Polyurethane elastomer pouring defect prediction-online repair method based on digital twinning-in-situ ultrasonic monitoring

Abstract

The invention relates to the technical field of material science and engineering, and discloses a digital twinning-in-situ ultrasonic monitoring-based polyurethane elastomer pouring defect prediction-on-line repairing method, which comprises the following steps of: establishing a viscosity-temperature-bubble multi-physics field twinning model, and arranging a module in the model; the ultrasonic probe real-time monitoring module collects data in the whole process, the acoustic signal processing module converts original radio-frequency signals into recognizable acoustic features, the AI recognition and analysis module intelligently analyzes defects, the pouring defect prediction module provides a basis for decision making, and the robot online repairing module executes repairing operation. The repair quality verification module carries out instant and quantitative effect evaluation on a repair result, the process parameter optimization feedback module optimizes the process from the source through data mining, the pouring production management module overall coordinates ordered operation and data interaction of all the modules, and finally the purposes of improving the product quality, the production efficiency and the material utilization rate are achieved. And the rejection rate and the labor cost are reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of material science and engineering technology, in particular to a polyurethane elastomer pouring defect prediction-online repair method based on digital twinning-in-situ ultrasonic monitoring. BACKGROUND

[0002] Polyurethane elastomers are widely used in key components such as industrial rollers, seals, tires, etc. due to their excellent wear resistance, tear resistance, and high elasticity. The product quality is highly dependent on the process stability of the pouring forming process. Any process fluctuation can easily produce internal defects such as bubbles, micro-cracks, and shrinkage, which can significantly reduce the mechanical properties and durability of the product, and even cause the product to be scrapped. Traditional quality control methods mainly rely on offline sampling after production (such as cutting sampling, X-ray detection or industrial CT scanning). This method is not only destructive and lagging, but also cannot cover all products, making it difficult to achieve real-time feedback and regulation during the production process, which has become a technical bottleneck restricting the improvement of the quality consistency of high-end polyurethane products.

[0003] In recent years, non-destructive testing techniques, especially ultrasonic testing, have been tried for online monitoring of high polymer materials. However, most existing technical solutions are limited to simple monitoring at a single stage, with obvious limitations: first, there is a lack of multi-station coordinated monitoring of the whole process of mixing-pouring-curing, which cannot capture the complete life cycle of defect generation; second, monitoring and decision-making are separated, usually only able to alarm but not to automatically identify defect types, quantify their sizes and evaluate their risk levels, and there is a lack of prediction ability for future evolution trends of defects; finally, and most importantly, existing systems generally form an "open loop", i.e. defects are found but cannot be immediately intervened, still relying on offline processing by human experience, which is inefficient and difficult to ensure effectiveness. How to build a fully automated closed-loop quality control system that integrates real-time sensing, intelligent diagnosis, forward-looking prediction, precise execution and quality verification, and fundamentally realize the paradigm shift from "passive inspection" to "active prediction and repair", is the core problem that needs to be solved in the current polyurethane precision pouring manufacturing field. SUMMARY

[0004] (I) Technical problems solved In view of the deficiencies of the prior art, the present application provides a polyurethane elastomer pouring defect prediction-online repair method based on digital twinning-in-situ ultrasonic monitoring, which has the advantages of real-time sensing, intelligent diagnosis, forward-looking prediction, precise execution and closed-loop optimization, and solves the problems of traditional offline detection lag, existing online monitoring technology only monitoring, lack of whole-process closed-loop quality control.

[0005] (II) Technical solutions To achieve the above objectives, the present invention provides the following technical solution: a method for predicting and repairing polyurethane elastomer casting defects based on digital twin-in-situ ultrasonic monitoring, comprising the following steps: Step 1: Establish a viscosity-temperature-bubble multiphysics twin model. The system includes a real-time ultrasonic probe monitoring module, an acoustic signal processing module, an AI recognition and analysis module, a casting defect prediction module, a robot online repair module, a repair quality verification module, a process parameter optimization feedback module, and a casting production management module. Step 2: The ultrasonic probe real-time monitoring module deploys ultrasonic probes on the top of the raw material mixing tank, above the casting mold, and at the bottom of the mold to monitor the production process status of polyurethane elastomer casting in real time. Step 3: The acoustic signal processing module preprocesses and extracts features from the raw radio frequency signals acquired by the ultrasonic probe, and transmits the acoustic feature data to the AI ​​recognition and analysis module. Step 4: The AI ​​recognition and analysis module incorporates a deep learning model. This model intelligently identifies defect features and calculates the sound intensity reflection coefficient based on the processed acoustic feature data. Crack depth and attenuation coefficient ; Step 5: The casting defect prediction module, based on the defect features and analysis results identified by AI, combined with the viscosity-temperature-bubble multiphysics twin model, predicts the defect evolution trend, classifies it into low, medium and high risk levels, and annotates the prediction results in the twin model in real time. Step Six: The robot's online repair module receives the predicted defect evolution trend and risk level, and performs robot path planning, glue injection volume calculation, and micro-reaction mixture point injection repair. Step 7: The repair quality verification module calls the ultrasonic probe and calculation formula to perform a second scan and calculation on the repaired area, and evaluates the difference before and after the repair. If it is unqualified, it feeds back to the robot online repair module to adjust the amount of glue or formula, and re-executes the repair and verification until it is qualified. Step 8: The process parameter optimization feedback module counts the occurrence rate of each defect, correlates the coefficient data calculated by AI with the production process parameters, analyzes the correlation of parameters, outputs optimized parameters through data fitting, pushes them to the casting equipment, records the changes in coefficients after optimization, and verifies the optimization effect. Step 9: The casting production management module coordinates the timing of each module, stores full-process data for traceability, provides a visual interface, supports setting probe parameters, viewing twin model defect annotations and coefficient change curves, and allows manual intervention in high-risk defect handling.

[0006] Preferably, the ultrasonic probe real-time monitoring module comprises a raw material mixing section monitoring unit, a casting molding section monitoring unit and a solidification initial stage section monitoring unit; the AI identification and analysis module comprises a bubble defect identification unit, a micro-crack identification unit and a shrinkage identification unit.

[0007] Preferably, the raw material mixing section monitoring unit acquires raw material mixing section data in real time by installing 2-3 groups of high-frequency ultrasonic probes on the top of a raw material stirring tank.

[0008] Preferably, the casting molding section monitoring unit acquires casting molding section data by installing a movable ultrasonic scanning frame above a casting mold, and the probe scans along the mold plane grid.

[0009] Preferably, the solidification initial stage section monitoring unit monitors solidification initial stage section data by installing a low-frequency ultrasonic probe at the bottom of the mold.

[0010] Preferably, the bubble defect identification unit calculates the sound intensity reflection coefficient , and the calculation formula is: In the formula, , the value of the sound intensity reflection coefficient is closer to 1, and the defect reflection is stronger, , respectively represent the acoustic impedance of the polyurethane material and the defect medium.

[0011] Preferably, the micro-crack identification unit calculates the crack depth , and the calculation formula is: In the formula, , represents the crack depth, , represents the propagation speed of ultrasonic waves in polyurethane, , represents the time difference of propagation on both sides of the crack, and represents the probe incidence angle.

[0012] Preferably, the shrinkage identification unit calculates the attenuation coefficient , and the calculation formula is: In the formula, , represents the attenuation coefficient, , represents the sound wave propagation distance, , represents the incident sound wave amplitude, , represents the received sound wave amplitude.

[0013] Preferably, the robot online repair module receives the predicted defect evolution trend and risk level, calibrates the end needle through laser positioning, formulates a repair scheme for different defects, such as bubble with low-viscosity prepolymer, micro-crack with toughening repair agent or shrinkage with high-density mixture, and combines defect size with acoustic intensity reflection coefficient When the acoustic intensity reflection coefficient is high, the glue injection compensation amount is increased, the glue injection amount is controlled through a high-precision metering pump, and after the repair is completed, ultrasonic preliminary reinspection is triggered.

[0014] Preferably, the repair quality verification module calls the ultrasonic probe to perform secondary scanning on the repaired area, collects and calculates the acoustic intensity reflection coefficient , crack depth and attenuation coefficient after repair, compares the data before and after repair, and when the acoustic intensity reflection coefficient is reduced by ≥90%, it is determined to be qualified; when the crack depth is reduced by ≥85%, it is determined to be qualified; and when the attenuation coefficient is reduced by ≥80%, it is determined to be qualified; otherwise, feedback is given to the robot module to adjust the glue injection amount or formula, and the repair and verification are re-executed until the repair is qualified.

[0015] Compared with the prior art, the present application provides a polyurethane elastomer pouring defect prediction-online repair method based on digital twin-in-situ ultrasonic monitoring, which has the following beneficial effects: 1. The present application realizes online detection and evolution trend prediction of internal defects (bubbles, micro-cracks and shrinkage) in the pouring process of polyurethane elastomer by constructing a viscosity-temperature-bubble multi-physical field twin model and integrating an ultrasonic probe real-time monitoring module, an AI intelligent identification and analysis module and a pouring defect prediction module. Ultimately, it achieves a fundamental change from passive inspection to active prediction, greatly improving the foresight and accuracy of quality control.

[0016] 2. The present application realizes millimeter-level positioning, microliter-level glue injection and second-level response through the collaborative work of the robot online repair module and the repair quality verification module to establish a full-automatic online repair closed loop of identification-positioning-repair-verification, realize differentiated and accurate repair and quantitative effect verification based on defect type and risk level, and effectively prevent defects from being missed and over-repaired.

[0017] 3. The present application realizes the traceability of production full-process data and the self-adaptive optimization of process parameters through the data integration and linkage of the process parameter optimization feedback module and the pouring production management module, achieves the leap from single-point defect repair to overall process improvement, continuously optimizes production parameters, ultimately reduces the defect occurrence rate from the source, forms a virtuous cycle of continuous improvement of production quality, and achieves the beneficial effects of greatly improving production efficiency and product consistency. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 The method flowchart of the present application. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0020] Please refer to Figure 1 , a polyurethane elastomer pouring defect prediction-online repair method based on digital twinning-in-situ ultrasonic monitoring, comprising the following steps: Step one, establish a viscosity-temperature-bubble multi-physical field twinning model, and set ultrasonic probe real-time monitoring module, acoustic signal processing module, AI identification and analysis module, pouring defect prediction module, robot online repair module, repair quality verification module, process parameter optimization feedback module and pouring production management module in the system; Step two, the ultrasonic probe real-time monitoring module deploys the ultrasonic probe at the top of the raw material stirring tank, above the pouring mold and at the bottom of the mold, and monitors the production process state of the polyurethane elastomer pouring in real time; Step three, the acoustic signal processing module pre-processes and extracts features from the original radio frequency (RF) signals collected by the ultrasonic probe, and transmits the acoustic feature data to the AI identification and analysis module; Step four, the AI identification and analysis module has a built-in deep learning model, which intelligently identifies defect features (type, size, coordinates) through the deep learning model, and calculates the sound intensity reflection coefficient , crack depth and attenuation coefficient based on the processed acoustic feature data; Step five, the casting defect prediction module predicts the defect evolution trend (such as bubble growth rate, crack propagation speed, and shrinkage diffusion range) based on the defect features and analysis results identified by AI, combined with the viscosity-temperature-bubble multi-physics twin model; according to the defect size (such as crack depth > 1mm), evolution rate and coefficient abnormality degree (such as attenuation coefficient higher than normal area by 30%), and divides low, medium and high risk levels, and marks the prediction results in real time in the twin model; Step six, the robot online repair module receives the predicted defect evolution trend and risk level, and performs path planning, glue injection amount calculation and micro-reaction mixture point injection repair of the robot; Step seven, the repair quality verification module calls the ultrasonic probe and calculation formula to perform secondary scanning and secondary calculation on the repaired area, and evaluates the difference before and after repair. If it is unqualified, it is fed back to the robot online repair module to adjust the glue injection amount or formula, and the repair and verification are re-executed until it is qualified; Step eight, the process parameter optimization feedback module calculates the incidence of each defect (bubble / micro-crack / shrinkage), correlates the coefficient data calculated by AI with the production process parameters (raw material stirring speed, pouring temperature, curing time), analyzes the parameter correlation (such as high stirring speed leading to high bubble reflection coefficient), outputs the optimized parameters (such as stirring speed from 500r / min to 450r / min) through data fitting, pushes them to the casting equipment and records the coefficient changes (such as bubble reflection coefficient mean decreased by 20%) after optimization, and verifies the optimization effect; Step nine, the casting production management module coordinates the timing of each module (ensures that the closed loop of monitoring-identification-prediction-repair-verification is ≤2min), stores the full-process data (including AI calculated coefficients, defect reports, repair records) for traceability (retained for ≥90 days); provides a visual interface to support setting probe parameters, viewing twin model defect labeling and coefficient change curve, and manually intervening in high-risk defect handling (such as emergency shutdown, adjusting repair parameters).

[0021] The advantages are that: through the ultrasonic probe real-time monitoring module, the acoustic signal processing module, the AI identification and analysis module, the pouring defect prediction module, the robot online repair module, the repair quality verification module, the process parameter optimization feedback module and the pouring production management module, among them, the ultrasonic probe real-time monitoring module is used for full-process, multi-station data acquisition, and provides original sensing data for the system; the acoustic signal processing module is used for converting the original radio frequency signal into identifiable acoustic characteristics, providing high-quality input for defect identification; the AI identification and analysis module is used for intelligent, accurate positioning, classification and quantitative analysis of defects; the pouring defect prediction module is used for prospectively judging the defect hazard degree and development trend based on a physical model, providing a basis for decision-making; the robot online repair module is used for performing accurate, adaptive automatic repair operation to eliminate defects; the repair quality verification module is used for instant, quantitative effect evaluation of the repair result, forming a quality closed loop; the process parameter optimization feedback module is used for optimizing the process from the source through data mining to reduce defect generation and realize self-evolution; the pouring production management module is used for overall coordination of the orderly operation and data interaction of each module, providing a man-machine interface, finally realizing the full-process automation and intelligentization of the defect "perception-identification-prediction-decision-execution-verification-optimization" in the polyurethane pouring production process, significantly improving product quality, production efficiency and material utilization rate, and reducing waste product rate and labor cost.

[0022] The ultrasonic probe real-time monitoring module includes a raw material mixing section monitoring unit, a pouring forming section monitoring unit and a solidification initial stage section monitoring unit; the AI identification and analysis module includes a bubble defect identification unit, a micro-crack identification unit and a shrinkage identification unit.

[0023] The raw material mixing section monitoring unit acquires raw material mixing section data in real time by installing 2-3 groups of high-frequency ultrasonic probes (center frequency 5-10 MHz) on the top of the raw material stirring tank, including ultrasonic echo signals, viscosity data, flow state data, temperature data, raw material mixing uniformity and initial bubble generation data of the mixed system.

[0024] The pouring forming section monitoring unit acquires pouring forming section data by installing a movable ultrasonic scanning frame (scanning accuracy ±0.1 mm) above the pouring mold, and the probe scans along the mold plane grid to obtain pouring layer thickness, flow front position, interface bonding quality, internal bubble distribution and local temperature field data.

[0025] The solidification initial stage section monitoring unit monitors solidification initial stage section data by installing a low-frequency ultrasonic probe (center frequency 1-2 MHz) at the bottom of the mold, including polyurethane viscosity dynamic change (indirectly represented by ultrasonic wave propagation speed) structure deformation data, residual stress distribution and micro-crack initiation signal during the solidification process.

[0026] The advantages are that the raw material mixing section monitoring unit, the pouring forming section monitoring unit and the initial curing section monitoring unit are arranged in the real-time monitoring module of the ultrasonic probe, the raw material mixing section monitoring unit performs online monitoring on the uniformity of raw material mixing and initial bubbles, and the quality is controlled from the source, the pouring forming section monitoring unit performs scanning monitoring on the flow state and macroscopic defects in the pouring process, and ensures that the forming process is controllable, the initial curing section monitoring unit performs monitoring on the curing reaction progress and the generation of microscopic defects, and captures the generation of late defects, finally, the key quality characteristics of the whole process of "mixing-pouring-curing" are seamlessly and continuously monitored by ultrasonic waves, a complete production state information chain is constructed, and a comprehensive and reliable data basis is provided for subsequent intelligent identification and decision-making.

[0027] The bubble defect recognition unit calculates the sound intensity reflection coefficient , and the calculation formula is: In the formula, represents the sound intensity reflection coefficient, the value is closer to 1, the defect reflection is stronger, and it is more likely to be a bubble or a hole, , respectively represent the acoustic impedance of the polyurethane material and the defect medium, and the acoustic impedance of the bubble is extremely low, resulting in extremely large.

[0028] The advantages are that the sound intensity reflection coefficient is calculated and used as the core acoustic criterion for distinguishing bubble / hole defects and the matrix, when the sound intensity reflection coefficient is close to 1, it is determined that there is a high acoustic impedance difference defect (such as a bubble) at this position, when the sound intensity reflection coefficient is in the middle range, it can assist in determining other inclusions, and when the sound intensity reflection coefficient is close to 0, it indicates that the sound wave is completely transmitted and the material continuity is good, and finally, the bubble defects are automatically identified and quantitatively evaluated with high sensitivity and high reliability.

[0029] The micro-crack recognition unit calculates the crack depth , and the calculation formula is: In the formula, represents the crack depth, represents the propagation speed of the ultrasonic wave in the polyurethane (which changes with temperature and needs to be calibrated in real time), represents the propagation time difference of the two sides of the crack, and represents the probe incident angle.

[0030] The advantages are that the crack depth the calculation of the attenuation coefficient, which is used as a key quantitative indicator for evaluating the severity of cracks and the urgency of repair, when the crack depth is shallow (e.g., <0.5 mm), it is classified as a low-risk defect and is included in the observation list, when the crack depth is deep (e.g., >1 mm) or rapidly expanding, it is immediately classified as a high-risk defect and triggers an emergency repair procedure, when the crack depth falls below the threshold after repair, it is determined to be a qualified repair, and finally achieves the effect of precise management of the entire life cycle of crack defects from detection, risk assessment to repair verification.

[0031] The shrinkage identification unit calculates the attenuation coefficient , whose calculation formula is: In the formula, represents the attenuation coefficient (dB / cm), represents the sound wave propagation distance (cm), represents the incident sound wave amplitude, represents the received sound wave amplitude, the area with shrinkage defects has enhanced scattering of ultrasonic waves, resulting in an abnormally high attenuation coefficient .

[0032] The advantage is that through the calculation of the attenuation coefficient , it is used as a sensitive indicator for evaluating the internal micro-compactness (e.g., shrinkage) of the material, when the attenuation coefficient is significantly higher than the normal surrounding area, indicating that there are a large number of scatterers (e.g., micro-pores) in the area, i.e., shrinkage defects, when the attenuation coefficient presents an abnormal distribution pattern, it can assist in determining the distribution range of shrinkage, when the attenuation coefficient returns to normal levels after repair, it indicates that the microstructure has been repaired and compacted, and finally achieves the effect of effective detection, positioning, and repair verification of microscopic shrinkage defects that are difficult to detect by the naked eye.

[0033] The robot online repair module receives the predicted defect evolution trend and risk level, calibrates the end needle (accuracy ±0.05 mm) through laser positioning; develops repair schemes for different defects (low-viscosity pre-polymer for bubbles, toughening repair agent for micro-cracks, and high-compactness mixture for shrinkage); combined with defect size and sound intensity reflection coefficient (e.g., increase the glue compensation amount if the sound intensity reflection coefficient is high), control the glue injection amount through a high-precision metering pump (minimum glue injection amount 0.01 mL), and trigger ultrasonic preliminary re-inspection after completion of repair.

[0034] The advantages are: the end needle is calibrated by the above-mentioned laser positioning (the precision is ±0.05 mm), the defect type is differentiated for repair, the glue injection amount is controlled through the sound intensity reflection coefficient linkage, the high-precision metering pump (the minimum glue injection amount is 0.01 mL), the repaired ultrasonic preliminary review is performed, and finally the industrial-level repair effect of micron-level positioning precision, milligram-level glue injection control and one-time repair qualified rate > 95% is achieved.

[0035] The repair quality verification module calls the ultrasonic probe to perform secondary scanning on the repaired area, collects and calculates the sound intensity reflection coefficient , crack depth and attenuation coefficient after repair, compares the data before and after repair, and when the sound intensity reflection coefficient is reduced by ≥90% (for example, the bubble reflection coefficient is reduced from 0.8 to ≤0.08), it is determined to be qualified; when the crack depth is reduced by ≥85% (for example, from 0.6 mm to ≤0.09 mm), it is determined to be qualified; when the attenuation coefficient is reduced by ≥80% (for example, from 4.2 dB / cm to ≤0.8 dB / cm), it is determined to be qualified, and if it is unqualified, it is fed back to the robot module to adjust the glue injection amount or formula, and the repair and verification are re-executed until it is qualified.

[0036] The advantages are: through the above-mentioned secondary scanning review, quantitative comparison of multiple parameters (sound intensity reflection coefficient, crack depth, attenuation coefficient), setting of clear mathematical qualified criteria (such as a reduction of 90%) and closed-loop feedback iteration mechanism of unqualified items, finally the quality verification is changed from subjective experience judgment to objective data driving to ensure that each repair point meets strict quantitative quality standards.

[0037] Summary: The present application realizes early detection and evolution prediction of defects by constructing a multi-physical field twin model of viscosity-temperature-bubble and combining in-situ ultrasonic monitoring and AI identification, realizes accurate online repair and immediate verification by a robot, forms an automatic closed loop of discovery-prediction-repair-verification, changes quality control from post-treatment to in-process intervention and pre-prediction, and finally suppresses the generation of defects from the source through process parameter optimization feedback, thereby greatly improving production efficiency and product yield.

[0038] Although embodiments of the present application have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the present application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for polyurethane elastomer casting defect prediction - online repair based on digital twin - in-situ ultrasonic monitoring, characterized in that, The method comprises the following steps: Step one, establish a viscosity-temperature-bubble multi-physics field twin model, the system is provided with an ultrasonic probe real-time monitoring module, an acoustic signal processing module, an AI identification and analysis module, a pouring defect prediction module, a robot online repair module, a repair quality verification module, a process parameter optimization feedback module and a pouring production management module; Step two, the ultrasonic probe real-time monitoring module deploys the ultrasonic probe at the top of the raw material mixing tank, above the pouring mold and at the bottom of the mold, and monitors the production process state of the polyurethane elastomer pouring in real time; Step three, the acoustic signal processing module pre-processes and extracts features from the original radio frequency signals collected by the ultrasonic probe, and transmits the acoustic feature data to the AI identification and analysis module; Step four, the AI recognition and analysis module is built-in with a deep learning model, which intelligently identifies defect features through the deep learning model, and calculates the sound intensity reflection coefficient based on the processed acoustic characteristic data , crack depth , and attenuation coefficient ; Step five, the pouring defect prediction module predicts the defect evolution trend based on the defect features and analysis results identified by AI, combines the viscosity-temperature-bubble multi-physics field twin model, divides the low, medium and high risk levels, and marks the prediction results in real time in the twin model; Step six, the robot online repair module receives the predicted defect evolution trend and risk level, plans the path of the robot, calculates the glue injection amount and points the micro-reaction mixture for repair; Step seven, the repair quality verification module calls the ultrasonic probe and calculation formula to perform secondary scanning and calculation on the repaired area, evaluates the difference before and after repair, and feeds back to the robot online repair module to adjust the glue injection amount or formula if unqualified, and re-executes the repair and verification until qualified; Step eight, the process parameter optimization feedback module calculates the occurrence rate of each defect, correlates the coefficient data calculated by AI with the production process parameters, analyzes the parameter correlation, outputs the optimized parameters through data fitting, pushes them to the pouring equipment and records the coefficient changes after optimization, and verifies the optimization effect; Step nine, the pouring production management module coordinates the timing of each module, stores the whole process data for traceability, provides a visual interface, supports setting probe parameters, viewing twin model defect labeling and coefficient change curve, and can manually intervene in high-risk defect treatment.

2. The digital twin based in-situ ultrasonic monitoring of polyurethane elastomer casting defects prediction - online repair method as claimed in claim 1, wherein: The ultrasonic probe real-time monitoring module comprises a raw material mixing section monitoring unit, a pouring forming section monitoring unit and a solidification initial stage section monitoring unit; the AI identification and analysis module comprises a bubble defect identification unit, a micro-crack identification unit and a shrinkage identification unit.

3. The digital twin - in-situ ultrasonic monitoring based polyurethane elastomer casting defect prediction - online repair method according to claim 2, characterized in that: The raw material mixing section monitoring unit installs 2-3 groups of high-frequency ultrasonic probes at the top of the raw material mixing tank to collect raw material mixing section data in real time.

4. The digital twin based in-situ ultrasonic monitoring of polyurethane elastomer casting defects prediction - online repair method as claimed in claim 2, wherein: The pouring forming section monitoring unit installs a movable ultrasonic scanning frame above the pouring mold, and the probe scans along the mold plane grid to obtain pouring forming section data.

5. The digital twin based in-situ ultrasonic monitoring of polyurethane elastomer casting defects prediction - online repair method as claimed in claim 2, wherein: The solidification initial stage section monitoring unit installs a low-frequency ultrasonic probe at the bottom of the mold to monitor the solidification initial stage section data.

6. The Digital Twin based in-situ ultrasonic monitoring of polyurethane elastomer casting defects prediction - online repair method as claimed in claim 2, wherein: The bubble defect recognition unit calculates the sound intensity reflection coefficient The calculation formula is: In the formula, represents the sound intensity reflection coefficient, the value of which is closer to 1, the stronger the defect reflection is, , respectively represent the acoustic impedance of the polyurethane material and the defect medium.

7. The Digital Twin based in-situ ultrasonic monitoring of polyurethane elastomer casting defects prediction - online repair method as claimed in claim 2, wherein: The micro crack recognition unit calculates the crack depth The calculation formula is: In the formula, represents the crack depth, represents the propagation speed of the ultrasonic wave in the polyurethane, Δt represents the difference in propagation time on both sides of the crack, and θ represents the probe incidence angle.

8. The digital twin - in-situ ultrasonic monitoring based polyurethane elastomer casting defect prediction - online repair method as claimed in claim 2, wherein: The shrinkage identification unit calculates an attenuation coefficient The calculation formula is: In the formula, denotes the attenuation coefficient, denotes the sound wave propagation distance, denotes the incident sound wave amplitude, denotes the received sound wave amplitude.

9. The Digital Twin based in-situ ultrasonic monitoring of polyurethane elastomer casting defects prediction - online repair method as claimed in claim 1, wherein: The robot online repair module receives the predicted defect evolution trend and risk level, calibrates the end needle through laser positioning, formulates repair schemes for different defects, such as bubble repair with low-viscosity prepolymer, micro-crack repair with toughening repair agent, or shrinkage repair with high-density mixture, and combines defect size and acoustic intensity reflection coefficient When the acoustic intensity reflection coefficient is high, the glue injection compensation amount is increased, the glue injection amount is controlled through a high-precision metering pump, and after the repair is completed, ultrasonic preliminary review is triggered.

10. The digital twin - in-situ ultrasonic monitoring based polyurethane elastomer casting defect prediction - online repair method as claimed in claim 1, wherein: The repair quality verification module calls the ultrasonic probe to perform secondary scanning on the repaired area, collects and calculates the acoustic intensity reflection coefficient of the repaired area , crack depth and attenuation coefficient , compares the data before and after the repair, and when the acoustic intensity reflection coefficient is reduced by ≥90%, it is determined to be qualified; when the crack depth is reduced by ≥85%, it is determined to be qualified; when the attenuation coefficient is reduced by ≥80%, it is determined to be qualified, and if not, feedback is given to the robot module to adjust the glue injection amount or formula, and the repair and verification are re-executed until the repair is qualified.

Images