CN118037721A

CN118037721A - Method and system for detecting defects of optical cable harness

Info

Publication number: CN118037721A
Application number: CN202410432356.6A
Authority: CN
Inventors: 何进; 李德健
Original assignee: Sichuan Jiawanguang Communication Co ltd
Current assignee: Sichuan Jiawanguang Communication Co ltd
Priority date: 2024-04-11
Filing date: 2024-04-11
Publication date: 2024-05-14
Anticipated expiration: 2044-04-11
Also published as: CN118037721B

Abstract

The embodiment of the application realizes efficient and accurate processing of a target optical cable harness detection task by dynamically adjusting and setting defect tolerance. Specifically, the first defect evaluation parameter and the second defect evaluation parameter are combined, so that the iterative defect tolerance is flexibly determined, and the detection process is more targeted and adaptive. In the detection and distribution stage, the feedback defect tolerance is utilized to identify the characteristic data of the target optical cable harness, so that the accuracy and timeliness of defect detection are effectively improved. Meanwhile, the trust and the defect index migration coefficient are adjusted based on the defect detection result, so that the self-learning and optimizing capacity of the detection system is further improved. Therefore, the defect detection efficiency and accuracy of the optical cable harness can be remarkably improved, powerful support can be provided for maintenance and management of the optical cable harness, the fault risk is reduced, and the system stability is improved.

Description

Method and system for detecting defects of optical cable harness

Technical Field

The application relates to the technical field of artificial intelligence, in particular to a method and a system for detecting defects of an optical cable harness.

Background

In the field of optical cable communications, the integrity of the optical cable harness is critical to ensure the quality and stability of signal transmission. However, during production, transportation, installation or use, the cable harness may suffer from various forms of damage or defects, such as surface scratches, indentations, breaks, etc. These defects may not only affect the transmission performance of the cable, but may also cause communication interruption or failure. Therefore, developing an efficient and accurate optical cable harness defect detection method has important significance for guaranteeing the reliability of a communication system.

Conventional methods of detecting defects in fiber optic cable harnesses typically rely on manual visual inspection or use of simple detection tools, which are not only inefficient, but are also susceptible to artifacts (e.g., fatigue, experience inadequacies) leading to missed or false detections. In recent years, with rapid development of computer vision and artificial intelligence technology, an optical cable harness defect identification network based on deep learning is becoming a research hotspot. The networks can automatically learn and extract the characteristics in the cable harness images, so that the defects are automatically identified and classified.

However, existing optical cable harness defect identification networks still have some problems in practical applications. For example, they typically use a fixed defect tolerance to determine if a cable harness is defective, and this approach cannot accommodate the different requirements for defect identification in different scenarios.

Disclosure of Invention

In view of the above, the present application aims to provide a method and a system for detecting defects of an optical cable harness.

According to a first aspect of the present application, there is provided a method for detecting defects of an optical cable harness, the method comprising:

Acquiring a first defect evaluation parameter corresponding to a target optical cable harness defect identification network and a second defect evaluation parameter generated by the target optical cable harness defect identification network aiming at a target optical cable harness detection task;

Dynamically adjusting the set defect tolerance according to the first defect evaluation parameter and the second defect evaluation parameter, and determining the iterative defect tolerance of the target optical cable harness detection task;

Aiming at the current round of detection and distribution stage of the target optical cable harness detection task, when the current defect index of the target optical cable harness detection task reaches the feedback defect tolerance corresponding to the current round of detection and distribution stage, carrying out characteristic data identification on the target optical cable harness detection task to generate a defect detection result, wherein the feedback defect tolerance is determined after dynamic adjustment on the iteration defect tolerance according to the defect index migration coefficient corresponding to the current round of detection and distribution stage and the trust degree of the optical cable detection characteristic data corresponding to the target optical cable harness detection task;

Based on the defect detection result, adjusting the trust and the defect index migration coefficient, and generating an adjusted trust and an adjusted defect index migration coefficient;

and according to the adjusted trust degree and the adjusted defect index migration coefficient, scheduling the target optical cable harness detection task to perform task execution distribution in the next detection distribution stage of the current detection distribution stage.

In one possible implementation manner of the first aspect, the target cable harness defect identification network includes: the method for determining the iterative defect tolerance of the target optical cable harness detection task includes the steps of:

dynamically adjusting the set defect tolerance according to a first defect evaluation parameter corresponding to each optical cable harness defect identification network and a second defect evaluation parameter corresponding to each optical cable harness defect identification network in the plurality of optical cable harness defect identification networks, and determining initial defect tolerance corresponding to each optical cable harness defect identification network;

And determining the iterative defect tolerance according to the initial defect tolerance corresponding to each of the plurality of optical cable harness defect identification networks.

In a possible implementation manner of the first aspect, the step of adjusting the defect indicator migration coefficient includes:

determining a task execution allocation rule corresponding to the next detection allocation stage based on the defect detection result;

and according to the preset defect index amplification corresponding to the task execution allocation rule, improving the defect index migration coefficient, and generating the adjusted defect index migration coefficient.

In a possible implementation manner of the first aspect, the step of adjusting the trust level includes:

When the defect detection result represents that the target optical cable harness detection task meets the verification condition, obtaining a blocking log of the optical cable detection characteristic data in a target past detection period, wherein the target past detection period comprises a plurality of unit detection periods;

Performing trust evaluation on the optical cable detection characteristic data aiming at each unit detection period according to the period weight parameter corresponding to each unit detection period in the plurality of unit detection periods and the blocking log in each unit detection period, and generating initial trust corresponding to each unit detection period;

And carrying out averaging processing on the initial trust degrees corresponding to the unit detection periods respectively to generate the adjusted trust degrees.

In a possible implementation manner of the first aspect, after the identifying feature data of the target optical cable harness detection task and generating a defect detection result, the method further includes:

when the defect detection result represents that the target optical cable harness detection task does not meet the verification condition, blocking the target optical cable harness detection task;

And adjusting a blocking log corresponding to the optical cable detection characteristic data according to the blocking operation.

In a possible implementation manner of the first aspect, the defect detection result includes: harness status tag information, the method further comprising:

when the wire harness state label information is not matched with the second defect evaluation parameter, adding the target optical cable wire harness detection task and the wire harness state label information to a first supervision training data sequence;

performing defect recognition learning on a preset optical cable harness defect recognition network according to the first supervision training data sequence to generate an initial optical cable harness defect recognition network;

acquiring an unsupervised training data sequence;

performing pseudo tag configuration on the unsupervised training data sequence according to the initial optical cable harness defect identification network to generate a second supervised training data sequence;

According to the second supervision training data sequence, performing adaptive parameter learning on a guided network corresponding to the initial optical cable harness defect recognition network, and generating an optical cable harness defect recognition network after the adaptive parameter learning;

And taking the optical cable harness defect recognition network after the adaptive parameter learning as an adjusted optical cable harness defect recognition network.

In a possible implementation manner of the first aspect, the performing adaptive parameter learning on the guided network corresponding to the initial cable harness defect identification network according to the second supervised training data sequence, and generating the cable harness defect identification network after the adaptive parameter learning includes:

Acquiring the supervision trust degree corresponding to each supervision training data in the second supervision training data sequence;

Sampling the second supervision training data sequence according to the supervision trust level to determine active training data and passive training data;

And performing defect recognition learning on the guided network according to the positive training data and the negative training data, and generating the optical cable harness defect recognition network after the adaptive parameter learning.

In a possible implementation manner of the first aspect, after the adaptively learning parameters is performed on the guided network corresponding to the initial cable harness defect identification network according to the second supervised training data sequence, and the cable harness defect identification network after adaptively learning parameters is generated, the method further includes:

Updating the initial optical cable harness defect identification network according to the optical cable harness defect identification network after the adaptive parameter learning, and generating an adjusted initial optical cable harness defect identification network;

According to the adjusted initial optical cable harness defect identification network, iteratively executing the obtained unsupervised training data sequence to the guided network corresponding to the initial optical cable harness defect identification network according to the second supervised training data sequence, and performing adaptive parameter learning on the guided network corresponding to the initial optical cable harness defect identification network to generate an adaptive parameter learning flow of the optical cable harness defect identification network after adaptive parameter learning until reaching preset adaptive parameter learning times;

and taking the current optical cable harness defect recognition network after adaptive parameter learning as the adjusted optical cable harness defect recognition network.

In a possible implementation manner of the first aspect, the method further includes:

And adjusting the first defect evaluation parameter according to the adjusted optical cable harness defect identification network.

According to a second aspect of the present application, there is provided a system for detecting a defect of an optical cable harness, the system comprising a machine-readable storage medium storing machine-executable instructions and a processor, the processor, when executing the machine-executable instructions, implementing the method for detecting a defect of an optical cable harness.

According to a third aspect of the present application, there is provided a computer-readable storage medium having stored therein computer-executable instructions that, when executed, implement the aforementioned method of detecting a defect of an optical cable harness.

According to any one of the aspects, the application has the technical effects that:

According to the embodiment of the application, the defect tolerance is dynamically adjusted and set, so that the target optical cable harness detection task is efficiently and accurately processed. Specifically, the first defect evaluation parameter and the second defect evaluation parameter are combined, so that the iterative defect tolerance is flexibly determined, and the detection process is more targeted and adaptive. In the detection and distribution stage, the feedback defect tolerance is utilized to identify the characteristic data of the target optical cable harness, so that the accuracy and timeliness of defect detection are effectively improved. Meanwhile, the trust and the defect index migration coefficient are adjusted based on the defect detection result, so that the self-learning and optimizing capacity of the detection system is further improved. Therefore, the defect detection efficiency and accuracy of the optical cable harness can be remarkably improved, powerful support can be provided for maintenance and management of the optical cable harness, the fault risk is reduced, and the system stability is improved.

In detail, the method and the device realize the dynamic adjustment of the set defect tolerance by acquiring the first defect evaluation parameter corresponding to the target optical cable harness defect identification network and the generated second defect evaluation parameter so as to determine the iteration defect tolerance of the target optical cable harness detection task. The dynamic adjustment mode can more accurately reflect the actual defect condition of the optical cable harness. In the current round of detection and distribution stage, when the current defect index reaches the feedback defect tolerance, characteristic data identification is carried out, and a defect detection result is generated. According to the method, the iterative defect tolerance is dynamically adjusted according to the defect index migration coefficient and the trust degree of the optical cable detection characteristic data, so that the accuracy and the sensitivity of defect detection are improved. Based on the defect detection result, the trust degree and the defect index migration coefficient can be adjusted, and the adjusted trust degree and defect index migration coefficient are generated. The dynamic adjustment mechanism is helpful for optimizing the defect detection process and improving the detection efficiency. And according to the adjusted trust degree and the defect index migration coefficient, the target optical cable harness detection task can be scheduled to perform task execution distribution in the next detection distribution stage of the round of detection distribution stage. The method is favorable for reasonably distributing detection resources, and further improves the efficiency and accuracy of detecting the defects of the optical cable harness. Therefore, through a dynamic adjustment mechanism and feedback-based characteristic data identification, the accuracy and the efficiency of detecting the defects of the optical cable harness are effectively improved.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the embodiments will be briefly described below, it being understood that the following drawings only illustrate some embodiments of the present application and therefore should not be considered as limiting the scope, and other related drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a schematic flow chart of a method for detecting defects of an optical cable harness according to an embodiment of the present application;

Fig. 2is a schematic diagram of a component structure of a system for detecting a defect of an optical cable harness according to an embodiment of the present application.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions of the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application, and it should be understood that the accompanying drawings in the present application are for the purpose of illustration and description only, and are not intended to limit the scope of the present application. In addition, it should be understood that the schematic drawings are not drawn to scale. A flowchart, as used in this disclosure, illustrates operations implemented in accordance with some embodiments of the present application. It should be understood that the operations of the flow diagrams may be implemented out of order and that steps without logical context may be performed in reverse order or concurrently. Furthermore, one skilled in the art, under the direction of this disclosure, may add at least one other operation to the flowchart and may destroy at least one operation from the flowchart.

In addition, the described embodiments are only some, but not all, embodiments of the application. The components of the embodiments of the present application generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the application, as presented in the figures, is not intended to limit the scope of the application, as claimed, but is merely representative of selected embodiments of the application. All other embodiments, which can be made by a person skilled in the art without making any inventive effort, correspond to the scope of the application according to the embodiments of the application.

Fig. 1 is a schematic flow chart illustrating a method and a system for detecting defects of an optical cable harness according to an embodiment of the present application, and it should be understood that in other embodiments, the sequence of part of steps in the method for detecting defects of an optical cable harness according to the present application may be shared with each other according to actual needs, or part of the steps may be omitted or maintained. The method for detecting the defects of the optical cable harness comprises the following detailed steps:

step S110, a first defect evaluation parameter corresponding to a target optical cable harness defect identification network and a second defect evaluation parameter generated by the target optical cable harness defect identification network for a target optical cable harness detection task are obtained.

In detail, the target cable harness defect recognition network is a neural network or algorithm model for recognizing cable harness defects, and potential defects can be recognized by learning and analyzing characteristics of images, data and the like of the cable harness. For example, the deep learning model can accurately identify defects such as breakage, wear, deformation, and the like from the images of the cable harness after training.

The first defect evaluation parameters refer to a set of parameters which are preset and are used for evaluating defects of the optical cable harness. These parameters are typically set based on industry standards, historical data, or expert experience and serve as a benchmark for evaluating defects. For example, the first defect review parameter may include specific numerical ranges of outer diameter, inner diameter, insulation thickness, conductor spacing, etc. of the fiber optic cable harness. For example, the standard range of outer diameters may be 4.5mm to 5.5mm.

The second defect evaluation parameters are a group of evaluation parameters which are generated in real time by the defect identification network and are used for indicating the current specific optical cable harness detection task. These parameters reflect the actual defect conditions of the current cable harness. For example, in actual inspection, the defect recognition network may analyze that the actual outer diameter is 5.7mm according to the image and data of the current cable harness, and exceeds the preset standard range, so this deviation value is used as a part of the second defect evaluation parameter.

Thus, in the present embodiment, as for the detection system of the defect of the optical cable harness, a server may be used. For example, the server receives a cable harness detection task that requires defect identification of a batch of newly produced cable harnesses. To perform this cable harness detection task, the server first accesses a database of parameters within it. In this parameter database, a plurality of preset parameters for the cable harness defect identification network are stored, including a first defect evaluation parameter, such as a standard value range of the diameter, the insulating layer thickness, the number of conductors, and the like of the cable harness.

Meanwhile, the server also calls a real-time data analysis module which performs preliminary analysis on the real-time detection data (such as images, electric signals and the like) of the current batch of optical cable harnesses and generates second defect evaluation parameters which may include the actually measured diameter, the insulation layer thickness change rate, the conductor spacing deviation and the like.

Step S120, dynamically adjusting the set defect tolerance according to the first defect evaluation parameter and the second defect evaluation parameter, and determining the iterative defect tolerance of the target optical cable harness detection task.

In detail, the set defect tolerance refers to an acceptable level of defect for the optical cable harness during production or inspection. Which is typically one or more specific values or ranges for determining whether the cable harness is acceptable. For example, the set defect tolerance may be that the thickness variation of the insulating layer is not more than 0.1mm, or that the conductor pitch variation is not more than 5%, or the like.

The iterative defect tolerance refers to new defect tolerance for a current detection task, which is obtained after dynamic adjustment, and reflects improvement and optimization of original tolerance based on latest data and conditions. For example, the defect tolerance of the thickness of the insulation layer in the original setting may be ±0.1mm, but after the dynamic adjustment, the iterative defect tolerance may be adjusted to ±0.15mm in consideration of the actual situation of the current batch of optical cable harnesses.

Thus, in this embodiment, the server now has two sets of defect review parameters: one set is a preset value (first defect evaluation parameter) based on the history data and the criteria, and the other set is an actual value (second defect evaluation parameter) based on the current real-time inspection data. To more accurately evaluate the quality of the batch of fiber optic cable harnesses, the server runs a defect tolerance adjustment algorithm.

The defect tolerance adjustment algorithm first compares the differences between the preset value and the actual value, and then dynamically adjusts the original defect tolerance according to the magnitude and direction of the differences and possible influencing factors (such as production lot, ambient temperature, etc.). For example, if the current lot of cable harness diameters are found to be generally small, the defect tolerance adjustment algorithm may reduce the defect tolerance in terms of diameter to more tightly control product quality.

Step S130, aiming at the present round of detection and distribution stage of the target optical cable harness detection task, when the current defect index of the target optical cable harness detection task reaches the feedback defect tolerance corresponding to the present round of detection and distribution stage, performing feature data identification on the target optical cable harness detection task to generate a defect detection result, wherein the feedback defect tolerance is determined after dynamically adjusting the iteration defect tolerance according to the defect index migration coefficient corresponding to the present round of detection and distribution stage and the trust degree of the optical cable detection feature data corresponding to the target optical cable harness detection task.

In the process of detecting the optical cable harness, the detection task may be performed in a plurality of stages. Each stage has a specific detection target and criteria, called a detection allocation stage. For example, a complete inspection task may include several stages such as preliminary visual inspection, detailed electrical performance testing, final quality assessment, and the like.

The current defect index is a numerical value or index reflecting the defect degree of the optical cable harness, which is measured or calculated in real time in the detection process. Such as the degree of breakage of the insulating layer, the resistance value of the conductor, etc., can be used as the current defect index.

The feedback defect tolerance is a dynamically determined defect tolerance for the current inspection allocation phase. The method is obtained by real-time adjustment according to the previous detection result and data so as to more accurately guide the current detection work. For example, if the previous inspection stage finds that the insulating layer is broken more, the tolerance to the breakage of the insulating layer may be reduced in the current stage, that is, the inspection standard may be improved.

The defect index migration coefficient is a coefficient for adjusting defect tolerance, and reflects the variation trend or correlation of defect indexes among different detection stages. For example, if the number of defects found in the previous inspection stage has a strong positive correlation with the number of defects in the current stage, the migration coefficient may be larger, meaning that the tolerance of the current stage needs to be tighter.

The trust degree of the optical cable detection characteristic data refers to the evaluation of the reliability and accuracy of the characteristic data (such as images, electric signals and the like) acquired in the optical cable beam detection process. The higher the confidence, the more truly the data will reflect the actual status of the cable harness. For example, if the detection device is calibrated and maintained well, the confidence in the data it acquires may be high; conversely, if the device ages or is not calibrated, the confidence level may decrease.

Thus, in this embodiment, after the defect tolerance is adjusted, the server starts to perform a specific detection task in the current round of detection allocation stage. For example, the surface of the cable harness may be scanned using image recognition techniques to identify possible defect features such as scratches, depressions, color anomalies, and the like. Meanwhile, the server can also detect the electrical properties of the optical cable harness, such as conductivity, insulativity and the like, through an electrical signal analysis technology.

These characteristic data and electrical performance data are combined and compared to the adjusted defect tolerance. When one or more of the metrics touches the feedback defect tolerance (which is a result of dynamically adjusting the iterative defect tolerance based on the defect metric migration coefficient and the confidence level of the current inspection data), the server immediately generates a detailed defect inspection report listing all the detected defect types, locations, severity, etc.

Step S140, based on the defect detection result, adjusting the confidence level and the defect indicator migration coefficient, and generating an adjusted confidence level and an adjusted defect indicator migration coefficient.

In this embodiment, based on the defect detection report generated in the previous step, the server re-evaluates and adjusts the trust level and the defect index migration coefficient of the current detection data. If the number of defects listed in the inspection report is found to be high and the severity is high, the server may decrease the confidence in the current batch of cable harnesses and increase the defect index migration coefficient to more tightly control the product quality in subsequent inspection.

Conversely, if the inspection report shows a smaller number of defects and a lower severity, the server may increase the confidence level and decrease the defect indicator migration factor to maintain or properly relax the control criteria for product quality in subsequent inspection.

And step S150, according to the adjusted trust degree and the adjusted defect index migration coefficient, scheduling the target optical cable harness detection task to perform task execution distribution in the next detection distribution stage of the current detection distribution stage.

In this embodiment, the server schedules the task in the next detection and allocation stage according to the adjusted trust level and the defect index migration coefficient. For example, the inspection resources (e.g., increasing or decreasing the number of inspection apparatuses, adjusting the inspection sequence, etc.) may be reallocated based on the new confidence and migration coefficients to ensure that defects in the fiber optic cable harness are more accurately identified and controlled in subsequent inspections. Meanwhile, the server can feed back the result and data of the round of detection to the production department and the quality management department for the production adjustment and the quality improvement.

Based on the steps, the embodiment of the application realizes the efficient and accurate processing of the target optical cable harness detection task by dynamically adjusting and setting the defect tolerance. Specifically, the first defect evaluation parameter and the second defect evaluation parameter are combined, so that the iterative defect tolerance is flexibly determined, and the detection process is more targeted and adaptive. In the detection and distribution stage, the feedback defect tolerance is utilized to identify the characteristic data of the target optical cable harness, so that the accuracy and timeliness of defect detection are effectively improved. Meanwhile, the trust and the defect index migration coefficient are adjusted based on the defect detection result, so that the self-learning and optimizing capacity of the detection system is further improved. Therefore, the defect detection efficiency and accuracy of the optical cable harness can be remarkably improved, powerful support can be provided for maintenance and management of the optical cable harness, the fault risk is reduced, and the system stability is improved.

In one possible embodiment, the target fiber optic cable harness defect identification network comprises: a plurality of fiber optic cable harness defect identification networks of different defect identification dimensions, step S120 may include:

Step S121, dynamically adjusting the set defect tolerance according to a first defect evaluation parameter corresponding to each optical cable harness defect recognition network and a second defect evaluation parameter corresponding to each optical cable harness defect recognition network in the plurality of optical cable harness defect recognition networks, and determining an initial defect tolerance corresponding to each optical cable harness defect recognition network.

Step S122, determining the iterative defect tolerance according to the initial defect tolerance corresponding to each of the plurality of optical cable harness defect identification networks.

In this embodiment, the server first constructs a plurality of cable harness defect identification networks for different defect identification dimensions. These networks may be focused on identifying different types of defects, such as surface damage, internal structural problems, electrical performance anomalies, and the like. Each network is specifically trained and optimized to achieve optimal recognition within its particular domain.

For each cable harness defect identification network, the server obtains corresponding first defect review parameters from an internal database or expert system. These parameters are preset based on industry standards, historical data, or expert experience for evaluating the cable harness for defects in various dimensions. For example, for a surface damage identification network, the first defect review parameters may include maximum scratch length, width, depth, etc. allowed.

When the target optical cable harness detection task is executed, the server collects relevant data of the current optical cable harness in real time, and generates second defect evaluation parameters through each defect identification network. Such data may come from image sensors, electrical performance testing equipment, or other online monitoring systems. For example, the surface damage recognition network may analyze the number, location and severity of defects such as scratches, pits, etc. that actually exist according to the real-time captured surface image of the cable harness.

The server dynamically adjusts the set defect tolerance of each optical cable harness defect identification network according to the acquired first defect evaluation parameter and second defect evaluation parameter. This process may involve comparing differences between preset and actual values, taking into account influencing factors such as production lot and environmental conditions, etc., and using algorithms or models to calculate and adjust the tolerance. The adjusted tolerance is referred to as the initial defect tolerance, which reflects the acceptable level of defects for each identified network at the current data and conditions.

Finally, the server comprehensively considers the initial defect tolerance corresponding to each of all the optical cable harness defect identification networks, and determines the iterative defect tolerance through a certain algorithm or decision logic. The iterative defect tolerance is a comprehensive index used for guiding the subsequent detection task allocation and result judgment. For example, if the initial defect tolerance of a certain identification network is lower (i.e. more stringent), then higher priority or more stringent criteria may be given in subsequent detection.

Through the process, the server can dynamically adjust the defect tolerance according to the actual conditions and data of different defect identification dimensions, so that the accuracy and the efficiency of detecting the optical cable harness are improved.

In one possible implementation manner, the step of adjusting the defect indicator migration coefficient includes:

and step A110, determining a task execution allocation rule corresponding to the next detection allocation stage based on the defect detection result.

Step a120, according to a preset defect index amplification corresponding to the task execution allocation rule, enhancing the defect index migration coefficient, and generating the adjusted defect index migration coefficient.

In this embodiment, after the detection of the defect of the optical cable harness in the current stage is completed, the server determines a task execution allocation rule in the next detection allocation stage according to the detection result. These rules may include prioritization of different types of defects, additional attention or omission to particular defects, and the amount of tasks assigned to different inspection devices, among others.

For example, if the detection result in the current stage shows a large number of defects of the surface damage type, the server may decide to increase the detection strength of the surface damage in the next stage, and adjust the detection priority of other types of defects accordingly. Such an adjustment may ensure that the critical defect problem of the cable harness can be more efficiently discovered and resolved with limited detection resources.

After determining task execution allocation rules of the next detection allocation stage, the server increases the defect index migration coefficient according to defect index amplification preset in the rules. The defect index amplification is set according to historical data, expert experience or industry standards and is used for guiding how to adjust the tolerance and the attention to the defects between different detection stages.

For example, if the task execution allocation rule specifies that the attention to the surface damage type defect needs to be greatly improved, the server searches for a preset defect index increase corresponding to the rule, and calculates a new defect index migration coefficient according to the preset defect index increase. The new migration coefficient may be increased by a fixed value or percentage based on the original migration coefficient to reflect the improvement of the attention to the surface damage type defect.

Finally, the server saves the calculated new defect index migration coefficient as the adjusted defect index migration coefficient used in the next detection and distribution stage. This adjusted mobility coefficient will directly affect the tolerance to defects of the optical cable beam and the setting of the detection standard in the next stage, thereby ensuring the accuracy and the high efficiency of the whole detection process.

Through the steps, the server can dynamically adjust the defect index migration coefficient according to the detection result of the current stage and the preset rule so as to adapt to the requirements and changes of different detection stages. The flexibility enables the server to always maintain efficient detection performance and accurate defect identification capability in the face of complex and variable cable harness defect problems.

In a possible implementation manner, the step of adjusting the trust level includes:

And step B110, when the defect detection result represents that the target optical cable harness detection task meets the verification condition, obtaining a blocking log of the optical cable detection characteristic data in a target past detection period, wherein the target past detection period comprises a plurality of unit detection periods.

And step B120, performing trust evaluation on the optical cable detection characteristic data according to the period weight parameter corresponding to each unit detection period in the plurality of unit detection periods and the blocking log in each unit detection period, and generating initial trust corresponding to each unit detection period.

And step B130, performing an averaging process on the initial trust degrees corresponding to the unit detection periods, and generating the adjusted trust degrees.

In this embodiment, after completing the task of detecting the target optical cable harness, the server verifies the generated defect detection result. When the verification result shows that a specific condition (such as defect type, number or severity reaches a preset threshold value) is met, the server triggers a trust adjustment flow.

As a first step of trust adjustment, the server accesses its storage system to obtain a blocking log of the cable detection feature data in the target past detection period. These logs record information about anomalies, data blocks, or failure events that occurred during the detection of the cable harness over a period of time (consisting of multiple unit detection cycles).

For example, if the server finds that a large number of outliers or missing values occur in the cable detection signature data within a certain unit detection period, the data within that period may be considered unreliable and recorded in the blocking log.

After obtaining the blocking log, the server evaluates the trust level of the optical detection characteristic data according to the period weight parameter corresponding to each unit detection period and the blocking log information in the period. The cycle weight parameter is preset according to the importance of each unit detection cycle and the influence degree of the whole detection result.

Confidence assessment may involve analysis and calculation of a number of factors, such as the proportion of outlier data, the amount of missing data, the frequency and severity of blocking events, etc. The server uses the data and information to evaluate the reliability of the cable test feature data in each unit test period and generate a corresponding initial confidence value.

For example, if a block log within a unit detection period shows that there are a number of serious data block events that have a significant impact on the overall detection result, the cable detection characteristic data within that period may be given a lower initial confidence value.

And finally, the server averages the initial trust degrees corresponding to all the unit detection periods. This process may include calculating an average, weighted average, or other statistic of all initial confidence levels to generate an adjusted confidence level value that comprehensively reflects the reliability of the cable test signature data over the entire target past test period.

This adjusted confidence value will be used in subsequent testing tasks to instruct the server on how to process and utilize the cable test signature data. For example, when the confidence level is low, the server may increase the steps of verifying and cleaning the data, so as to reduce the risk of misjudgment caused by unreliable data; when the trust level is higher, the server may trust more and use the data for defect identification and analysis.

In a possible implementation manner, after the step S130, the method further includes:

and step C110, when the defect detection result represents that the target optical cable harness detection task does not meet the verification condition, blocking the target optical cable harness detection task.

And step C120, adjusting a blocking log corresponding to the optical cable detection characteristic data according to the blocking operation.

In this embodiment, after the server performs the task of detecting the target cable harness and generates the defect detection results, it is further verified whether the results satisfy the preset verification conditions. These validation conditions may be set based on industry standards, historical data, expert experience, or specific business requirements to ensure accuracy and reliability of the test results.

If the defect detection results fail to meet these verification conditions, the server will not process or analyze the results further, but rather will block the target cable harness detection task directly. Such blocking operations may include suspending execution of a currently detected task, marking the task as invalid or abnormal, removing the task from the normal processing queue, and so forth.

For example, if the verification condition requires that the false alarm rate in the defect detection result must be below a certain threshold value and the actual result is above the threshold value, the server determines that the detection task does not satisfy the verification condition and performs a blocking operation on the detection task.

The server can adjust the blocking log associated with the optical cable detection characteristic data according to the specific type and reason of the blocking operation while the blocking operation is carried out. These logs record anomalies that occur during the detection process, events that do not meet validation conditions, and other problems that require attention or processing.

Adjusting the blocking log may include adding new log entries, updating the state or information of existing entries, or associating related log entries with particular detection tasks and results, etc. The purpose of this is to keep a complete record of the execution of the detection task and the processing of the results, facilitating subsequent problem tracking, troubleshooting, and performance optimization.

For example, when the server performs blocking operation on a certain detection task that does not meet the verification condition, it may add a new entry in the blocking log, and record the identifier of the task, the execution time, the verification condition that is not met, and the specific cause that causes blocking. At the same time, it may also update other log entries associated with the task to reflect the latest state and processing results of the task.

Through such a process flow, the server can ensure that effective measures can be taken in time to prevent further propagation and processing of erroneous results in the face of detection tasks that do not satisfy the verification conditions, while also being able to retain sufficient log information for subsequent problem analysis and resolution.

In one possible embodiment, the defect detection result includes: harness status tag information, the method further comprising:

and step D110, adding the target optical cable harness detection task and the harness state label information to a first supervision training data sequence when the harness state label information is not matched with the second defect evaluation parameter.

In this embodiment, after the server performs the task of detecting the target optical cable harness and generates the defect detection result, the results include harness status tag information for identifying different statuses (such as normal, damaged, aged, etc.) of the optical cable harness. Meanwhile, the server also generates a second defect evaluation parameter according to the actual detection data, and the second defect evaluation parameter is used for quantitatively evaluating the defect condition of the optical cable harness.

When the server finds that the harness status tag information of a certain target cable harness does not match the second defect-assessment parameter, this means that there may be a case where the tag information is wrong or inconsistent. To handle this, the server will add the target cable harness detection task to the first supervisory training data sequence along with the corresponding harness status tag information. This sequence is used to collect training samples with tag information for later use in supervised learning.

For example, if a cable harness is marked as "normal" but the second defect review parameter indicates that there is a significant sign of damage or aging, the server will add the detection task and label information for that harness to the first supervisory training data sequence for further analysis and processing.

And step D120, performing defect recognition learning on a preset optical cable harness defect recognition network according to the first supervision training data sequence, and generating an initial optical cable harness defect recognition network.

After the first supervision training data sequence is collected, the server uses the data to perform defect identification learning on a preset optical cable harness defect identification network. This process may involve deep learning, neural network training, etc., to enable the network to more accurately identify defects in the cable harness through training.

For example, the server may use a deep learning model such as a Convolutional Neural Network (CNN) as a preset optical cable harness defect identification network, and use tag information in a first supervision training data sequence and a corresponding optical cable harness image as training samples, and train network parameters through an optimization technology such as a back propagation algorithm, so that the type and degree of defects of the optical cable harness can be more accurately identified.

After training, the server obtains an initial optical cable harness defect identification network which has a certain defect identification capability.

Step D130, an unsupervised training data sequence is acquired.

And step D140, performing pseudo tag configuration on the unsupervised training data sequence according to the initial optical cable harness defect identification network, and generating a second supervised training data sequence.

In order to further improve the performance of the cable harness defect identification network, the server also acquires an unsupervised training data sequence, wherein the data is actual detection data without tag information. The server then predicts these unsupervised training data sequences using the initial cable harness defect identification network and configures them with pseudo tags.

The pseudo tag is a temporary tag generated from the predicted result of the network and is used for representing the predicted result of the network on the unlabeled data. By configuring the unsupervised training data with pseudo tags, the server can convert them into tagged training samples, thereby expanding the training data set and improving the generalization ability of the network.

For example, the server may predict cable harness images in an unsupervised training data sequence using an initial cable harness defect recognition network and configure each image with a pseudo tag (e.g., "normal", "damaged", etc.) based on the prediction. The server may then add these pseudo-tagged images to the second supervised training data sequence.

And D150, performing adaptive parameter learning on the guided network corresponding to the initial optical cable harness defect recognition network according to the second supervision training data sequence, and generating an optical cable harness defect recognition network after the adaptive parameter learning.

After the pseudo tag is configured and the second supervisory training data sequence is generated, the server uses the data to adaptively learn parameters for the guided network corresponding to the initial cable harness defect identification network. The guided network is a network similar to the initial network structure but with adjustable parameters, and can train under the guidance of the initial network, and adjust the parameters according to the training result to optimize the performance.

Adaptive parameter learning may involve various optimization algorithms and training techniques, such as gradient descent, random gradient descent (SGD), adam, etc. Through this process, the guided network can learn more about the characteristics and patterns of the defect identification of the cable harness and promote its ability to identify different defect conditions.

And step D160, taking the optical cable harness defect recognition network after the adaptive parameter learning as an adjusted optical cable harness defect recognition network.

And finally, the server takes the guided network after the adaptive parameter learning as an adjusted optical cable harness defect identification network to be used in a subsequent optical cable harness detection task. The adjusted network has stronger generalization capability and higher recognition accuracy due to more training and learning.

In one possible implementation, the step D150 may include:

And step D151, acquiring the supervision trust degree corresponding to each supervision training data in the second supervision training data sequence.

In the process of adaptive parameter learning by the server, firstly, the supervision and trust degree of each supervision and training data (namely, data with pseudo labels) in the second supervision and training data sequence needs to be acquired. The supervision confidence is an index for measuring the reliability and accuracy of each supervision training data, and can be determined according to the source, quality, confidence of the pseudo tag and other factors of the data.

For example, for some supervised training data annotated by experienced specialists, the server may give higher degree of supervised trust; for data with automatic labeling or unknown sources, lower supervision trust may be given. In addition, the server may also automatically evaluate the supervised confidence level of each supervised training data using some statistical method or machine learning algorithm.

And step D152, sampling the second supervision training data sequence according to the supervision trust degree, and determining active training data and passive training data.

After acquiring the supervised confidence level of each supervised training data, the server samples the second sequence of supervised training data based on the confidence levels to determine positive and negative training data for adaptive parameter learning.

Positive training data refers to data that is highly supervised, considered reliable and accurate, and that will be used to guide the guided network in forward learning, i.e., learning how to correctly identify the defects of the cable harness. Negative training data refers to data that has low supervisory confidence and may be subject to errors or uncertainties, which will be used to guide the guided network through negative learning, i.e., learning how to avoid false recognition or to deal with uncertainties.

For example, the server may set a supervised confidence threshold, with data above the threshold being positive training data and data below the threshold being negative training data. Or the server may employ more complex sampling strategies such as hierarchical sampling or weighted sampling based on a distribution of supervisory confidence levels.

And D153, performing defect identification learning on the guided network according to the positive training data and the negative training data, and generating the optical cable harness defect identification network after the adaptive parameter learning.

Finally, the server will use the positive training data and the negative training data to perform defect identification learning on the guided network. In this process, the guided network will attempt to fit the correct tag in the positive training data while avoiding fitting the wrong or uncertain tag in the negative training data.

In particular, the server may employ a strategy of semi-supervised learning or weakly supervised learning to train the guided network. For example, for positive training data, the server may employ standard supervised learning methods (e.g., cross entropy loss functions) to train the network; for passive training data, the server may employ special penalty functions or regularization terms to guide the network to avoid false identifications or to handle uncertainties (e.g., using tag smoothing techniques, introducing noisy tags, etc.).

Through such a training process, the guided network will be able to learn more effective features and patterns regarding the identification of cable harness defects and promote its ability to identify different defect conditions. Finally, the server obtains an optical cable harness defect identification network with adaptive parameter learning, and the network has stronger generalization capability and higher identification accuracy.

In a possible embodiment, after the step D150, the method further includes:

And E110, updating the initial optical cable harness defect identification network according to the optical cable harness defect identification network after the adaptive parameter learning, and generating an adjusted initial optical cable harness defect identification network.

After the server generates an adaptive parameter learned optical cable harness defect identification network through adaptive parameter learning, the learned network is then used to update the initial optical cable harness defect identification network. The purpose of the update is to integrate the optimizations and improvements obtained during the adaptive parameter learning process into the original network, thereby improving its performance.

For example, the server may copy parameters such as weights and offsets in the network after adaptive parameter learning to the initial network, or combine parameters of the two networks using some fusion strategy. Through the updating, the initial optical cable harness defect identification network can inherit the advantages of the network after the adaptive parameter learning, and has stronger defect identification capability.

And E120, iteratively executing the acquisition of the unsupervised training data sequence to the guided network corresponding to the initial cable harness defect recognition network according to the adjusted initial cable harness defect recognition network, and performing adaptive parameter learning on the guided network corresponding to the initial cable harness defect recognition network according to the second supervised training data sequence, so as to generate an adaptive parameter learning flow of the cable harness defect recognition network after adaptive parameter learning until reaching preset adaptive parameter learning times.

After the initial cable harness defect identification network is updated, the server iteratively executes an adaptive parameter learning process. This means that the server will repeatedly perform a series of steps from the acquisition of the unsupervised training data sequence to the generation of the adaptive parameter learned cable harness defect identification network.

The aim of iteratively executing the adaptive parameter learning process is to continuously optimize and improve the performance of the optical cable harness defect identification network through a plurality of iterations. In each iteration, the server generates new pseudo tags by using the new unsupervised training data sequence, and performs adaptive parameter learning based on the pseudo tags, so as to continuously improve the recognition capability of the network.

For example, after a first iteration, the server may obtain a slightly improved performance fiber optic cable harness defect identification network. Then, in a second iteration, the server will use the new unsupervised training data sequence to further refine the network. Through a number of such iterations, the server's cable harness defect identification network will be able to gradually approach optimal performance.

In order to avoid unlimited iterations and to ensure the convergence of the algorithm, the server sets a preset number of adaptive parameter learning times. When the number of iterations reaches this preset value, the server will stop the iteration and perform the next operation.

For example, if the preset adaptive parameter learning number set by the server is 10 times, the server will terminate the adaptive parameter learning flow after completing the 10 th iteration. The optical cable harness defect identification network at this time is the final network after multiple optimization and improvement.

And E130, taking the current optical cable harness defect recognition network after adaptive parameter learning as the adjusted optical cable harness defect recognition network.

After the iteration is terminated, the server takes the current optical cable harness defect identification network after adaptive parameter learning as an adjusted network. The adjusted network is the final result obtained by the server through repeated iteration adaptive parameter learning flow, and has stronger defect identification capability and higher accuracy.

For example, if the server obtains a network for identifying defects in the cable harness that is excellent in performance in the last iteration, then this network will be considered an adjusted network and used in subsequent cable harness detection tasks. Through the processing flow, the server can continuously improve the performance and the capability of identifying the defects of the optical cable harness.

In one possible implementation manner, the embodiment may adjust the first defect evaluation parameter according to the adjusted defect identification network of the optical cable harness.

For example, after the server completes adaptive parameter learning of the cable harness defect identification network and generates an adjusted network, the first defect review parameter is then adjusted based on the adjusted network. The first defect review parameter is a set of parameters for evaluating the extent of defects in the cable harness, which may include indicators of various aspects of defect size, shape, location, etc.

The purpose of the server to adjust the first defect evaluation parameters is to make the parameters more in line with the actual defect situation of the optical cable harness, thereby improving the accuracy and reliability of defect evaluation. The adjustment is based on the characteristics and modes learned by the adjusted optical cable harness defect identification network, and the characteristics and modes reflect the distribution and rules of the actual optical cable harness defects.

Specifically, the server may adjust the first defect review parameter by:

Analyzing the adjusted network performance: firstly, the server performs performance analysis on the adjusted optical cable harness defect identification network, wherein the performance analysis comprises indexes such as identification accuracy, recall rate, F1 score and the like. These indicators can reflect the network's ability to identify on different types of defects.

Determining key defect characteristics: based on the performance analysis results, the server may determine which features are critical to identifying cable harness defects. These key features may include specific textures, shape changes, color anomalies, and the like.

Adjusting parameter settings: the server adjusts the setting of the first defect review parameter based on the determined key defect signature. For example, if the network performs well in identifying a type of defect, the server may increase the weight of the evaluation parameters associated with that type of defect; conversely, if the network does not identify well on a certain type of defect, the server may decrease the weight of the corresponding parameter or adjust its threshold.

Verifying the adjustment effect: after the first defect-assessment parameter is adjusted, the server uses the actual cable harness image or sample data to verify the adjustment effect. By comparing the defect evaluation results before and after adjustment, the server can evaluate the rationality and effectiveness of parameter adjustment.

Iterative optimization: if the verification result shows that the adjusted parameter setting does not reach the expected effect, the server may perform iterative optimization, adjust the parameter setting again and re-verify until a satisfactory defect evaluation result is obtained.

Through the adjustment process, the server can ensure that the first defect evaluation parameter and the adjusted optical cable harness defect recognition network keep consistent, so that the performance and reliability of the whole defect recognition and evaluation system are improved.

The detection system 100 of the defect of the optical cable harness shown in fig. 2 includes: a processor 1001 and a memory 1003. The processor 1001 is coupled to the memory 1003, such as via a bus 1002. Optionally, the system 100 for detecting defects in a fiber optic cable harness may further include a transceiver 1004, where the transceiver 1004 may be used for data interaction between the server and other servers, such as transmission of data and/or reception of data, etc. It should be noted that, the transceiver 1004 is not limited to one in actual dispatching, and the structure of the system 100 for detecting defects of the optical cable harness is not limited to the embodiment of the present application.

The Processor 1001 may be a CPU (Central Processing Unit ), general purpose Processor, DSP (DIGITAL SIGNAL Processor, data signal Processor), ASIC (Application SpecificIntegrated Circuit ), FPGA (Field Programmable GATE ARRAY, field programmable gate array) or other programmable logic device, transistor logic device, hardware component, or any combination thereof. Which may implement or perform the various exemplary logic blocks, modules and circuits described in connection with this disclosure. The processor 1001 may also be a combination that implements computing functionality, such as a combination comprising one or more microprocessors, a combination of a DSP and a microprocessor, or the like.

Bus 1002 may include a path to transfer information between the components. Bus 1002 may be a PCI (PERIPHERAL COMPONENT INTERCONNECT, peripheral component interconnect standard) bus, or an EISA (ExtendedIndustry Standard Architecture ) bus, or the like. The bus 1002 may be divided into an address bus, a data bus, a control bus, and the like. For ease of illustration, only one thick line is shown in fig. 2, but not only one bus or one type of bus.

The Memory 1003 may be, but is not limited to, ROM (Read Only Memory) or other type of static storage device that can store static information and instructions, RAM (Random Access Memory ) or other type of dynamic storage device that can store information and instructions, EEPROM (ELECTRICALLY ERASABLEPROGRAMMABLE READ ONLY MEMORY ), CD-ROM (Compact DiscRead Only Memory, compact disc Read Only Memory) or other optical disk storage, optical disk storage (including compact discs, laser discs, optical discs, digital versatile discs, blu-ray discs, etc.), magnetic disk storage media, other magnetic storage devices, or any other medium that can be used to carry or store program code and that can be Read by a computer.

The memory 1003 is used for storing program codes for executing the embodiments of the present application and is controlled to be executed by the processor 1001. The processor 1001 is configured to execute the program code stored in the memory 1003 to implement the steps shown in the foregoing method embodiment.

Embodiments of the present application provide a computer readable storage medium having program code stored thereon, which when executed by a processor, implements the steps of the foregoing method embodiments and corresponding content.

It should be understood that, although various operation steps are indicated by arrows in the flowcharts of the embodiments of the present application, the order in which these steps are implemented is not limited to the order indicated by the arrows. In some implementations of embodiments of the application, the implementation steps in the flowcharts may be performed in other orders based on demand, unless explicitly stated herein. Furthermore, some or all of the steps in the flowcharts may include a plurality of sub-steps or a plurality of stages, depending on the actual implementation scenario. Some or all of these sub-steps or phases may be performed at the same time, or each of these sub-steps or phases may be performed at different times, respectively. In the case of different execution timings, the execution order of the sub-steps or stages may be flexibly configured based on requirements, which is not limited by the embodiment of the present application.

The foregoing is merely an optional implementation manner of some of the implementation scenarios of the present application, and it should be noted that, for those skilled in the art, other similar implementation manners according to the technical idea of the present application may be adopted without departing from the technical idea of the solution of the present application, which is also within the protection scope of the embodiments of the present application.

Claims

1. A method for detecting defects in an optical cable harness, the method comprising:

2. The method for detecting a cable harness defect according to claim 1, wherein the target cable harness defect identification network comprises: the method for determining the iterative defect tolerance of the target optical cable harness detection task includes the steps of:

3. The method for detecting defects of an optical cable harness according to claim 1, wherein the step of adjusting the defect index migration coefficient comprises:

4. The method for detecting defects of an optical cable harness according to claim 1, wherein the step of adjusting the degree of trust comprises:

5. The method for detecting defects of a fiber optic cable harness according to claim 1, wherein after the performing feature data identification on the target fiber optic cable harness detection task to generate a defect detection result, the method further comprises:

6. The method for detecting defects of an optical cable harness according to claim 1, wherein the defect detection result includes: harness status tag information, the method further comprising:

acquiring an unsupervised training data sequence;

7. The method for detecting defects of an optical cable harness according to claim 6, wherein the performing adaptive parameter learning on the guided network corresponding to the initial optical cable harness defect identification network according to the second supervision training data sequence, and generating the optical cable harness defect identification network after the adaptive parameter learning comprises:

8. The method for detecting defects of an optical cable harness according to claim 6, wherein after performing adaptive parameter learning on a guided network corresponding to the initial optical cable harness defect identification network according to the second supervisory training data sequence and generating an optical cable harness defect identification network after the adaptive parameter learning, the method further comprises:

9. The method for detecting a defect in a fiber optic cable harness of claim 6, further comprising:

10. A system for detecting a defect in an optical cable harness comprising a processor and a computer readable storage medium storing machine executable instructions that when executed by the processor implement the method for detecting a defect in an optical cable harness of any one of claims 1-9.