Quality detection method and system for rubber product

Abstract

The invention discloses a rubber product quality detection method and system, and relates to the related technical field of rubber quality detection.The method comprises the steps that rubber products are conveyed to a target detection table, and data acquisition of the rubber products is executed; inputting the wave impedance distribution data and the surface heat distribution gradient field into a defect causal atlas channel to execute phase inversion fusion processing, establishing a cross-modal coupling phase tensor, and identifying deep defect response characteristics by solving a phase difference matrix; activating a pressure spectrum sensing unit to apply control pressure, reading a local mechanical response spectrum, sending the local mechanical response spectrum to a cross-modal calibration layer, and executing joint error authentication; and performing quality detection report according to a self-calibration defect mapping result. The technical problems that in the prior art, multi-modal sensing data fusion efficiency is low, deep defects are difficult to comprehensively recognize, and consequently the quality detection precision and reliability of rubber products are insufficient are solved, and the technical effects that cross-modal cooperative detection is achieved, and the precision and reliability of defect recognition results are improved are achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of rubber quality detection, and particularly relates to a quality detection method and system for rubber products. BACKGROUND

[0002] As an important industrial material, the quality of rubber products is directly related to the safety, reliability and service life of products. However, during the production process of rubber products, internal defects such as bubbles, inclusions and delamination, and surface defects such as cracks and deformation may be introduced due to material unevenness, process parameter fluctuations or external environmental factors, which not only reduces the mechanical properties and durability of the products, but also may cause serious safety accidents. Traditional quality detection mainly relies on manual visual inspection, hardness testing or tensile testing, which has the limitations of strong subjectivity, low efficiency and difficulty in detecting deep defects. With the development of sensing technology, ultrasonic detection and infrared thermal imaging are introduced into rubber product detection. Ultrasonic detection can identify internal defects by analyzing the propagation characteristics of sound waves in materials, while infrared thermal imaging can reveal potential abnormalities by capturing surface temperature distribution. However, ultrasonic detection has insufficient sensitivity to surface defects, and infrared technology is greatly affected by environmental interference and difficult to quantify deep defect features. Therefore, how to realize efficient fusion of multi-modal sensing data has become the key to improving the detection accuracy of rubber products.

[0003] Therefore, in the related art, there is a technical problem of low fusion efficiency of multi-modal sensing data, difficulty in comprehensively identifying deep defects, resulting in insufficient quality detection accuracy and reliability of rubber products. SUMMARY

[0004] The present application provides a quality detection method and system for rubber products, which solves the technical problem of low fusion efficiency of multi-modal sensing data, difficulty in comprehensively identifying deep defects, resulting in insufficient quality detection accuracy and reliability of rubber products in the prior art, and achieves the technical effect of realizing cross-modal collaborative detection and improving the accuracy and reliability of defect identification results.

[0005] The application provides a quality detection method for rubber products, which comprises the following steps: when the rubber products are conveyed to a target detection table, activating an ultrasonic coherent sensing unit and an infrared spectrum sensing unit to respectively perform data acquisition of the rubber products, and establishing wave impedance distribution data and surface heat distribution gradient field; inputting the wave impedance distribution data and the surface heat distribution gradient field into a defect causal graph channel, based on the defect causal graph channel, performing phase inversion fusion processing on the ultrasonic coherent signal and the infrared spectrum temperature signal, establishing a cross-modal coupling phase tensor, identifying deep defect response characteristics by solving a phase difference matrix; activating a pressure spectrum sensing unit to apply a control pressure to the rubber products, reading a local mechanical response spectrum, sending the local mechanical response spectrum and the deep defect response characteristics to a cross-modal calibration layer to perform joint error authentication, and generating a self-calibration defect mapping result; and performing quality detection and reporting according to the self-calibration defect mapping result.

[0006] In a possible implementation, the quality detection method for rubber products further performs the following processing: activating a data processing layer, performing complex wavelet phase inversion on the wave impedance distribution data, extracting phase drift rate and amplitude offset characteristics of local wave impedance, and performing time series gradient convolution on the surface heat distribution gradient field to extract phase response and temperature curvature distribution of surface heat flow; constructing a cross-domain phase-amplitude joint feature set after extracting the results, inputting the cross-domain phase-amplitude joint feature set into a causal correlation inference layer, calculating the coupling strength between wave impedance disturbance and temperature gradient anomaly by Granger causal analysis and mutual information entropy weight constraint, and establishing a three-element coupling matrix containing phase drift rate, amplitude offset and temperature curvature; selecting nodes with coupling strength exceeding a calibration threshold in the three-element coupling matrix as defect activation nodes, performing directed propagation by using a phase topology propagation operator to form a defect causal graph; in the evolution process of the defect causal graph, performing node connection weight correction by cross-modal phase tensor residual error to perform evolution dynamic self-calibration; after calculating the phase difference matrix, identifying deep defect response characteristics according to the calculation results and the self-calibrated defect causal graph.

[0007] In a possible implementation, the quality detection method for rubber products further performs the following processing: reading geometric structure, stress distribution and functional requirement information of the rubber products to establish a basic information data set; performing importance analysis of rubber spatial positions by using the basic information data set to configure a position importance matrix; establishing a calibration threshold of position nodes according to the position importance matrix and the functional requirements; and identifying mapping position coupling strength in the three-element coupling matrix by using the calibration threshold to construct defect activation nodes.

[0008] In a possible implementation, the quality detection method of the rubber product further performs the following processing: after the abnormal positioning of the deep defect response feature configuration, performing mechanical response anomaly analysis based on the local mechanical response spectrum for abnormal positioning, and extracting mapping abnormal features; performing mechanical auxiliary authentication of the corresponding deep defect response feature based on the mapping abnormal features, and generating a self-calibration defect mapping result.

[0009] In a possible implementation, the quality detection method of the rubber product further performs the following processing: reading a quality detection data set of a batch of rubber products, and extracting common defect features; identifying defect deviation trends using the common defect features, and constructing additional early warning signals; after the self-calibration defect mapping result is enhanced by the additional early warning signals, performing early warning reporting, and configuring a quality monitoring attention factor based on the additional early warning signals; and performing attention data collection and quality detection identification of the same batch of rubber products through the quality monitoring attention factor.

[0010] In a possible implementation, the quality detection method of the rubber product further performs the following processing: configuring a double-level causal reasoning network, the double-level causal reasoning network including a local reasoning layer and a global reasoning layer; inputting the self-calibration defect mapping result into the local reasoning layer, performing local defect feature modeling, and establishing a local reasoning result; sending the local reasoning result, wave impedance distribution data, and surface thermal distribution gradient field to the global reasoning layer, performing global defect propagation prediction, and establishing a prediction fitting result; after the self-calibration defect mapping result is compensated according to the prediction fitting result, performing quality detection reporting.

[0011] In a possible implementation, the quality detection method of the rubber product further performs the following processing: configuring a detection shunting strategy using the self-calibration defect mapping result, and configuring a shunting signal of the rubber product; after the self-calibration defect mapping result is bound with a unique identification code of the rubber product, performing shunting transmission processing based on the shunting signal.

[0012] This application also provides a quality inspection system for rubber products, the system comprising: a data acquisition module, used to activate an ultrasonic coherent sensing unit and an infrared spectral sensing unit to perform data acquisition of the rubber product after it is transferred to the target inspection station, and to establish wave impedance distribution data and surface thermal distribution gradient field; a fusion processing module, used to input the wave impedance distribution data and surface thermal distribution gradient field into a defect causal map channel, and based on the defect causal map channel, to perform phase inversion fusion processing on the ultrasonic coherent signal and the infrared spectral temperature signal to establish a cross-modal coupled phase tensor, and to identify deep defect response characteristics by solving the phase difference matrix; a joint error authentication module, used to activate a pressure spectrum sensing unit to apply control pressure to the rubber product, read the local mechanical response spectrum, send the local mechanical response spectrum and the deep defect response characteristics to a cross-modal calibration layer, perform joint error authentication, and generate a self-calibrated defect mapping result; and a quality inspection reporting module, used to report the quality inspection based on the self-calibrated defect mapping result.

[0013] The proposed method and system for quality inspection of rubber products involves transmitting rubber products to a target inspection station for data acquisition. Impedance distribution data and surface thermal gradient field are input into the defect causality map channel for phase inversion fusion processing. A cross-modal coupled phase tensor is established, and the response characteristics of deep defects are identified by solving the phase difference matrix. A pressure spectrum sensing unit is activated to apply control pressure, and the local mechanical response spectrum is read and sent to the cross-modal calibration layer for joint error verification. Quality inspection results are reported based on the self-calibrated defect mapping. This method solves the technical problems of low efficiency in multimodal sensor data fusion and difficulty in comprehensively identifying deep defects in existing technologies, leading to insufficient accuracy and reliability in rubber product quality inspection. It achieves cross-modal collaborative detection, improving the accuracy and reliability of defect identification results. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings of the embodiments of this disclosure will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0015] Figure 1 This is a schematic diagram of the quality testing method for rubber products provided in the embodiments of this application.

[0016] Figure 2 This is a schematic diagram of the structure of a quality inspection system for rubber products provided in an embodiment of this application.

[0017] Figure labeling: Data acquisition module 10, fusion processing module 20, joint error authentication module 30, quality inspection reporting module 40. Detailed Implementation

[0018] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0019] This application provides a method for quality testing of rubber products, such as... Figure 1 As shown, the method includes: Step S100: After the rubber product is transferred to the target inspection station, the ultrasonic coherent sensing unit and the infrared spectrum sensing unit are activated to perform data acquisition of the rubber product and establish wave impedance distribution data and surface thermal distribution gradient field.

[0020] Preferably, the rubber product is precisely transported to the target testing station to ensure consistent position and conditions for each test. Then, the ultrasonic coherent sensing unit and the infrared spectral sensing unit are activated to perform data acquisition on the rubber product. The ultrasonic coherent sensing unit is a functional component that uses ultrasonic waves to acquire the amplitude and phase information of ultrasonic waves. The ultrasonic coherent sensing unit emits known high-frequency ultrasonic waves with stable phase relationships into the rubber product. The waves propagate inside the rubber product and are reflected, scattered, and transmitted when they encounter different media interfaces such as a uniform rubber matrix, bubbles, impurities, or layered structures. The ultrasonic signals returned from different depths and positions inside the rubber product are received and their phase and amplitude changes are analyzed to reconstruct the distribution of the wave impedance value inside the rubber product in three-dimensional space, so as to reflect the uniformity of the internal structure of the rubber material. The wave impedance is the product of the material density and the propagation speed of the sound wave in the material. Different materials or different states of the same material, such as uneven density or defects, have different wave impedances. The wave impedance of defects such as bubbles and cavities differs significantly from that of normal rubber areas, thus appearing as abnormal points or abnormal areas in the distribution.

[0021] Preferably, the infrared spectrum sensing unit is a functional component that uses infrared radiation thermal imaging to detect the infrared radiation naturally emitted or stimulated emitted by the surface of an object to measure its temperature. During detection, a brief thermal excitation is actively applied to the rubber product, such as a flash lamp or hot air, or the inherent temperature distribution during the production process is directly detected. The infrared radiation image of the surface is captured by an infrared camera to obtain the surface thermal distribution of the rubber product, that is, a two-dimensional image formed by the temperature values ​​of each point on the surface of the rubber product. Then, a surface thermal distribution gradient field is constructed to describe the direction and intensity of the temperature change of the surface of the rubber product in space, such as the rate of temperature increase or decrease from one point to another. Internal defects such as delamination, uneven thickness, and foreign objects, as well as surface defects, can change the thermal conductivity of the material, thereby generating abnormal hot spots, cold spots, or specific temperature change patterns on the surface, making the anomalies easier to identify.

[0022] Step S200: Input the wave impedance distribution data and surface thermal distribution gradient field into the defect causal map channel. Based on the defect causal map channel, perform phase inversion fusion processing on the ultrasonic coherent signal and the infrared temperature signal to establish a cross-modal coupled phase tensor. By solving the phase difference matrix, identify the response characteristics of deep defects.

[0023] Preferably, the wave impedance distribution data and the surface thermal distribution gradient field are input into the defect causality map channel for processing. The defect causality map channel is used to establish the causal relationship between the ultrasonic signal and the infrared signal. Then, phase inversion fusion processing is performed on the ultrasonic coherent signal and the infrared temperature signal, that is, the wave impedance distribution data and the surface thermal distribution gradient field are converted into the phase domain, which is more sensitive to small defects. Here, phase refers to the periodic position of the waveform generated by the ultrasonic wave and the thermal wave under periodic thermal excitation, which can reveal subtle features that the amplitude information cannot reflect. Inversion is to infer and reconstruct the defect that caused the signal by the noisy and interference signal received by the sensor. Then, the ultrasonic coherent signal and the infrared temperature signal are aligned and fused, that is, the phase field of the ultrasonic signal and the phase field of the infrared thermal signal are spatially correlated to obtain a common internal defect model that can most reasonably explain the two sets of phase data from different physical sources.

[0024] Preferably, through phase inversion fusion processing, a cross-modal coupled phase tensor is output, which is a matrix containing phase information of two different modes, ultrasound and infrared, and their interrelated coupling strength. This matrix is ​​used to characterize the degree to which acoustic phase anomalies and thermal phase anomalies interact and corroborate each other at each point in space. Then, the phase difference matrix is ​​solved based on the cross-modal coupled phase tensor to quantify the difference between the actual observed phase coupling relationship and the phase coupling relationship predicted by the defect-free benchmark model. This transforms the fused signal into a region map highlighting anomalies. Regions with significantly non-zero values ​​in the difference matrix are suspected regions that deviate severely from the ideal state. Finally, deep defect response characteristics are identified, including defects located inside rubber products, such as internal bubbles, inclusions, and deep delamination.

[0025] Furthermore, step S200 also includes step S210, activating the data processing layer, performing complex wavelet phase inversion on the wave impedance distribution data, extracting the phase drift rate and amplitude shift features of the local wave impedance, and performing temporal gradient convolution on the surface thermal distribution gradient field to extract the phase response and temperature curvature distribution of the surface heat flow; step S220, after constructing a cross-domain phase-amplitude joint feature set from the extraction results, inputting the cross-domain phase-amplitude joint feature set into the causal correlation inference layer, and calculating the relationship between wave impedance perturbation and temperature gradient anomaly through Grapman causal analysis and mutual information entropy weight constraints. The coupling strength is determined, and a ternary coupling matrix including phase drift rate, amplitude offset, and temperature curvature is established. In step S230, nodes with coupling strength exceeding the calibration threshold are selected as defect activation nodes in the ternary coupling matrix, and directed propagation is performed using the phase topology propagation operator to form a defect causal graph. In step S240, during the evolution of the defect causal graph, node connection weight correction is performed through cross-modal phase tensor residuals to perform evolutionary dynamic self-calibration. In step S250, after calculating the phase difference matrix, deep defect response characteristics are identified based on the calculation results and the self-calibrated defect causal graph.

[0026] Preferably, the wave impedance distribution data is input into the data processing layer for complex wavelet phase inversion. Specifically, the wave impedance signal is decomposed simultaneously on both spatial and frequency scales using complex wavelet transform, providing both phase and amplitude information. Then, the local phase information obtained after complex wavelet decomposition is used for inverse deduction to extract the phase drift rate and amplitude shift characteristics of the local wave impedance. The phase drift rate is used to quantify the rate of change or gradient of the local phase relative to the defect-free reference model; a sharp phase drift may indicate a defect boundary. The amplitude shift characteristics are used to quantify the attenuation or enhancement of local signal energy; for example, bubbles cause significant attenuation of ultrasonic amplitude.

[0027] Preferably, the surface thermal distribution gradient field is input into the data processing layer for temporal gradient convolution. That is, the surface thermal distribution gradient field is convolved using gradient operators to enhance and extract the dynamic pattern of temperature field changes over time, thereby extracting the phase response and temperature curvature distribution of surface heat flow. Specifically, if the thermal excitation is periodic, the phase response of surface heat flow refers to the phase delay of surface temperature fluctuations relative to the thermal excitation source. Internal defects will change the heat wave propagation path, resulting in phase delay. If it is pulsed thermal excitation, the phase response of surface heat flow is a phase representation of the heat flow diffusion rate. The temperature curvature distribution is used to describe the curvature of the surface of the temperature field. Curvature can accurately locate the core region of thermal anomalies and eliminate the interference of smooth background temperature. For example, the center curvature of hot spots or cold spots is very high.

[0028] Preferably, the extracted results are combined to construct a cross-domain phase-amplitude joint feature set, containing multiple higher-order features extracted from acoustic and thermal data that are more sensitive to defects. This cross-domain phase-amplitude joint feature set is then input into a causal association inference layer for Grappman causality analysis. Grappman causality analysis is used to infer whether one variable has a direct causal effect on another variable, i.e., to determine whether an ultrasonic anomaly at one location directly causes a thermal anomaly at another location, or vice versa, to distinguish between genuine coupling anomalies and coincidences. Mutual information is used to measure the amount of shared information between two variables. Entropy weights are assigned based on the degree of feature change. The strength of the statistical association between acoustic and thermal anomalies is then quantified and assigned a confidence weight through mutual information entropy weight constraints, i.e., determining the coupling strength between impedance perturbation and temperature gradient anomaly, characterizing the reliability of the causal association between acoustic and thermal anomalies at this point. Finally, a ternary coupling matrix is ​​established, where each element represents a point in space, containing phase drift rate, amplitude offset, temperature curvature, and a comprehensive coupling strength value.

[0029] Preferably, the calibration threshold is used to determine whether the coupling strength is significant enough to be considered a defect. It is dynamically adjusted according to the importance of different locations on the rubber product. For example, the bending part of the rubber seal ring with the greatest stress or the weakest joint in the structure has higher importance. The calibration threshold is lowered in high importance areas to improve detection sensitivity, while the calibration threshold is increased in non-critical areas to suppress false alarms. Then, the coupling strength of each point in the ternary coupling matrix is ​​compared with the calibration threshold, and the nodes with coupling strength exceeding the calibration threshold are selected as defect activation nodes. Next, the phase topology propagation operator is used to perform directed propagation defect influence simulation analysis. That is, based on physical laws such as stress concentration transmission and heat diffusion path, the directed propagation simulation starts from the defect activation node and proceeds along the direction in which the phase field change in the material is most natural and most in line with physical laws. New suspected defect nodes are connected to form a defect causal map, where the nodes are suspected defect points and the edges represent the propagation path and causal relationship of the defect influence.

[0030] Preferably, during the evolution of the defect causal graph, node connection weight correction is performed through cross-modal phase tensor residuals. Specifically, the predicted acoustic and thermal phase signals are calculated based on the defect causal graph, and then compared with the actual signals received from the sensors. The difference is calculated as the cross-modal phase tensor residual. If the cross-modal phase tensor residual of a certain connection edge is large, it indicates that the corresponding propagation path of the current defect causal graph does not match the actual situation. Therefore, the weight of that connection is automatically weakened or deleted to ensure that the defect causal graph can self-adjust and correct itself based on actual data feedback, completing the dynamic self-calibration of the evolution. Then, the deviation of the phase at each position from the defect-free baseline state is calculated to determine the phase difference matrix to identify abnormal regions. Combined with the self-calibrated defect causal graph, regions in the identified abnormal regions that are also located in the key causal path of the defect causal graph are identified, and finally, the deep defect response characteristics are determined.

[0031] Furthermore, step S230 also includes step S231, reading the geometric structure, stress distribution, and functional requirements information of the rubber product to establish a basic information dataset; step S232, using the basic information dataset to perform an importance analysis of the spatial location of the rubber and configuring a location importance matrix; step S233, establishing calibration thresholds for location nodes based on the location importance matrix and functional requirements; and step S234, using the calibration thresholds to identify the mapping position coupling strength in the ternary coupling matrix and constructing defect activation nodes.

[0032] Preferably, information on the geometric structure, stress distribution, and functional requirements of the rubber product is acquired and combined to form a basic information dataset. The geometric structure includes data such as the position, thickness, and curvature of each spatial point on the rubber product. The stress distribution is the result of finite element simulation analysis, used to show the parts of the rubber product that bear the greatest mechanical stress under normal working conditions, such as the lip contact area of ​​a sealing ring or the bending area of ​​a part. The functional requirements information refers to the key functional areas of the rubber product, such as the main sealing surface of an oil seal, the mounting holes of a shock-absorbing pad, and the effective sensing area of ​​a diaphragm. Based on the basic information dataset, an importance analysis of the spatial position of the rubber is performed, that is, an importance score is calculated for each point on the surface of the rubber product. If the geometric structure is complex, the working stress is high, and it is located in a key functional area, the failure risk and negative impact on the product function are greater, and the importance score is higher. This generates a position importance matrix that corresponds one-to-one with the three-dimensional space of the rubber product, where each element represents the importance of that spatial position.

[0033] Preferably, different judgment criteria are set for different locations of the rubber product based on the location importance matrix and functional requirements. Specifically, a lower threshold is used in high-importance areas to improve detection sensitivity and avoid missed detections; a higher threshold is used in low-importance areas, and only when the sensor signal shows a very strong anomaly is it judged as a defect, in order to suppress false alarms and improve detection efficiency. Then, calibration thresholds for location nodes are established. Then, the calibration thresholds are used to identify the coupling strength of the mapped locations in the ternary coupling matrix, that is, each location in the ternary coupling matrix is ​​compared with the corresponding adaptive calibration threshold in the location importance matrix, and the nodes whose coupling strength exceeds the calibration threshold are identified as defect activation nodes. For example, for the same coupling strength X, if it is higher than the calibration threshold of location A in the high-importance area, then location A is identified as a defect activation node; if it is lower than the calibration threshold of location B in the low-importance area, then location B is not identified as a defect activation node.

[0034] Step S300: Activate the pressure spectrum sensing unit to apply control pressure to the rubber product, read the local mechanical response spectrum, send the local mechanical response spectrum and the deep defect response characteristics to the cross-modal calibration layer, perform joint error authentication, and generate self-calibrated defect mapping results.

[0035] Preferably, the pressure spectrum sensing unit is a functional component capable of precisely controlling and measuring the pressure applied to a rubber product and the resulting deformation. For example, it could be a pressure head or plate with high-precision force and displacement sensors. The pressure spectrum sensing unit applies controlled pressure to the rubber product and records the local mechanical response spectrum during the pressure application process. This spectrum may include a force-displacement curve describing the relationship between the applied pressure and the deformation of the rubber product, a stress-strain curve derived from the force-displacement curve, local stiffness / flexibility, and creep or relaxation behavior after the pressure is maintained for a period of time. Then, the local mechanical response spectrum and deep defect response characteristics are integrated and input into a cross-modal calibration layer for joint error authentication, including verification of abnormal mechanical behavior and false alarm identification and rejection. Specifically, for suspicious products with deep defect response characteristics... For defective regions, the focus is on verifying whether these regions also exhibit anomalies in their local mechanical response spectra. For example, does a region predicted as having internal voids show a significant decrease in local stiffness under pressure? Does a region predicted as having delamination defects show abnormal rebound hysteresis or permanent deformation upon pressure release? If the mechanical response corresponding to the defective region predicted by acoustic and thermal analysis is also abnormal, the confidence level of the defect is increased. If a region is marked as abnormal by acoustic and thermal features, but its mechanical response is no different from that of the surrounding healthy material, the defect may be a false alarm due to sensor noise or harmless inhomogeneity of the material itself. In this case, the confidence level of the defect in that region is reduced or it is removed. Finally, a self-calibrated defect mapping result is generated, containing accurate defect distribution, thereby ensuring improved reliability and credibility of rubber product quality inspection.

[0036] Furthermore, step S300 also includes step S310, after configuring the anomaly location using the deep defect response features, performing mechanical response anomaly analysis based on the local mechanical response spectrum to extract the mapping anomaly features; step S320, performing mechanical auxiliary authentication based on the mapping anomaly features corresponding to the deep defect response features to generate a self-calibrated defect mapping result.

[0037] Preferably, anomaly localization is performed based on the response characteristics of deep defects. This involves delineating multiple suspected defect areas requiring close attention within the rubber product, then retrieving the local mechanical response spectrum measured by the pressure spectrum sensing unit in the suspected defect areas, and performing anomaly analysis of the mechanical response for anomaly localization. This includes comparing the mechanical behavior of the suspected defect areas with the mechanical responses of the surrounding known normal areas, and then quantifying and extracting mapping anomaly features that can characterize the abnormal mechanical properties of the area, such as local stiffness reduction, residual deformation, creep rate, stress concentration factor, etc.

[0038] Preferably, mechanical-assisted authentication of corresponding deep defect response features is performed according to the mapping anomaly features, that is, a one-to-one correlation judgment is made between the mapping anomaly features and the deep defect response features. Specifically, if a region is marked as an internal cavity by the deep defect response features, and its mapping anomaly features show a significant local stiffness decrease and large residual deformation, the defect confidence of this region is marked as extremely high; if a region is marked as abnormal by the deep defect response features, but its mapping anomaly features show that all mechanical parameters such as its stiffness and resilience have no statistical difference from the normal region, it is determined that the acoustic-thermal signals may be interfered by noise, harmless material inhomogeneity, etc., and the defect in this region is very likely a false alarm, and its defect confidence will be greatly reduced or reset to zero; finally, a self-calibrated defect mapping result is generated, and each defect point in its defect distribution is attached with a confidence level verified by mechanics, so as to achieve automatic verification and calibration of the defect prediction results and ensure the reliability and credibility of the quality inspection results.

[0039] Step S400, perform quality inspection reporting according to the self-calibrated defect mapping result.

[0040] Step S400 further includes step S410, reading the quality inspection data set of rubber products in the batch, and extracting common defect features; step S420, using the common defect features to identify the defect deviation trend and constructing additional warning signals; step S43, after using the additional warning signals to enhance the warning of the self-calibrated defect mapping result, perform warning reporting, and configure quality monitoring attention factors based on the additional warning signals; step S440, collect the attention data of rubber products in the same batch and perform quality inspection identification through the quality monitoring attention factors.

[0041] Preferably, all the quality inspection results of rubber products in the same production batch are pooled to form a batch quality inspection data set, and data analysis is performed on it based on cluster analysis to extract common defect features, that is, regularly recurring defect patterns, for example, most defects are internal bubbles, the bubbles are concentrated in the same specific area of the rubber product, the size and shape of the defects are similar, etc. Then, time series statistical analysis is performed on the common defect features to judge whether the production process is developing in an adverse direction, and further identify the defect deviation trend. For example, the current rubber products are qualified but the average size of the internal bubbles is slowly increasing hour by hour, and the proportion of defects concentrated in a specific area is gradually increasing. When the defect deviation trend is identified, additional warning signals are generated to indicate an increased risk of mass defects in rubber products caused by out-of-control production lines.

[0042] Preferably, additional early warning signals are used to enhance the early warning of self-calibrated defect mapping results. This involves combining macroscopic batch early warnings with microscopic individual product inspection results. When reporting the inspection results of a product, if the defect characteristics of that product highly match the common defect characteristics of the batch early warning, a high-level warning label is added to the report. For example, this defect indicates that it conforms to the current batch risk trend and requires high attention. Furthermore, quality monitoring attention factors are configured based on the additional early warning signals to specify the parameters that need to be monitored. For example, spatial attention factors focus inspection resources on identified high-frequency defect areas; feature attention factors improve the detection sensitivity for specific types of defects such as microbubbles; and process attention factors check specific production equipment or process parameters related to defect trends. The sensor's operating mode is adjusted through these quality monitoring attention factors to perform data acquisition and quality inspection identification of rubber products in the same batch. For example, if the early warning is about microbubbles in a certain area, the ultrasonic sensor uses a higher frequency probe for that area to improve resolution. Computational resources and analysis are focused on features related to the early warning. For example, for bubble warnings, the identification threshold for amplitude offset features is specifically optimized to ensure sensitivity and accuracy for such defects.

[0043] Furthermore, step S400 also includes step S450, configuring a two-level causal inference network, the two-level causal inference network including a local inference layer and a global inference layer; step S460, inputting the self-calibrated defect mapping result into the local inference layer, performing local defect feature modeling, and establishing local inference results; step S470, sending the local inference results, wave impedance distribution data, and surface thermal distribution gradient field to the global inference layer, performing global defect propagation prediction, and establishing prediction fitting results; step S480, after compensating the self-calibrated defect mapping result according to the prediction fitting result, performing quality inspection reporting.

[0044] Preferably, a two-level causal inference network is configured, including a local inference layer and a global inference layer. The local inference layer is used to analyze the microscopic properties of a single defect, while the global inference layer is used to analyze the macroscopic interaction between the defect and the overall product structure and stress field. The self-calibrated defect mapping results are input into the local inference layer to perform local defect feature modeling, that is, to perform a refined analysis of each independent defect and extract its microscopic features, which may include defect type, size and shape, crack sharpness, and defect interface properties, etc., and then to quantitatively assess the severity of each defect, thereby outputting the local inference results. Then, the local inference results, wave impedance distribution data, and surface thermal distribution gradient field are sent to the global inference layer to perform global defect propagation prediction, that is, to simulate the stress concentration of defects, whether defects extend, the mutual influence of adjacent defects, and the influence of weak areas of defects when rubber products are subjected to stress in actual use, and then predict and output the potential risks and evolution paths of the existing defects in the future, and determine the prediction fitting results. Finally, the self-calibrated defect mapping results are compensated based on the prediction fitting results. That is, the self-calibrated defect mapping results are superimposed with the prediction fitting results to correct the severity of the defects. Finally, the quality inspection report is executed. For example, the severity level and risk priority of cracks that are currently very small in size but are predicted to expand rapidly are greatly increased; conversely, the risk level of voids that are currently large in size but are judged to be very stable and will never expand is appropriately reduced.

[0045] Furthermore, step S400 also includes step S490, configuring a detection diversion strategy using the self-calibrated defect mapping result, and configuring the diversion signal of the rubber product; step S4100, binding the self-calibrated defect mapping result with the unique identifier of the rubber product, and performing diversion transmission processing based on the diversion signal.

[0046] Preferably, based on preset business rules and quality standards, a detection diversion strategy is configured. Specifically, products without any defects are diverted to the qualified product channel; products with minor, repairable defects such as small surface scratches are diverted to the rework area channel; products with severe, irreparable defects such as large internal delamination are diverted to the scrap channel; and products with different defect characteristics are diverted to the process reanalysis channel. The self-calibrated defect mapping results are then converted into simple, executable diversion instructions, and diversion signals for rubber products are configured. For example, signal 1 represents qualified, signal 2 represents rework, signal 3 represents scrap, and signal 4 represents process reanalysis. The unique identifier of a rubber product refers to its RFID tag or laser-engraved serial number. After mapping and binding this identifier to the self-calibrated defect mapping results, diversion and transmission processing is performed according to the diversion signals. That is, the sorting mechanism at the end of the production line receives the diversion signals and moves the rubber products with unique identifiers to the corresponding conveyor channels and material bin areas, thereby realizing full lifecycle quality traceability management from detection to sorting and improving the accuracy and reliability of defect identification results.

[0047] In the above text, refer to Figure 1 A method for quality inspection of rubber products according to embodiments of the present invention is described in detail. Next, reference will be made to... Figure 2 A quality inspection system for rubber products according to an embodiment of the present invention is described.

[0048] The quality inspection system for rubber products according to embodiments of the present invention addresses the technical problems in the prior art, such as low efficiency of multimodal sensor data fusion and difficulty in comprehensively identifying deep defects, leading to insufficient accuracy and reliability in the quality inspection of rubber products. It achieves the technical effect of realizing cross-modal collaborative detection and improving the accuracy and reliability of defect identification results. Figure 2 As shown, the quality inspection system for rubber products includes: a data acquisition module 10, a fusion processing module 20, a joint error authentication module 30, and a quality inspection reporting module 40.

[0049] The data acquisition module 10 is used to activate the ultrasonic coherent sensing unit and the infrared spectrum sensing unit to perform data acquisition of the rubber product after it is transferred to the target inspection station, and to establish wave impedance distribution data and surface thermal distribution gradient field. The fusion processing module 20 is used to input the wave impedance distribution data and surface thermal distribution gradient field into the defect causality map channel, and to perform phase inversion fusion processing on the ultrasonic coherent signal and the infrared spectrum temperature signal based on the defect causality map channel, to establish a cross-modal coupled phase tensor, and to identify the response characteristics of deep defects by solving the phase difference matrix. The joint error certification module 30 is used to activate the pressure spectrum sensing unit to apply control pressure to the rubber product, read the local mechanical response spectrum, send the local mechanical response spectrum and the response characteristics of deep defects to the cross-modal calibration layer, perform joint error certification, and generate a self-calibrated defect mapping result. The quality inspection reporting module 40 is used to report the quality inspection based on the self-calibrated defect mapping result.

[0050] The specific configuration of the fusion processing module 20 will be described in detail below. The fusion processing module 20 further includes: activating the data processing layer, performing complex wavelet phase inversion on the wave impedance distribution data, extracting the phase drift rate and amplitude shift features of the local wave impedance, performing temporal gradient convolution on the surface thermal distribution gradient field, and extracting the phase response and temperature curvature distribution of the surface heat flow; constructing a cross-domain phase-amplitude joint feature set from the extraction results, and inputting the cross-domain phase-amplitude joint feature set into the causal association inference layer, calculating the coupling strength between wave impedance disturbance and temperature gradient anomaly through Grapman causal analysis and mutual information entropy weight constraints, and establishing a ternary coupling matrix containing phase drift rate, amplitude shift, and temperature curvature; selecting nodes with coupling strength exceeding the calibration threshold in the ternary coupling matrix as defect activation nodes, performing directed propagation using the phase topology propagation operator to form a defect causal map; performing node connection weight correction through cross-modal phase tensor residuals to perform evolutionary dynamic self-calibration; and identifying deep defect response features based on the calculation results and the self-calibrated defect causal map after calculating the phase difference matrix.

[0051] The specific configuration of the fusion processing module 20 will be described in detail below. The fusion processing module 20 further includes: reading the geometric structure, stress distribution, and functional requirements information of the rubber product to establish a basic information dataset; using the basic information dataset to perform an importance analysis of the rubber's spatial location and configuring a location importance matrix; establishing calibration thresholds for location nodes based on the location importance matrix and functional requirements; and using the calibration thresholds to identify the mapping position coupling strength in the ternary coupling matrix and construct defect activation nodes.

[0052] The specific configuration of the joint error authentication module 30 will be described in detail below. The joint error authentication module 30 further includes: configuring anomaly localization using the deep defect response characteristics, performing mechanical response anomaly analysis based on the local mechanical response spectrum to extract mapping anomaly characteristics; performing mechanical auxiliary authentication based on the mapping anomaly characteristics corresponding to the deep defect response characteristics to generate a self-calibrated defect mapping result.

[0053] The specific configuration of the quality inspection reporting module 40 will be described in detail below. The quality inspection reporting module 40 further includes: reading the quality inspection dataset of a batch of rubber products and extracting common defect features; using the common defect features to identify defect offset trends and constructing additional early warning signals; using the additional early warning signals to enhance the early warning of the self-calibrated defect mapping results, executing an early warning report, and configuring quality monitoring attention factors based on the additional early warning signals; and performing attention data collection and quality inspection identification of the same batch of rubber products through the quality monitoring attention factors.

[0054] The specific configuration of the quality inspection reporting module 40 will be described in detail below. The quality inspection reporting module 40 further includes: configuring a two-level causal inference network, which includes a local inference layer and a global inference layer; inputting the self-calibrated defect mapping result into the local inference layer to perform local defect feature modeling and establish a local inference result; sending the local inference result, wave impedance distribution data, and surface thermal distribution gradient field to the global inference layer to perform global defect propagation prediction and establish a prediction fitting result; and after compensating the self-calibrated defect mapping result based on the prediction fitting result, performing quality inspection reporting.

[0055] The specific configuration of the quality inspection reporting module 40 will be described in detail below. The quality inspection reporting module 40 further includes: configuring a detection diversion strategy using the self-calibrated defect mapping result, configuring a diversion signal for the rubber product; binding the self-calibrated defect mapping result with the unique identifier of the rubber product, and performing diversion transmission processing based on the diversion signal.

[0056] The rubber product quality inspection system provided in this embodiment of the invention can execute the rubber product quality inspection method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects for executing the method.

[0057] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for quality inspection of rubber products, characterized in that, The method includes: When the rubber product is transferred to the target inspection station, the ultrasonic coherent sensing unit and the infrared spectrum sensing unit are activated to perform data acquisition of the rubber product and establish wave impedance distribution data and surface thermal distribution gradient field. The wave impedance distribution data and surface thermal distribution gradient field are input into the defect causality map channel. Based on the defect causality map channel, phase inversion fusion processing is performed on the ultrasonic coherent signal and the infrared temperature signal to establish a cross-modal coupled phase tensor. By solving the phase difference matrix, the response characteristics of deep defects are identified. The pressure spectrum sensing unit is activated to apply control pressure to the rubber product, the local mechanical response spectrum is read, and the local mechanical response spectrum and the deep defect response characteristics are sent to the cross-modal calibration layer to perform joint error authentication and generate self-calibrated defect mapping results. Quality inspection reports are generated based on the self-calibrated defect mapping results.

2. The quality inspection method for rubber products as described in claim 1, characterized in that, The wave impedance distribution data and the surface thermal distribution gradient field are input into the defect causality map channel, including: Activate the data processing layer, perform complex wavelet phase inversion on the wave impedance distribution data, extract the phase drift rate and amplitude shift characteristics of the local wave impedance, perform temporal gradient convolution on the surface heat distribution gradient field, and extract the phase response and temperature curvature distribution of the surface heat flow. After constructing a cross-domain phase-amplitude joint feature set from the extracted results, the cross-domain phase-amplitude joint feature set is input into the causal correlation inference layer. Through Grapman causal analysis and mutual information entropy weight constraints, the coupling strength between wave impedance disturbance and temperature gradient anomaly is calculated, and a ternary coupling matrix including phase drift rate, amplitude offset and temperature curvature is established. In the ternary coupling matrix, nodes with coupling strength exceeding the calibration threshold are selected as defect activation nodes, and directed propagation is performed using the phase topology propagation operator to form a defect causal graph. During the evolution of the defect causal graph, node connection weights are corrected through cross-modal phase tensor residuals, and dynamic self-calibration of the evolution is performed. After calculating the phase difference matrix, the response characteristics of deep defects are identified based on the calculation results and the self-calibrated defect causal map.

3. The quality inspection method for rubber products as described in claim 2, characterized in that, In the ternary coupling matrix, nodes with coupling strength exceeding a calibrated threshold are selected as defect activation nodes, including: Read the geometric structure, stress distribution, and functional requirements information of rubber products to establish a basic information dataset; The importance of the spatial location of rubber is analyzed using the aforementioned basic information dataset, and a location importance matrix is ​​configured. Establish calibration thresholds for location nodes based on the aforementioned location importance matrix and functional requirements; The calibration threshold is used to identify the coupling strength of the mapping position in the ternary coupling matrix, and defect activation nodes are constructed.

4. The quality inspection method for rubber products as described in claim 1, characterized in that, The local mechanical response spectrum and the deep defect response characteristics are sent to the cross-modal calibration layer to perform joint error authentication, including: After configuring the anomaly location using the deep defect response characteristics, the mechanical response anomaly analysis of the anomaly location is performed based on the local mechanical response spectrum to extract the mapping anomaly characteristics. Based on the mapped anomaly features, perform mechanical-assisted authentication of the corresponding deep defect response features to generate self-calibrated defect mapping results.

5. The quality inspection method for rubber products as described in claim 1, characterized in that, The quality inspection report based on the self-calibrated defect mapping results also includes: Read the quality inspection dataset of batch rubber products and extract common defect features; The common defect characteristics are used to identify defect deviation trends and construct additional early warning signals; After enhancing the self-calibration defect mapping result with the additional early warning signal, an early warning is issued, and a quality monitoring attention factor is configured based on the additional early warning signal. The quality monitoring focus factors are used to collect focus data and identify quality issues in the same batch of rubber products.

6. The quality inspection method for rubber products as described in claim 1, characterized in that, The quality inspection report based on the self-calibrated defect mapping results also includes: Configure a two-level causal reasoning network, which includes a local reasoning layer and a global reasoning layer; The self-calibrated defect mapping result is input into the local inference layer to perform local defect feature modeling and establish local inference results. The local inference results, wave impedance distribution data, and surface thermal distribution gradient field are sent to the global inference layer to perform global defect propagation prediction and establish prediction fitting results. After compensating the self-calibrated defect mapping results based on the predicted fitting results, a quality inspection report is generated.

7. The quality inspection method for rubber products as described in claim 1, characterized in that, The quality inspection report based on the self-calibrated defect mapping results also includes: The detection shunting strategy is configured using the self-calibrated defect mapping results, and the shunting signal for the rubber product is configured accordingly. After binding the self-calibrated defect mapping result with the unique identifier of the rubber product, the split transmission process is performed based on the split signal.

8. A quality inspection system for rubber products, characterized in that, The system is used to implement the quality inspection method for rubber products according to any one of claims 1 to 7, and the system comprises: The data acquisition module is used to activate the ultrasonic coherent sensing unit and the infrared spectrum sensing unit to perform data acquisition of the rubber product after the rubber product is transferred to the target inspection station, and to establish wave impedance distribution data and surface thermal distribution gradient field. The fusion processing module is used to input the wave impedance distribution data and surface thermal distribution gradient field into the defect causal map channel. Based on the defect causal map channel, it performs phase inversion fusion processing on the ultrasonic coherent signal and the infrared temperature signal to establish a cross-modal coupled phase tensor. By solving the phase difference matrix, it identifies the response characteristics of deep defects. The joint error authentication module is used to activate the pressure spectrum sensing unit to apply control pressure to the rubber product, read the local mechanical response spectrum, send the local mechanical response spectrum and the deep defect response characteristics to the cross-modal calibration layer, perform joint error authentication, and generate self-calibrated defect mapping results. The quality inspection reporting module is used to report quality inspection results based on the self-calibrated defect mapping results.

Images