A production line yield intelligent management method and system

Abstract

The embodiment of the application provides a production line yield intelligent management method and system, relates to the technical field of production line yield intelligent management, and comprises the following steps: collecting operation data of a production line equipment, analyzing the operation data, identifying equipment micro-vibration, establishing a conduction path model of the equipment micro-vibration, establishing a mapping relationship between vibration conduction intensity and product defects according to the conduction path model, predicting product defects caused by the equipment micro-vibration according to the equipment micro-vibration and the mapping relationship, performing cross-validation on the predicted product defects, identifying hidden product defects, performing multi-source consistency checking and reverse deduction verification on online detection data input into a yield prediction system to identify online detection data distortion, and calibrating core parameters of the yield prediction system according to the predicted product defect information and the identified hidden product defect information. The application can improve the intelligent management level and prediction accuracy of the production line yield.

Description

Technical Field

[0001] This application relates to the field of intelligent management technology for production line yield, and in particular to an intelligent management method and system for production line yield. Background Technology

[0002] In modern industrial production, intelligent management systems effectively monitor and adjust process parameters to maintain high yield rates through rule libraries that generate process parameter adjustment suggestions. However, during long-term, high-intensity production operations, critical components of precision manufacturing equipment may experience extremely minor structural vibration frequency shifts, resulting in continuous, low-frequency micro-vibrations. These micro-vibrations exceed the monitoring thresholds of conventional sensors and are not identified as fault signals by the intelligent management system. These micro-vibrations are transmitted through the equipment base and factory floor to adjacent precision assembly molds, causing slow wear on critical mating surfaces and slight deviations in key product dimensions or clearances. Simultaneously, these micro-vibrations can also be transmitted to precision metering units, interfering with weighing sensors and causing milligram-level random errors in auxiliary material dispensing, compromising the accuracy of raw material proportions, and ultimately leading to abnormal physical or chemical properties of the product. To avoid production line shutdowns due to a large backlog of "edge-defect" products, the production line quality manager may relax the classification threshold of the defect image recognition system. This adjustment causes products with slight dimensional deviations to be incorrectly marked as "qualified" by the system, meaning the online inspection data received by the intelligent management system no longer reflects the true physical quality of the product. The input of this "optimized" data into the yield prediction module causes a deviation in the module's core weight parameters, which greatly reduces its ability to predict the actual yield of the production line, and may even lead to incorrect judgments. Summary of the Invention

[0003] This application aims to address at least one of the technical problems existing in the prior art. To this end, this application proposes an intelligent management method and system for production line yield, aiming to improve the intelligent management level and prediction accuracy of production line yield.

[0004] In a first aspect, embodiments of this application provide an intelligent management method for production line yield, including:

[0005] Collect operating data from production line equipment and analyze the operating data to identify micro-vibrations in the equipment;

[0006] Based on the identified device micro-vibrations, a transmission path model for the device micro-vibrations is established;

[0007] Based on the transmission path model, establish the mapping relationship between vibration transmission intensity and product defects;

[0008] Predict product defects caused by the equipment micro-vibration based on the equipment micro-vibration and the mapping relationship;

[0009] For predicted product defects, cross-validation is performed by combining manual re-inspection results and independent sensor data that are unaffected by vibration to identify masked product defects;

[0010] Multi-source consistency verification and reverse derivation verification are performed on the online detection data input into the yield prediction system to identify distortion of the online detection data;

[0011] When the online detection data is distorted, the core parameters of the yield prediction system are calibrated based on the predicted product defect information and the identified masked product defect information.

[0012] According to some embodiments of this application, when the online detection data is distorted, the step of calibrating the core parameters of the yield prediction system based on the predicted product defect information and the identified masked product defect information includes:

[0013] Identify conflict information and sources of conflict information between the predicted product defects and the identified masked product defects;

[0014] Determine the confidence level of conflicting information and the severity of its impact on the core functions of the product;

[0015] Based on the source of the conflicting information, the confidence level of the conflicting information, and the severity of its impact on the core functions of the product, the conflicting information is prioritized and a decision is made.

[0016] Based on the ruling result, the conflict information is fused and corrected to obtain the fused and corrected defect information;

[0017] Using the fused and corrected defect information, combined with the online detection data, the true value of the original data is deduced in reverse.

[0018] The core parameters of the yield prediction system are calibrated based on the true values.

[0019] According to some embodiments of this application, the steps of determining the confidence level of conflict information and the severity of the impact of conflict information on the core functions of the product include:

[0020] Calculate the confidence level of conflict information based on the degree of conformity between the inspector's judgment and the actual performance, as well as the results of blind testing;

[0021] Track the subsequent performance of the released product, and when a fault caused by conflicting information is discovered, determine the severity of the impact of the source of the conflicting information on the core function of the product based on the fault.

[0022] According to some embodiments of this application, the step of tracking the subsequent performance of the released product and, upon discovering a fault related to conflict information, determining the severity of the impact of the source of the conflict information on the core functions of the product based on the actual failure situation includes:

[0023] A fault attribution map is established for each released product. The fault attribution map includes the product's process parameters on the production line, online detection data, conflict information during adjudication, and environmental exposure data collected subsequently.

[0024] When a fault is found in the released product, the fault path backtracking mechanism is activated, and the fault tracing result is obtained by tracing the production process parameters and defect information that are strongly correlated with the fault type according to the fault attribution map.

[0025] Analyze the environmental conditions and operating conditions at the time of the failure, and compare them with information that was not identified as conflicting at the time of adjudication, so as to eliminate external failure factors that are not directly related to the conflicting information at the time of adjudication and obtain the result of non-conflict factor elimination.

[0026] Based on the traceability results and the elimination results of non-conflict factors, the actual failure situation is confirmed;

[0027] Based on the actual failure scenarios, determine the severity of the impact of the conflict information source on the core functions of the product.

[0028] According to some embodiments of this application, the step following the identification of conflict information and the source of conflict information between the predicted product defect and the identified masked product defect includes:

[0029] Identify the interdependencies between conflicting information sources;

[0030] Quantify the strength and direction of the interdependence;

[0031] Based on the quantified interdependencies, the weights of the affected conflict information sources are adjusted to correct the conflict information sources.

[0032] According to some embodiments of this application, the step of calculating the confidence level of conflict information based on the inspector's judgment of conformity with actual performance and blind test results includes:

[0033] When inspectors make defect judgments, their judgment results should be recorded.

[0034] The inspector's judgment is compared with the performance data of the product obtained by independent sensors in subsequent production or actual use to evaluate the degree of conformity between the inspector's judgment and the actual performance of the product.

[0035] When the sample size of blind testing is insufficient to cover all defect types, based on historical defect distribution data and the current production line status, identify defect types that are not covered by blind testing but have a significant impact on the core functions of the product.

[0036] For the defect types not covered by blind testing, virtual blind testing results are generated by analyzing historical data of such defects and combining the prediction results of physical models related to such defects.

[0037] The inspector's individual confidence level is updated based on the degree of conformity between the inspector's judgment and the actual performance of the product, as well as the results of the virtual blind test.

[0038] Based on the individual confidence level, adjust the confidence level calibration factor and calculate the confidence level of conflict information.

[0039] According to some embodiments of this application, the step of identifying defect types that are not covered by blind testing but have a significant impact on the core functions of the product, based on historical defect distribution data and the current production line status, when the blind testing sample size is insufficient to cover all defect types, includes:

[0040] Receive historical defect distribution data and obtain cleaned and supplemented historical defect distribution data through a multi-source cross-validation mechanism;

[0041] Real-time monitoring of fluctuation patterns in key sensor data from the production line allows for the identification of abnormal states detected in real time.

[0042] The historical defect distribution data after cleaning and filling is correlated with the abnormal states detected in real time to obtain the correlation analysis results;

[0043] The defect impact factor is quantified based on the severity of the impact of historical defects on the core functions of the product, and the defect impact factor is corrected based on the type and intensity of the transient disturbances in the current production line to obtain the corrected defect impact factor.

[0044] Based on the correlation analysis results and the revised defect impact factor, identify defect types that were not covered by blind testing and have a significant impact on the core functions of the product.

[0045] According to some embodiments of this application, the step of quantifying the defect impact factor based on the severity of the impact of historical defects on the core functions of the product, and correcting the defect impact factor based on the type and intensity of the transient disturbances in the current production line to obtain the corrected defect impact factor includes:

[0046] By establishing a multi-dimensional defect risk assessment model, the defect impact factors are comprehensively quantified to obtain a comprehensive quantitative result. The multi-dimensional factors include the frequency of occurrence of historical defects, repair costs, security risks, and user complaint rates.

[0047] For long-tail defects that occur at extremely low frequencies in the historical defects, the influence factors of the long-tail defects are evaluated by analogy based on the physical similarity between the long-tail defects and known high-impact defects, as well as the results of simulation tests under extreme working conditions, to obtain the long-tail defect analogy evaluation results.

[0048] For the unrecorded defects in the historical defects, by analyzing the physical relationship between the generation mechanism of the unrecorded defects and the core functions of the product, and combining the expert's judgment on the potential failure modes that the unrecorded defects may cause in specific application scenarios, the preliminary impact factor assessment results of the new defects are obtained.

[0049] Based on the comprehensive quantitative results, the long-tail defect analogy evaluation results, and the preliminary impact factor evaluation results of the novel defect, the quantified defect impact factor is obtained.

[0050] The quantified defect impact factor is corrected based on the type and intensity of transient disturbances in the current production line.

[0051] According to some embodiments of this application, the step of correcting the quantified defect impact factor based on the type and intensity of the current production line transient disturbance includes:

[0052] Identify the type and intensity of transient disturbances in the current production line;

[0053] The interaction between the transient disturbances is analyzed to obtain the comprehensive effect of the transient disturbances on the defect influence factor;

[0054] Based on the combined effect of the transient disturbance on the defect impact factor, the physical response of sensor data under the transient disturbance is analyzed, the combined impact of the transient disturbance on product performance or structural stability is inferred, and a correction factor is obtained.

[0055] Based on the correction factor, and considering the differences in the impact of transient disturbances on different defect types, the quantified defect impact factor is corrected.

[0056] Secondly, embodiments of this application provide an intelligent management system for production line yield, comprising:

[0057] The data acquisition and analysis module is used to collect operating data of the production line equipment and analyze the operating data to identify micro-vibrations of the equipment.

[0058] The transmission path model establishment module is used to establish a transmission path model of the identified device micro-vibrations based on the identified device micro-vibrations.

[0059] The mapping relationship establishment module is used to establish a mapping relationship between vibration transmission intensity and product defects based on the transmission path model.

[0060] The defect prediction module is used to predict product defects caused by the micro-vibration of the equipment based on the mapping relationship;

[0061] The defect identification module is used to identify masked product defects by cross-validating the results of manual re-inspection with independent data for predicted product defects.

[0062] The data distortion identification module is used to perform multi-source consistency verification and reverse derivation verification on the online detection data input to the yield prediction system in order to identify distortion in the prior detection data;

[0063] The calibration module is used to calibrate the core parameters of the yield prediction system based on the predicted product defect information and the identified masked product defect information when the online detection data is distorted.

[0064] The technical solution according to the embodiments of this application has at least the following beneficial effects: The intelligent management method for production line yield rate disclosed in this application collects the operating data of production line equipment and identifies equipment micro-vibrations, thereby establishing a transmission path model of equipment micro-vibrations and a mapping relationship between vibration transmission intensity and product defects, which can accurately predict product defects caused by equipment micro-vibrations. For the predicted product defects, cross-validation is performed by combining manual re-inspection results and independent sensor data to effectively identify masked product defects, solving the problem in the prior art that hidden defects caused by equipment micro-vibrations are difficult to detect. Furthermore, this application performs multi-source consistency verification and reverse derivation verification on the online detection data input to the yield rate prediction system, which can identify online detection data distortion, and when data distortion occurs, calibrates the core parameters of the yield rate prediction system based on the predicted product defect information and the identified masked product defect information. This application, through comprehensive and in-depth analysis and processing of equipment micro-vibration, product defects, and online detection data, can penetrate the distorted data appearance, identify the types of hidden defects, and attribute them together with visible defects to difficult-to-detect physical roots, thereby providing an effective and fundamental solution that significantly improves the intelligent management level and prediction accuracy of production line yield.

[0065] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0066] The accompanying drawings are used to provide a further understanding of the technical solutions of this application and constitute a part of the specification. They are used together with the embodiments of this application to explain the technical solutions of this application and do not constitute a limitation on the technical solutions of this application.

[0067] Figure 1A flowchart illustrating an intelligent management method for production line yield rate provided in one embodiment of this application;

[0068] Figure 2 This is a schematic diagram of an intelligent management system for production line yield rate provided in one embodiment of this application. Detailed Implementation

[0069] To make the objectives, technical methods, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0070] It should be noted that the meaning of "multiple" (or "more than") in the description of the embodiments of this application refers to two or more, and "greater than," "less than," "exceeding," etc. are understood to exclude the number itself, while "above," "below," "within," etc. are understood to include the number itself. If "first," "second," etc. are used in the description, they are only for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated or the order of the technical features indicated.

[0071] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent the existence of A alone, the simultaneous existence of A and B, or the existence of B alone. A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of singular or plural items. For example, at least one of a, b, and c can represent: the existence of a alone, the existence of b alone, the existence of c alone, the simultaneous existence of a and b, the simultaneous existence of a and c, the simultaneous existence of b and c, or the simultaneous existence of a, b, and c, where a, b, and c can be single or multiple.

[0072] In the description of this application, unless otherwise expressly defined, terms such as "setup," "installation," and "connection" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this application in conjunction with the specific content of the technical solution.

[0073] Based on the above, this application proposes an intelligent management method and system for production line yield, aiming to improve the intelligent management level and prediction accuracy of production line yield.

[0074] The intelligent production line yield management method provided in this application can be applied to a terminal, a server, or software running on either a terminal or a server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, etc.; the server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms; the software can be an application that implements the intelligent production line yield management method, but is not limited to the above forms.

[0075] This application can be applied to numerous general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via communication networks. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices. It should be noted that in various specific embodiments of this invention, when processing is required based on data related to the characteristics of an object (e.g., user attributes or sets of attribute information), permission or consent from the corresponding object is obtained first, and the collection, use, and processing of this data comply with relevant laws and standards. Furthermore, when the embodiments of the present invention need to obtain the attribute information of an object, they will obtain the separate permission or separate consent of the corresponding object through pop-up windows or redirection to a confirmation page. After obtaining the separate permission or separate consent of the corresponding object, they will then obtain the relevant data of the object necessary for the embodiments of the present invention to operate normally.

[0076] See Figure 1 , Figure 1This is a flowchart illustrating an intelligent production line yield management method according to an embodiment of this application. The intelligent production line yield management method provided in this embodiment includes, but is not limited to, steps S110 to S170, which will be described in detail below.

[0077] Step S110: Collect operating data of the production line equipment and analyze the operating data to identify micro-vibrations of the equipment;

[0078] Step S120: Based on the identified equipment micro-vibrations, establish a transmission path model for the equipment micro-vibrations;

[0079] Step S130: Based on the transmission path model, establish the mapping relationship between vibration transmission intensity and product defects;

[0080] Step S140: Predict product defects caused by equipment micro-vibration based on equipment micro-vibration and mapping relationship;

[0081] Step S150: For the predicted product defects, cross-validate the results of manual re-inspection and independent sensor data that are not affected by vibration to identify the masked product defects.

[0082] Step S160: Perform multi-source consistency verification and reverse derivation verification on the online detection data input to the yield prediction system to identify distortion of the online detection data;

[0083] Step S170: When online detection data is distorted, calibrate the core parameters of the yield prediction system based on the predicted product defect information and the identified masked product defect information.

[0084] It should be noted that "equipment micro-vibration" refers to extremely small vibrations generated by production equipment during operation, which are difficult to detect by conventional sensors. Although these vibrations are weak, their long-term accumulation can negatively impact product quality. "Transmission path model" refers to a mathematical or physical model describing how equipment micro-vibrations propagate through physical media (such as equipment base, factory floor) to other equipment or product production stages. "Vibration transmission intensity" quantifies the degree of attenuation or amplification of micro-vibrations along the transmission path. "Product defect" refers to any problem that occurs during product production that does not meet quality standards. "Mapping relationship" refers to the association rules between vibration transmission intensity and specific product defect types and their severity. "Online inspection data" refers to product quality inspection data collected in real time on the production line. "Multi-source consistency verification" refers to verifying the consistency between data by comparing data from different sensors or detection systems. "Reverse derivation verification" refers to deducing the rationality of input data from known results to identify data distortion. The "yield prediction system" is an intelligent system that predicts the yield rate of the production line based on historical and real-time data; its "core parameters" are key weights or model coefficients that affect the accuracy of the prediction. This method is deployed in an industrial IoT platform or a manufacturing execution system (MES) and is implemented by integrating various sensors, data analysis modules, and decision support systems.

[0085] In one embodiment, operational data of the production line equipment is collected and analyzed to identify equipment micro-vibrations. This data can be collected by installing high-sensitivity accelerometers or laser displacement sensors at key locations on the production equipment. These sensors can capture minute vibration signals generated by the equipment during normal operation. The collected raw data is then input into a data analysis module. By analyzing the frequency characteristics, amplitude variations, or time-domain waveforms of the data, specific patterns or abnormal signals related to the equipment micro-vibrations can be identified. Next, based on the identified equipment micro-vibrations, a transmission path model of these vibrations is established. After identifying the equipment micro-vibrations, it is necessary to further understand how these vibrations affect other equipment or products on the production line. For example, additional vibration sensors can be deployed at different locations on the production line (such as equipment bases, molds, metering units, etc.) to simultaneously collect vibration data from these locations. Then, using this multi-point data, combined with physical modeling methods such as finite element analysis (FEA) or modal analysis, a transmission path model of the equipment micro-vibrations is constructed. For example, a three-dimensional finite element model can be established to simulate the process of equipment micro-vibrations being transmitted from the factory floor to the precision assembly mold, and the energy loss along different transmission paths can be quantified. Secondly, based on the transmission path model, a mapping relationship between vibration transmission intensity and product defects is established. After clarifying the transmission path of equipment micro-vibrations, it is necessary to further explore the specific impact of these micro-vibrations on product quality. Through statistical analysis, regression analysis, or pattern recognition, vibrations at different transmission intensities are correlated with corresponding product defects, thereby establishing a mapping relationship between vibration transmission intensity and product defects. Furthermore, when the system identifies micro-vibration in a certain piece of equipment, it first uses the transmission path model to calculate the transmission intensity of this micro-vibration in each key production stage. Then, these transmission intensities are input into the established mapping relationship model to predict the types, probabilities, and potential severity of product defects that may be caused by these micro-vibrations.

[0086] Furthermore, for predicted product defects, cross-validation is performed using manual re-inspection results and independent sensor data unaffected by vibration to identify masked product defects. Since some defects may be masked due to human intervention in the production line, such as adjusting detection thresholds, cross-validation is necessary to identify these masked defects. For example, for products predicted by the system to be defective, experienced human inspectors can be assigned to re-inspect and record the results. Simultaneously, independent sensors unaffected by equipment micro-vibrations (e.g., high-precision dimensional measurement using laser interferometers or material composition detection using chemical analyzers) are deployed to obtain accurate product quality data. By comparing the prediction results, manual re-inspection results, and independent sensor data, defects that the online inspection system failed to identify but actually exist—i.e., masked product defects—can be discovered. Moreover, multi-source consistency checks and reverse derivation verification are performed on the online inspection data input to the yield prediction system to identify online inspection data distortion. To ensure the accuracy of the yield prediction system, strict quality control of the input data is required. Significant differences between data from different sources indicate potential data distortion. Simultaneously, known production process parameters and product design specifications can be used to reverse-engineer and verify online inspection data. For example, based on the product's design dimensions and material properties, a reasonable range for online inspection data can be derived. If the actual inspection data exceeds this range, the data is considered potentially distorted. Finally, when online inspection data is distorted, the core parameters of the yield prediction system are calibrated based on predicted product defect information and identified masked product defect information. When online inspection data distortion is detected, traditional yield prediction systems may give incorrect predictions. In this case, more reliable defect information is needed to calibrate the system. For example, if it is found that the online inspection data underestimates the incidence of dimensional deviation defects, the weight parameters related to dimensional deviations in the prediction system can be increased to make it more sensitive to such defects.

[0087] It's important to clarify that identifying conflicting information between predicted and concealed product defects, and the sources of this conflicting information, involves cross-validating the predicted product defects caused by equipment micro-vibrations with those identified through manual re-inspection and independent sensor data unaffected by vibration. This process identifies inconsistencies in defect type, location, severity, and other information, tracing the origins of these inconsistencies in the prediction or identification process. For example, a prediction system might indicate a microcrack defect in a certain area, but manual re-inspection or independent sensors might not have detected it, or might have detected a different type of defect, thus creating a conflict. Determining the confidence level of the conflicting information and the severity of its impact on the product's core functionality can be understood as conducting reliability and importance assessments of the identified conflicting information. Confidence level measures the accuracy and reliability of the conflicting information itself, and is assessed through historical data, model accuracy, and sensor calibration status. The severity level focuses on the potential negative impact that the defect involved in the conflicting information might have on the product's core functions, performance, security, or user experience if it were to occur. In practical applications, conflicting information is prioritized based on its source, confidence level, and the severity of its impact on the product's core functions. The resulting decision assigns a priority to each conflicting piece of information, taking into account the reliability of its source, its own confidence level, and the degree of its impact on the product's core functions. For example, conflicting information originating from a high-precision independent sensor and significantly impacting the product's core functions may have a higher priority than conflicting information originating from a low-confidence prediction model and having a smaller impact. The decision guides the subsequent defect information fusion and correction process. Furthermore, the conflicting information is fused and corrected based on the decision, resulting in fused and corrected defect information. This involves integrating and adjusting conflicting defect information using methods such as weighted averaging, majority voting, and expert system rules, based on the priority decision, to form a more accurate and reliable comprehensive defect information. Therefore, using the fused and corrected defect information, combined with online inspection data, to reverse-engineer the true value of the original data means combining the fused and corrected defect information, which is closer to reality, with the currently distorted online inspection data. Through reverse engineering or data inversion algorithms, it is inferred that the original production line data should present its true state without data distortion. This helps to compensate for information gaps or errors caused by distortion in online inspection data. Finally, calibrating the core parameters of the yield prediction system based on the true values ​​means using the reverse-engineered true values ​​of the original data as a reference benchmark to adjust and optimize the key parameters of the yield prediction system.

[0088] In one embodiment, suppose a production line manufacturing precision electronic components experiences data distortion in its online inspection system due to sensor drift. At this time, a yield prediction system predicts a batch of products may have internal structural defects based on equipment micro-vibration, while manual re-inspection and independent ultrasonic sensors identify surface scratches in another batch, with some overlap between the two batches. First, the system identifies the conflict between the two defect information ("internal structural defects" and "surface scratches") and traces their origins to the "micro-vibration prediction model" and "manual re-inspection / ultrasonic sensor," respectively. Next, the system determines the confidence level of these conflicting information. For example, the micro-vibration prediction model might have an 80% confidence level, while manual re-inspection and ultrasonic sensors, being direct detection methods, might have a 95% confidence level. Simultaneously, the system assesses the severity of these defects' impact on the product's core functionality. For example, internal structural defects might cause product failure at high temperatures, categorized as high severity; while surface scratches only affect appearance, categorized as medium severity. Then, the system prioritizes these defects based on this information. Because ultrasonic sensors and manual inspections have higher confidence levels and represent more severe internal structural defects, the system may prioritize the internal structural defect information identified by ultrasonic sensors and correct conflicting parts of the micro-vibration prediction model. For surface scratches, due to their relatively low severity, the system may combine prediction and inspection results with a weighted fusion. Subsequently, based on the decision, the system fuses and corrects conflicting information, resulting in a defect list that integrates high-confidence and high-severity information. Using this fused and corrected defect information, combined with the currently distorted online inspection data, the system reverse-engineers the true defect distribution and performance parameters of these electronic components under distortion-free conditions. Finally, based on these reverse-engineered true values, the system adjusts the core parameters of the yield prediction system, for example, increasing the sensitivity to internal structural defects and decreasing the weight of minor surface scratches, thereby enabling the calibrated yield prediction system to more accurately assess product quality, even if the online inspection data still has a certain degree of distortion.

[0089] It should be noted that "the degree of conformity between inspector's judgment and actual performance" refers to the degree of matching between the human inspector's judgment of product defects and the performance data obtained by independent sensors in subsequent production stages or actual use. By comparing the inspector's judgment with the product's actual performance, the accuracy and reliability of manual inspection can be evaluated. "Blind test results" refer to the results obtained by independently testing the product without informing the testers of its origin or pre-set defect information. Their purpose is to provide an objective and unbiased evaluation benchmark. By combining the conformity of the inspector's judgment with the blind test results, the reliability of conflicting information can be quantitatively evaluated, thus obtaining the confidence level of the conflicting information. Furthermore, "tracking the subsequent performance of released products" refers to the continuous monitoring of products that have passed production line testing and been put into use, collecting their operational data and fault information in the actual application environment. When a fault is found in a released product, and this fault can be traced back to conflicting information identified during the production process, the severity of the impact of the source of the conflicting information on the core function of the product can be determined based on the nature, severity, and impact on the product's functionality. For example, if the defect involved in the conflicting information causes the core function of the product to completely fail, its severity is high; if it only affects secondary functions or appearance, its severity is relatively low.

[0090] It's important to note that establishing a fault attribution map involves creating a comprehensive data archive for each product released from the production line. This map records not only various process parameters during production, such as temperature, pressure, and speed, but also online monitoring data acquired by various sensors on the production line, as well as conflict information identified during the conflict resolution phase. Furthermore, the map continuously collects data on the product's exposure in its actual operating environment, such as ambient temperature, humidity, vibration, and shock, to comprehensively track the product's lifecycle performance. When a released product fails during subsequent use, a fault path tracing mechanism is activated. This mechanism utilizes the pre-established fault attribution map to conduct in-depth analysis of the failed product. By comparing the fault type with the production process parameters and defect information recorded in the map, production links or potential defects strongly correlated with the fault type can be traced, thus obtaining fault tracing results. Further, to ensure the accuracy of fault analysis, it is also necessary to analyze the specific environmental conditions and operating conditions at the time of the fault. This information will be compared with external factors that were not identified as conflicts during the conflict resolution phase. This process aims to eliminate external failure factors that are not directly related to the conflicting information known at the time of adjudication, such as user error or extreme environmental conditions, thereby obtaining a result of eliminating non-conflicting factors. Ultimately, based on the failure tracing results and the results of eliminating non-conflicting factors, the actual failure situation of the product can be comprehensively confirmed. Based on this actual failure situation, the severity of the impact of a specific source of conflicting information on the core functions of the product can be accurately determined.

[0091] It's important to clarify that identifying the interdependencies between conflicting information sources means that after acquiring predicted product defect information and identified masked product defect information, the system not only identifies the information itself but also further analyzes whether there are mutual influences or correlations between the different sources that generated this information (e.g., equipment micro-vibration prediction models, manual re-inspection, independent sensor data, etc.). For example, the equipment micro-vibration prediction results may have some correlation with independent sensor data at a specific production stage, or the accuracy of manual re-inspection may be implied by the condition of a specific piece of equipment. This step aims to construct a network or graph reflecting the complex relationships between these sources. Furthermore, quantifying the strength and direction of interdependencies means that after identifying interdependencies, these relationships need to be quantitatively described. Strength can be expressed as indicators such as correlation coefficients, mutual information, and causal strength, reflecting the degree to which one source influences another. Direction indicates which source influences which source; for example, A influences B, not the other way around. This can be modeled and calculated through statistical analysis, machine learning models (such as Bayesian networks, causal inference models), or expert knowledge. The purpose is to provide a precise basis for subsequent weight adjustments. Therefore, adjusting the weights of affected conflicting information sources based on the quantified interdependencies to correct the conflicting information sources means that once the strength and direction of the interdependencies between conflicting information sources are clarified, the system dynamically adjusts the weights of these sources in the conflicting information fusion process. For example, if a conflicting information source is found to have low confidence and is strongly influenced by another high-confidence source, then during fusion, the weight of the low-confidence source can be appropriately reduced, or corrected according to the degree of its influence. This adjustment ensures that when fusing conflicting information, the true contribution and reliability of each source can be more accurately reflected, thereby correcting the representation of the original conflicting information sources.

[0092] It should be noted that the inspector's judgment refers to the subjective or objective assessment made by the inspector during manual re-inspection regarding the presence, type, and severity of defects in the product. These judgments are recorded in real-time or in batches as the basis for subsequent evaluation. Independent sensor data refers to data unaffected by equipment micro-vibrations that objectively reflects the actual performance or defect status of the product, such as functional test data and durability test data of the final product. By comparing the inspector's subjective judgment with this objective data, the accuracy of the inspector's judgment can be quantified, i.e., its degree of conformity with the actual performance of the product. In this case, it is necessary to utilize historical defect data and real-time production line status data, and through data analysis and pattern recognition, proactively identify important but insufficiently tested defect types. Specifically, virtual blind testing results are a method to compensate for the shortcomings of actual blind testing through data simulation and model prediction. For defect types lacking actual blind testing data, their historical occurrence patterns and influencing factors can be analyzed in depth, and combined with physical models describing the defect's generation mechanism, their performance under different conditions can be predicted, thereby generating simulated test results. The inspector's individual confidence level reflects the accuracy and reliability of that inspector's judgment on different defect types. By combining actual conformity data and virtual blind test results, the professional judgment ability of each inspector can be assessed and updated more comprehensively and dynamically. Therefore, the confidence level calibration factor is a parameter used to correct the original confidence level calculation result. By introducing the inspector's individual confidence level, different weights or corrections can be assigned to conflicting information from different sources, making the final calculated confidence level of conflicting information more accurate and reliable.

[0093] In one embodiment, suppose an electronics production line needs to inspect circuit boards for microcracks during production. Traditionally, quality assessment is performed through manual visual inspection and a portion of blind test samples. However, due to the diverse types of microcracks and the extremely low frequency of some types, blind test samples often cannot cover all potential microcrack types. Specifically, when an inspector visually inspects a circuit board and determines the presence of a certain microcrack, their judgment is recorded. Subsequently, the circuit board undergoes a high-precision scan using a separate X-ray inspection device to obtain its actual microcrack state. By comparing the inspector's judgment with the X-ray inspection results, the inspector's compliance in microcrack detection can be calculated. Furthermore, if a specific type of microcrack (e.g., an extremely fine crack appearing at a specific solder joint) is found to be absent or extremely infrequent in blind test samples, the system will predict the propagation trend and failure probability of this type of microcrack under different stress conditions based on the occurrence patterns of this type of microcrack in historical production data, its correlation with soldering process parameters, and a fatigue physics model of the circuit board material. This will generate a virtual blind test result for this type of microcrack. Subsequently, the system will comprehensively consider the inspector's judgment conformity on other common microcrack types and the virtual blind test result generated for the aforementioned specific microcrack type, dynamically updating the inspector's individual confidence level in microcrack detection. Finally, when calculating the confidence level of conflict information involving microcracks, the system will adjust the corresponding confidence level calibration factor based on the updated inspector's individual confidence level. This ensures that the final conflict information confidence level more accurately reflects the actual situation, especially when facing rare or difficult-to-detect defects, providing a reliable evaluation basis and guiding the yield prediction system to perform more precise parameter calibration.

[0094] It should be noted that receiving historical defect distribution data refers to the system collecting information such as the frequency, type, and cause of product defects over a past period from various sources (e.g., production databases, quality management systems, customer feedback systems, etc.). Multi-source cross-validation can be understood as comparing defect information from different data sources, such as comparing automatic inspection data from the production line with manual quality inspection records, to ensure the accuracy and completeness of the data. Cleaned and supplemented historical defect distribution data refers to defect statistics after data preprocessing, removing duplicate, erroneous, or inconsistent data, and appropriately filling in missing data; its purpose is to provide a reliable historical background of defects. Real-time monitoring of fluctuation patterns in key production line sensor data refers to continuously collecting real-time data from key equipment on the production line (such as temperature sensors, pressure sensors, vibration sensors, etc.) and analyzing the trends, abnormal peaks, or periodic fluctuations of this data to identify anomalies that may indicate defect occurrence. Real-time monitored abnormal states refer to sensor data patterns that deviate significantly from normal production states, identified through preset thresholds or machine learning models; its purpose is to promptly detect potential production problems. The correlation analysis involves linking historical defect distribution data (after cleaning and supplementation) with real-time monitored anomalies to obtain correlation analysis results. This refers to exploring the potential connection between the occurrence of historical defects and real-time production line anomalies using data mining and statistical analysis methods. For example, does a certain abnormal vibration pattern frequently accompany a specific defect? ​​The correlation analysis results can be understood as a data report or model revealing the correlation between specific production line anomalies and specific defect types. The defect impact factor is quantified based on the severity of the impact of historical defects on the core functions of the product. This defect impact factor is then corrected based on the type and intensity of current production line transient disturbances, resulting in a corrected defect impact factor. The defect impact factor is a quantitative indicator that measures the degree of damage caused by different types of defects to core functions such as product performance, reliability, safety, and user experience. Current production line transient disturbances refer to short-lived, non-continuous abnormal events during production, such as voltage fluctuations, sudden changes in air pressure, and brief equipment stalls. Correcting the defect impact factor involves dynamically adjusting the weight or value of the defect impact factor based on the potential impact of real-time transient disturbances on product defects to more accurately reflect the defect risk under the current production conditions. Based on the correlation analysis results and the revised defect impact factors, identify defect types that were not covered by blind testing and have a significant impact on the core functions of the product.

[0095] It should be noted that when quantifying the impact factors of defects, a multi-dimensional defect risk assessment model is first established to comprehensively quantify these factors. This multi-dimensional assessment model considers multiple dimensions, including the frequency of occurrence of historical defects, repair costs, safety risks, and user complaint rates, aiming to comprehensively assess the potential impact of known defects. Among these, the frequency of occurrence reflects the prevalence of defects; repair costs measure the economic investment required to resolve defects; safety risks assess the potential harm that defects may cause to users or the system; and the user complaint rate directly reflects the impact of defects on user experience and satisfaction. By comprehensively considering these dimensions, a more comprehensive and objective quantitative result can be obtained. Furthermore, for long-tail defects with extremely low occurrence frequencies in historical defects, due to their data sparsity, it is difficult to accurately assess them directly using statistical models. Therefore, this application adopts an analogical assessment method. Specifically, by analyzing the physical similarities between long-tail defects and known high-impact defects, such as their commonalities in materials, structures, stress distribution, or failure mechanisms, and combining this with simulation test results under extreme conditions, the potential impact factors of long-tail defects are inferred. For example, if a long-tail defect has a similar crack propagation mechanism to a known high-impact defect, its failure behavior under extreme conditions can be evaluated through simulation testing, thus allowing for a reasonable analogy of its impact factor. This yields a long-tail defect analogy evaluation result. Furthermore, for unrecorded defects in historical data—i.e., entirely new or unrecorded defect types—this application preliminarily assesses their impact by analyzing the physical correlation between their generation mechanism and the core functions of the product. For instance, by conducting physical analysis on failed samples of newly discovered defects, combined with expertise in materials science and structural mechanics, the potential impact of the defect on the product's core functions such as strength, conductivity, or sealing can be inferred. Simultaneously, by combining the judgments of domain experts regarding the potential failure modes that this unrecorded defect may trigger in specific application scenarios, such as expert experience and fault tree analysis, a preliminary impact factor evaluation result for the novel defect is obtained. After obtaining the comprehensive quantitative results, the long-tail defect analogy evaluation results, and the preliminary impact factor evaluation results for the novel defect, these results are integrated to obtain a more comprehensive and accurate quantified defect impact factor. Finally, based on the type and intensity of transient disturbances in the current production line, the quantified defect impact factor is corrected to ensure that the defect impact factor can dynamically adapt to real-time changes in the production line.

[0096] It should be noted that identifying the type and intensity of transient disturbances on the current production line refers to real-time monitoring and analysis of data from various sensors on the production line, such as vibration sensors, temperature sensors, pressure sensors, and current and voltage sensors, to detect and classify various transient, non-periodic abnormal fluctuations or events occurring on the production line, and quantify their intensity. Analyzing the interactions between transient disturbances to obtain their combined effect on defect influencing factors can be understood as, after identifying multiple transient disturbances, further evaluating whether there are complex relationships among these disturbances, such as synergistic enhancement, mutual cancellation, or series transmission. For example, two micro-vibrations may resonate at a specific frequency, thereby amplifying their impact on product quality; or a temperature transient may change the physical properties of a material, making it more sensitive to subsequent mechanical shocks. In practical applications, based on the combined effect of transient disturbances on defect impact factors, the physical response of sensor data under transient disturbances is analyzed to infer the comprehensive impact of transient disturbances on product performance or structural stability, resulting in a correction factor. This means that after clarifying the combined effect of transient disturbances, it is necessary to combine physical models and historical data to deeply analyze the specific change patterns of production line sensor data under these combined disturbances. Based on these inferred combined effects, a correction factor can be calculated. This correction factor will be used to adjust the originally quantified defect impact factor to more accurately reflect the actual risks brought about by transient disturbances. Considering the differences in the impact of transient disturbances on different defect types, the quantified defect impact factor is corrected according to the correction factor. This means that when applying the correction factor, it is not possible to simply apply a uniform correction to all defect types. Different transient disturbances may have different sensitivities or impact paths on different types of defects. Therefore, it is necessary to establish a mapping relationship or weighting mechanism to differentiate the application of the correction factor according to its degree of influence on specific defect types. For example, for vibration-related defects, the correction factor may be assigned a higher weight; while for temperature-related defects, it may be assigned a lower weight, or different correction functions may be used.

[0097] In one embodiment, it is assumed that two main transient disturbances exist on a precision electronic component production line: periodic micro-vibrations generated by the high-speed movement of the placement machine and instantaneous temperature fluctuations in the reflow oven during the heating / cooling phase. First, the system identifies the type and intensity of the placement machine's micro-vibrations using a high-precision accelerometer and the type and intensity of the reflow oven's temperature fluctuations using a thermocouple array. Next, the system analyzes the interaction between these two transient disturbances. For example, historical data and physical models reveal that when the placement machine's micro-vibrations and specific temperature fluctuations in the reflow oven occur simultaneously, the risk of microvoids forming inside the solder joints may increase significantly—a synergistic effect. The system quantifies this synergistic effect to obtain its combined influence on the microvoid defect factor of the solder joints. Then, based on this combined effect, the system analyzes the physical response of sensor data under these transient disturbances. For example, it monitors the instantaneous temperature distribution in the solder joint area using an infrared thermal imager and detects changes in the acoustic properties inside the solder joints using an ultrasonic sensor. Based on these physical responses, the system infers the combined impact of specific combinations of micro-vibrations and temperature fluctuations on the structural stability (e.g., shear strength) and electrical properties (e.g., contact resistance) of the weld joint, and calculates a correction factor accordingly. Finally, based on this correction factor, the system considers the differences in the impact of these two transient disturbances on different defect types. For example, for weld joint micro-void defects, the correction factor is assigned a higher weight for correction; while for defects that may be caused by other factors (e.g., material purity), the correction factor may be assigned a lower weight or not corrected at all.

[0098] See Figure 2 , Figure 2 This is a schematic diagram of a production line yield intelligent management system provided in one embodiment of this application. The production line yield intelligent management system 200 includes:

[0099] The data acquisition and analysis module 210 is used to collect operating data of the production line equipment and analyze the operating data to identify micro-vibrations of the equipment;

[0100] The transmission path model establishment module 220 is used to establish a transmission path model of the equipment micro-vibration based on the identified equipment micro-vibration.

[0101] The mapping relationship establishment module 230 is used to establish the mapping relationship between vibration transmission intensity and product defects based on the transmission path model;

[0102] Defect prediction module 240 is used to predict product defects caused by equipment micro-vibration based on mapping relationship;

[0103] The defect identification module 250 is used to identify masked product defects by cross-validating the results of manual re-inspection and independent data in response to predicted product defects.

[0104] The data distortion identification module 260 is used to perform multi-source consistency verification and reverse derivation verification on the online detection data input to the yield prediction system in order to identify the distortion of prior detection data;

[0105] The calibration module 270 is used to calibrate the core parameters of the yield prediction system based on the predicted product defect information and the identified masked product defect information when online detection data is distorted.

[0106] It should be noted that the information interaction and execution process between the above modules are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, which will not be repeated here.

[0107] Those skilled in the art will understand that all or some of the steps and systems in the methods disclosed above can be implemented as software, firmware, hardware, and suitable combinations thereof. Some or all of the physical components can be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, as is known to those skilled in the art, communication media typically include computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0108] The foregoing has provided a detailed description of the preferred embodiments of this application. However, this application is not limited to the above-described embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application. All such equivalent modifications or substitutions are included within the scope defined in this application.

Claims

1. A method for intelligent management of production line yield rate, characterized in that, include: Collect operating data from production line equipment and analyze the operating data to identify micro-vibrations in the equipment; Based on the identified device micro-vibrations, a transmission path model for the device micro-vibrations is established; Based on the transmission path model, establish the mapping relationship between vibration transmission intensity and product defects; Predict product defects caused by the equipment micro-vibration based on the equipment micro-vibration and the mapping relationship; For predicted product defects, cross-validation is performed by combining manual re-inspection results and independent sensor data that are unaffected by vibration to identify masked product defects; Multi-source consistency verification and reverse derivation verification are performed on the online detection data input into the yield prediction system to identify distortion of the online detection data; When the online detection data is distorted, the core parameters of the yield prediction system are calibrated based on the predicted product defect information and the identified masked product defect information.

2. The method according to claim 1, characterized in that, When the online detection data is distorted, the steps for calibrating the core parameters of the yield prediction system based on the predicted product defect information and the identified masked product defect information include: Identify conflict information and sources of conflict information between the predicted product defects and the identified masked product defects; Determine the confidence level of conflicting information and the severity of its impact on the core functions of the product; Based on the source of the conflicting information, the confidence level of the conflicting information, and the severity of its impact on the core functions of the product, the conflicting information is prioritized and a decision is made. Based on the ruling result, the conflict information is fused and corrected to obtain the fused and corrected defect information; Using the fused and corrected defect information, combined with the online detection data, the true value of the original data is deduced in reverse. The core parameters of the yield prediction system are calibrated based on the true values.

3. The method according to claim 2, characterized in that, The steps for determining the confidence level of conflicting information and the severity of the impact of conflicting information on the core functions of the product include: Calculate the confidence level of conflict information based on the degree of conformity between the inspector's judgment and the actual performance, as well as the results of blind testing; Track the subsequent performance of the released product, and when a fault caused by conflicting information is discovered, determine the severity of the impact of the source of the conflicting information on the core function of the product based on the fault.

4. The method according to claim 3, characterized in that, The steps of tracking the subsequent performance of the released product and, upon discovering a fault caused by conflicting information, determining the severity of the impact of the source of the conflicting information on the core functions of the product include: A fault attribution map is established for each released product. The fault attribution map includes the product's process parameters on the production line, online detection data, conflict information during adjudication, and environmental exposure data collected subsequently. When a fault is found in the released product, the fault path backtracking mechanism is activated, and the fault tracing result is obtained by tracing the production process parameters and defect information that are strongly correlated with the fault type based on the fault attribution map. Analyze the environmental conditions and operating conditions at the time of the failure, and compare them with information that was not identified as conflicting at the time of adjudication, so as to eliminate external failure factors that are not directly related to the conflicting information at the time of adjudication and obtain the result of non-conflict factor elimination. Based on the fault tracing results and the non-conflict factor elimination results, the actual failure situation is confirmed; Based on the actual failure scenarios, determine the severity of the impact of the conflict information source on the core functions of the product.

5. The method according to claim 2, characterized in that, The steps following the identification of conflict information between the predicted product defect and the identified masked product defect, and the source of the conflict information, include: Identify the interdependencies between conflicting information sources; Quantify the strength and direction of the interdependence; Based on the quantified interdependencies, the weights of the affected conflict information sources are adjusted to correct the conflict information sources.

6. The method according to claim 3, characterized in that, The steps for calculating the confidence level of conflict information, based on the inspector's judgment and the actual performance, as well as the results of blind testing, include: When inspectors make defect judgments, their judgment results should be recorded. The inspector's judgment is compared with the performance data of the product obtained by independent sensors in subsequent production or actual use to evaluate the degree of conformity between the inspector's judgment and the actual performance of the product. When the sample size of blind testing is insufficient to cover all defect types, based on historical defect distribution data and the current production line status, identify defect types that are not covered by blind testing but have a significant impact on the core functions of the product. For the defect types not covered by blind testing, virtual blind testing results are generated by analyzing historical data of such defects and combining the prediction results of physical models related to such defects. The inspector's individual confidence level is updated based on the degree of conformity between the inspector's judgment and the actual performance of the product, as well as the results of the virtual blind test. Based on the individual confidence level, adjust the confidence level calibration factor and calculate the confidence level of conflict information.

7. The method according to claim 6, characterized in that, When the blind testing sample size is insufficient to cover all defect types, the steps for identifying defect types that are not covered by blind testing but have a significant impact on the core functions of the product, based on historical defect distribution data and the current production line status, include: Receive historical defect distribution data and obtain cleaned and supplemented historical defect distribution data through a multi-source cross-validation mechanism; Real-time monitoring of fluctuation patterns in key sensor data from the production line allows for the identification of abnormal states detected in real time. The historical defect distribution data after cleaning and filling is correlated with the abnormal states detected in real time to obtain the correlation analysis results; The defect impact factor is quantified based on the severity of the impact of historical defects on the core functions of the product, and the defect impact factor is corrected based on the type and intensity of the transient disturbances in the current production line to obtain the corrected defect impact factor. Based on the correlation analysis results and the revised defect impact factor, identify defect types that were not covered by blind testing and have a significant impact on the core functions of the product.

8. The method according to claim 7, characterized in that, The steps of quantifying the defect impact factor based on the severity of the impact of historical defects on the core functions of the product, and correcting the defect impact factor based on the type and intensity of the transient disturbances in the current production line to obtain the corrected defect impact factor include: By establishing a multi-dimensional defect risk assessment model, the defect impact factors are comprehensively quantified to obtain a comprehensive quantitative result. The multi-dimensional factors include the frequency of occurrence of historical defects, repair costs, security risks, and user complaint rates. For long-tail defects that occur at extremely low frequencies in the historical defects, the influence factors of the long-tail defects are evaluated by analogy based on the physical similarity between the long-tail defects and known high-impact defects, as well as the results of simulation tests under extreme working conditions, to obtain the long-tail defect analogy evaluation results. For unrecorded defects among the historical defects, the physical relationship between the generation mechanism of the unrecorded defects and the core functions of the product is analyzed, and combined with the expert's judgment on the potential failure modes that the unrecorded defects may cause in specific application scenarios, the preliminary impact factor assessment results of the new defects are obtained. Based on the comprehensive quantitative results, the long-tail defect analogy evaluation results, and the preliminary impact factor evaluation results of the novel defect, the quantified defect impact factor is obtained. The quantified defect impact factor is corrected based on the type and intensity of transient disturbances in the current production line.

9. The method according to claim 8, characterized in that, The step of correcting the quantified defect impact factor based on the type and intensity of the current production line transient disturbance includes: Identify the type and intensity of transient disturbances in the current production line; The interaction between the transient disturbances is analyzed to obtain the comprehensive effect of the transient disturbances on the defect influence factor; Based on the combined effect of the transient disturbance on the defect impact factor, the physical response of sensor data under the transient disturbance is analyzed, the combined impact of the transient disturbance on product performance or structural stability is inferred, and a correction factor is obtained. Based on the correction factor, and considering the differences in the impact of transient disturbances on different defect types, the quantified defect impact factor is corrected.

10. A production line yield intelligent management system, characterized in that, include: The data acquisition and analysis module is used to collect operating data of the production line equipment and analyze the operating data to identify micro-vibrations of the equipment. The transmission path model establishment module is used to establish a transmission path model of the identified device micro-vibrations based on the identified device micro-vibrations. The mapping relationship establishment module is used to establish a mapping relationship between vibration transmission intensity and product defects based on the transmission path model. The defect prediction module is used to predict product defects caused by the micro-vibration of the equipment based on the mapping relationship; The defect identification module is used to cross-validate the results of manual re-inspection with independent sensor data that is unaffected by vibration, in order to identify hidden product defects. The data distortion identification module is used to perform multi-source consistency verification and reverse derivation verification on the online detection data input to the yield prediction system in order to identify online detection data distortion. The calibration module is used to calibrate the core parameters of the yield prediction system based on the predicted product defect information and the identified masked product defect information when the online detection data is distorted.

Images