correlation coefficient exceeds the preset correlation threshold, it is determined that the interlayer interface is in a state of rapid shear stress accumulation, and the accumulation start time and accumulation rate are recorded. By integrating the cumulative rate data of each interlayer interface and combining it with the ambient humidity readings, the cumulative rate is corrected to obtain the corrected interlayer shear stress growth trend.

[0009] Furthermore, in step S4, identifying the initial risk location of the layered cracks in the edge region includes: A clustering algorithm is used to perform joint clustering processing on the edge micro-displacement readings and the interlayer shear stress growth trend, dividing the edge region into several stress and displacement feature clusters; Calculate the cluster center coordinates and the dispersion of data points within each feature cluster, and mark feature clusters with dispersion exceeding a preset dispersion threshold as high-risk clusters; Extract the corresponding spatial coordinates from the high-risk clusters, and combine the environmental humidity readings to correct the risk level of each coordinate for humidity influence. A list of potential debonding locations is generated based on the corrected location coordinates. The list includes the spatial coordinates of the risk locations and the corresponding risk level identifiers.

[0010] Furthermore, the method also includes: The step of dynamically adjusting the debonding risk assessment threshold based on the porosity distribution characteristics is as follows: Extract the porosity gradient value corresponding to each risk location in the potential debonding location list; Calculate the distribution variance of the porosity gradient value. When the distribution variance exceeds a preset variance threshold, increase the value of the debonding risk judgment threshold. The increased debonding risk assessment threshold is applied to the risk location identification process at subsequent time points to achieve dynamic updating of the threshold.

[0011] Furthermore, it also includes: The crack propagation rate is extracted from the edge micro-displacement readings, and the extent of debonding region propagation is evaluated, specifically as follows: Obtain the edge micro-displacement reading sequence of each risk location in the potential debonding location list at consecutive time nodes; The edge micro-displacement reading sequence is subjected to time difference processing to calculate the displacement change per unit time, thereby obtaining the crack propagation rate. Based on the crack propagation rate and the spatial coordinates of the crack initiation position, the radial distance of crack propagation is calculated. The radial distance is converted into the expanded area of ​​the debonding region to obtain the expanded range of the debonding region.

[0012] Furthermore, the prediction of interlayer bonding mass degradation path in step S5 includes: A decision tree algorithm is used to construct an interlayer bonding quality decay prediction model. The input features of the prediction model include the real-time shrinkage rate reading, the interlayer interface shear stress reading, the edge micro-displacement reading, and the debonding region expansion range. The input features are organized on a time series to form a multidimensional time series feature vector; The multidimensional temporal feature vectors are classified using a decision tree algorithm, and the interlayer bonding quality is divided into multiple attenuation levels. Based on the decay level sequence at each time point, the quality decay path in the future time period is predicted. When the predicted path indicates that the quality will reach the failure level, an early warning signal is generated.

[0013] Furthermore, the joint clustering processing of the edge micro-displacement readings and the interlayer shear stress growth trend using a clustering algorithm specifically includes: The edge micro-displacement readings are combined with the interlaminar shear stress growth trend to form a two-dimensional feature vector; The two-dimensional feature vectors are normalized to eliminate dimensional differences; Set the number of clusters, and use the K-means algorithm to iteratively calculate the position of the center point of each cluster until the cluster centers converge. Calculate the Euclidean distance from each data point to its cluster center, and mark data points whose distance exceeds a preset distance threshold as outliers. The spatial location corresponding to the outlier is the high-risk location.

[0014] Furthermore, the dynamic adjustment of the debonding risk judgment threshold also includes: Establish a mapping function between porosity gradient values ​​and debonding risk assessment thresholds; According to the mapping function, when an increase in porosity gradient value is detected, the debonding risk judgment threshold is increased by a preset proportional coefficient. Record the time point and adjustment range of each threshold adjustment to form a dynamic threshold adjustment history record; Based on the historical records of the threshold dynamic adjustment, the correlation between the threshold adjustment frequency and the final quality of the billet is analyzed, and the parameters of the mapping function are optimized.

[0015] Furthermore, the generation of the warning signal also includes: A visualization report is generated based on the quality degradation path. The visualization report plots a trend curve with time on the horizontal axis and interlayer combined quality degradation level on the vertical axis. Mark the deformation monitoring data values ​​and corresponding debonding risk levels at each key time point on the trend curve; When the trend curve predicts that the quality will reach the failure level within a preset time window, the multi-level alarm mechanism of the molding monitoring system is triggered. The multi-level alarm mechanism outputs alarm signals of different levels based on the predicted remaining time to reach the failure level; the shorter the remaining time, the higher the alarm level.

[0016] The present invention has the following beneficial effects: This invention discloses a real-time deformation analysis method for ceramic products during the forming stage. By deploying a shear stress probe and an edge micro-displacement acquisition device, shrinkage rate, interlayer interface shear stress, edge micro-displacement, and environmental humidity data are acquired in real time from adjacent layers of a multi-layer composite body, forming a complete deformation monitoring record for the forming stage. Based on this, the porosity distribution of each layer of clay is identified and its blocking strength on moisture migration channels is assessed, determining the moisture migration blocking level under high pressure. This level is compared with real-time shrinkage rate and shear stress readings to quantify the cumulative increase in interlayer shear stress during the drying stage, obtaining the growth amount. Then, combined with environmental humidity, a clustering algorithm is used to identify the probability of layered crack initiation in the edge region and form a list of potential initiation locations. Based on this list and porosity distribution, the debonding risk judgment threshold is dynamically adjusted. Next, the crack propagation rate is extracted from the edge micro-displacement to assess the expansion range of the debonding area. Finally, the debonding expansion range, interlayer shear stress, and shrinkage rate data are integrated, and a decision tree algorithm is used to predict the overall interlayer bonding quality decay path, generating a visual report and triggering an alarm. The core of this invention is to solve the problem of hidden interlayer debonding and delamination cracks caused by differences in porosity, uneven moisture migration and accumulation of shear stress in multi-layer composite green bodies during the forming and drying process. It realizes real-time perception, dynamic risk quantification and early warning of the whole link from microscopic blocking mechanism to macroscopic bonding quality, which greatly improves the quality stability and yield of complex multi-layer ceramic products in the forming stage. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the overall steps of a method for real-time analysis of deformation during the forming stage of ceramic products, as described in this embodiment of the invention. Figure 2 This is a detailed flowchart of step S2 in an embodiment of the present invention; Figure 3 This is a detailed flowchart of step S3 in an embodiment of the present invention. Detailed Implementation

[0018] The following specific examples illustrate the implementation of this disclosure. Those skilled in the art can easily understand other advantages and effects of this disclosure from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. This disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this disclosure. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0019] This embodiment provides an optimal implementation of a method for real-time monitoring of interlayer bonding quality during the forming stage of ceramic products. This method addresses the problems of interlayer debonding and delamination cracks caused by differences in clay porosity during the forming and drying stages of multi-layer composite bodies. Through the collection and intelligent analysis of multi-dimensional deformation monitoring data, it achieves real-time early warning of interlayer bonding quality. The following is in conjunction with the appendix... Figures 1 to 3 This embodiment will be described in detail.

[0020] like Figure 1 As shown, in step S1, deformation monitoring data of the multi-layer composite preform during the forming and drying stages are obtained. The deformation monitoring data includes real-time shrinkage rate readings between adjacent layers, interlayer interface shear stress readings, edge micro-displacement readings, and ambient humidity readings.

[0021] Furthermore, during the molding stage, the molding pressure value applied to the blank is recorded in real time by the pressure sensor on the press; before the drying stage, the residual stress value at the end of molding is recorded as the initial internal compressive stress; the molding pressure reading is added to the residual stress value to obtain the high pressure resultant force parameter.

[0022] In the forming process of multi-layer composite green bodies, a dedicated sensor array system is first deployed at key monitoring locations on the green body. Specifically, a thin-film shear stress probe is pre-embedded at the interface between adjacent layers. This probe uses the piezoresistive sensing principle; when relative slippage or shearing occurs between layers, the strain-sensitive element inside the probe generates a change in resistance, which is converted into a voltage signal by a Wheatstone bridge circuit, and then converted into a digital signal by a high-precision analog-to-digital converter. Through a pre-established calibration curve, the voltage value is mapped to a shear stress value in megapascals, thereby obtaining the shear stress reading at the interlayer interface. This embedded deployment method ensures that the probe is in close contact with the interface between the clay layers, enabling real-time capture of stress changes caused by relative movement between layers under forming pressure and during drying shrinkage, providing first-hand mechanical data for subsequent evaluation of the interlayer bonding state.

[0023] Furthermore, a laser displacement sensor array is fixedly installed at the edge region of the billet. These sensors employ the triangulation principle, emitting laser beams towards the edge surface of the billet and calculating the minute displacement at the edge by receiving the positional changes of the reflected light spots on a photosensitive array. The sensors achieve micrometer-level measurement accuracy, capable of capturing minute edge deformations caused by uneven moisture evaporation during the drying stage. The system continuously records the displacement values ​​at each measuring point at a preset sampling frequency, forming a time series of edge micro-displacement readings. Since the edge region is often where interlayer stress concentration and crack initiation first occur, these readings are of significant indicative value for identifying potential debonding risks. Time-series analysis of the edge micro-displacement readings can reveal abrupt displacement changes or accelerated growth patterns, which often indicate the initiation and propagation of internal cracks.

[0024] Meanwhile, to monitor the shrinkage behavior of each layer of the billet in real time, contact or non-contact thickness measuring devices are arranged at multiple representative locations within the billet plane. The contact devices utilize the principle of a linear variable differential transformer, reflecting layer thickness changes by measuring the displacement of the probe; the non-contact devices employ ultrasonic or eddy current sensing technology, estimating thickness through changes in signal propagation time or sensing characteristics. The system continuously records the thickness evolution data of each layer during the forming and drying processes. The difference between the current thickness and the initial thickness is divided by the initial thickness and multiplied by a percentage factor to obtain the real-time shrinkage rate reading for each layer. This reading directly reflects the degree of volume shrinkage of the clay during the dehydration process. The inconsistent shrinkage rates caused by differences in porosity among different layers are the fundamental cause of interlayer shear stress accumulation. Through correlation analysis between real-time shrinkage rate readings and interlayer interface shear stress readings, the causal relationship between shrinkage incoordination and stress accumulation can be revealed.

[0025] In addition, integrated temperature and humidity sensors are deployed around the billet to collect real-time humidity readings in the forming workshop and drying kiln. Ambient humidity is a key external factor affecting the rate of moisture evaporation from the clay. High humidity slows surface water loss and reduces the shrinkage gradient; low humidity accelerates surface drying, potentially increasing the difference in shrinkage between the surface and inner layers. By integrating ambient humidity readings with other deformation monitoring data, the modulating effect of the external environment can be considered when assessing interlayer bonding quality, thereby improving the accuracy of risk assessment. The output signals from all sensors are aggregated to a central data acquisition system via industrial Ethernet or wireless transmission modules. The system assigns a precise timestamp to each data set, forming a multi-dimensional, time-synchronized deformation monitoring dataset.

[0026] Through the coordinated operation of the aforementioned sensor array, complete deformation monitoring data was obtained, including real-time shrinkage rate readings between adjacent layers, interlayer interface shear stress readings, edge micro-displacement readings, and environmental humidity readings. This comprehensive monitoring method, involving multiple physical quantities and spatial locations, overcomes the limitations of traditional single-parameter monitoring and lays a solid foundation for subsequent data-driven interlayer bonding quality analysis.

[0027] Through real-time, multi-dimensional data acquisition, the dynamic deformation characteristics of multi-layer composite preforms during the forming and drying stages can be fully captured, providing reliable data support for accurately identifying porosity distribution differences, assessing the strength of moisture migration blocking, and predicting debonding risks.

[0028] Step S2: Identify the porosity distribution characteristics of each layer of mud based on the deformation monitoring data, and evaluate the blocking strength of the porosity distribution on the interlayer moisture migration channels by combining the correlation between the real-time shrinkage rate reading and the interlayer interface shear stress reading, and determine the moisture migration blocking level under high pressure.

[0029] like Figure 2 As shown, step S2, identifying the porosity distribution characteristics of each layer of mud, includes: extracting the shrinkage rate difference value of each layer from the deformation monitoring data, identifying the layer with higher shrinkage rate as a low-porosity layer with higher density, and identifying the layer with lower shrinkage rate as a high-porosity layer with higher looseness; calculating the porosity gradient value between adjacent layers, and determining that there is a risk of water migration channel blockage at the corresponding interlayer interface when the porosity gradient value exceeds a preset gradient threshold; and determining the porosity distribution characteristics of each interlayer interface based on the distribution range of the porosity gradient value and the fluctuation amplitude of the interlayer interface shear stress reading.

[0030] Based on the deformation monitoring data collected in step S1, the porosity distribution characteristics of each layer of clay are first identified. During the ceramic body forming process, the porosity of the clay refers to the proportion of pore volume to the total volume within the material. This parameter directly affects the material's density, moisture storage capacity, and drying shrinkage behavior. Because clay layers with different formulations or different laying processes have different initial porosities, this difference leads to inconsistencies in the water loss rate and shrinkage degree of each layer during the drying stage, thereby causing interlayer stress mismatch.

[0031] Specifically, real-time shrinkage rate readings of each layer at the same time point are extracted from deformation monitoring data, and the difference in shrinkage rate between adjacent layers is calculated. This difference is obtained by subtracting the shrinkage rate of the lower layer from the shrinkage rate of the upper layer; a positive value indicates that the upper layer shrinks faster, and a negative value indicates the opposite. According to materials science principles, the shrinkage rate of clay during the drying process is positively correlated with its initial porosity: loose layers with higher porosity contain more free water and capillary water, resulting in greater volume shrinkage after water loss; while dense layers with lower porosity contain relatively less water and shrink less. Based on this mapping relationship, layers with higher shrinkage rate readings are identified as high-porosity layers with higher porosity, and layers with lower shrinkage rate readings are identified as low-porosity layers with higher density. This method of inferring microstructure from macroscopic deformation avoids the need for destructive sampling tests on the green body during the forming stage, achieving non-destructive real-time identification of porosity distribution.

[0032] Furthermore, the shrinkage rate difference between adjacent layers is calculated and compared with a preset shrinkage threshold. When the shrinkage rate difference exceeds the preset shrinkage threshold, it is determined that there is a significant difference in porosity between the upper and lower layers, and the porosity layer type of each layer needs to be further determined. Specifically, among adjacent layers where the shrinkage rate difference exceeds the threshold, the layer with the higher shrinkage rate is determined to be a high-porosity layer, and the layer with the lower shrinkage rate is determined to be a low-porosity layer. When the shrinkage rate difference does not exceed the preset shrinkage threshold, it is determined that the porosity difference between adjacent layers is small, and the risk of interlayer moisture migration blockage is low.

[0033] Specifically, the preset shrinkage threshold is determined based on the characteristics of the clay and process parameters, with a typical range of 0.5%-2.0%. The specific method for determining the value is as follows: a drying experiment is conducted on several batches of green bodies, the shrinkage rate data of each layer is recorded, the distribution of the difference in shrinkage rate between adjacent layers of the batches that experienced interlayer debonding is statistically analyzed, and a critical value that can effectively distinguish between normal batches and debonding batches is selected as the threshold. Typically, the 25th percentile of the distribution of the difference in shrinkage rate between debonding batches is taken.

[0034] Furthermore, the porosity gradient value between adjacent layers is calculated. This gradient value is defined as the difference between the estimated porosity values ​​of the upper and lower layers divided by the interlayer distance, quantifying the rate of change of porosity in space. When the porosity gradient value exceeds a preset gradient threshold, it means that there is a significant difference in the microstructure of the two adjacent layers. This structural abrupt change will have a significant impact on the interlayer moisture migration channels. Specifically, if the lower layer is a dense layer with low porosity and the upper layer is a loose layer with high porosity, then under the action of molding pressure, the pore channels of the dense layer will be further compressed or even closed, forming a physical barrier to the downward migration of moisture. During drying, the moisture in the upper loose layer cannot effectively diffuse downward through the dense layer, but can only evaporate to the upper surface, causing the upper layer to shrink rapidly due to water loss, while the lower layer shrinks later, resulting in a relative displacement trend between the two and inducing shear stress at the interface. Therefore, when the porosity gradient value exceeds the preset gradient threshold, it is determined that there is a risk of moisture migration channel blockage at the corresponding interlayer interface. This threshold is determined based on actual production experience or experimental verification, and is usually set at a gradient level that can significantly affect the interlayer moisture balance.

[0035] Subsequently, based on the distribution range of porosity gradient values ​​and the fluctuation amplitude of interlayer interface shear stress readings, the porosity distribution characteristics of each interlayer interface were further determined. For the entire multi-layer composite body, the porosity gradient values ​​of all interlayer interfaces were statistically analyzed, and their distribution histograms or cumulative distribution curves were plotted to identify the concentrated intervals and outliers of the gradient values. Simultaneously, the time series of shear stress readings for each interlayer interface were extracted, and the stress fluctuation amplitude, i.e., the difference or standard deviation between the maximum and minimum stress values, was calculated. A comparative analysis of the porosity gradient distribution and stress fluctuation amplitude revealed a correlation: interfaces with larger porosity gradients also exhibited correspondingly larger shear stress fluctuation amplitudes, because the uncoordinated shrinkage caused by structural abrupt changes resulted in drastic stress changes during the drying process. Through this correlation analysis, not only can the high or low porosity of each layer of clay be identified, but also which interlayer interfaces, due to uneven porosity distribution, become weak points with stress concentration, providing spatially resolved structural characteristic information for subsequent risk assessment.

[0036] Based on the identification of porosity distribution characteristics, the blocking strength of porosity distribution on interlayer moisture migration channels is further evaluated. Moisture migration channels refer to the diffusion paths of moisture formed by the pore network within the clay. In a multi-layered composite body, ideally, moisture should be able to flow freely between layers to maintain overall humidity balance. However, when the porosity of a layer is significantly lower than that of adjacent layers, the pore connectivity of that layer deteriorates, increasing the resistance to moisture penetration and thus blocking the migration channels. The assessment of the blocking strength requires comprehensive consideration of the magnitude of the porosity gradient, the initial value of the interlayer interface shear stress, and the influence of ambient humidity.

[0037] The specific evaluation process is as follows: First, based on the identified porosity distribution characteristics, a blocking index BI is calculated for each interlayer interface. The formula for calculating this index is: BI=G×(1+α×τ / τ0)×(1-β×(H-H0) / H0) Wherein, G is the porosity gradient value (unit: 1 / mm), τ is the interlayer interface shear stress reading (unit: MPa), τ0 is the standard reference stress value (typical value is 1.0 MPa), H is the current ambient humidity reading (unit: %RH), H0 is the standard reference humidity value (typical value is 60%RH), α is the stress correction factor (typical value range 0.1-0.3), and β is the humidity correction factor (typical value range 0.2-0.5).

[0038] In the above formula, the first term G reflects the fundamental contribution of the porosity gradient to the blocking effect; the second term (1+α×τ / τ0) is a stress correction factor. Higher interfacial stress indicates that the interface is subjected to greater compressive force, further compressing the pore space and enhancing the blocking effect. Therefore, when this factor is greater than 1, the blocking index increases; the third term (1-β×(H-H0) / H0) is a humidity correction factor. When the ambient humidity is higher than the standard value, surface evaporation slows down, and water has more time to diffuse, weakening the blocking effect. Therefore, this factor decreases the blocking index; when the ambient humidity is lower than the standard value, the blocking effect increases. Through the calculation of the above formula, the quantitative value of the blocking strength of the water migration channel at each interlayer interface is obtained.

[0039] Finally, the moisture migration blocking level under high pressure was determined. During the forming stage, the preform is subjected to external pressure from the press and internal compressive stress generated by interlayer bonding, which together form a high pressure effect. This effect significantly compresses the pore space of the low-porosity layer, further enhancing its blocking effect on moisture migration. To quantify this effect, the calculated blocking strength value is multiplied by the high pressure effect parameter to obtain the corrected blocking strength. Based on the range of the corrected blocking strength, moisture migration blocking is divided into several levels, such as low blocking, medium blocking, and high blocking. A low blocking level indicates that moisture can flow relatively smoothly between layers with small differences in interlayer shrinkage; a medium blocking level indicates that there is some migration barrier, requiring attention; a high blocking level means that moisture migration is severely restricted, the risk of uncoordinated interlayer shrinkage is high, and it is prone to debonding and cracking.

[0040] Further, the corrected blocking strength under the combined high-pressure force is calculated. The previously obtained blocking index is multiplied by the normalized high-pressure force coefficient, which is defined as the current high-pressure force value divided by the standard molding pressure value (e.g., 10 MPa). When the high-pressure force coefficient is greater than 1, it indicates that the pressure is higher than the standard value, pore compression is more obvious, and the blocking strength is enhanced; when the coefficient is less than 1, the blocking strength is correspondingly weakened. After obtaining the corrected blocking strength value, the moisture migration blocking level is determined according to the preset level classification standard (e.g., 0-0.3 is low blocking, 0.3-0.7 is medium blocking, and 0.7-1.0 is high blocking).

[0041] Determining the level of moisture migration inhibition provides crucial structural input parameters for subsequent assessment of interlayer shear stress accumulation trends.

[0042] By establishing a correlation between macroscopic deformation monitoring data and microscopic porosity distribution characteristics, non-destructive identification of internal structural differences in multilayer composite preforms was achieved. Furthermore, by evaluating the blocking strength of porosity distribution on moisture migration channels, the intrinsic mechanism of interlayer moisture imbalance under high-pressure combined forces was revealed, providing a theoretical basis for accurately predicting stress accumulation and debonding risk during the drying stage. This cross-scale analysis method, from material microstructure to macroscopic mechanical behavior, significantly improves the scientific rigor and predictive ability of interlayer bonding quality monitoring.

[0043] Step S3: Correlation analysis is performed on the moisture migration blocking level, the real-time shrinkage rate reading, and the interlayer interface shear stress reading to evaluate the cumulative increase in interlayer shear stress caused by the uncoordinated shrinkage of each layer during the drying stage, and to obtain the growth trend of interlayer shear stress.

[0044] like Figure 3As shown, step S3, evaluating the cumulative increase of interlayer shear stress, includes: acquiring interlayer interface shear stress readings at multiple time points during the drying process, calculating the incremental shear stress value between adjacent time points; matching the incremental shear stress value with the real-time shrinkage rate reading at the corresponding time point, and analyzing the correlation coefficient between shrinkage rate change and shear stress increment; when the correlation coefficient exceeds a preset correlation threshold, determining that the interlayer interface is in a state of rapid shear stress accumulation, and recording the accumulation start time point and accumulation rate; integrating the accumulation rate data of each interlayer interface, and correcting the accumulation rate in conjunction with the environmental humidity reading to obtain the corrected interlayer shear stress growth trend.

[0045] After obtaining the moisture migration inhibition level, this level was used as a key structural parameter and correlated with real-time shrinkage rate readings and interlayer interface shear stress readings to assess the cumulative increase in interlayer shear stress caused by uneven shrinkage among layers during the drying stage. The core idea of ​​this correlation analysis is that the moisture migration inhibition level determines the ease of achieving interlayer moisture balance, and the imbalance of moisture balance directly leads to differences in the shrinkage rates of each layer, thereby causing a continuous accumulation of shear stress at the interlayer interface.

[0046] Specifically, a mapping relationship is first established between the level of moisture migration blockage and the degree of shrinkage incoordination. For interlayer interfaces classified as high-blockage, moisture cannot be effectively exchanged between the upper and lower layers. During drying, the upper layer (usually a high-porosity layer) shrinks rapidly due to water loss, while the lower layer (usually a low-porosity layer) shrinks slowly, resulting in a significant difference in shrinkage rates. This difference in shrinkage rates can be obtained by extracting the real-time shrinkage rate readings of the corresponding layers and subtracting them. The greater the difference in shrinkage rates, the stronger the relative displacement trend between the two layers, and the greater the shear stress generated at the interface. For interfaces with medium-blockage levels, the degree of shrinkage incoordination is moderate; while for interfaces with low-blockage levels, because moisture can migrate relatively freely, the shrinkage of each layer tends to be synchronous, and the increase in shear stress is relatively gradual.

[0047] Furthermore, interlayer interface shear stress readings were acquired at multiple time points during the drying process. Throughout the drying cycle, the output signal of the shear stress probe was collected at fixed time intervals (e.g., every minute or dynamically adjusted according to the drying rate) to form a stress-time curve. The shear stress increment between adjacent time points was calculated, i.e., the stress reading at the current moment was subtracted from the stress reading at the previous moment. This increment reflects the magnitude of stress change per unit time and is a direct indicator for assessing the stress accumulation rate. If the increment is consistently positive and large, it indicates that stress is accumulating rapidly; if the increment is close to zero or fluctuates, it indicates that the stress is in a relatively stable state.

[0048] Subsequently, the shear stress increment value was matched with the real-time shrinkage rate reading at the corresponding time point for analysis. For each time point, the shrinkage rate reading (or the difference in shrinkage rate between adjacent layers) and the shear stress increment value at that moment were extracted to form a data pair. Correlation analysis was used in statistics to calculate the correlation coefficient between the change in shrinkage rate and the increase in shear stress. This correlation coefficient quantifies the strength of the linear association between the two; a coefficient value close to 1 indicates a strong positive correlation, meaning that the increase in shear stress increment also increases as the difference in shrinkage rate increases; a coefficient value close to 0 indicates no significant correlation between the two. In the drying scenario of multi-layer composite preforms, theoretically, shrinkage incoordination and stress accumulation should show a significant positive correlation.

[0049] When the calculated correlation coefficient exceeds a preset correlation threshold (e.g., 0.7 or 0.8), the interlayer interface is determined to be in a state of rapid shear stress accumulation. This determination implies that the increase in interfacial stress is mainly driven by shrinkage incompatibility, rather than other random factors. At this point, the accumulation start time is recorded, i.e., the moment when the correlation coefficient first exceeds the threshold, marking the transition of the interface from a stable state to a risk accumulation stage. Simultaneously, based on the magnitude of the shear stress increment, the accumulation rate, i.e., the average increase in stress per unit time, is calculated. A higher accumulation rate indicates faster stress accumulation and a more rapid increase in the risk of debonding.

[0050] Next, the cumulative rate data of each interlayer interface are integrated. Since multilayer composite green bodies may contain multiple interlayer interfaces, the accumulation rate at each interface may differ. By summarizing the cumulative rate data of all interfaces, the key interfaces with the highest cumulative rates are identified. These interfaces often correspond to locations with the largest porosity gradients and the most severe moisture migration blockages. Simultaneously, the cumulative rate is corrected based on ambient humidity readings. Changes in ambient humidity affect the evaporation rate of the green body surface, thereby altering the internal moisture gradient and shrinkage rate. During periods of high humidity, evaporation slows down, reducing shrinkage inconsistency, and the cumulative rate should be adjusted downwards accordingly; during periods of low humidity, evaporation accelerates, and the cumulative rate should be adjusted upwards. The correction method can involve multiplying the cumulative rate by a correction factor based on humidity deviation, calculated from the difference between the actual humidity and the standard humidity.

[0051] Finally, the corrected trend of interlaminar shear stress growth was obtained. This trend was plotted as a curve with time on the horizontal axis and the corrected accumulation rate or accumulated stress value on the vertical axis. The slope of the curve reflects the dynamic process of stress accumulation: steep slopes indicate a rapid accumulation phase requiring close monitoring; gentle slopes indicate that the stress is stabilizing. Real-time monitoring of the trend curve allows for early identification of moments when the stress is about to reach the material's interfacial strength limit, providing a basis for decision-making regarding intervention measures (such as adjusting the drying temperature and humidity curve, pausing drying for balancing, etc.).

[0052] By performing multi-dimensional correlation analysis between the structural parameter of moisture migration blocking level and real-time deformation monitoring data, the dynamic mechanism of interlayer shear stress accumulation during the drying stage was revealed. Through quantifying the accumulation increase, identifying the accumulation initiation time point, and calculating the accumulation rate, a detailed characterization of the stress evolution process was achieved. Combined with environmental humidity correction, the accuracy of accumulation trend prediction was further improved, enabling the monitoring system to issue early warnings before the stress reaches a dangerous level, effectively reducing the probability of debonding and cracking, and ensuring the forming quality of multilayer composite preforms.

[0053] Step S4: Based on the interlaminar shear stress growth trend, combined with the edge micro-displacement reading and the ambient humidity reading, identify the starting risk location of the delamination crack in the edge region and generate a list of potential debonding locations.

[0054] Step S4, identifying the initial risk location of the layered cracks in the edge region, includes: using a clustering algorithm to jointly cluster the edge micro-displacement readings and the interlayer shear stress growth trend, dividing the edge region into several stress and displacement feature clusters; calculating the cluster center coordinates and the dispersion of data points within each feature cluster, marking feature clusters with dispersion exceeding a preset dispersion threshold as high-risk clusters; extracting the corresponding spatial coordinates from the high-risk clusters, and correcting the position coordinates for humidity influence based on the environmental humidity readings; generating a list of potential debonding locations based on the corrected position coordinates, the list containing the spatial coordinates of the risk locations and the corresponding risk level identifiers.

[0055] The clustering algorithm employs the K-means clustering method, specifically including: combining the edge micro-displacement readings with the interlayer shear stress growth trend into a two-dimensional feature vector; normalizing the two-dimensional feature vector to eliminate dimensional differences; setting the number of clusters, iteratively calculating the center point position of each cluster using the K-means algorithm until the cluster centers converge; calculating the Euclidean distance from each data point to its respective cluster center, and marking data points whose distance exceeds a preset distance threshold as outliers, the spatial location corresponding to the outlier being a high-risk location.

[0056] Based on the interlaminar shear stress growth trend obtained in step S3, and combined with edge micro-displacement readings and ambient humidity readings, the initiation risk locations of delamination cracks in the edge region are identified. The edge region is the area with the most significant stress concentration in multilayer composite preforms. Due to the lack of constraint from surrounding material, interlaminar shrinkage incompatibility is more easily transformed into visible displacements and cracks at this location. Therefore, accurately identifying high-risk locations in the edge region is crucial for preventing debonding and crack propagation.

[0057] Specifically, a clustering algorithm is first used to jointly cluster the edge micro-displacement readings and the interlaminar shear stress growth trend. Clustering is an unsupervised learning method that divides data into several groups based on their inherent similarity. Data points within each group have similar characteristics, while there are significant differences between groups. In this application, the micro-displacement reading of each edge monitoring point is used as the first-dimensional feature, and the corresponding interlaminar shear stress growth trend (e.g., cumulative rate or cumulative increment) is used as the second-dimensional feature, forming a two-dimensional feature vector. Since the unit of micro-displacement is usually micrometers, while the unit of shear stress is megapascals, the numerical scales of the two differ significantly. To avoid features with larger numerical values ​​dominating the clustering results, the two-dimensional feature vector needs to be normalized. Normalization methods can include min-max normalization, which linearly maps each feature value to the interval between 0 and 1, or Z-score standardization, which converts the feature values ​​into a distribution with a mean of 0 and a standard deviation of 1. Normalization eliminates the dimensional differences, ensuring that the two types of features have equal weight in the clustering process.

[0058] Subsequently, the K-means clustering method was used to cluster the normalized feature vectors. The core idea of ​​the K-means algorithm is to divide the data points into K clusters through iterative optimization, minimizing the sum of squared distances from each data point to the cluster center within each cluster. After convergence, all edge monitoring points were divided into K feature clusters, each cluster representing a typical stress-displacement combination pattern.

[0059] Furthermore, the cluster center coordinates and the dispersion of data points within each characteristic cluster are calculated. The cluster center coordinates are the average values ​​of all data points in the cluster across the dimensions of micro-displacement and stress growth, representing the typical characteristics of the cluster. The dispersion of data points within a cluster can be measured by calculating the variance within the cluster or the standard deviation of the distance from the data point to the cluster center. Clusters with lower dispersion indicate that the data points within the cluster are close to each other, and the stress and displacement states in this region are relatively consistent, usually corresponding to the normal region of the billet. Clusters with higher dispersion indicate that the data points within the cluster are more dispersed, with some abnormal points deviating far from the cluster center. Such anomalies often foreshadow local stress abrupt changes or abnormal displacement growth, which are precursors to crack initiation. Characteristic clusters with dispersion exceeding a preset dispersion threshold are marked as high-risk clusters. This threshold can be determined based on statistical analysis of historical debonding cases and set at a level that can effectively distinguish between normal fluctuations and abnormal risks.

[0060] Spatial coordinates are extracted from high-risk clusters. Each edge monitoring point has its spatial coordinates (e.g., positions along the length and width of the billet) recorded during deployment. For data points marked as high-risk clusters, their corresponding spatial coordinates are extracted to form a preliminary list of high-risk locations. Simultaneously, humidity-based corrections are applied to the location coordinates based on ambient humidity readings. Local differences in ambient humidity (e.g., uneven humidity distribution in different areas of the drying kiln) can lead to different drying rates at different locations within the same billet, thus affecting the distribution of crack initiation locations. The humidity-based correction method is as follows: based on humidity readings near each edge monitoring point, a humidity deviation value (i.e., the difference between the actual humidity and the average humidity) is calculated. For low-humidity areas, due to faster drying, the risk weight of this area is increased, and its priority is raised when generating the list; for high-humidity areas, the risk weight is decreased. The corrected location coordinates not only contain spatial information but also incorporate the influence of environmental factors, making the identification of risk locations more accurate.

[0061] Furthermore, the specific implementation of the K-means algorithm can be further refined to identify outliers. After clustering, the Euclidean distance from each data point to its cluster center is calculated. The Euclidean distance is defined as the square root of the sum of the squares of the differences in each dimension of the feature vector. Data points whose distance exceeds a preset distance threshold are marked as outliers. Although these outliers are assigned to a cluster, their characteristics differ significantly from the cluster center, which is likely due to abrupt changes in data caused by local stress concentration or crack initiation. The spatial location corresponding to the outlier is a high-risk location and needs to be highlighted in the list of potential debonding locations.

[0062] Finally, a list of potential debonding locations is generated based on the corrected location coordinates. This list is presented in tabular or data structure form, with each record containing: the spatial coordinates of the edge monitoring point (e.g., X, Y coordinates or number), the corresponding micro-displacement reading, stress growth trend value, the identifier of the characteristic cluster to which it belongs, the degree of dispersion within the cluster, the humidity correction factor, and a comprehensive risk level identifier. The risk level identifier is divided into three levels: high risk, medium risk, and low risk, determined comprehensively based on multiple factors such as the cluster dispersion, the distance to anomalies, and the weighted values ​​after humidity correction. High-risk locations are listed at the top of the list for priority processing. This list provides spatial location data for subsequent dynamic threshold adjustment, crack propagation rate extraction, and mass degradation prediction.

[0063] By combining multidimensional deformation monitoring data with intelligent clustering algorithms, automated and quantitative identification of the initiation risk locations of layered cracks in edge regions was achieved. The K-means clustering method effectively discovers the inherent grouping structure of the data, distinguishing normal and abnormal areas and avoiding the subjectivity and uncertainty of human experience-based judgment. Combining this with environmental humidity correction further enhances the accuracy and adaptability of the identification. The generated list of potential debonding locations provides clear targets for the molding monitoring system, enabling limited monitoring and intervention resources to be precisely deployed to the most needed locations, significantly improving the efficiency and effectiveness of debonding and crack prevention.

[0064] Furthermore, in this embodiment of the application, the method further includes: a step of dynamically adjusting the debonding risk judgment threshold according to the porosity distribution characteristics, specifically: extracting the porosity gradient value corresponding to each risk location in the potential debonding location list; calculating the distribution variance of the porosity gradient value; when the distribution variance exceeds a preset variance threshold, increasing the value of the debonding risk judgment threshold; and applying the increased debonding risk judgment threshold to the risk location identification process at subsequent time nodes to achieve dynamic updating of the threshold.

[0065] The dynamic adjustment of the debonding risk judgment threshold further includes: establishing a mapping function between the porosity gradient value and the debonding risk judgment threshold; according to the mapping function, when an increase in the porosity gradient value is detected, increasing the debonding risk judgment threshold according to a preset proportional coefficient; recording the time node and adjustment range of each threshold adjustment to form a dynamic adjustment history record of the threshold; and based on the dynamic adjustment history record of the threshold, analyzing the correlation between the threshold adjustment frequency and the final quality of the billet, and optimizing the parameters of the mapping function.

[0066] After generating a list of potential debonding locations, to further improve the accuracy and adaptability of risk assessment, the debonding risk assessment threshold needs to be dynamically adjusted based on the porosity distribution characteristics. The debonding risk assessment threshold is a critical value used to distinguish between normal and risky states, such as the shear stress threshold or micro-displacement threshold. In actual production, different batches of green bodies and different clay formulations may lead to differences in porosity distribution; using a fixed threshold may result in misjudgments or omissions. A dynamic adjustment mechanism can adaptively optimize the threshold based on the actual structural characteristics of the current green body, improving the robustness of the monitoring system.

[0067] Specifically, the porosity gradient values ​​corresponding to each risk location are first extracted from the list of potential debonding sites. These gradient values, calculated in step S2, reflect the porosity change rate of each interlayer interface. For edge locations identified as high-risk, the corresponding interlayer interface often has a large porosity gradient. The porosity gradient values ​​of all risk locations in the list are extracted to form a gradient value sequence.

[0068] Further, the variance of the porosity gradient value sequence is calculated. Variance is a statistic that measures the dispersion of data, and is calculated as the average of the squares of the differences between each data point and the mean. If the variance is small, it indicates that the porosity gradient values ​​at each risk location are relatively similar, and the structural differences in the billet are relatively uniform, allowing for a relatively uniform risk assessment standard. If the variance is large, it indicates significant differences in the porosity gradient at different locations, with structural abrupt changes at some locations far exceeding those at others. In this case, continuing to use a fixed threshold may lead to an underestimation of the risk at high gradient locations. When the variance exceeds a preset variance threshold, the current billet is determined to have strong structural heterogeneity, requiring an increase in the debonding risk assessment threshold. Increasing the threshold aims to improve the sensitivity of risk assessment, ensuring that locations with significant structural abrupt changes can be identified as high-risk in a timely manner, avoiding omissions.

[0069] Subsequently, the increased debonding risk threshold was applied to the risk location identification process at later time points. During the drying stage, as time progresses, the monitoring parameters continuously change, requiring the system to constantly update the risk assessment. Using the dynamically adjusted threshold, newly acquired edge micro-displacement readings, shear stress readings, etc., are assessed to identify newly emerging risk locations or changes in the level of existing risk locations. This dynamic threshold update mechanism enables the monitoring system to adaptively adjust the judgment criteria based on the real-time state of the billet, achieving a transformation from static monitoring to intelligent monitoring.

[0070] Furthermore, to further optimize the dynamic adjustment strategy, a mapping function between the porosity gradient value and the debonding risk assessment threshold is established. This function can take the form of linear mapping, piecewise linear mapping, or nonlinear mapping. For example, when using linear mapping, the threshold adjustment amount is set to be equal to the average porosity gradient value multiplied by a preset proportional coefficient. According to this mapping function, when an increase in the porosity gradient value is detected, the system automatically increases the debonding risk assessment threshold according to the proportional coefficient. The proportional coefficient can be selected based on statistical analysis of historical production data, choosing a parameter value that maximizes the accuracy of risk identification.

[0071] In actual operation, the time point and adjustment range of each threshold adjustment are recorded, forming a dynamic threshold adjustment history. This history is stored in time series form, including the threshold before adjustment, the threshold after adjustment, the porosity gradient variance that triggered the adjustment, environmental humidity, and other relevant parameters. By accumulating a large number of adjustment records, in-depth data mining can be performed to analyze the correlation between the frequency of threshold adjustments and the final quality of the billet. For example, statistics show that batches with frequent threshold adjustments often correspond to billets with large fluctuations in porosity distribution, and these billets also have a higher final debonding rate; while batches with fewer threshold adjustments have better structural stability and a higher quality pass rate. Based on this correlation analysis, the parameters of the mapping function can be further optimized, such as adjusting the value of the proportional coefficient, or introducing a nonlinear term to more finely characterize the relationship between the gradient and the threshold, thereby improving the effectiveness of the dynamic adjustment mechanism.

[0072] By introducing a dynamic threshold adjustment mechanism, the debonding risk assessment criteria can be adaptively optimized based on the actual structural characteristics of the billet and real-time changes in the production process. This adaptive mechanism overcomes the limitations of fixed thresholds, significantly reduces the probability of false positives and false negatives, and improves the applicability of the monitoring system to different batches and different formulations of billets. Simultaneously, by establishing a mapping function and accumulating historical adjustment records, a data foundation is provided for the continuous optimization of threshold parameters, enabling the monitoring system to self-learn and self-evolve, further enhancing the intelligence level of interlayer quality monitoring.

[0073] Furthermore, the method also includes extracting the crack propagation rate from the edge micro-displacement readings and evaluating the expansion range of the debonding region. Specifically, this involves: obtaining the edge micro-displacement reading sequence of each risk location in the potential debonding location list at continuous time nodes; performing time difference processing on the edge micro-displacement reading sequence to calculate the displacement change per unit time and obtain the crack propagation rate; calculating the radial distance of crack propagation based on the crack propagation rate and the spatial coordinates of the crack initiation location; and converting the radial distance into the expansion area of ​​the debonding region to obtain the expansion range of the debonding region.

[0074] After identifying potential debonding locations and dynamically adjusting the risk assessment threshold, the crack propagation rate is further extracted from edge micro-displacement readings, and the extent of debonding region expansion is evaluated. This step aims to quantify the dynamic process of crack initiation and propagation, providing key parameters for predicting the final failure time and extent of the billet.

[0075] Specifically, the process begins by obtaining a sequence of edge micro-displacement readings at consecutive time points for each risk location in the potential debonding location list. For each high-risk location identified in the list, the micro-displacement data of its corresponding monitoring point are traced throughout the drying process. These data are arranged in chronological order to form a time series, for example, [d1, d2, d3, ..., d...]. ], where d This represents the micro-displacement reading at the i-th time point. This sequence fully records the displacement evolution process at the material edge at the crack initiation location.

[0076] Furthermore, the edge micro-displacement reading sequence is subjected to time difference processing. Time difference refers to calculating the difference between micro-displacement readings at adjacent time points, i.e., Δd. =d +1 -d This difference reflects the displacement increment at the edge location within the time interval Δt. Dividing the displacement increment by the time interval yields the displacement change per unit time, i.e., the crack propagation rate v. =Δd / Δt. The unit of this rate is usually micrometers per minute or micrometers per hour. By calculating the difference over the entire sequence, a series of propagation rate values ​​are obtained, reflecting the rate of change in the crack propagation process. If the propagation rate continues to increase, it indicates that the crack is propagating at an accelerated pace, and the risk of debonding increases sharply; if the rate tends to stabilize or decrease, it indicates that the propagation has temporarily slowed down.

[0077] Subsequently, based on the fracture propagation rate and the spatial coordinates of the fracture initiation location, the radial distance of the fracture propagation is calculated. Assuming the fracture propagates from the initiation point in all directions, the propagation pattern can be approximated as circular or elliptical. For the simplified circular propagation model, the radial distance r can be obtained by multiplying the cumulative propagation rate by time, i.e., r = Σ(v ×Δt). For more complex propagation patterns, the radial distance in each direction can be calculated based on the differences in micro-displacement readings in different directions, forming an asymmetric propagation profile. Combining the spatial coordinates (x0, y0) of the fracture initiation position, a spatial distribution map of fracture propagation can be drawn, marking the position of the fracture front.

[0078] Finally, the radial distance is converted into the expanded area of ​​the debonding region. For a circular expansion, the area A = π × r 2 For elliptical expansion, the area A = π × a × b, where a and b are the radii of the major and minor axes, respectively. By calculating the expansion area, a quantitative index of the debonding region's expansion range is obtained. This index directly reflects the severity of the debonding damage: the larger the area, the wider the affected area of ​​the billet, and the worse the structural integrity. Simultaneously, the expansion range can be compared with the total area of ​​the billet to calculate the debonding percentage, for example, 5% or 10% of the total billet area. When the debonding percentage exceeds a certain critical value, the billet is deemed unable to meet quality requirements and needs to be discarded or reworked.

[0079] By analyzing the time series of edge micro-displacement readings, a quantitative description of the crack propagation dynamics was achieved. The extracted propagation rate parameters revealed the evolution of debonding damage, providing a basis for predicting the remaining life of the billet. The assessed propagation range provides a clear quantitative standard for quality assessment, avoiding the subjectivity and delays of relying on manual visual inspection. This data-driven dynamic monitoring and quantitative assessment method significantly improves the ability to detect and address debonding defects early, effectively reducing scrap rates and production losses.

[0080] Step S5: Integrate the deformation monitoring data according to the potential debonding location list, predict the interlayer bonding quality decay path of the multilayer composite preform, generate an early warning signal and output it to the molding monitoring system.

[0081] Step S5, predicting the quality decay path of interlayer bonding, includes: constructing an interlayer bonding quality decay prediction model using a decision tree algorithm; the input features of the prediction model include the real-time shrinkage rate reading, the interlayer interface shear stress reading, the edge micro-displacement reading, and the debonding region expansion range; organizing the input features in a time series to form a multi-dimensional time-series feature vector; classifying the multi-dimensional time-series feature vector using the decision tree algorithm to divide the interlayer bonding quality into multiple decay levels; predicting the quality decay path in the future time period based on the decay level sequence at each time node; and generating a warning signal when the predicted path indicates that the quality will reach the failure level.

[0082] The process of generating an early warning signal and outputting it to the molding monitoring system further includes: generating a visualization report based on the quality decay path; plotting a trend curve with time as the horizontal axis and the interlayer bonding quality decay level as the vertical axis; marking the deformation monitoring data values ​​and corresponding debonding risk levels at each key time node on the trend curve; triggering a multi-level alarm mechanism of the molding monitoring system when the trend curve predicts that the quality will reach the failure level within a preset time window; and outputting alarm signals of different levels based on the remaining time before the predicted failure level is reached, with a higher alarm level for shorter remaining time.

[0083] After obtaining the list of potential debonding locations, the dynamically adjusted risk assessment threshold, and key parameters such as crack propagation rate and debonding area expansion range, the final step is to integrate all deformation monitoring data, predict the interlayer bonding quality decay path of the multilayer composite preform, and generate an early warning signal to output to the forming monitoring system to achieve closed-loop intelligent monitoring and early warning.

[0084] A decision tree algorithm is used to construct a prediction model for interlayer bonding quality degradation. The training process of the decision tree algorithm includes: collecting deformation monitoring data of historical batches of billets; obtaining interlayer bonding quality grade labels through sampling destructive testing after the billets are dried, forming a training sample set; or, when the system is first deployed, setting initial decision rules based on the theory of mechanics of materials, gradually accumulating training samples through manual labeling, and iteratively optimizing model parameters.

[0085] For scenarios with existing historical production data, deformation monitoring data of historical batches of billets are collected, and after drying, the interlayer bonding quality is graded through mechanical tests (such as interlayer shear strength tests) or dissection observation to form a training sample set for model learning.

[0086] Decision trees are a classic supervised learning algorithm that maps input features to output categories by constructing tree-structured decision rules. In this application, the input features of the prediction model include real-time shrinkage rate readings, interlayer interface shear stress readings, edge micro-displacement readings, and the extent of debonding. These features cover multiple dimensions of information regarding the macroscopic deformation, microscopic stress, local damage, and damage propagation of the billet during forming and drying, comprehensively reflecting the interlayer bonding state. To train the decision tree model, historical production data needs to be collected first. Deformation monitoring data for each batch of billets during the drying process needs to be recorded, and after drying, the interlayer bonding quality needs to be graded through mechanical tests (such as interlayer shear strength tests) or dissection observation. The quality grades can be divided into five levels: excellent, good, medium, poor, and failed, corresponding to different bonding strength ranges or debonding degrees. The input features of the historical data and the corresponding quality grade labels are combined to form a training sample set for model learning.

[0087] Furthermore, the input features are organized over time to form a multi-dimensional temporal feature vector. Since the degradation of inter-layer bonding quality is a dynamic process, features at a single moment may not fully reflect trend information. Therefore, for each prediction time t, not only are the feature values ​​at time t used, but also the feature values ​​from previous times (e.g., t-1, t-2, ..., tn) are introduced to form a higher-dimensional temporal feature vector. For example, if there are four types of input features, considering the current time and the two previous times, the temporal feature vector contains 4 × 3 = 12 dimensions. This temporal organization method enables the decision tree model to learn the temporal evolution patterns of features, such as the continuous growth trend of shear stress and the accelerated changes in micro-displacements, thereby improving prediction accuracy.

[0088] The multi-dimensional time-series feature vector is classified using a decision tree algorithm to divide the interlayer bonding quality into multiple decay levels. For each time node in the drying process, deformation monitoring data is collected in real time, a time-series feature vector is constructed, input into a trained decision tree model, and the current quality level is output. As time progresses, the continuously output level sequence constitutes a discrete trajectory of quality decay. Based on the decay level sequence at each time node, the quality decay path in future time periods is predicted. Prediction methods can employ trend extrapolation: analyzing the changing patterns of existing level sequences, such as a downward trend from "Excellent" → "Good" → "Medium," inferring that the next stage might reach the "Poor" level. Alternatively, time series prediction techniques, such as Markov chains or recurrent neural networks, can be combined to probabilistically predict the level evolution over several future steps. When the predicted path indicates that the quality will reach a failure level at some future time, a serious risk of debonding in the billet is determined, and the system immediately generates a warning signal.

[0089] The generation and output of early warning signals is the final step in the entire monitoring process. Based on the predicted mass decay path, the system generates a visual report. This report plots a trend curve with time on the horizontal axis and the combined mass decay level between layers on the vertical axis. Each point on the curve corresponds to the mass level at a given time point, and the points are connected by broken lines or stepped lines to form a clear decay trajectory. To enhance the information content of the report, deformation monitoring data values ​​at each key time point are marked on the trend curve. For example, at a certain moment, the shear stress is marked as 2.5 MPa, the micro-displacement as 120 μm, and the debonding area as 8 cm². 2 The label also indicates the risk level of debonding as "medium risk." This labeling allows users to intuitively understand the specific reasons behind the quality degradation.

[0090] Furthermore, when the trend curve predicts that the quality will reach the failure level within a preset time window (e.g., within the next 10 or 30 minutes), the system triggers a multi-level alarm mechanism in the molding monitoring system. This multi-level alarm mechanism outputs different levels of alarm signals based on the remaining time before the predicted failure level is reached. Specifically, if the remaining time is long (e.g., more than 30 minutes), a low-level warning alarm is issued, alerting operators with a flashing yellow indicator light or a notification message; if the remaining time is moderate (e.g., 10 to 30 minutes), a medium-level warning alarm is issued, alerting operators with an orange indicator light and a buzzer, suggesting adjustments such as lowering the drying temperature or increasing the ambient humidity; if the remaining time is very short (e.g., less than 10 minutes), a high-level emergency alarm is issued, forcibly intervening with a red indicator light, a high-pitched buzzer, and automatic shutdown of the drying equipment to prevent further deterioration of the green body leading to complete failure. Simultaneously, the alarm information is transmitted in real-time to the display screen and mobile terminal in the central control room via the molding monitoring system's communication interface, ensuring that relevant personnel can be promptly informed and respond.

[0091] The predictive model built using the decision tree algorithm enables intelligent prediction of the quality decay path between layers, transforming passive post-event detection into proactive pre-event warning. The introduction of multi-dimensional temporal feature vectors allows the model to capture the dynamic characteristics of quality evolution, significantly improving prediction accuracy and robustness. Visualized reports, presented in an intuitive graphical and data-driven manner, help operators quickly understand the current state and future trends, supporting scientific decision-making. A multi-level alarm mechanism responds according to urgency, avoiding interference from frequent false alarms while ensuring timely handling of serious risks. The deployment of the entire early warning system achieves a complete closed loop from data acquisition, feature extraction, pattern recognition to decision support, significantly improving the quality monitoring level during the multi-layer composite body forming stage, effectively reducing the incidence of debonding and crack defects, and ensuring the yield and production efficiency of ceramic products.

[0092] Specifically, the preset thresholds involved in the embodiments of this application are detailed as follows: The preset gradient threshold is determined based on actual production experience or experimental verification. For common ceramic green bodies, the porosity gradient value between layers is calculated by measuring the porosity of several batches of green bodies (using Archimedes' drainage method or mercury intrusion porosimetry), and the gradient value distribution of batches that experienced debonding is statistically analyzed. The 25th percentile is then selected as the threshold.

[0093] The preset relevant threshold is determined based on statistical significance. The higher the threshold, the stronger the correlation between shrinkage rate and stress increment is required to determine a rapid accumulation state, and the stricter the determination. The lower the threshold, the more lenient the determination, and it can be adjusted according to the frequency of debonding in actual production.

[0094] The preset discrete threshold is determined based on the statistical distribution of the variance within the cluster. Typically, the monitoring data of normal batch billets are clustered, the variance of each cluster is statistically analyzed, and 1.5 times the upper quartile is taken as the discrete threshold.

[0095] The preset distance threshold is determined based on the statistical characteristics of the cluster radius. The calculation method is as follows: calculate the average distance from all data points in each cluster to the cluster center, and take twice the average distance as the anomaly judgment threshold.

[0096] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. The present invention has been described in detail with reference to preferred embodiments. Those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications and substitutions should be covered within the scope of the claims of the present invention.

[0097] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0098] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between apparatuses or units through some interfaces, and may be electrical, mechanical, or other forms.

[0099] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0100] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0101] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method for real-time analysis of deformation during the forming stage of ceramic products, characterized in that, Includes the following steps: Step S1: Obtain deformation monitoring data of the multi-layer composite preform during the forming and drying stages. The deformation monitoring data includes real-time shrinkage rate readings between adjacent layers, interlayer interface shear stress readings, edge micro-displacement readings, and ambient humidity readings. Step S2: Identify the porosity distribution characteristics of each layer of mud based on the deformation monitoring data, and evaluate the blocking strength of the porosity distribution on the interlayer moisture migration channel by combining the correlation between the real-time shrinkage rate reading and the interlayer interface shear stress reading, and determine the moisture migration blocking level under the action of high pressure. Step S3: Correlation analysis is performed on the moisture migration blocking level, the real-time shrinkage rate reading, and the interlayer interface shear stress reading to evaluate the cumulative increase of interlayer shear stress caused by the uncoordinated shrinkage of each layer during the drying stage, and to obtain the growth trend of interlayer shear stress. Step S4: Based on the interlaminar shear stress growth trend, combined with the edge micro-displacement reading and the ambient humidity reading, identify the initial risk location of the delamination crack in the edge region and generate a list of potential debonding locations. Step S5: Integrate the deformation monitoring data according to the potential debonding location list, predict the interlayer bonding quality decay path of the multilayer composite preform, generate an early warning signal and output it to the molding monitoring system.

2. The method for real-time analysis of deformation during the forming stage of ceramic products according to claim 1, characterized in that, The step S2, which identifies the porosity distribution characteristics of each layer of clay, includes: Extract the real-time shrinkage rate readings of each layer from the deformation monitoring data, calculate the shrinkage rate difference between adjacent layers, and determine the porosity layer type based on the comparison result of the shrinkage rate difference with the preset shrinkage threshold. The porosity layer type includes high porosity layer and low porosity layer. Calculate the porosity gradient value between adjacent layers. When the porosity gradient value exceeds a preset gradient threshold, it is determined that there is a risk of water migration channel blockage at the corresponding interlayer interface. Based on the distribution range of the porosity gradient values ​​and the fluctuation range of the interlayer interface shear stress readings, the porosity distribution characteristics of each interlayer interface are determined.

3. The method for real-time analysis of deformation during the forming stage of ceramic products according to claim 1, characterized in that, The evaluation of the cumulative increase in interlaminar shear stress in step S3 includes: Obtain interlayer interface shear stress readings at multiple time points during the drying process, and calculate the shear stress increment between adjacent time points; The shear stress increment value is matched with the real-time shrinkage rate reading at the corresponding time point, and the correlation coefficient between the shrinkage rate change and the shear stress increment is analyzed. When the correlation coefficient exceeds the preset correlation threshold, it is determined that the interlayer interface is in a state of rapid shear stress accumulation, and the accumulation start time and accumulation rate are recorded. By integrating the cumulative rate data of each interlayer interface and combining it with the ambient humidity readings, the cumulative rate is corrected to obtain the corrected interlayer shear stress growth trend.

4. The method for real-time analysis of deformation during the forming stage of ceramic products according to claim 1, characterized in that, Step S4, which identifies the initial risk location of layered cracks in the edge region, includes: A clustering algorithm is used to perform joint clustering processing on the edge micro-displacement readings and the interlayer shear stress growth trend, dividing the edge region into several stress and displacement feature clusters; Calculate the cluster center coordinates and the dispersion of data points within each feature cluster, and mark feature clusters with dispersion exceeding a preset dispersion threshold as high-risk clusters; Extract the corresponding spatial coordinates from the high-risk clusters, and combine the environmental humidity readings to correct the risk level of each coordinate for humidity influence. A list of potential debonding locations is generated based on the corrected location coordinates. The list includes the spatial coordinates of the risk locations and the corresponding risk level identifiers.

5. The method for real-time analysis of deformation during the forming stage of ceramic products according to claim 2, characterized in that, Also includes: The step of dynamically adjusting the debonding risk assessment threshold based on the porosity distribution characteristics is as follows: Extract the porosity gradient value corresponding to each risk location in the potential debonding location list; Calculate the distribution variance of the porosity gradient value. When the distribution variance exceeds a preset variance threshold, increase the value of the debonding risk judgment threshold. The increased debonding risk assessment threshold is applied to the risk location identification process at subsequent time points to achieve dynamic updating of the threshold.

6. The method for real-time analysis of deformation during the forming stage of ceramic products according to claim 4, characterized in that, Also includes: The crack propagation rate is extracted from the edge micro-displacement readings, and the extent of debonding region propagation is evaluated, specifically as follows: Obtain the edge micro-displacement reading sequence of each risk location in the potential debonding location list at consecutive time nodes; The edge micro-displacement reading sequence is subjected to time difference processing to calculate the displacement change per unit time, thereby obtaining the crack propagation rate. Based on the crack propagation rate and the spatial coordinates of the crack initiation position, the radial distance of crack propagation is calculated. The radial distance is converted into the expanded area of ​​the debonding region to obtain the expanded range of the debonding region.

7. The method for real-time analysis of deformation during the forming stage of ceramic products according to claim 6, characterized in that, The step S5, predicting the interlayer bonding mass degradation path, includes: A decision tree algorithm is used to construct an interlayer bonding quality decay prediction model. The input features of the prediction model include the real-time shrinkage rate reading, the interlayer interface shear stress reading, the edge micro-displacement reading, and the debonding region expansion range. The input features are organized on a time series to form a multidimensional time series feature vector; The multidimensional temporal feature vectors are classified using a decision tree algorithm, and the interlayer bonding quality is divided into multiple attenuation levels. Based on the decay level sequence at each time point, the quality decay path in the future time period is predicted. When the predicted path indicates that the quality will reach the failure level, an early warning signal is generated.

8. The method for real-time analysis of deformation during the forming stage of ceramic products according to claim 4, characterized in that, The method of using a clustering algorithm to jointly cluster the edge micro-displacement readings and the interlayer shear stress growth trend specifically includes: The edge micro-displacement readings are combined with the interlaminar shear stress growth trend to form a two-dimensional feature vector; The two-dimensional feature vectors are normalized to eliminate dimensional differences; Set the number of clusters, and use the K-means algorithm to iteratively calculate the position of the center point of each cluster until the cluster centers converge. Calculate the Euclidean distance from each data point to its cluster center, and mark data points whose distance exceeds a preset distance threshold as outliers. The spatial location corresponding to the outlier is the high-risk location.

9. The method for real-time analysis of deformation during the forming stage of ceramic products according to claim 5, characterized in that, The dynamic adjustment of the debonding risk assessment threshold also includes: Establish a mapping function between porosity gradient values ​​and debonding risk assessment thresholds; According to the mapping function, when an increase in porosity gradient value is detected, the debonding risk judgment threshold is increased by a preset proportional coefficient. Record the time point and adjustment range of each threshold adjustment to form a dynamic threshold adjustment history record; Based on the historical records of the threshold dynamic adjustment, the correlation between the threshold adjustment frequency and the final quality of the billet is analyzed, and the parameters of the mapping function are optimized.

10. The method for real-time analysis of deformation during the forming stage of ceramic products according to claim 7, characterized in that, The generation of the early warning signal also includes: A visualization report is generated based on the quality degradation path. The visualization report plots a trend curve with time on the horizontal axis and interlayer combined quality degradation level on the vertical axis. Mark the deformation monitoring data values ​​and corresponding debonding risk levels at each key time point on the trend curve; When the trend curve predicts that the quality will reach the failure level within a preset time window, the multi-level alarm mechanism of the molding monitoring system is triggered. The multi-level alarm mechanism outputs alarm signals of different levels based on the predicted remaining time to reach the failure level; the shorter the remaining time, the higher the alarm level.

Images