A fault detection model training method and device, a fault detection method, electronic equipment and a readable storage medium

Abstract

The application provides a training method and device of a fault detection model, a fault detection method, electronic equipment and a readable storage medium. Embodiments of the application acquire fault information of a display substrate and maintenance data corresponding to the display substrate, input the fault information into a fuzzy neural network to generate a maintenance scheme through the fuzzy neural network, and train the fuzzy neural network according to the maintenance data and the maintenance scheme to obtain a fault detection model. In this way, embodiments of the application can continuously acquire fault information and learn the processing mode for faults, thereby realizing identification and processing of more faults and helping to improve the identification effect of the faults existing in the display substrate.

Description

Technical Field

[0001] This invention relates to the field of deep learning technology, and in particular to a training method, apparatus, fault detection method, electronic device, and readable storage medium for a fault detection model. Background Technology

[0002] Display substrates and other products may have manufacturing defects such as localized short circuits or general short circuits during the production process. These defective substrates typically require inspection and repair using methods such as laser cutting. However, existing repair methods are not very effective at identifying defective display substrates. Summary of the Invention

[0003] This invention provides a training method, apparatus, fault detection method, electronic device, and readable storage medium for a fault detection model, in order to solve the problem of poor recognition performance for display substrates.

[0004] In a first aspect, embodiments of the present invention provide a method for training a fault detection model, comprising the following steps:

[0005] Obtain fault information of the display substrate and corresponding maintenance data;

[0006] The fault information is input into a fuzzy neural network to generate a repair plan.

[0007] The fuzzy neural network is trained based on the maintenance data and the maintenance plan to obtain a fault detection model.

[0008] In some embodiments, the fault information includes an image of the display substrate, and the step of inputting the fault information into a fuzzy neural network to generate a repair plan through the fuzzy neural network includes:

[0009] The image of the display substrate is input into the fuzzy neural network;

[0010] The repair scheme obtained from the output of the fuzzy neural network includes at least one of the following: laser type used for repair, cutting method, and film layer to be cut.

[0011] In some embodiments, the image on the display substrate includes one or more of a short-circuit fault image, an open-circuit fault image, and a foreign object fault image.

[0012] In some embodiments, training the fuzzy neural network to obtain a fault detection model based on the maintenance data and the maintenance plan includes:

[0013] The repair plan and the repair data are compared to generate difference data;

[0014] The parameters of the fuzzy neural network are adjusted based on the difference data;

[0015] The steps of generating differential data and adjusting the parameters of the fuzzy neural network are executed iteratively. When the accuracy of the differential data is less than a preset threshold or the number of iterations reaches a preset threshold, the fuzzy neural network is used as the trained fault detection model.

[0016] In some embodiments, before acquiring the fault information of the display substrate and the repair data corresponding to the display substrate, the method further includes:

[0017] When the fuzzy neural network cannot determine the fault corresponding to the target product, a judgment result for the target product is obtained;

[0018] If the determination result includes determining that the target product has a faulty display substrate, then the step of obtaining the fault information of the display substrate and the repair data corresponding to the display substrate is executed.

[0019] In some embodiments, the fuzzy neural network includes a preamble network, which includes a first input layer, a fuzzification layer, a fuzzy specification calculation layer and a first output layer arranged sequentially.

[0020] The first input layer is used to pass input information to the fuzzification layer, the fuzzification layer is used to calculate the membership degree of multiple input components included in the input information, the fuzzy specification calculation layer includes multiple nodes for representing fuzzy specifications, the fuzzy specification calculation layer is used to calculate the fitness of each fuzzy specification according to the membership degree of the input components, and the first output layer is used to implement the normalization calculation of the fitness of the fuzzy specifications.

[0021] In some embodiments, the fuzzy neural network further includes a consequent network, which includes a second input layer, a processing layer, and a second output layer arranged sequentially.

[0022] The second input layer is used to obtain the input information and input it into the processing layer. The processing layer includes multiple limiting rules and is used to calculate the results corresponding to each limiting rule. The second output layer is used to generate a maintenance plan corresponding to the input information based on the output results of the processing layer and the output results of the preceding network.

[0023] Secondly, embodiments of the present invention provide a training apparatus for a fault detection model, comprising:

[0024] The acquisition module is used to acquire fault information of the display substrate and maintenance data corresponding to the display substrate;

[0025] An input module is used to input the fault information into a fuzzy neural network to generate a repair plan through the fuzzy neural network;

[0026] The training module is used to train the fuzzy neural network to obtain a fault detection model based on the maintenance data and the maintenance plan.

[0027] Thirdly, embodiments of the present invention provide a fault detection method, comprising the following steps:

[0028] Obtain product information for the product to be repaired;

[0029] The product information is input into the fault detection model to obtain a repair plan, wherein the fault detection model is trained by the training method of the fault detection model described in any one of the first aspects.

[0030] Fourthly, embodiments of the present invention provide an electronic device, including: a transceiver, a memory, a processor, and a program stored in the memory and executable on the processor; the processor is configured to read the program in the memory to implement the steps of the training method for the fault detection model as described in any of the first aspects.

[0031] Fifthly, embodiments of the present invention provide a readable storage medium for storing a program, which, when executed by a processor, implements the steps of the training method for the fault detection model as described in any of the first aspects.

[0032] This invention acquires fault information of a display substrate and corresponding repair data; inputs the fault information into a fuzzy neural network to generate a repair plan; and trains the fuzzy neural network based on the repair data and the repair plan to obtain a fault detection model. In this way, this invention can continuously acquire fault information and learn how to handle faults, thereby enabling the identification and handling of more faults and improving the identification effect of faults existing in the display substrate. Attached Figure Description

[0033] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 This is a flowchart of a training method for a fault detection model according to an embodiment of the present invention;

[0035] Figure 2This is a structural diagram of a fuzzy neural network in one embodiment of the present invention;

[0036] Figure 3 This is a structural diagram of a training device for a fault detection model according to an embodiment of the present invention. Detailed Implementation

[0037] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] This invention provides a method for training a fault detection model.

[0039] like Figure 1 As shown, in one embodiment, the training method including the fault detection model includes the following steps:

[0040] Step 101: Obtain fault information of the display substrate and maintenance data corresponding to the display substrate.

[0041] In this embodiment, a display product is used as an example for illustration. Specifically, the display product may be a display substrate. During the manufacturing process, some film layers of the display substrate may have some material residue or excessive material removal due to process and other factors. This may cause some abnormal phenomena such as short circuits or open circuits in some structures of the display substrate. Accordingly, these display substrates need to be repaired.

[0042] In some embodiments, prior to step 101, the method further includes:

[0043] When the fuzzy neural network cannot determine the fault corresponding to the target product, a judgment result for the target product is obtained;

[0044] If the determination result includes determining that the target product has a faulty display substrate, then the step of obtaining the fault information of the display substrate and the repair data corresponding to the display substrate is executed.

[0045] It should be understood that for identifiable display substrates, repairs can be performed directly based on the generated repair plan. If an effective repair plan cannot be generated directly using the fault detection system, manual inspection and repair by operators are usually required.

[0046] The judgment result of the target product refers to the judgment result of the human. For example, the target product may be judged as a faulty display substrate, or it may be judged as a normal product.

[0047] When the target product is determined to have a faulty display substrate, the operator needs to manually repair the display substrate. During this process, the operator's specific operation is called repair data. This repair data may include, but is not limited to, the type of laser used, the cutting method, and the film layers to be cut.

[0048] Step 102: Input the fault information into the fuzzy neural network to generate a repair plan through the fuzzy neural network.

[0049] After obtaining fault information and maintenance data, the fault information is input into the fuzzy neural network, which can then output a corresponding maintenance plan.

[0050] Fuzzy neural networks refer to neural networks obtained by combining fuzzy theory and neural networks. They combine the advantages of neural networks and fuzzy theory, integrating learning, association, recognition and information processing, which helps to improve learning and processing efficiency.

[0051] In some embodiments, the fault information includes an image of the display substrate.

[0052] In this embodiment, images of display substrates with different faults can be processed uniformly, which helps to improve the convenience of processing.

[0053] In some embodiments, images of display substrates with different faults can be differentiated. Specifically, the images of the display substrates include one or more of short-circuit fault images, open-circuit fault images, and foreign object fault images. In practice, different faults existing in the display substrates can be identified and further processed separately, thereby obtaining targeted repair solutions and improving the accuracy of the generated repair solutions.

[0054] The step of inputting the fault information into a fuzzy neural network to generate a repair plan through the fuzzy neural network includes:

[0055] The image of the display substrate is input into the fuzzy neural network;

[0056] The repair scheme obtained from the output of the fuzzy neural network includes at least one of the following: laser type used for repair, cutting method, and film layer to be cut.

[0057] like Figure 2As shown, in one embodiment, the fuzzy neural network includes a premise network, which includes a first input layer (1-1st layer), a fuzzification layer (1-2nd layer), a fuzzy specification calculation layer (1-3rd layer), and a first output layer (1-4th layer) arranged sequentially.

[0058] like Figure 2 As shown, the first input layer is used to pass the input information 1, x1, ..., xn to the fuzzing layer.

[0059] The fuzzing layer is used to calculate the membership degree of multiple input components included in the input information. In one embodiment, the multiple input components 1, x1, ... xn included in the input information can be calculated by the following formula (1).

[0060]

[0061] In formula (1) above, c is the center of the function, and b is the width of the function. To calculate the obtained membership degree, i.e., the corresponding Figure 2 middle

[0062] The fuzzy specification computation layer includes multiple nodes representing fuzzy specifications, where each node represents a fuzzy specification. The fuzzy specification computation layer is used to calculate the fitness α of each fuzzy specification based on the membership degree of the input components. m .

[0063] The first output layer is used to perform normalization calculation of the fitness of the fuzzy specification, and its number of nodes is the same as the number of fuzzy rule calculation layers. This first output layer can perform normalization calculation of the fitness using the following formula (2) to obtain the normalization calculation result.

[0064]

[0065] Please continue reading. Figure 2 In some embodiments, the fuzzy neural network further includes a latter network, which includes a second input layer (2-1st layer), a processing layer (2-2nd layer), and a second output layer (2-3rd layer) arranged sequentially.

[0066] In one embodiment, the consequent network is defined using an "if-then" rule, as shown in formula (3).

[0067] R k :

[0068] The second input layer is used to obtain input information for the processing layer, where the input value of the 0th node is 1 to provide a constant term.

[0069] The processing layer includes multiple constraint rules, and is used to calculate the results corresponding to each constraint rule.

[0070] The second output layer is used to generate a maintenance plan y corresponding to the input information based on the output results of the processing layer and the output results of the preceding network.

[0071] Step 103: Train the fuzzy neural network to obtain a fault detection model based on the maintenance data and the maintenance plan.

[0072] In some embodiments, training the fuzzy neural network to obtain a fault detection model based on the maintenance data and the maintenance plan includes:

[0073] The repair plan and the repair data are compared to generate difference data;

[0074] The parameters of the fuzzy neural network are adjusted based on the difference data;

[0075] The steps of generating differential data and adjusting the parameters of the fuzzy neural network are executed iteratively. When the accuracy of the differential data is less than a preset threshold or the number of iterations reaches a preset threshold, the fuzzy neural network is used as the trained fault detection model.

[0076] The image data of the display substrate is used as the input information of the fuzzy neural network. After the input information is processed by the fuzzy neural network, the output signal Y is obtained as the repair plan, while the actual repair data is recorded as Y1. By comparing Y and Y1, the difference between the repair data and the repair plan can be obtained.

[0077] In one embodiment, the difference data e can be calculated using the following formula (4).

[0078]

[0079] in, This is the expected value of Y1, which can be calculated using the mean or estimated value of Y1.

[0080] After obtaining the difference data, the weights of each node in the fuzzy neural network are further adjusted through an optimization algorithm so that the output signal Y continuously approaches the maintenance data Y1.

[0081] In one embodiment, the weights of nodes can be optimized using the following formulas (5) to (7).

[0082]

[0083]

[0084]

[0085] In the above formulas, c is the function center in formula (1), b is the function width in formula (1), and p is the processing result of the second input layer for the input information in the consequent network. β and ζ are both preset coefficients, and the superscripts of c, b, and p represent the number of optimizations.

[0086] The above formulas (5) to (7) can be used to optimize the weights of each node in the training network.

[0087] After the difference data e meets certain accuracy requirements or a certain number of iterations, the training of the fuzzy neural network is completed. At this point, the fuzzy neural network is used as a fault detection model.

[0088] In this way, the embodiments of the present invention can continuously acquire fault information and learn how to handle faults, thereby enabling the identification and handling of more faults and helping to improve the identification effect of faults existing in the display substrate.

[0089] This invention provides a training device for a fault detection model.

[0090] like Figure 3 As shown, in one embodiment, the training device 300 for the fault detection model includes:

[0091] The acquisition module 301 is used to acquire fault information of the display substrate and maintenance data corresponding to the display substrate;

[0092] The input module 302 is used to input the fault information into the fuzzy neural network so as to generate a maintenance plan through the fuzzy neural network;

[0093] The training module 303 is used to train the fuzzy neural network to obtain a fault detection model based on the maintenance data and the maintenance plan.

[0094] In some embodiments, the fault information includes an image of the display substrate, and the input module 302 includes:

[0095] An input submodule is used to input the image of the display substrate into the fuzzy neural network;

[0096] The output submodule is used to obtain the repair scheme output by the fuzzy neural network. The repair scheme includes at least one of the following: the type of laser used for repair, the cutting method, and the film layer to be cut.

[0097] In some embodiments, the image on the display substrate includes one or more of a short-circuit fault image, an open-circuit fault image, and a foreign object fault image.

[0098] In some embodiments, the training module 303 includes:

[0099] The difference data generation submodule is used to compare the repair plan and the repair data to generate difference data;

[0100] An adjustment submodule is used to adjust the parameters of the fuzzy neural network based on the difference data;

[0101] The training submodule is used to iteratively execute the steps of generating differential data and adjusting the parameters of the fuzzy neural network. When the accuracy of the differential data is less than a preset threshold or the number of iterations reaches a preset threshold, the fuzzy neural network is used as the trained fault detection model.

[0102] In some embodiments, it also includes:

[0103] The judgment result acquisition module is used to acquire a judgment result for the target product when the fuzzy neural network cannot determine the fault corresponding to the target product.

[0104] The acquisition module 301 is specifically used to perform the step of acquiring the fault information of the display substrate and the maintenance data corresponding to the display substrate if the determination result includes the result that the target product is determined to have a faulty display substrate.

[0105] In some embodiments, the fuzzy neural network includes a preamble network, which includes a first input layer, a fuzzification layer, a fuzzy specification calculation layer and a first output layer arranged sequentially.

[0106] The first input layer is used to pass input information to the fuzzification layer, the fuzzification layer is used to calculate the membership degree of multiple input components included in the input information, the fuzzy specification calculation layer includes multiple nodes for representing fuzzy specifications, the fuzzy specification calculation layer is used to calculate the fitness of each fuzzy specification according to the membership degree of the input components, and the first output layer is used to implement the normalization calculation of the fitness of the fuzzy specifications.

[0107] In some embodiments, the fuzzy neural network further includes a consequent network, which includes a second input layer, a processing layer, and a second output layer arranged sequentially.

[0108] The second input layer is used to obtain the input information and input it into the processing layer. The processing layer includes multiple limiting rules and is used to calculate the results corresponding to each limiting rule. The second output layer is used to generate a maintenance plan corresponding to the input information based on the output results of the processing layer and the output results of the preceding network.

[0109] This invention provides a fault detection method, comprising the following steps:

[0110] Obtain product information for the product to be repaired;

[0111] The product information is input into the fault detection model to obtain a repair plan, wherein the fault detection model is trained using the aforementioned fault detection model training method.

[0112] Since the fault detection model used is trained using the fault detection model training method in the above-mentioned fault detection model training method embodiment, it can achieve at least all of the above-mentioned technical effects, which will not be elaborated here.

[0113] This invention provides an electronic device, including: a transceiver, a memory, a processor, and a program stored in the memory and executable on the processor; the processor is configured to read the program in the memory to implement the steps of the training method for a fault detection model as described in any of the first aspects.

[0114] This invention provides a readable storage medium for storing a program that, when executed by a processor, implements the steps of the training method for the fault detection model described above.

[0115] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A training method for a fault detection model, characterized in that, Includes the following steps: Obtain fault information of the display substrate and corresponding maintenance data; The fault information is input into a fuzzy neural network to generate a repair plan. The fault information includes images of display substrates with different faults. Inputting the fault information into the fuzzy neural network includes inputting the images of the display substrates with different faults into the fuzzy neural network. The fuzzy neural network is trained based on the maintenance data and the maintenance plan to obtain a fault detection model; The step of training the fuzzy neural network to obtain a fault detection model based on the maintenance data and the maintenance plan includes: The repair plan and the repair data are compared to generate difference data; wherein, the difference data is obtained by the following formula: ； in, To obtain the expected value of the actual maintenance data Y1, after calculating the difference data, the weights of each node in the fuzzy neural network are adjusted using an optimization algorithm; the weights of each node are optimized using the following formula: ； ； ； c represents the function center in the membership calculation formula of the fuzzification layer in the fuzzy neural network, which is used to calculate the membership of multiple input components included in the input information; b represents the function width in the membership calculation formula of the fuzzification layer in the fuzzy neural network; and p represents the processing result of the second input layer for the input information in the consequent network. and All are preset coefficients, and the superscripts of c, b, and p represent the number of optimizations; The parameters of the fuzzy neural network are adjusted based on the difference data; The steps of generating differential data and adjusting the parameters of the fuzzy neural network are executed iteratively. When the accuracy of the differential data is less than a preset threshold or the number of iterations reaches a preset threshold, the fuzzy neural network is used as the trained fault detection model.

2. The method according to claim 1, characterized in that, The process of generating a maintenance plan using the fuzzy neural network includes: The repair scheme obtained from the output of the fuzzy neural network includes at least one of the following: laser type used for repair, cutting method, and film layer to be cut.

3. The method according to claim 2, characterized in that, The images on the display substrate include one or more of the following: short circuit fault images, open circuit fault images, and foreign object fault images.

4. The method according to claim 1, characterized in that, Before acquiring the fault information of the display substrate and the corresponding repair data, the method further includes: When the fuzzy neural network cannot determine the fault corresponding to the target product, a judgment result for the target product is obtained; If the determination result includes determining that the target product has a faulty display substrate, then the step of obtaining the fault information of the display substrate and the repair data corresponding to the display substrate is executed.

5. The method according to any one of claims 1 to 4, characterized in that, The fuzzy neural network includes a preamble network, which includes a first input layer, a fuzzification layer, a fuzzy specification calculation layer, and a first output layer arranged sequentially. The first input layer is used to pass input information to the fuzzification layer, the fuzzification layer is used to calculate the membership degree of multiple input components included in the input information, the fuzzy specification calculation layer includes multiple nodes for representing fuzzy specifications, the fuzzy specification calculation layer is used to calculate the fitness of each fuzzy specification according to the membership degree of the input components, and the first output layer is used to implement the normalization calculation of the fitness of the fuzzy specifications.

6. The method according to claim 5, characterized in that, The fuzzy neural network further includes a consequent network, which includes a second input layer, a processing layer, and a second output layer arranged sequentially. The second input layer is used to obtain the input information and input it into the processing layer. The processing layer includes multiple limiting rules and is used to calculate the results corresponding to each limiting rule. The second output layer is used to generate a maintenance plan corresponding to the input information based on the output results of the processing layer and the output results of the preceding network.

7. A training apparatus for a fault detection model, used to execute the training method for the fault detection model as described in any one of claims 1-6, characterized in that, include: The acquisition module is used to acquire fault information of the display substrate and maintenance data corresponding to the display substrate; An input module is used to input the fault information into a fuzzy neural network to generate a repair plan through the fuzzy neural network. The fault information includes images of display substrates with different faults. The input module is specifically used to input the images of display substrates with different faults into the fuzzy neural network. The training module is used to train the fuzzy neural network to obtain a fault detection model based on the maintenance data and the maintenance plan.

8. A fault detection method, characterized in that, Includes the following steps: Obtain product information for the product to be repaired; The product information is input into the fault detection model to obtain a repair plan, wherein the fault detection model is trained by the fault detection model training method according to any one of claims 1 to 6.

9. An electronic device, comprising: A transceiver, a memory, a processor, and a program stored in the memory and executable on the processor; characterized in that the processor is configured to read the program in the memory to implement the steps of the training method for the fault detection model as described in any one of claims 1 to 6.

10. A readable storage medium for storing a program, characterized in that, When the program is executed by the processor, it implements the steps of the training method for the fault detection model as described in any one of claims 1 to 6.

Images