Adaptive optimization method and system for connector production technology

Abstract

The invention relates to the technical field of artificial intelligence, in particular to a self-adaptive optimization method and system for a connector production skill, and the method comprises the steps: monitoring a production technology assembly line of a connector, and obtaining a connector product set and a monitoring result sequence; performing product performance test on the connector product set to obtain product test data; performing flow anomaly identification on the monitoring result sequence to obtain an anomaly identification result; performing fault type prediction on the anomaly recognition result by using a pre-trained anomaly-fault bidirectional detection model to obtain a predicted fault type; and when the test fault type is different from the predicted fault type, performing anomaly identification on the fault product monitoring result sequence by using the anomaly-fault bidirectional detection model to obtain a predicted anomaly identification result for assisting in improving a production technology assembly line. According to the invention, the problem that a large amount of manpower and time need to be consumed for exploration and debugging when the production process of the current connector is modified can be solved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a self-adaptive optimization method and system for connector production technology. Background Art

[0002] With the development and popularization of electronic devices, USB or type-C connectors have become important tools for data or power transmission between electronic devices. These connectors are affected by the production process and are prone to substandard products, which may cause data loss or damage to electronic devices when connected.

[0003] In order to improve the production process of connectors, product samples are taken regularly and tested through experiments to obtain test results. Then, according to the test results, the numerical control parameters when manufacturing the product are modified to complete the modification of the production process. This process modification depends on subjective human evaluation and is also lagging. Only when new quality problems arise or the number of defective products is large will the production process be adjusted. The modified production process can only solve the current problem, but cannot achieve the optimal state. Once the production materials or production equipment change, the production process needs to be re-planned, which consumes a lot of manpower costs. Summary of the invention

[0004] The present invention provides an adaptive optimization method and system for connector production technology, the main purpose of which is to solve the problem that the production process of connectors such as USB or type-C is difficult to approach perfection, and when materials and equipment change, modifying the production process requires a lot of manpower and time for exploration and debugging.

[0005] To achieve the above-mentioned purpose, the present invention provides an adaptive optimization method for connector production technology, comprising:

[0006] Obtain a production technology assembly line of a target model connector, monitor the production technology assembly line using a pre-built edge detection device, and obtain a connector product set and a monitoring result sequence corresponding to each connector product in the connector product set within a preset time when the production technology assembly line is started;

[0007] According to a preset test strategy, a product performance test is performed on the connector product set to obtain product test data, and according to the product test data, a faulty product is extracted from the connector product set, and a test fault type of the faulty product is obtained;

[0008] Performing process anomaly identification on the monitoring result sequence of the target product in the connector product set to obtain an anomaly identification result, and marking the anomaly identification result on the target product;

[0009] Using a pre-trained anomaly-fault bidirectional detection model, predicting the fault type of the anomaly recognition result to obtain a predicted fault type;

[0010] Determining whether the test fault type corresponds to the predicted fault type;

[0011] When the test fault type corresponds to the predicted fault type, outputting the production step name corresponding to the process abnormality identification result to a pre-built list to be improved;

[0012] When the test fault type is different from the predicted fault type, performing a monitoring data tracing operation on the faulty product to obtain a monitoring result sequence of the faulty product;

[0013] Using the anomaly-fault bidirectional detection model, anomaly identification is performed on the faulty product monitoring result sequence to obtain a predicted anomaly identification result;

[0014] Identify the production step name corresponding to the predicted abnormality recognition result, and transform the predicted abnormality recognition result into a predicted improvement enhancement direction according to a preset improvement strategy;

[0015] The production step name and the predicted improvement enhancement direction are saved in the list to be improved.

[0016] Optionally, the monitoring result sequence includes production step names, machine numerical control parameters, and product scan images corresponding to each production step in the production technology assembly line, wherein the product scan images include photographed images and infrared scanning structure images.

[0017] Optionally, performing a product performance test on the connector product set according to a preset test strategy to obtain product test data includes:

[0018] According to a preset test strategy, using a pre-built simulation service, product simulation is performed on the monitoring result sequence to obtain a simulation model, and a test process based on a repetitive test type item is performed on the simulation model to obtain a simulation test result;

[0019] Performing a test process based on a one-time detection type item on the connector product set to obtain a test result;

[0020] The simulation test results and the experimental test results are uniformly output to obtain product test data.

[0021] Optionally, the performing process anomaly identification on the monitoring result sequence of the target product in the connector product set to obtain an anomaly identification result includes:

[0022] Extracting a target product from the connector product set, and extracting machine numerical control parameters and product scanning images in the monitoring result sequence of the target product;

[0023] Using the product scan image and the machine numerical control parameters, performing an abnormal operation comparison operation based on the operation purpose and operation result of the machine on the monitoring result sequence to obtain an abnormal operation recognition result;

[0024] Acquire a standard product structure diagram of the target product, use the product scan image, and perform structural distribution anomaly recognition on the target product according to the standard product structure diagram to obtain a structural anomaly recognition result;

[0025] The operation abnormality recognition result and the structure abnormality recognition result are output to obtain an abnormality recognition result.

[0026] Optionally, the using a pre-trained anomaly-fault bidirectional detection model to predict the fault type of the anomaly recognition result to obtain the predicted fault type includes:

[0027] Using a pre-trained anomaly-fault bidirectional detection model, a feature extraction operation is performed on the anomaly recognition result to obtain an anomaly feature vector;

[0028] Performing feature recognition on the abnormal feature vector to obtain feature recognition results, and grouping the feature recognition results into preset abnormal types to obtain feature clusters corresponding to each abnormal type;

[0029] Fully connected classification judgment is performed on each of the feature clusters to obtain a fault probability distribution value corresponding to a preset fault type set, and a predicted fault type is extracted from the fault type set based on the fault probability distribution value.

[0030] Optionally, before predicting the fault type of the abnormality identification result using the pre-trained abnormality-fault bidirectional detection model to obtain the predicted fault type, the method further includes:

[0031] Get the pre-built production exception type set and failure type set:

[0032] F = {f1, f2, ..., f m}

[0033] A={a1,a2,...,a n}

[0034] In the formula, F represents the set of production anomaly types, m represents the number of types in the set of production anomaly types, and f mrepresents the mth type of production anomaly type in the production anomaly type set, A represents the fault type set; n represents the number of types in the production anomaly type set, a n represents the nth type of fault in the set of fault types;

[0035] Obtain sample data including normal production process data and abnormal production process data, wherein each production process data in the sample data is also equipped with a corresponding fault type:

[0036] D={(x i ,y fi ,y ai )}

[0037] In the formula, D represents the sample data, x i represents the i-th production process data in the sample data, y fi Indicates the fault type corresponding to the i-th production process data in the sample data, y ai Indicates the production anomaly type corresponding to the i-th production process data in the sample data;

[0038] Obtain an anomaly-fault bidirectional detection model including two output layers and a generative adversarial network, and configure the production anomaly type set and the fault type set as standard output types of the two output layers, respectively, wherein the two output layers are a process anomaly prediction output layer and a fault judgment output layer, and the generative adversarial network includes a generator and a discriminator;

[0039] Extracting a target sample from the sample data in turn, performing network forward operation on the target sample using the anomaly-fault bidirectional detection model, and obtaining predicted sample faults and predicted sample production anomalies;

[0040] Using the pre-built model overall loss function, according to the real fault type and real production abnormality type corresponding to the target sample in the sample data, the loss value of the predicted sample fault and the predicted sample production abnormality is calculated to obtain the overall loss value. The model overall loss function is expressed as:

[0041] L 整体 =L total +αL G +βL D

[0042] Where, L 整体 Expressed as the overall loss function of the model, L tota1 represents the output layer loss function of the process anomaly prediction output layer and the fault judgment output layer, L G represents the loss function of the generator G, L Drepresents the loss function of the discriminator D, α and β represent weight parameters for balancing the loss function of the output layer and the loss function of the generative adversarial network;

[0043] Among them, L total It is expressed as:

[0044] L total =λL cls +(1-λ)L reg

[0045]

[0046] Where, L total represents the output layer loss function, L cls represents the classification loss function of the fault judgment output layer, L reg represents the regression loss function of the process anomaly prediction output layer, λ is the weight parameter; y i represents the predicted sample fault corresponding to the i-th target sample, represents the actual fault type corresponding to the i-th target sample, and N represents the number of types in the fault type set; a i Indicates that the predicted sample corresponding to the i-th target sample produces an abnormality, represents the actual production anomaly type corresponding to the i-th target sample, and M represents the number of samples of the sample data;

[0047] Among them, L G It is expressed as:

[0048]

[0049] In the formula, represents the generated data calculated by the generator for the predicted sample failure and the predicted sample production anomaly, P g represents the standard normal distribution of the generator, represents the discriminant result of the discriminator on the generated data, x represents the real fault type and the real production abnormality type, P r Representing the standard normal distribution of the real fault type and the real production abnormality type;

[0050] According to the gradient descent algorithm, the overall loss value is minimized to obtain the network model parameters when the overall loss value is the smallest, and the network reverse parameters of the anomaly-fault bidirectional detection model are updated according to the network model parameters to obtain an updated anomaly-fault bidirectional detection model;

[0051] Determining whether the overall loss value converges;

[0052] When the overall loss value has not converged, returning to the above step of extracting a target sample from the sample data in sequence, and iteratively optimizing the update anomaly-fault bidirectional detection model;

[0053] When the overall loss value converges, the training process is stopped to obtain a trained anomaly-fault bidirectional detection model.

[0054] Optionally, after obtaining the product test data, the method further includes:

[0055] Saving the product test data into a pre-built model feedback database;

[0056] According to the preset time interval frequency, the anomaly-fault bidirectional detection model is trained using the model feedback database to obtain an optimized anomaly-fault bidirectional detection model.

[0057] Optionally, after marking the abnormality identification result on the target product, the method further includes:

[0058] Supervising the abnormality recognition result within a preset supervision time period to obtain an abnormality rate;

[0059] Determining whether the abnormality rate is greater than a preset warning threshold;

[0060] When the abnormal rate is greater than the warning threshold, an alarm prompt message is generated.

[0061] Optionally, the step of converting the predicted abnormality identification result into a predicted improvement enhancement direction according to a preset improvement strategy includes:

[0062] According to a preset improvement strategy, the standard product structure diagram is compared with the predicted abnormality recognition result to obtain difference steps and difference values;

[0063] The difference value is inverted to obtain the value to be improved, the difference step is converted into the step to be improved, and the predicted improvement and enhancement direction is obtained according to the value to be improved and the step to be improved.

[0064] To achieve the above object, the present invention also provides an adaptive optimization system for connector production technology, comprising:

[0065] A process monitoring module is used to obtain the production technology assembly line of the target model connector, monitor the production technology assembly line using a pre-built edge detection device, and obtain a connector product set and a monitoring result sequence corresponding to each connector product in the connector product set within a preset time of starting the production technology assembly line;

[0066] A product testing module, configured to perform product performance testing on the connector product set according to a preset test strategy to obtain product test data, extract faulty products from the connector product set according to the product test data, obtain the test fault type of the faulty product, and perform process anomaly identification on the monitoring result sequence of the target product in the connector product set to obtain anomaly identification results, and mark the anomaly identification results on the target product;

[0067] A fault prediction module is used to use a pre-trained anomaly-fault bidirectional detection model to predict the fault type of the anomaly recognition result, obtain the predicted fault type, and determine whether the test fault type corresponds to the predicted fault type, and when the test fault type corresponds to the predicted fault type, output the production step name corresponding to the process anomaly recognition result to a pre-built list of items to be improved;

[0068] The module for identifying points to be improved is used to perform monitoring data tracing operations on the faulty product when the test fault type is different from the predicted fault type to obtain a sequence of monitoring results of the faulty product, and to use the abnormality-fault bidirectional detection model to perform abnormality identification on the sequence of monitoring results of the faulty product to obtain a predicted abnormality identification result, and to identify the production step name corresponding to the predicted abnormality identification result, and according to a preset improvement strategy, to convert the predicted abnormality identification result into a predicted improvement enhancement direction, and to save the production step name and the predicted improvement enhancement direction in the list to be improved.

[0069] In order to solve the above problem, the present invention further provides an electronic device, the electronic device comprising:

[0070] a memory storing at least one instruction; and

[0071] The processor executes the instructions stored in the memory to implement the above-mentioned adaptive optimization method of connector production technology.

[0072] In order to solve the above problems, the present invention also provides a computer-readable storage medium, in which at least one instruction is stored. The at least one instruction is executed by a processor in an electronic device to implement the above-mentioned adaptive optimization method of connector production technology.

[0073] The present invention solves the problem described in the background technology. The present invention first monitors the steps of the connector production technology assembly line by means of digital monitoring and image monitoring through edge detection equipment. When the connector production starts, a monitoring result sequence of each connector will be obtained. The present invention first compares the image recognition results of the numerical control parameters with the actual production results to complete the abnormality recognition and obtain the abnormality recognition results. These abnormality recognition results can effectively solve the problems of uneven product packaging injection molding, wrong or asymmetric installation positions of parts, etc. Then, the abnormality recognition results are predicted by a pre-trained abnormality-fault bidirectional detection model to obtain the predicted fault type. When the predicted fault type is the same as the test fault type actually detected, it indicates that only the step where the abnormality occurs needs to be improved. When the predicted fault type is different from the test fault type, it indicates that there are deficiencies in the technical process that need to be repaired. The abnormality-fault bidirectional detection model has two output layers, which can find out the deficiencies that may occur in the technical process according to the test fault type, and can also predict the fault type of the product according to the deficiencies in the technical process. When training the model, the present invention also adds a generative adversarial network to further increase the mapping relationship between process abnormalities and fault types. Therefore, the present invention can solve the problem that the production process of connectors such as USB or type-C is difficult to reach perfection, and when materials and equipment change, modifying the production process requires a lot of manpower and time for exploration and debugging. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 A schematic flow chart of a method for adaptively optimizing connector production techniques provided by an embodiment of the present invention;

[0075] Figure 2 A functional module diagram of a self-adaptive optimization system for connector production technology provided by an embodiment of the present invention;

[0076] Figure 3 A schematic structural diagram of an electronic device for implementing the adaptive optimization method for connector production technology provided by one embodiment of the present invention.

[0077] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0078] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0079] The embodiment of the present application provides an adaptive optimization method for connector production technology. The execution subject of the adaptive optimization method for connector production technology includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided by the embodiment of the present application. In other words, the adaptive optimization method for connector production technology can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc.

[0080] Reference Figure 1 FIG. 1 is a flow chart of a method for adaptively optimizing connector production technology provided by an embodiment of the present invention. In this embodiment, the method for adaptively optimizing connector production technology includes:

[0081] S1. Obtain the production technology assembly line of the target model connector, use the pre-built edge detection equipment to monitor the production technology assembly line, and obtain the connector product set and the monitoring result sequence corresponding to each connector product in the connector product set within the preset time of starting the production technology assembly line.

[0082] In an embodiment of the present invention, the target model connector is a USB or type-C type connector, and the production technology assembly line includes material selection, injection molding, stamping, electroplating, assembly and welding, etc. These steps can be replaced following technological advances.

[0083] Furthermore, the edge detection device includes an optical camera, an infrared camera, a data extraction service interface, etc. Among them, the captured image obtained by the optical camera can check whether the product's appearance is deformed, has burrs, or is missing, and the infrared scanning structure image obtained by the infrared camera can check whether the structure inside the connector is properly positioned, whether the plastic is evenly wrapped, etc., and the data extraction service interface directly extracts the machine numerical control parameters controlled at each step, which is convenient for comparison with the product scanning image to check the difference between the ideal value and the actual operation value of the machine.

[0084] Specifically, in an embodiment of the present invention, after the production technology assembly line is monitored by the edge detection device, when the assembly line is started, batch production can be carried out within a preset time period to obtain a set of connector products, and the edge detection device will obtain a monitoring result sequence for each connector product, and the monitoring result sequence includes the production step name, machine numerical control parameters, and product scanned images corresponding to each production step in the production technology assembly line, wherein the product scanned image includes a photographed image and an infrared scanning structure image.

[0085] S2. According to a preset test strategy, perform product performance testing on the connector product set to obtain product test data, and extract faulty products from the connector product set based on the product test data, and obtain the test fault type of the faulty product.

[0086] Among them, the testing strategy refers to the planning strategy for testing items such as stamping quality testing, electroplating layer testing, welding quality testing, usage times testing, electrical performance testing and mechanical performance testing on products through both actual testing and model prediction.

[0087] In detail, in an embodiment of the present invention, the product performance test is performed on the connector product set according to a preset test strategy to obtain product test data, including:

[0088] According to a preset test strategy, using a pre-built simulation service, product simulation is performed on the monitoring result sequence to obtain a simulation model, and a test process based on a repetitive test type item is performed on the simulation model to obtain a simulation test result;

[0089] Performing a test process based on a one-time detection type item on the connector product set to obtain a test result;

[0090] The simulation test results and the experimental test results are uniformly output to obtain product test data.

[0091] In the embodiment of the present invention, a simulation test is performed using a monitoring result sequence as a data basis and a BIM service as a simulation service, wherein the BIM (Building Information Modeling) technology is a revolutionary method for architectural design, construction and management, which can use digital information to design the physical and functional characteristics of each building or infrastructure project, including three-dimensional modeling, life cycle management, performance analysis testing and other functions.

[0092] In the embodiment of the present invention, the simulation model constructed by BIM service executes the complex, repetitive and long-term test items of the product. For simple one-time tests, such as sealing and waterproofing, the actual product can be tested. Finally, the two test results are uniformly output to obtain the final product test data.

[0093] The present invention also extracts faulty products from the connector product set based on the product test data, and marks the test fault type of each faulty product.

[0094] S3. Perform process anomaly identification on the monitoring result sequence of the target product in the connector product set to obtain an anomaly identification result, and mark the anomaly identification result on the target product.

[0095] In detail, in an embodiment of the present invention, the process anomaly identification is performed on the monitoring result sequence of the target product in the connector product set to obtain the anomaly identification result, including:

[0096] Extracting a target product from the connector product set, and extracting machine numerical control parameters and product scanning images in the monitoring result sequence of the target product;

[0097] Using the product scan image and the machine numerical control parameters, performing an abnormal operation comparison operation based on the operation purpose and operation result of the machine on the monitoring result sequence to obtain an abnormal operation recognition result;

[0098] Acquire a standard product structure diagram of the target product, use the product scan image, and perform structural distribution anomaly recognition on the target product according to the standard product structure diagram to obtain a structural anomaly recognition result;

[0099] The operation abnormality recognition result and the structure abnormality recognition result are output to obtain an abnormality recognition result.

[0100] The standard product structure diagram refers to a scanned image of a fully qualified product that has undergone various tests, and also includes externally shot images and structure scanned images, which are used to provide a standardized structure for the product.

[0101] According to the above steps, the embodiment of the present invention performs abnormality detection from two aspects: on the one hand, by comparing the machine numerical control parameters and the product scanned image, check whether the machine execution capability is abnormal; on the other hand, by comparing the standard product structure diagram with the product scanned image, check the structural distribution of the target product, such as the uniformity of the plastic packaging, the sealing, the relative position distribution of each part, etc. The abnormality identification step in the embodiment of the present invention performs abnormality comparison in each step of the production technology assembly line.

[0102] The embodiment of the present invention obtains an abnormality identification result through process abnormality identification. When the abnormality identification result is a null value, a no-abnormality mark is provided. When there is content in the abnormality identification, the abnormal step name and the abnormal point in the monitoring result sequence are marked together on the target product, which can facilitate the subsequent direct extraction of defective products in the connector product set.

[0103] Furthermore, in an embodiment of the present invention, after marking the abnormality identification result on the target product, the method further includes:

[0104] Supervising the abnormality recognition result within a preset supervision time period to obtain an abnormality rate;

[0105] Determining whether the abnormality rate is greater than a preset warning threshold;

[0106] When the abnormal rate is greater than the warning threshold, an alarm prompt message is generated.

[0107] In the embodiment of the present invention, if the abnormality rate is less than the warning threshold value of 1%, it can be considered that the product quality is unqualified due to accidental factors. If the abnormality rate is greater than 1%, it can be determined that a failure has occurred in the production technology assembly line and an alarm prompt is required.

[0108] S4. Using a pre-trained anomaly-fault bidirectional detection model, predict the fault type of the anomaly recognition result to obtain a predicted fault type.

[0109] In an embodiment of the present invention, the anomaly-fault bidirectional detection model is a neural network model including two output layers and a generative adversarial network, which is used to judge the fault type according to the production anomaly type of the production technology assembly line, or to infer the production anomaly type according to the fault type detected.

[0110] In detail, in the embodiment of the present invention, the use of the pre-trained anomaly-fault bidirectional detection model to predict the fault type of the anomaly recognition result to obtain the predicted fault type includes:

[0111] Using a pre-trained anomaly-fault bidirectional detection model, a feature extraction operation is performed on the anomaly recognition result to obtain an anomaly feature vector;

[0112] Performing feature recognition on the abnormal feature vector to obtain feature recognition results, and grouping the feature recognition results into preset abnormal types to obtain feature clusters corresponding to each abnormal type;

[0113] Fully connected classification judgment is performed on each of the feature clusters to obtain a fault probability distribution value corresponding to a preset fault type set, and a predicted fault type is extracted from the fault type set based on the fault probability distribution value.

[0114] In the embodiment of the present invention, firstly, each abnormal problem in the production technology assembly line is identified through feature extraction and feature recognition to obtain each production abnormality type. Among them, each production abnormality type is the most important m labels among the abnormality types extracted by the principal component analysis method, such as [injection molding, processing, assembly...], so that each abnormality type is summarized into the m labels. In addition, the present invention also summarizes the fault type through feature engineering to obtain n labels, such as [water wading, leakage, loss resistance...].

[0115] Then, in the embodiment of the present invention, through the mapping relationship between the [production anomaly type] label and the [fault type] label, the feature clusters corresponding to each production anomaly type are fully connected to obtain the fault probability distribution value of the fault type. In the embodiment of the present invention, based on the detection experience in actual business scenarios, the fault type with a score greater than a certain threshold, such as 80%, is extracted as the predicted fault type.

[0116] Further, in an embodiment of the present invention, before predicting the fault type of the abnormality recognition result using the pre-trained abnormality-fault bidirectional detection model to obtain the predicted fault type, the method further includes:

[0117] Get the pre-built production exception type set and failure type set:

[0118] F = {f1, f2, ..., f m}

[0119] A={a1,a2,...,a n}

[0120] In the formula, F represents the set of production anomaly types, m represents the number of types in the set of production anomaly types, and f m represents the mth type of production anomaly type in the production anomaly type set, A represents the fault type set; n represents the number of types in the production anomaly type set, a n represents the nth type of fault in the set of fault types;

[0121] Obtain sample data including normal production process data and abnormal production process data, wherein each data in the sample data also includes a fault mark:

[0122] D={(x i ,y fi ,y ai )}

[0123] In the formula, D represents the sample data, x i represents the i-th production process data in the sample data, y fi Indicates the fault type corresponding to the i-th production process data in the sample data, y ai Indicates the production anomaly type corresponding to the i-th production process data in the sample data;

[0124] Obtain an anomaly-fault bidirectional detection model with two output layers and a generative adversarial network, and configure the production anomaly type set and the fault type set as standard output types of the two output layers, wherein the two output layers are a process anomaly prediction output layer and a fault judgment output layer, and the generative adversarial network includes a generator and a discriminator;

[0125] Extracting a target sample from the sample data in turn, performing network forward operation on the target sample using the anomaly-fault bidirectional detection model, and obtaining predicted sample faults and predicted sample production anomalies;

[0126] Using the pre-built model overall loss function, according to the real fault type and real production abnormality type corresponding to the target sample in the sample data, the loss value of the predicted sample fault and the predicted sample production abnormality is calculated to obtain the overall loss value. The model overall loss function is expressed as:

[0127] L 整体 =L total +αL G +βL D

[0128] Where, L 整体 Expressed as the overall loss function of the model, L total represents the output layer loss function of the process anomaly prediction output layer and the fault judgment output layer, L G represents the loss function of the generator G, L D represents the loss function of the discriminator D, α and β represent weight parameters for balancing the loss function of the output layer and the loss function of the generative adversarial network;

[0129] Among them, L total It is expressed as:

[0130] L total =λL cls +(1-λ)L reg

[0131]

[0132] Where, L total represents the output layer loss function, L cls represents the classification loss function of the fault judgment output layer, L reg represents the regression loss function of the process anomaly prediction output layer, λ is the weight parameter; y i represents the predicted sample fault corresponding to the i-th target sample, represents the actual fault type corresponding to the i-th target sample, and N represents the number of types in the fault type set; a i Indicates that the predicted sample corresponding to the i-th target sample produces an abnormality, represents the actual production anomaly type corresponding to the i-th target sample, and M represents the number of samples of the sample data;

[0133] Among them, L G It is expressed as:

[0134]

[0135] In the formula, represents the generated data calculated by the generator for the predicted sample failure and the predicted sample production anomaly, P g represents the standard normal distribution of the generator, represents the discriminant result of the discriminator on the generated data, x represents the real fault type and the real production abnormality type, P r Representing the standard normal distribution of the real fault type and the real production abnormality type;

[0136] According to the gradient descent algorithm, the overall loss value is minimized to obtain the network model parameters when the overall loss value is the smallest, and the parameters of the anomaly-fault bidirectional detection model are updated according to the network model parameters to obtain an updated anomaly-fault bidirectional detection model;

[0137] Determining whether the overall loss value converges;

[0138] When the overall loss value has not converged, returning to the above step of extracting a target sample from the sample data in sequence, and iteratively optimizing the update anomaly-fault bidirectional detection model;

[0139] When the overall loss value converges, the training process is stopped to obtain a trained anomaly-fault bidirectional detection model.

[0140] Specifically, in the embodiment of the present invention, the sample data includes three parameters: production process data, fault type, and production abnormality type. The feature extraction network of the model can perform feature extraction, and when it reaches the output layer, the output layer will only extract the relevant feature vectors of two of the parameters for analysis. For example, the process abnormality prediction output layer will only use the relevant feature vectors of the two parameters of the production process data and the fault type to calculate the predicted sample production abnormality.

[0141] In the embodiment of the present invention, the two output layers are respectively a process anomaly prediction output layer and a fault judgment output layer. The process anomaly prediction output layer is a regression prediction network, while the fault judgment output layer is a classification judgment network. total =λL cls +(1-λ)L reg The loss function of is more common. total =λL cls +(1-λ)L reg L cls and L reg They are also their respective commonly used loss functions.

[0142] In the embodiment of the present invention, in order to improve the mapping relationship between the production anomaly type set and the fault type set, the training process is restricted by means of a generative adversarial network. The generative adversarial network is used to generate more diverse and possible synthetic data to train our multi-task model, thereby improving the generalization ability and robustness of the model.

[0143] Finally, the progress of model training is controlled by the convergence of the overall loss value. When the overall loss value converges, it indicates that the model has a large improvement effect and can still be trained; when the overall loss value converges, it indicates that the improvement effect of the model is not large. In order to avoid model robustness, the training process is stopped to obtain a trained anomaly-fault bidirectional detection model.

[0144] Furthermore, in another embodiment of the present invention, after obtaining the product test data, the method further includes:

[0145] Saving the product test data into a pre-built model feedback database;

[0146] According to the preset time interval frequency, the anomaly-fault bidirectional detection model is trained using the model feedback database to obtain an optimized anomaly-fault bidirectional detection model.

[0147] According to the above steps, the present invention can maintain continuous optimization of the abnormality-fault bidirectional detection model to avoid the degradation of connector product quality caused by changes over time, changes in machine life, changes in materials and other factors.

[0148] S5, determining whether the test fault type corresponds to the predicted fault type;

[0149] When the test fault type corresponds to the predicted fault type, S6, output the production step name corresponding to the process abnormality identification result to the pre-built list to be improved.

[0150] In an embodiment of the present invention, when the test fault type corresponds to the predicted fault type, it indicates that the defective product is produced because of the abnormal step corresponding to the abnormal identification result in the monitoring result sequence, so that the production step name corresponding to the process abnormal identification result can be directly added to the pre-built list for improvement, waiting for professional technicians to debug or repair.

[0151] When the test fault type is different from the predicted fault type, S7, performing a monitoring data tracing operation on the faulty product to obtain a monitoring result sequence of the faulty product.

[0152] In the embodiment of the present invention, when the test fault type is different from the predicted fault type, it indicates that in addition to observable problems with the production process, there are also invisible technical problems that need to be improved.

[0153] S8. Using the anomaly-fault bidirectional detection model, perform anomaly identification on the faulty product monitoring result sequence to obtain a predicted anomaly identification result.

[0154] In the embodiment of the present invention, the fault type in the fault product monitoring result sequence is first identified through the anomaly-fault bidirectional detection model, and then outputted through the process anomaly prediction output layer of the anomaly-fault bidirectional detection model to obtain the predicted anomaly identification result. For example, it is predicted that the diameter packaging value of the data line in the first stamping step in the production technology assembly line needs to be increased by 0.2mm.

[0155] S9. Identify the production step name corresponding to the predicted abnormality recognition result, and transform the predicted abnormality recognition result into a predicted improvement enhancement direction according to a preset improvement strategy.

[0156] In detail, in the embodiment of the present invention, the predicted abnormality identification result is converted into a predicted improvement enhancement direction according to a preset improvement strategy, including:

[0157] According to a preset improvement strategy, the standard product structure diagram is compared with the predicted abnormality recognition result to obtain difference steps and difference values;

[0158] The difference value is inverted to obtain the value to be improved, the difference step is converted into the step to be improved, and the predicted improvement and enhancement direction is obtained according to the value to be improved and the step to be improved.

[0159] The improvement strategy refers to a percentage improvement strategy that selects different improvement methods according to different steps. For example, the stamping step can change the hardness of the plastic material and the magnitude of injection molding, while the electroplating step can change the thickness and uniformity of the electroplating layer, which may affect the moisture resistance of the electroplating layer. The operation of each step is different, and the improvement technology and unit are also different. Therefore, a percentage improvement method is used to improve each step separately.

[0160] Specifically, in the embodiment of the present invention, if the diameter packaging value of the data line in the first stamping step needs to be increased by 0.2mm according to the predicted abnormal identification result, but the abnormal identification result only shows that 0.1mm is missing, then if the machine numerical control parameter of the stamping step shows that 1mm of plastic is attached. The present invention transforms the difference step into a step to be improved. In the actual working process, not only the 0.1mm shown in the abnormal identification result needs to be supplemented, but also the thickness needs to be increased by 0.1mm to achieve the best hardness performance.

[0161] Therefore, the final predicted improvement and enhancement direction is to first improve the equipment execution problem so that the machine will not lose the 0.1mm thickness, and then in the machine CNC parameters of the stamping step of the production technology assembly line, change the attached 1mm plastic to attached 1.1mm plastic, which can completely improve the hardness performance problem of the faulty product.

[0162] S10, saving the production step name and the predicted improvement and enhancement direction into the list to be improved.

[0163] In the embodiment of the present invention, the production step name and the predicted improvement and enhancement direction, as well as the abnormality recognition result generated in step S2 can all be saved in the list to be improved, waiting for professional technicians to debug.

[0164] The present invention solves the problem described in the background technology. The present invention first monitors the steps of the connector production technology assembly line by means of digital monitoring and image monitoring through edge detection equipment. When the connector production starts, a monitoring result sequence of each connector will be obtained. The present invention first compares the image recognition results of the numerical control parameters with the actual production results to complete the abnormality recognition and obtain the abnormality recognition results. These abnormality recognition results can effectively solve the problems of uneven product packaging injection molding, wrong or asymmetric installation positions of parts, etc. Then, the abnormality recognition results are predicted by a pre-trained abnormality-fault bidirectional detection model to obtain the predicted fault type. When the predicted fault type is the same as the test fault type actually detected, it indicates that only the step where the abnormality occurs needs to be improved. When the predicted fault type is different from the test fault type, it indicates that there are deficiencies in the technical process that need to be repaired. The abnormality-fault bidirectional detection model has two output layers, which can find out the deficiencies that may occur in the technical process according to the test fault type, and can also predict the fault type of the product according to the deficiencies in the technical process. When training the model, the present invention also adds a generative adversarial network to further increase the mapping relationship between process abnormalities and fault types. Therefore, the present invention can solve the problem that the production process of connectors such as USB or type-C is difficult to reach perfection, and when materials and equipment change, modifying the production process requires a lot of manpower and time for exploration and debugging.

[0165] like Figure 2 , which is a functional module diagram of a self-adaptive optimization system for connector production technology provided by an embodiment of the present invention.

[0166] The self-adaptive optimization system 100 for connector production technology of the present invention can be installed in an electronic device. According to the functions to be implemented, the self-adaptive optimization system 100 for connector production technology can include a process monitoring module 101, a product testing module 102, a fault prediction module 103 and a point identification module for improvement 104. The module of the present invention can also be called a unit, which refers to a series of computer program segments that can be executed by an electronic device processor and can complete fixed functions, which are stored in the memory of the electronic device.

[0167] The process monitoring module 101 is used to obtain the production technology pipeline of the target model connector, use the pre-built edge detection equipment to monitor the production technology pipeline, and obtain the connector product set and the monitoring result sequence corresponding to each connector product in the connector product set within the preset time of starting the production technology pipeline;

[0168] The product testing module 102 is used to perform product performance testing on the connector product set according to a preset test strategy to obtain product test data, extract faulty products from the connector product set according to the product test data, obtain the test fault type of the faulty product, and perform process anomaly identification on the monitoring result sequence of the target product in the connector product set to obtain anomaly identification results, and mark the anomaly identification results on the target product;

[0169] The fault prediction module 103 is used to use the pre-trained anomaly-fault bidirectional detection model to predict the fault type of the anomaly recognition result, obtain the predicted fault type, and determine whether the test fault type corresponds to the predicted fault type, and when the test fault type corresponds to the predicted fault type, output the production step name corresponding to the process anomaly recognition result to the pre-built list of things to be improved;

[0170] The module 104 for identifying points to be improved is used to perform monitoring data tracing operations on the faulty product when the test fault type is different from the predicted fault type to obtain a sequence of monitoring results of the faulty product, and to use the anomaly-fault bidirectional detection model to perform anomaly identification on the sequence of monitoring results of the faulty product to obtain a predicted anomaly identification result, and to identify the production step name corresponding to the predicted anomaly identification result, and according to a preset improvement strategy, to convert the predicted anomaly identification result into a predicted improvement enhancement direction, and to save the production step name and the predicted improvement enhancement direction in the list to be improved.

[0171] In detail, the modules in the adaptive optimization system 100 for connector production technology in the embodiment of the present invention are used in the same manner as described above. Figure 1The adaptive optimization method of connector production technology described in the invention has the same technical means and can produce the same technical effects, so it will not be repeated here.

[0172] like Figure 3 , is a schematic diagram of the structure of an electronic device for implementing a method for adaptively optimizing connector production techniques provided by an embodiment of the present invention.

[0173] The electronic device 1 may include a processor 10, a memory 11 and a bus 12, and may also include a computer program stored in the memory 11 and executable on the processor 10, such as a program for a method of adaptively optimizing connector production techniques.

[0174] Among them, the memory 11 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, mobile hard disk, multimedia card, card-type memory (for example: SD or DX memory, etc.), magnetic memory, disk, optical disk, etc. The memory 11 can be an internal storage unit of the electronic device 1 in some embodiments, such as a mobile hard disk of the electronic device 1. The memory 11 can also be an external storage device of the electronic device 1 in other embodiments, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device 1. Further, the memory 11 also includes an internal storage unit of the electronic device 1 and an external storage device. The memory 11 can not only be used to store application software and various types of data installed in the electronic device 1, such as the code of the adaptive optimization method program of the connector production technology, but also can be used to temporarily store data that has been output or is to be output.

[0175] The processor 10 may be composed of an integrated circuit in some embodiments, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and combinations of various control chips, etc. The processor 10 is the control core (Control Unit) of the electronic device, and uses various interfaces and lines to connect various components of the entire electronic device, and executes or executes programs or modules stored in the memory 11 (such as a program for an adaptive optimization method for connector production technology, etc.), and calls data stored in the memory 11 to execute various functions of the electronic device 1 and process data.

[0176] The bus 12 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus 12 may be divided into an address bus, a data bus, a control bus, etc. The bus 12 is configured to realize connection and communication between the memory 11 and at least one processor 10, etc.

[0177] Figure 3 Only an electronic device with components is shown, and those skilled in the art will understand that Figure 3 The structure shown does not constitute a limitation on the electronic device 1, and may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0178] For example, although not shown, the electronic device 1 may also include a power source (such as a battery) for supplying power to each component. Preferably, the power source may be logically connected to the at least one processor 10 through a power management device, so that the power management device can realize functions such as charging management, discharging management, and power consumption management. The power source may also include any components such as one or more DC or AC power sources, recharging devices, power failure detection circuits, power converters or inverters, power status indicators, etc. The electronic device 1 may also include a variety of sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.

[0179] Furthermore, the electronic device 1 may also include a network interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device 1 and other electronic devices.

[0180] Optionally, the electronic device 1 may further include a user interface, which may be a display, an input unit (such as a keyboard), or a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an OLED (Organic Light-Emitting Diode) touch device. The display may also be appropriately referred to as a display screen or a display unit, which is used to display information processed in the electronic device 1 and to display a visual user interface.

[0181] It should be understood that the embodiment is for illustration only and the scope of the patent application is not limited to this structure.

[0182] The program of the method for self-adaptive optimization of connector production technology stored in the memory 11 of the electronic device 1 is a combination of multiple instructions. When running in the processor 10, it can achieve:

[0183] Obtain a production technology assembly line of a target model connector, monitor the production technology assembly line using a pre-built edge detection device, and obtain a connector product set and a monitoring result sequence corresponding to each connector product in the connector product set within a preset time when the production technology assembly line is started;

[0184] According to a preset test strategy, a product performance test is performed on the connector product set to obtain product test data, and according to the product test data, a faulty product is extracted from the connector product set, and a test fault type of the faulty product is obtained;

[0185] Performing process anomaly identification on the monitoring result sequence of the target product in the connector product set to obtain an anomaly identification result, and marking the anomaly identification result on the target product;

[0186] Using a pre-trained anomaly-fault bidirectional detection model, predicting the fault type of the anomaly recognition result to obtain a predicted fault type;

[0187] Determining whether the test fault type corresponds to the predicted fault type;

[0188] When the test fault type corresponds to the predicted fault type, outputting the production step name corresponding to the process abnormality identification result to a pre-built list to be improved;

[0189] When the test fault type is different from the predicted fault type, performing a monitoring data tracing operation on the faulty product to obtain a monitoring result sequence of the faulty product;

[0190] Using the anomaly-fault bidirectional detection model, anomaly identification is performed on the faulty product monitoring result sequence to obtain a predicted anomaly identification result;

[0191] Identify the production step name corresponding to the predicted abnormality recognition result, and transform the predicted abnormality recognition result into a predicted improvement enhancement direction according to a preset improvement strategy;

[0192] The production step name and the predicted improvement enhancement direction are saved in the list to be improved.

[0193] Specifically, the specific implementation method of the processor 10 for the above instructions can refer to Figures 1 to 3 The description of the relevant steps in the corresponding embodiments will not be repeated here.

[0194] Furthermore, if the module / unit integrated in the electronic device 1 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. The computer-readable storage medium can be volatile or non-volatile. For example, the computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, and a read-only memory (ROM).

[0195] The present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor of an electronic device, the computer program can implement:

[0196] Obtain a production technology assembly line of a target model connector, monitor the production technology assembly line using a pre-built edge detection device, and obtain a connector product set and a monitoring result sequence corresponding to each connector product in the connector product set within a preset time when the production technology assembly line is started;

[0197] According to a preset test strategy, a product performance test is performed on the connector product set to obtain product test data, and according to the product test data, a faulty product is extracted from the connector product set, and a test fault type of the faulty product is obtained;

[0198] Performing process anomaly identification on the monitoring result sequence of the target product in the connector product set to obtain an anomaly identification result, and marking the anomaly identification result on the target product;

[0199] Using a pre-trained anomaly-fault bidirectional detection model, predicting the fault type of the anomaly recognition result to obtain a predicted fault type;

[0200] Determining whether the test fault type corresponds to the predicted fault type;

[0201] When the test fault type corresponds to the predicted fault type, outputting the production step name corresponding to the process abnormality identification result to a pre-built list to be improved;

[0202] When the test fault type is different from the predicted fault type, performing a monitoring data tracing operation on the faulty product to obtain a monitoring result sequence of the faulty product;

[0203] Using the anomaly-fault bidirectional detection model, anomaly identification is performed on the faulty product monitoring result sequence to obtain a predicted anomaly identification result;

[0204] Identify the production step name corresponding to the predicted abnormality recognition result, and transform the predicted abnormality recognition result into a predicted improvement enhancement direction according to a preset improvement strategy;

[0205] The production step name and the predicted improvement enhancement direction are saved in the list to be improved.

[0206] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, systems and methods can be implemented in other ways. For example, the system embodiments described above are only illustrative, and actual implementation may have other division methods.

[0207] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0208] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0209] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0210] In addition, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices stated in the system claim can also be implemented by one unit or device through software or hardware. The second and other words are used to indicate names, but not to indicate any particular order.

[0211] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. An adaptive optimization method for connector production technology, characterized in that: The method comprises: Obtain a production technology assembly line of a target model connector, monitor the production technology assembly line using a pre-built edge detection device, and obtain a connector product set and a monitoring result sequence corresponding to each connector product in the connector product set within a preset time when the production technology assembly line is started; According to a preset test strategy, a product performance test is performed on the connector product set to obtain product test data, and according to the product test data, a faulty product is extracted from the connector product set, and a test fault type of the faulty product is obtained; Performing process anomaly identification on the monitoring result sequence of the target product in the connector product set to obtain an anomaly identification result, and marking the anomaly identification result on the target product; Using a pre-trained anomaly-fault bidirectional detection model, predicting the fault type of the anomaly recognition result to obtain a predicted fault type; Determining whether the test fault type corresponds to the predicted fault type; When the test fault type corresponds to the predicted fault type, outputting the production step name corresponding to the process abnormality identification result to a pre-built list to be improved; When the test fault type is different from the predicted fault type, performing a monitoring data tracing operation on the faulty product to obtain a monitoring result sequence of the faulty product; Using the anomaly-fault bidirectional detection model, anomaly identification is performed on the faulty product monitoring result sequence to obtain a predicted anomaly identification result; Identify the production step name corresponding to the predicted abnormality recognition result, and transform the predicted abnormality recognition result into a predicted improvement enhancement direction according to a preset improvement strategy; The production step name and the predicted improvement enhancement direction are saved in the list to be improved.

2. The method for adaptive optimization of connector production technology according to claim 1, characterized in that: The monitoring result sequence includes the production step name, machine numerical control parameters, and product scanned images corresponding to each production step in the production technology assembly line, wherein the product scanned image includes a photographed image and an infrared scanned structure image.

3. The method for adaptive optimization of connector production technology according to claim 2, characterized in that: The step of performing a product performance test on the connector product set according to a preset test strategy to obtain product test data includes: According to a preset test strategy, using a pre-built simulation service, product simulation is performed on the monitoring result sequence to obtain a simulation model, and a test process based on a repetitive test type item is performed on the simulation model to obtain a simulation test result; Performing a test process based on a one-time detection type item on the connector product set to obtain a test result; The simulation test results and the experimental test results are uniformly output to obtain product test data.

4. The method for adaptive optimization of connector production technology according to claim 3, characterized in that: The performing process anomaly identification on the monitoring result sequence of the target product in the connector product set to obtain an anomaly identification result includes: Extracting a target product from the connector product set, and extracting machine numerical control parameters and product scanning images in the monitoring result sequence of the target product; Using the product scan image and the machine numerical control parameters, performing an abnormal operation comparison operation based on the operation purpose and operation result of the machine on the monitoring result sequence to obtain an abnormal operation recognition result; Acquire a standard product structure diagram of the target product, use the product scan image, and perform structural distribution anomaly recognition on the target product according to the standard product structure diagram to obtain a structural anomaly recognition result; The operation abnormality recognition result and the structure abnormality recognition result are output to obtain an abnormality recognition result.

5. The method for adaptive optimization of connector production technology according to claim 4, characterized in that: The method of using the pre-trained anomaly-fault bidirectional detection model to predict the fault type of the anomaly recognition result to obtain the predicted fault type includes: Using a pre-trained anomaly-fault bidirectional detection model, a feature extraction operation is performed on the anomaly recognition result to obtain an anomaly feature vector; Performing feature recognition on the abnormal feature vector to obtain feature recognition results, and grouping the feature recognition results into preset abnormal types to obtain feature clusters corresponding to each abnormal type; Fully connected classification judgment is performed on each of the feature clusters to obtain a fault probability distribution value corresponding to a preset fault type set, and a predicted fault type is extracted from the fault type set based on the fault probability distribution value.

6. The method for adaptively optimizing connector production technology according to claim 5, characterized in that: Before predicting the fault type of the abnormality identification result by using the pre-trained abnormality-fault bidirectional detection model to obtain the predicted fault type, the method further includes: Get the pre-built production exception type set and failure type set: F={f1,f2,…,f m } <h2 style=";text-align:left;direction:ltr">A = {a1,a2,…,a<h2 style=";text-align:left;direction:ltr"> n <h2 style=";text-align:left;direction:ltr">} In the formula, F represents the set of production anomaly types, m represents the number of types in the set of production anomaly types, and f m represents the mth type of production anomaly type in the production anomaly type set, A represents the fault type set; n represents the number of types in the production anomaly type set, a n represents the nth type of fault in the set of fault types; Obtain sample data including normal production process data and abnormal production process data, wherein each production process data in the sample data is also equipped with a corresponding fault type: D={(x i ,and fi ,and ai )} In the formula, D represents the sample data, x i represents the i-th production process data in the sample data, y fi Indicates the fault type corresponding to the i-th production process data in the sample data, y ai Indicates the production anomaly type corresponding to the i-th production process data in the sample data; Obtain an anomaly-fault bidirectional detection model including two output layers and a generative adversarial network, and configure the production anomaly type set and the fault type set as standard output types of the two output layers, respectively, wherein the two output layers are a process anomaly prediction output layer and a fault judgment output layer, and the generative adversarial network includes a generator and a discriminator; Extracting a target sample from the sample data in turn, performing network forward operation on the target sample using the anomaly-fault bidirectional detection model, and obtaining predicted sample faults and predicted sample production anomalies; Using the pre-built model overall loss function, according to the real fault type and real production abnormality type corresponding to the target sample in the sample data, the loss value of the predicted sample fault and the predicted sample production abnormality is calculated to obtain the overall loss value. The model overall loss function is expressed as: L 整体 =L total +αL G +βL D Where, L 整体 Expressed as the overall loss function of the model, L total represents the output layer loss function of the process anomaly prediction output layer and the fault judgment output layer, L G represents the loss function of the generator G, L D represents the loss function of the discriminator D, α and β represent weight parameters for balancing the loss function of the output layer and the loss function of the generative adversarial network; Among them, L total It is expressed as: THE total =λL cls +(1―λ)L reg Where, L total represents the output layer loss function, L cls represents the classification loss function of the fault judgment output layer, L reg represents the regression loss function of the process anomaly prediction output layer, λ is the weight parameter; y i represents the predicted sample fault corresponding to the i-th target sample, represents the actual fault type corresponding to the i-th target sample, and N represents the number of types in the fault type set; a i Indicates that the predicted sample corresponding to the i-th target sample produces an abnormality, represents the actual production anomaly type corresponding to the i-th target sample, and M represents the number of samples of the sample data; Among them, L G It is expressed as: In the formula, represents the generated data calculated by the generator for the predicted sample failure and the predicted sample production anomaly, P g represents the standard normal distribution of the generator, represents the discriminant result of the discriminator on the generated data, x represents the real fault type and the real production abnormality type, P r Representing the standard normal distribution of the real fault type and the real production abnormality type; According to the gradient descent algorithm, the overall loss value is minimized to obtain the network model parameters when the overall loss value is the smallest, and the network reverse parameters of the anomaly-fault bidirectional detection model are updated according to the network model parameters to obtain an updated anomaly-fault bidirectional detection model; Determining whether the overall loss value converges; When the overall loss value has not converged, returning to the above step of extracting a target sample from the sample data in sequence, and iteratively optimizing the update anomaly-fault bidirectional detection model; When the overall loss value converges, the training process is stopped to obtain a trained anomaly-fault bidirectional detection model.

7. The method for adaptively optimizing connector production technology according to claim 6, characterized in that: After obtaining the product test data, the method further includes: Saving the product test data into a pre-built model feedback database; According to the preset time interval frequency, the anomaly-fault bidirectional detection model is trained using the model feedback database to obtain an optimized anomaly-fault bidirectional detection model.

8. The method for adaptively optimizing connector production technology according to claim 7, characterized in that: After marking the abnormality identification result on the target product, the method further includes: Supervising the abnormality recognition result within a preset supervision time period to obtain an abnormality rate; Determining whether the abnormality rate is greater than a preset warning threshold; When the abnormal rate is greater than the warning threshold, an alarm prompt message is generated.

9. The method for adaptively optimizing connector production technology according to claim 8, characterized in that: The method of converting the predicted abnormality identification result into a predicted improvement enhancement direction according to a preset improvement strategy includes: According to a preset improvement strategy, the standard product structure diagram is compared with the predicted abnormality recognition result to obtain difference steps and difference values; The difference value is inverted to obtain the value to be improved, the difference step is converted into the step to be improved, and the predicted improvement and enhancement direction is obtained according to the value to be improved and the step to be improved.

10. An adaptive optimization system for connector production technology, characterized in that: The system comprises: A process monitoring module is used to obtain the production technology assembly line of the target model connector, monitor the production technology assembly line using a pre-built edge detection device, and obtain a connector product set and a monitoring result sequence corresponding to each connector product in the connector product set within a preset time of starting the production technology assembly line; A product testing module, configured to perform product performance testing on the connector product set according to a preset test strategy to obtain product test data, extract faulty products from the connector product set according to the product test data, obtain the test fault type of the faulty product, and perform process anomaly identification on the monitoring result sequence of the target product in the connector product set to obtain anomaly identification results, and mark the anomaly identification results on the target product; A fault prediction module is used to use a pre-trained anomaly-fault bidirectional detection model to predict the fault type of the anomaly recognition result, obtain the predicted fault type, and determine whether the test fault type corresponds to the predicted fault type, and when the test fault type corresponds to the predicted fault type, output the production step name corresponding to the process anomaly recognition result to a pre-built list of items to be improved; The module for identifying points to be improved is used to perform monitoring data tracing operations on the faulty product when the test fault type is different from the predicted fault type to obtain a sequence of monitoring results of the faulty product, and to use the abnormality-fault bidirectional detection model to perform abnormality identification on the sequence of monitoring results of the faulty product to obtain a predicted abnormality identification result, and to identify the production step name corresponding to the predicted abnormality identification result, and according to a preset improvement strategy, to convert the predicted abnormality identification result into a predicted improvement enhancement direction, and to save the production step name and the predicted improvement enhancement direction in the list to be improved.

Images