Production line fault location method, device and readable storage medium

Abstract

The application discloses a production line fault positioning method and device and a readable storage medium, and relates to the technical field of production line monitoring. The production line fault positioning method comprises the following steps: acquiring a trained tree model, wherein the trained tree model is obtained by training sample production data of a target product to be positioned; determining first risk probabilities of each production station of the target product according to split node information of the trained tree model; acquiring prior risk probabilities of each production station in each production station, and correcting the prior risk probabilities to obtain posterior risk probabilities of each production station; determining second risk probabilities of each production station according to the first risk probabilities and the posterior risk probabilities; and generating corresponding risk station information and risk station information according to the first risk probabilities and the second risk probabilities. The application improves the applicability of the fault positioning technology for the production line.

Description

Technical Field

[0001] This invention relates to the field of production line monitoring technology, and in particular to a method, equipment and readable storage medium for locating production line faults. Background Technology

[0002] With industrial development, the application of production lines has become increasingly widespread. When products are manufactured on an assembly line, multiple production stations are responsible for the manufacturing and assembly of each component of the product. Production lines generally also include a testing station centrally located at the end of the production line to inspect the products. The testing items at the testing station cover the production process of multiple production stations.

[0003] Therefore, when a defective product is detected at the inspection station, the cause of the defect may come from one of the multiple production stations, or even from one of the multiple production positions at that station, making it difficult to locate the fault on the production line.

[0004] Existing fault location technologies typically require accumulating a certain amount of fault data and judgment results, manually establishing rules, or generating rules using algorithms. However, when faced with new fault causes, existing rules become inapplicable, requiring lengthy adjustments or relearning, meaning that existing fault location technologies have low applicability. Summary of the Invention

[0005] The main objective of this invention is to provide a method, apparatus, equipment, and computer-readable storage medium for locating faults in a production line, aiming to solve the technical problem of low applicability of existing fault location technologies.

[0006] To achieve the above objectives, in a first aspect, the present invention provides a production line fault location method, the production line fault location method comprising:

[0007] Obtain a trained tree model, which is trained using sample production data of the target product for which the fault is to be located;

[0008] Based on the split node information of the trained tree model, determine the first risk probability of each production station of the target product.

[0009] Obtain the prior risk probability of each production station within each production station, and correct the prior risk probability to obtain the posterior risk probability of each production station.

[0010] The second risk probability of each production station is determined based on the first risk probability and the posterior risk probability.

[0011] Based on the first risk probability and the second risk probability, generate corresponding risk station information and risk workstation information.

[0012] Based on the above technical solution, a trained tree model is obtained, which is trained from sample production data of the target product whose fault needs to be located. The first risk probability of each production station of the target product is determined based on the split node information of the trained tree model. Since only the data feature with the largest gain (i.e., the production station in the sample production data) can be used as a split node during the tree model training process, the first risk probability of each production station of the target product, i.e., the probability that the defective product is caused by the production station, can be determined based on the split node information. Then, the prior risk probability of each production workstation within each production station can be obtained and corrected to obtain the posterior risk probability of each production workstation. Based on the first risk probability and the posterior risk probability, the second risk probability of each production workstation is determined. This allows the posterior risk probability to be estimated from the prior risk probability of each production workstation within each production station, and the second risk probability of each production workstation, i.e., the probability that the defective product is caused by the production workstation, can be determined by combining the first risk probability. Based on the first and second risk probabilities, corresponding risk station information and risk workstation information are generated. This enables the location of risky workstations and risky work positions without requiring additional process parameter information or the accumulation of fault data, effectively improving the applicability of fault location technology.

[0013] According to the first aspect, prior to the step of obtaining the trained tree model, the method includes:

[0014] Acquire sample production data and an untrained tree model, where the sample production data includes each production station the target product passes through, each production station within each production station the target product passes through, and the test results of the target product.

[0015] Using each production station as data feature, each production workstation within each production station as feature value, and the detection result of the target product as data label, the untrained tree model is trained to obtain a trained tree model.

[0016] Based on the above technical solution, sample production data and an untrained tree model are acquired. The sample production data includes the production stations the target product passed through, the production workstations within those stations, and the test results of the target product. Then, using each production station as a data feature, each production workstation as a feature value, and the target product's test result as a data label, the untrained tree model is trained to obtain a trained tree model. Thus, the trained tree model can accurately determine the test result for any target product in the sample data. Based on the split node information of this trained tree model, the importance of each production station to the test result is determined.

[0017] According to the first aspect, or any implementation of the first aspect above, the steps for obtaining sample production data include:

[0018] In response to the non-conforming product query command, extract the non-conforming product information corresponding to the non-conforming product from the original production data;

[0019] Based on the information on non-conforming products, determine the production time range of the non-conforming products, extract the production information of qualified products within the production time range from the original production data, and use the non-conforming products and qualified products as target products, and the information on non-conforming products and the production information of qualified products as target production information.

[0020] The target production information is encoded according to the preset encoding rules to obtain the corresponding sample production data.

[0021] Based on the above technical solution, in response to a non-conforming product query command, the non-conforming product information corresponding to the non-conforming product is extracted from the original production data. Then, based on the non-conforming product information, the production time range of the non-conforming product is determined, and the production information of the conforming products within that production time range is extracted from the original production data. Both the non-conforming and conforming products are used as target products, and the non-conforming product information and conforming product production information are used as target production information. Therefore, this embodiment uses all products within the production time range that generated the non-conforming product as target products and determines the target production information for each target product. This ensures the integrity and consistency of the sample production data used for training the model.

[0022] According to the first aspect, or any implementation of the first aspect above, before the step of encoding the target production information according to a preset encoding rule to obtain the corresponding sample production data, the method further includes:

[0023] Obtain sub-material production information for the target product;

[0024] The target production information and sub-material production information are cascaded to synthesize the corresponding product-related production process information, and the product-related production process information is used as the new product production information.

[0025] Based on the above technical solution, by acquiring the production information of sub-materials of the target product, and then cascading the target production information and sub-material production information to synthesize the corresponding product-related production process information, this product-related production process information is used as new product production information. This allows for fault location even for the production processes of sub-materials that are not on the same production line as the target product, effectively improving the comprehensiveness and accuracy of the fault location method.

[0026] According to the first aspect, or any implementation of the first aspect above, before the step of encoding the target production information according to a preset encoding rule to obtain the corresponding sample production data, the method further includes:

[0027] Based on the testing items for the target product, determine the relevant workstations and work positions corresponding to the testing items;

[0028] Based on the relevant workstations and work positions, the target production information is filtered, and each production workstation corresponding to the relevant workstation and each production work position corresponding to the relevant work position in the target production information is retained to obtain new target production information.

[0029] Based on the above technical solution, the relevant workstations and work positions corresponding to the testing items of the target product are determined. Then, based on the relevant workstations and work positions, the target production information is filtered, retaining the production workstations corresponding to the relevant workstations and the production work positions corresponding to the relevant work positions, thus obtaining new target production information.

[0030] According to the first aspect, or any implementation of the first aspect above, after the step of encoding the target production information according to a preset encoding rule to obtain the corresponding sample production data, the following steps are included:

[0031] Obtain the probability of the target product passing through each production station for the first time;

[0032] Remove the production stations whose first pass probability is lower than a preset threshold from the sample production data to obtain new sample production data.

[0033] Based on the above technical solution, by obtaining the first pass probability of the target product through each production station, production stations with a first pass probability lower than a preset threshold are removed from the sample production data to obtain new sample production data. This further simplifies the sample production data, improves model training efficiency, reduces interference from redundant information, and can also greatly improve the accuracy of the trained tree model, thereby improving the accuracy of locating risky stations and risky work positions based on the trained tree model.

[0034] According to the first aspect, or any implementation of the first aspect above, after the step of encoding the target production information according to a preset encoding rule to obtain the corresponding sample production data, the following steps are included:

[0035] Obtain the second probability of the target product passing through each workstation in each production station;

[0036] From the sample production data, retain any production station that the second probability characterization indicates must have been passed through, and obtain new sample production data.

[0037] Based on the above technical solution, by obtaining the second probability of the target product passing through each workstation within each production station, and retaining the production station that is guaranteed to be passed through by the second probability from the sample production data, new sample production data is obtained. This further simplifies the sample production data, improves the model training efficiency, reduces interference from redundant information, and can also greatly improve the accuracy of the trained tree model, thereby improving the accuracy of locating risk stations and risk workstations based on the trained tree model.

[0038] According to the first aspect, or any implementation of the first aspect above, after the step of encoding the target production information according to a preset encoding rule to obtain the corresponding sample production data, the following steps are included:

[0039] Obtain the correlation coefficients between each pair of production stations;

[0040] Remove any one of the two production stations whose correlation coefficients indicate a corresponding relationship from the sample production data to obtain new sample production data.

[0041] Based on the above technical solution, by obtaining the correlation coefficients between each pair of production stations, any one of the two production stations whose correlation coefficients indicate a corresponding relationship is removed from the sample production data, resulting in new sample production data. This further simplifies the sample production data, improves model training efficiency, reduces interference from redundant information, and can also greatly improve the accuracy of the trained tree model, thereby improving the accuracy of locating risky stations and risky work positions based on the trained tree model.

[0042] Based on the first aspect, or any implementation of the first aspect above, the steps for determining the first risk probability of each production station of the target product based on the split node information of the trained tree model include:

[0043] Based on the split node information, determine the number of split participations for each production station;

[0044] The first risk probability of each production station is determined by summing the number of split participations.

[0045] Based on the above technical solution, the number of splitting participations of each production station is determined according to the splitting node information; the first risk probability of each production station is determined by summing the number of splitting participations. This embodiment can calculate the first risk probability of each production station using only the splitting node information, thus determining the probability that non-conforming products are caused by each production station. This eliminates the need for additional process parameter information and the accumulation of fault data, improving the convenience of determining the first risk probability of each production station and effectively enhancing the applicability of fault location technology.

[0046] The steps of obtaining the prior risk probability of each production station within each production workstation, and correcting the prior risk probability to obtain the posterior risk probability of each production station, according to the first aspect or any implementation of the first aspect above, include:

[0047] Obtain the prior risk probability and additional risk probability of each production station within each production station. The additional risk probability includes the overall non-conformance rate of the target product and the probability of the product passing through each production station.

[0048] The prior risk probability is corrected based on the overall non-conformance rate and the product pass probability to obtain the posterior risk probability of each production station.

[0049] Based on the above technical solution, the prior risk probability and additional risk probability of each production station within each production workstation are obtained. The additional risk probability includes the overall defect rate of the target product and the product pass probability at each production station. Then, the prior risk probability is corrected based on the overall defect rate and product pass probability to obtain the posterior risk probability of each production station. This makes the posterior risk probability closer to the true risk probability of the production station.

[0050] According to the first aspect, or any implementation of the first aspect above, the steps for determining the second risk probability of each production station based on the first risk probability and the posterior risk probability include:

[0051] Determine the station risk probability of each production station corresponding to each production workstation in the first risk probability;

[0052] The second risk probability of each production station is obtained by comparing the posterior risk probability of each production station with the corresponding station risk probability.

[0053] Based on the above technical solution, the station risk probability of each production workstation in the first risk probability is determined, and the second risk probability of each production workstation is obtained based on the posterior risk probability of each production workstation and the corresponding station risk probability. Thus, the posterior risk probability is further corrected using the first risk probability, thereby improving the accuracy of the second risk probability of each production workstation.

[0054] Based on the first aspect, or any implementation of the first aspect above, the steps for generating corresponding risk workstation information and risk work position information based on the first risk probability and the second risk probability include:

[0055] Based on the first risk probability, each production station is sorted to generate a corresponding risk station order, and the risk station order is used as risk station information.

[0056] Based on the second risk probability, each production station is sorted to generate a corresponding risk station order, and the risk station order is used as risk station information.

[0057] Based on the above technical solution, each production station and each production workstation is sorted according to the first risk probability and the second risk probability, respectively, to generate the corresponding risk station order and risk workstation order. This makes it easier for users to troubleshoot through the risk station order and risk workstation order, thereby improving the efficiency of production line fault location.

[0058] In a second aspect, the present invention provides a production line fault location device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the production line fault location method described above.

[0059] The second aspect and any implementation thereof correspond to the first aspect and any implementation thereof, respectively. The technical effects of the second aspect and any implementation thereof are similar to those of the first aspect and any implementation thereof, and will not be repeated here.

[0060] Thirdly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform a production line fault location method as described in the first aspect or any of the possible implementations of the first aspect.

[0061] The third aspect and any implementation thereof correspond to the first aspect and any implementation thereof, respectively. The technical effects of the third aspect and any implementation thereof are similar to those of the first aspect and any implementation thereof, and will not be repeated here.

[0062] Fourthly, embodiments of the present invention provide a computer program including instructions for executing the production line fault location method in the first aspect and any possible implementation thereof.

[0063] The fourth aspect and any implementation thereof correspond to the first aspect and any implementation thereof, respectively. The technical effects of the fourth aspect and any implementation thereof are similar to those of the first aspect and any implementation thereof, and will not be repeated here. Attached Figure Description

[0064] Figure 1This is a schematic diagram of a production line scenario according to an embodiment of the present invention;

[0065] Figure 2 This is a schematic diagram of the structure of a production line fault location device in the hardware operating environment involved in the embodiments of the present invention;

[0066] Figure 3 This is a flowchart illustrating the first embodiment of the production line fault location method of the present invention;

[0067] Figure 4 This is a flowchart illustrating the second embodiment of the production line fault location method of the present invention;

[0068] Figure 5 This is a flowchart illustrating the third embodiment of the production line fault location method of the present invention.

[0069] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0070] 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. 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.

[0071] In this article, the term "and / or" is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0072] In the specification and claims of this invention, the terms "first" and "second," etc., are used to distinguish different objects, not to describe a specific order of objects. For example, "first target object" and "second target object," etc., are used to distinguish different target objects, not to describe a specific order of target objects.

[0073] In embodiments of the present invention, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" or "for example" in embodiments of the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0074] In the description of the embodiments of the present invention, unless otherwise stated, "multiple" means two or more. For example, multiple processing units means two or more processing units; multiple systems means two or more systems.

[0075] To ensure clarity and conciseness in the description of the following embodiments, a brief introduction to the implementation scheme of a production line fault location method is first given:

[0076] With industrial development, production lines have become increasingly common. When products are manufactured on an assembly line, multiple production stations are responsible for manufacturing and assembling the various components of the product. (Refer to...) Figure 1 , Figure 1 This is a schematic diagram of a production line according to an embodiment of the present invention. Production station 1, production station 2, and so on up to production station n are arranged sequentially according to the product's manufacturing process. Within each production station, production workstations are arranged in parallel, responsible for simultaneous production and manufacturing. Figure 1 Production station 1 includes production workstations 11, 12, 13, up to 1a; production station 2 includes production workstations 21, 22, 23, up to 2b… Production station n includes production workstations n1, n2, n3, up to nz. Production stations are distinguished by numbers, and workstations within the same production station are also distinguished by numbers, thus each production station and workstation on the production line has a unique identifier. The distribution of products among production stations follows an independent probability distribution, and the product distribution between production stations does not affect each other.

[0077] Production lines typically also include testing stations centrally located at the end of the line to inspect products for functionality, performance, and appearance. The testing stations cover the production processes of multiple production stations.

[0078] Therefore, when a defective product is detected at the inspection station, the cause of the defect may come from one of the multiple production stations, or even from one of the multiple production positions at that station, making it difficult to locate the fault on the production line.

[0079] Existing fault location techniques typically identify process-related causes of product nonconformity, such as changes in process parameters or equipment status. However, these techniques require additional information about the relationship between the process and workstations / stations to pinpoint the fault, rather than directly locating it at a specific workstation or station. Furthermore, existing methods are usually applied to a single workstation or piece of equipment. When the inspection project involves a process encompassing multiple workstations, the aforementioned relationships are often difficult to establish accurately, leading to inaccurate fault location.

[0080] Furthermore, existing fault location technologies typically require accumulating a certain amount of fault data and judgment results, manually establishing rules, or generating rules using algorithms. However, when faced with new causes of faults, the existing rules no longer apply, requiring lengthy adjustments or relearning, meaning that existing fault location technologies have low applicability.

[0081] This invention presents a production line fault location method that quantifies the feature importance of a tree model trained on sample production data of the target product with the fault to be located. This allows for the determination of the importance of each production station and workstation in causing the defective product, thus identifying the risk stations and workstations. Therefore, this invention does not require additional process parameter information or the accumulation of fault data. By quantifying the feature importance of a tree model trained on sample production data of the target product with the fault to be located, the location of risk stations and workstations can be achieved, effectively improving the applicability of fault location technology.

[0082] In some embodiments, a trained tree model is obtained, which is trained from sample production data of the target product whose fault is to be located. Based on the split node information of the trained tree model, the first risk probability of each production station of the target product is determined. Since only the data feature with the largest gain (i.e., the production station in the sample production data) can be used as a split node during the training process of the tree model, the first risk probability of each production station of the target product, i.e., the probability that the defective product is caused by the production station, can be determined based on the split node information. Then, the prior risk probability of each production workstation within each production station can be obtained, and the prior risk probability can be corrected to obtain the posterior risk probability of each production workstation. Based on the first risk probability and the posterior risk probability, the second risk probability of each production workstation is determined, so that the posterior risk probability is estimated from the prior risk probability of each production workstation within each production station, and the second risk probability of each production workstation is determined by combining it with the first risk probability, i.e., the probability that the defective product is caused by the production workstation. Based on the first risk probability and the second risk probability, corresponding risk station information and risk workstation information are generated. This enables the location of risky workstations and risky work positions without requiring additional process parameter information or the accumulation of fault data, effectively improving the applicability of fault location technology.

[0083] Reference Figure 2 , Figure 2 This is a schematic diagram of the production line fault location equipment structure in the hardware operating environment involved in the embodiments of the present invention.

[0084] Specifically, production line fault location equipment can be MR equipment, PC (Personal Computer), tablet computer, portable computer or server, etc.

[0085] like Figure 2 As shown, the production line fault location equipment may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen and an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk drive. The memory 1005 may also optionally be a storage device independent of the aforementioned processor 1001.

[0086] Those skilled in the art will understand that Figure 2 The structure shown does not constitute a limitation on the fault location equipment for the production line, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0087] like Figure 2 As shown, the memory 1005, which serves as a storage medium, may include an operating system, a data storage module, a network communication module, a user interface module, and computer programs.

[0088] exist Figure 2 In the production line fault location device shown, the network interface 1004 is mainly used for data communication with other devices; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and memory 1005 in the production line fault location device of the present invention can be set in the production line fault location device. The production line fault location device calls the computer program stored in the memory 1005 through the processor 1001 and executes the production line fault location method provided in the embodiment of the present invention.

[0089] It should be understood that the above description is merely an example provided to better understand the technical solution of this embodiment, and is not intended to be the only limitation of this embodiment.

[0090] The following is combined Figure 3 The flowchart shown is a first embodiment of the production line fault location method, which provides a detailed explanation of the production line fault location method.

[0091] See Figure 3 An embodiment of the production line fault location method of the present invention provides a production line fault location method, which includes:

[0092] Step S100: Obtain the trained tree model, wherein the trained tree model is trained from the sample production data of the target product whose fault is to be located;

[0093] In this embodiment, it should be noted that the tree model can be a tree model constructed using a tree model algorithm as the base learner, employing either a serialization method or a parallelization method. Serialization methods include XGBoost and lightGBM algorithms, while parallelization methods include Random Forest and GBDT (Gradient Boosting Decision Tree) algorithms. In this embodiment, the XGBoost algorithm from the serialization method or the Random Forest algorithm from the parallelization method is preferred.

[0094] Furthermore, it should be noted that the trained tree model is trained using sample production data of the target product whose fault is to be located. The target product can include both qualified and unqualified products, and the sample production data for the target product includes at least the production stations the target product passed through, the production workstations within those stations, and the test results of the target product.

[0095] Therefore, after it is necessary to locate the faults in the production stations and workstations of the target product, a trained tree model can be obtained. The trained tree model can be trained in real time from the sample production data of the target product whose faults are to be located, or it can be trained in advance from the sample production data of the target product whose faults are to be located.

[0096] Step S200: Determine the first risk probability of each production station of the target product based on the split node information of the trained tree model.

[0097] In this embodiment, it should be noted that when selecting split nodes during the training process, the untrained tree model calculates the gain of the split node based on the feature value (production station) in the data features (i.e., production stations), and selects the production station with the largest gain among all production stations as the split node to generate child nodes. Therefore, when defective products are concentratedly generated from a certain production station of a certain production station, that production station has the maximum gain and is selected as the split node to generate child nodes.

[0098] In addition, it should be noted that the split node information includes at least the number of times each production station of the target product participates in the split (i.e., the number of times each production station acts as a split node in the trained tree model), and may also include the gain when the production station acts as a split node.

[0099] Therefore, based on the split node information of the trained tree model, the first risk probability of each production station for the target product can be determined. The first risk probability characterizes the probability that a production station is responsible for defective products; a higher first risk probability indicates a concentrated distribution of defective products within the production station. For example, the ratio of the number of split participations of a production station to the sum of the number of split participations of all production stations can be used as the first risk probability of that production station.

[0100] Step S200, which determines the first risk probability of each production station for the target product based on the split node information of the trained tree model, includes:

[0101] Step S210: Determine the number of times each production station participates in splitting based on the splitting node information;

[0102] Step S220: Determine the first risk probability of each production station based on the sum of the number of split participations and the number of split participations.

[0103] In this embodiment, feature importance analysis is used to calculate the first risk probability of each production station. As mentioned before, the production station is the data feature, and the quantification of feature importance can be achieved by counting the number of times the production station acts as a splitting node in the trained tree model to partition attributes. Therefore, the number of splitting participations of each production station can be determined based on the splitting node information. Since only the data feature with the largest gain (i.e., the production station) can serve as a splitting node, the more splitting parameters a production station has, the higher its first risk probability. The calculation method for the first risk probability of a production station is as follows:

[0104]

[0105] Among them, P i For the first risk probability of the production station, w i w represents the number of split participations of production station i. x Let N represent the number of split participations for each production station, where N is the total number of production stations. The first risk probability for a production station is obtained by dividing the number of split participations for each production station by the sum of the number of split participations for all production stations.

[0106] This embodiment determines the number of splitting participations for each production station based on splitting node information; and determines the first risk probability for each production station by summing the number of splitting participations. This embodiment can calculate the first risk probability for each production station using only splitting node information, thus determining the probability that non-conforming products are caused by each production station. This eliminates the need for additional process parameter information and the accumulation of fault data, improving the convenience of determining the first risk probability for each production station and effectively enhancing the applicability of fault location technology.

[0107] Step S300: Obtain the prior risk probability of each production station within each production station, and correct the prior risk probability to obtain the posterior risk probability of each production station.

[0108] In this embodiment, it should be noted that the prior risk probability can be the defect rate of the products that have passed through the production station in the target product. For example, if there are 100 target products in total, and 60 products have passed through the production station, including 6 defective products, then the prior risk probability of that production station is 6 / 60 = 10%.

[0109] This embodiment can obtain the prior risk probability of each production station by statistically analyzing the non-conformity rate of the products passing through each production station. Then, the prior risk probability can be corrected by additional information (such as the overall non-conformity rate of the target product, the product passing probability of each production station, etc.) to obtain the posterior risk probability of each production station, making the posterior risk probability closer to the true risk probability of the production station.

[0110] The step S300, which involves obtaining the prior risk probability of each production station within each production workstation and correcting the prior risk probability to obtain the posterior risk probability of each production station, includes:

[0111] Step S310: Obtain the prior risk probability and additional risk probability of each production station in each production station, wherein the additional risk probability includes the overall non-conforming rate of the target product and the product passing probability of each production station.

[0112] Step S320: Based on the overall non-conformance rate and the product pass probability, the prior risk probability is corrected to obtain the posterior risk probability of each production station.

[0113] In this embodiment, it should be noted that the comparative analysis of production stations within the production workstation can be obtained through Bayesian estimation to determine the probability of subsequent risk of non-conforming products caused by the production station.

[0114] Therefore, this embodiment can obtain the prior risk probability and additional risk probability of each production station within each production workstation. The additional risk probability includes the overall defect rate of the target product and the product pass probability at each production station. Then, using Bayesian estimation, the prior risk probability is corrected based on the overall defect rate and product pass probability to obtain the posterior risk probability of each production station. The calculation method for the posterior risk probability is as follows:

[0115]

[0116] In the formula, P(j|f) represents the probability that a defective product f in the target product is generated by production station j, i.e., the posterior risk probability; P(j) represents the probability that a product from production station j in production station i passes through; P(f) represents the probability that a defective product accounts for a total of all target products, i.e., the overall defect rate of the target product; and P(f|j) represents the prior risk probability of production station j.

[0117] This embodiment obtains the prior risk probability and additional risk probability of each production station within each production workstation. The additional risk probability includes the overall non-conformity rate of the target product and the product pass probability at each production station. The prior risk probabilities are then corrected based on the overall non-conformity rate and product pass probability to obtain the posterior risk probability of each production station. This makes the posterior risk probability closer to the true risk probability of the production station.

[0118] Step S400: Determine the second risk probability for each production station based on the first risk probability and the posterior risk probability;

[0119] Understandably, the second risk probability can characterize the probability that a defective product is caused by a production station. However, the second risk probability is not only the posterior risk probability of the production station itself, but is also influenced by the first risk probability of the production workstation to which that station belongs. Therefore, the posterior risk probability can be further corrected based on the first risk probability to obtain the second risk probability of each production station, thereby improving the accuracy of the second risk probability for each production station.

[0120] Step S400, which determines the second risk probability of each production station based on the first risk probability and the posterior risk probability, includes:

[0121] Step S410: Determine the station risk probability of each production station corresponding to each production workstation in the first risk probability.

[0122] Step S420: Obtain the second risk probability of each production station based on the posterior risk probability of each production station and the corresponding station risk probability.

[0123] In this embodiment, it should be noted that the production station includes at least one production workstation, which is included within the production station. Therefore, the second risk probability is not only the posterior risk probability of the production workstation itself, but is also affected by the first risk probability of the production station to which the production workstation belongs.

[0124] It is understandable that the first risk probability includes the risk probability of each production station, that is, the probability that the production station will cause defective products.

[0125] This embodiment first determines the station risk probability of each production workstation corresponding to each production workstation in the first risk probability. Based on the posterior risk probability of each production workstation and the corresponding station risk probability, the second risk probability of each production workstation is obtained by correcting the posterior risk probability of each production workstation using the station risk probability corresponding to each production workstation. The calculation method for the second risk probability of a production workstation is as follows:

[0126] P i，j =P i ·P(j|f)

[0127] In the formula, P i,j Let P(j|f) represent the second risk probability of production station j within production station i, and let P(j|f) represent the posterior risk probability of production station j. i This represents the risk probability of production station i.

[0128] This embodiment determines the station risk probability of each production workstation in the first risk probability and obtains the second risk probability of each production workstation based on the posterior risk probability of each production workstation and the corresponding station risk probability. Thus, the posterior risk probability is further corrected using the first risk probability, thereby improving the accuracy of the second risk probability for each production workstation.

[0129] Step S500: Generate corresponding risk station information and risk workstation information based on the first risk probability and the second risk probability.

[0130] In this embodiment, production workstations with a first risk probability higher than a first risk threshold can be designated as risk workstations to generate corresponding risk workstation information. Similarly, production workstations with a second risk probability higher than a second risk threshold can be designated as risk workstations to generate corresponding risk workstation information.

[0131] In this embodiment, the production stations can also be sorted according to the order of the first risk probability to obtain the corresponding risk station order as risk station information. Similarly, the production workstations can be sorted according to the order of the second risk probability to obtain the corresponding risk workstation order as risk workstation information.

[0132] Alternatively, the above two implementation methods can be combined. First, production stations with a first risk probability higher than a first risk threshold are designated as risk stations, and these risk stations are sorted according to their first risk probabilities to obtain their corresponding risk station order as risk station information. Then, production stations with a second risk probability higher than a second risk threshold are designated as risk stations, and these risk stations are sorted according to their second risk probabilities to obtain their corresponding risk station order as risk station information.

[0133] This allows risk station and workstation information to be output to pre-defined user devices (such as smartphones, tablets, and laptops) and displayed visually, such as by converting the information into text or charts. Users can then use this information to identify the production stations and workstations responsible for product defects.

[0134] Step S500, which generates corresponding risk station information and risk workstation information based on the first risk probability and the second risk probability, includes:

[0135] Step S510: Sort each production station according to the first risk probability, generate the corresponding risk station order, and use the risk station order as risk station information.

[0136] Step S520: Based on the second risk probability, sort each production station to generate the corresponding risk station order, and use the risk station order as risk station information.

[0137] After obtaining the first risk probability and the second risk probability, the production stations can be sorted based on the first risk probability to generate a corresponding risk station order, which is then used as the risk station information. Then, based on the second risk probability, the production workstations can be sorted to generate a corresponding risk workstation order, which is then used as the risk workstation information.

[0138] Therefore, in this embodiment, each production station and each production workstation is sorted according to the first risk probability and the second risk probability, respectively, to generate the corresponding risk station order and risk workstation order. This makes it easier for users to troubleshoot through the risk station order and risk workstation order, thereby improving the efficiency of production line fault location.

[0139] In the first embodiment of the present invention, a trained tree model is obtained, wherein the trained tree model is trained from sample production data of the target product whose fault is to be located. Based on the split node information of the trained tree model, the first risk probability of each production station of the target product is determined. Since only the data feature with the largest gain (i.e., the production station in the sample production data) can be used as a split node during the training process of the tree model, the first risk probability of each production station of the target product, i.e., the probability that the defective product is caused by the production station, can be determined based on the split node information. Furthermore, the prior risk probability of each production workstation within each production station can be obtained, and the prior risk probability can be corrected to obtain the posterior risk probability of each production workstation. Based on the first risk probability and the posterior risk probability, the second risk probability of each production workstation is determined, so that the posterior risk probability is estimated from the prior risk probability of each production workstation within each production station, and the second risk probability of each production workstation is determined by combining the first risk probability, i.e., the probability that the defective product is caused by the production workstation. Based on the first risk probability and the second risk probability, corresponding risk station information and risk workstation information are generated. This enables the location of risky workstations and risky work positions without requiring additional process parameter information or the accumulation of fault data, effectively improving the applicability of fault location technology.

[0140] The following is combined Figure 4 The flowchart shown is a second embodiment of the production line fault location method, which provides a detailed explanation of the production line fault location method.

[0141] See Figure 4 In another embodiment of the production line fault location method of the present invention, before the step of obtaining the trained tree model in step S100, the method includes:

[0142] Step A10: Obtain sample production data and untrained tree model, wherein the sample production data includes each production station that the target product passes through, each production station in each production station that the target product passes through, and the test results of the target product.

[0143] Step A20: Using each production station as data feature, each production workstation in each production station as feature value, and the detection result of the target product as data label, train the untrained tree model to obtain a trained tree model.

[0144] In this embodiment, it should be noted that the untrained tree model is a model constructed using a tree model algorithm as the base learner and employing a sequential or parallelization method. Furthermore, for sample production data from different sources, the hyperparameters of the untrained tree model can be adjusted automatically to obtain the untrained tree model with the best classification results. For example, the optimal hyperparameters of the untrained tree model can first be obtained using a Bayesian optimization method. These optimal hyperparameters are then used as fixed hyperparameters under the current conditions of the sample production data. Finally, the fixed hyperparameters are input into the untrained tree model for model training.

[0145] In this embodiment, it should also be noted that the sample production data includes each production station the target product passes through, each production workstation within each production station, and the test results of the target product. The test results are the results obtained after the target product is tested at the test station. It is understood that, as input data to the untrained tree model, the sample production data needs to undergo standardization processing, that is, it needs to be encoded and transformed into a form that the untrained tree model can recognize. The sample production data includes sample data and data labels. The sample data can include a two-dimensional matrix consisting of each production station the target product passes through and each production workstation within each production station. The data labels are the test results of the target product and can be a set {0,1}, with qualified products represented by 0 and unqualified products by 1.

[0146] Therefore, this embodiment trains an untrained tree model by using each production station as data features, each production workstation within each station as feature values, and the detection result of the target product as a data label. This allows the untrained tree model to utilize the distribution of qualified and unqualified products across the stations and workstations to determine the detection result of the target product and output a judgment result: whether the target product is qualified or unqualified. This results in a trained tree model. It can be understood that the trained tree model can accurately determine the detection result of any target product in the sample data.

[0147] Understandably, when an untrained tree model selects split nodes during training, it calculates the gain of each split node based on its feature value (production station) within the data features (i.e., production stations), and selects the production station with the highest gain among all production stations as its child node. When defective products are concentratedly generated from a specific production station within a particular production station, that production station has the highest gain and is selected as a child node for the split node. In other words, if the gain of a specific production station within a production station that causes a concentration of defective products is consistently at its maximum, it will be selected multiple times as a child node for the split node. This characteristic can be used to identify the production station causing the concentration of defective products.

[0148] In this embodiment, sample production data and an untrained tree model are acquired. The sample production data includes each production station the target product passes through, each production workstation within each production station, and the test results of the target product. Then, each production station is used as a data feature, each production workstation within each production station is used as a feature value, and the detection result of the target product is used as a data label to train the untrained tree model, resulting in a trained tree model. Therefore, the trained tree model can accurately determine the detection result of any target product in the sample data. Based on the split node information of this trained tree model, the importance of each production station to the detection result is determined.

[0149] Step A10, the step of obtaining sample production data, includes:

[0150] Step A11: In response to the non-conforming product query command, extract the non-conforming product information corresponding to the non-conforming product from the original production data;

[0151] Step A12: Based on the information of non-conforming products, determine the production time range of non-conforming products, extract the production information of qualified products within the production time range from the original production data, and use non-conforming products and qualified products as target products, and use the information of non-conforming products and the production information of qualified products as target production information.

[0152] Step A13: Encode the target production information according to the preset coding rules to obtain the corresponding sample production data.

[0153] In this embodiment, it should be noted that the original production data may include information such as product number, production station, the production station number the product passed through, the time of entering the production station, the time of leaving the production station, the production workstation number the product passed through, the time of entering the production workstation, and the time of leaving the production workstation. The non-conforming product query instruction may include user-defined query conditions for non-conforming products, or a specified product number. The query conditions may include specifying the testing station, the start and end range of the product testing time, and the non-conforming test items of the non-conforming product. Specifying a product number may include a list of non-conforming product numbers with the same non-conforming test items at the same testing station.

[0154] This embodiment, in response to a non-conforming product query command, extracts non-conforming product information from the original production data. This information may include the production stations the non-conforming product passed through, the production workstations it passed through, the production time range (i.e., the time range from the non-conforming product entering the first production station to the completion of production), the inspection stations it passed through, the inspection results, the non-conforming inspection items, and the times the non-conforming product entered and left the inspection stations. Based on this information, the production time range of the non-conforming products is determined. Within this time range, the production information of the conforming products is extracted from the original production data. Both the non-conforming and conforming products are designated as target products, and their information is used as target production information. Therefore, this embodiment treats all products within the production time range that generated the non-conforming products as target products and determines their target production information. This target production information includes at least the production stations the target product passed through, the production workstations within those stations, and the test results of the target product.

[0155] To convert the target production information into a form recognizable by the untrained tree model, it can be encoded according to preset encoding rules to obtain corresponding sample production data. For example, the station encoding rule in the preset encoding rules can be described as follows: based on the process sequence of the target product, production stations are numbered starting from 0 and incrementing sequentially until all production stations are represented by non-negative integer numbers from 0 to N. The production station encoding rule within each production station is described as follows: production stations are numbered starting from 0 and incrementing sequentially until all production stations are represented by non-negative integer numbers from 0 to M. Thus, the product's production process data is converted into coded data represented by non-negative integers, serving as sample data in the sample production data. The specific format of the sample data can be a two-dimensional matrix composed of production stations and their respective production stations.

[0156] Prior to step A13, which involves encoding the target production information according to a preset encoding rule to obtain the corresponding sample production data, the method further includes:

[0157] Step B10: Obtain the sub-material production information for the target product;

[0158] Step B20 involves cascading the target production information and sub-material production information to synthesize the corresponding product-related production process information, and using the product-related production process information as the new product production information.

[0159] It is understandable that the target product may require sub-materials from other production lines to complete its production. Therefore, if the target product is defective, it may be due to the production station or workstation in the production process of the sub-materials of the defective product.

[0160] Therefore, when the target product and its sub-materials are not produced on the same production line, it is necessary to associate the target product's production information with the corresponding sub-materials' production information. By obtaining the target product's sub-materials' production information, and then cascading the target production information and sub-materials' production information according to the target product's technological process sequence, the corresponding product-related production process information is synthesized, and this product-related production process information is used as the new product production information.

[0161] This embodiment acquires the production information of sub-materials for the target product, then cascades the target production information and sub-material production information to synthesize corresponding product-related production process information, and uses this product-related production process information as new product production information. This allows for fault location even for the production processes of sub-materials that are not on the same production line as the target product, effectively improving the comprehensiveness and accuracy of the fault location method.

[0162] Prior to step A13, which involves encoding the target production information according to a preset encoding rule to obtain the corresponding sample production data, the method further includes:

[0163] Step C10: Based on the testing items of the target product, determine the relevant workstations and work positions corresponding to the testing items;

[0164] Step C20: Based on the relevant workstations and work positions, filter the target production information, retain each production workstation corresponding to the relevant workstation and each production work position corresponding to the relevant work position in the target production information, and obtain new target production information.

[0165] It is understandable that, since the target product usually passes through production stations and workstations that are unrelated to the inspection items in the inspection station, there will be a lot of redundant information in the target production information, which will affect the model training efficiency.

[0166] In this embodiment, by pre-setting the mapping relationship between inspection items and workstations / stations, the relevant workstations and stations corresponding to the inspection items of the target product can be queried. Then, based on the relevant workstations and stations, the target production information can be filtered, retaining the production workstations corresponding to the relevant workstations and the production stations corresponding to the relevant stations, thus obtaining new target production information. Therefore, this embodiment removes production workstations and stations unrelated to the inspection items of the target product, greatly reducing redundant information in the target production information, effectively improving model training efficiency, reducing interference from redundant information, and significantly improving the accuracy of the trained tree model, thereby improving the accuracy of locating risky workstations and stations based on the trained tree model.

[0167] The following is combined Figure 5 The flowchart shown is a third embodiment of the production line fault location method, which provides a detailed explanation of the production line fault location method.

[0168] See Figure 5 In another embodiment of the production line fault location method of the present invention, after step A13, which encodes the target production information according to a preset encoding rule to obtain the corresponding sample production data, the method further includes:

[0169] Step D10: Obtain the first pass probability of the target product through each production station;

[0170] Step D20: Remove the production stations whose first pass probability is lower than a preset threshold from the sample production data to obtain new sample production data.

[0171] Step E10: Obtain the second probability of the target product passing through each workstation in each production station.

[0172] Step E20: Retain any production station that the second probability characterization indicates must have been passed through from the sample production data to obtain new sample production data.

[0173] Step F10: Obtain the correlation coefficients between each pair of production workstations;

[0174] Step F20: Remove any one of the two production stations whose correlation coefficients indicate a corresponding relationship from the sample production data to obtain new sample production data.

[0175] In this embodiment, it should be noted that after step A13, at least one of the following operations can be performed: the validity test operation of steps D10 and D20, the repeatability test operation of steps E10 and E20, and the correlation test operation of steps F10 and F20. The execution order of the validity test operation, the repeatability test operation, and the correlation test operation can be set according to specific needs, and this embodiment does not impose any restrictions on this.

[0176] It is understandable that not all target products may pass through each production station. Therefore, steps D10 to D20 include: calculating the ratio between the number of target products passing through each production station and the total number of target products, thus obtaining the first probability of a target product passing through each production station (e.g., 10%, 15%, 20%). Then, production stations with a first probability of passing through that station lower than a preset threshold are removed from the sample production data to obtain new sample production data.

[0177] The rules for validity testing are as follows:

[0178]

[0179] Among them, P i,pass Let P be the probability of the target product passing through production station i for the first time. When the probability of the first passing is greater than or equal to a preset threshold P, threshold If the probability of a target product passing through production station i is high (i.e., a large number of target products have passed through production station i), then production station i can be retained. If the probability of passing through is less than the preset threshold Pthreshold, then production station i has a low probability of passing through (i.e., only a small number of target products have passed through production station i), and production station i has little impact on the detection results of the target products. Therefore, production station i can be deleted.

[0180] Further, before step D20, the process may include: obtaining the proportion of defective products passing through the target product at production stations with a first passing probability lower than a preset threshold; if the proportion of defective products is greater than a preset threshold (e.g., 30%, 40%, 50%), then retaining the production stations with the first passing probability lower than the preset threshold and the production workstations under those stations in the sample production data; if the proportion of defective products is not greater than the preset threshold, then executing step D20: deleting the production stations with the first passing probability lower than the preset threshold from the sample production data to obtain new sample production data. This embodiment uses the proportion of defective products passing through the target product at production stations with a first passing probability lower than the preset threshold to determine whether a production station has an impact on defective products. Therefore, if the proportion of defective products is greater than the preset threshold, the production stations with the first passing probability lower than the preset threshold in the sample production data are retained, thus avoiding the simplification of the sample production data and ensuring the accuracy of locating risky stations and workstations.

[0181] Understandably, in most cases, all target products will pass through every production station, but not all target products will necessarily pass through every production workstation within a production station. Therefore, when all target products will inevitably pass through a certain production workstation within some production stations, the risk probability of these production stations is actually the same. Steps E10 to E20 include: calculating the ratio between the number of target products passing through each production workstation in each production station and the total number of target products, thus obtaining the second probability of a target product passing through each workstation in each production station. Then, from the sample production data, the second probability of passing through represents any production station that will inevitably be passed through, obtaining new sample production data.

[0182] The rules for repeatability testing are as follows:

[0183]

[0184] Among them, P i,j Let P be the probability that the target product passes through production station j within production station i for the second time. It should be noted that P... i,j =1 indicates that it must be passed through. Since all production stations whose second pass probability indicates that they must be passed through have the same impact (i.e., risk probability) on the test results of the target product, this embodiment will retain any production station whose second pass probability indicates that it must be passed through from the sample production data, and then delete the other production stations whose second pass probability indicates that they must be passed through from the sample production data.

[0185] It is understandable that when a target product passes through a certain production station in a production station, it must pass through a corresponding production station in another production station. This indicates a correspondence between the two, meaning their risk probabilities are the same. For example, if the target product passes through production station 52 in production station 5, it must pass through production station 73 in production station 7. Therefore, it can be considered that there is a correspondence between production stations 5 and 7. Steps F10 to F20 include: calculating the correlation coefficient between each pair of production stations using the Spearman method. The correlation coefficient is calculated as follows:

[0186]

[0187] Where P(i,j) is the correlation coefficient between production station i and production station j, and N is the number of production workstations in the production station. x Let j be the encoded value of production station x in production station i (i.e., the encoded value corresponding to the production station after encoding the target production information according to the preset encoding rules). x This is the coded value of production station x in production station j. This is the average of the encoded values ​​of all production stations in production station i. This is the average of the encoded values ​​of all production stations in production station j.

[0188] Then, remove any one of the two production stations whose correlation coefficients indicate a corresponding relationship from the sample production data to obtain new sample production data.

[0189] The rules for correlation testing are as follows:

[0190]

[0191] P(i,j) is the correlation coefficient between the production station in production station i and the production station in production station j for the target product. P(i,j) = 1 indicates that a correspondence exists. When there is a correspondence between the production stations in the two production stations, the impact (i.e., risk probability) of the two production stations on the test results of the target product is actually the same. Therefore, in this embodiment, when P(i,j) = 1, either production station i or production station j is retained; when P(i,j) ≠ 1, both production stations i and j are retained.

[0192] This embodiment obtains new sample production data by acquiring the first probability of the target product passing through each production station, deleting production stations with a first probability lower than a preset threshold from the sample production data, and / or acquiring the second probability of the target product passing through each workstation within each production station, retaining any production station whose second probability indicates a certain passage from the sample production data, and / or acquiring the correlation coefficient between each pair of production stations, deleting any one of the two production stations whose correlation coefficient indicates a corresponding relationship from the sample production data. This further simplifies the sample production data, improves model training efficiency, reduces interference from redundant information, and significantly improves the accuracy of the trained tree model, thereby improving the accuracy of locating risky production stations and workstations based on the trained tree model.

[0193] In addition, this embodiment of the invention also proposes a computer storage medium on which a computer program is stored. When the computer program is executed by a processor, it implements the operations in the production line fault location method provided in the above embodiment. The specific steps will not be described in detail here.

[0194] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0195] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0196] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0197] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for locating faults on a production line, characterized in that, The production line fault location process does not require the introduction of additional production line process parameter information, and it can complete the fault location of workstations and work positions without the need for prior accumulation of fault sample data. The production line fault location method includes: Obtain a trained tree model, wherein the trained tree model is trained from sample production data of the target product for which the fault is to be located; Based on the split node information, determine the number of split participations for each production station; The first risk probability of each production station is determined based on the sum of the number of split participations. Obtain the prior risk probability and additional risk probability of each production station within each production station, wherein the additional risk probability includes the overall non-conforming rate of the target product and the product passing through each production station, wherein the prior risk probability is the non-conforming rate of the target product passing through the production station. The prior risk probability is corrected based on the overall non-conformance rate and the product pass probability to obtain the posterior risk probability of each production station. Determine the station risk probability of each production station corresponding to each production workstation in the first risk probability, wherein the first risk probability includes the station risk probability corresponding to each production station, and the first risk probability is the probability that the production station will cause non-conforming products. The second risk probability of each production workstation is obtained based on the posterior risk probability of each workstation and the corresponding workstation risk probability. Based on the first risk probability and the second risk probability, corresponding risk workstation information and risk work position information are generated.

2. The production line fault location method according to claim 1, characterized in that, Prior to the step of obtaining the trained tree model, the method includes: Acquire sample production data and an untrained tree model, wherein the sample production data includes each production station through which the target product passes, each production workstation in each of the production stations through which the target product passes, and the test results of the target product. Using each of the aforementioned production stations as data features, each production workstation within each of the aforementioned production stations as feature values, and the detection results of the target product as data labels, the untrained tree model is trained to obtain a trained tree model.

3. The production line fault location method as described in claim 2, characterized in that, The step of obtaining sample production data includes: In response to the non-conforming product query command, extract the non-conforming product information corresponding to the non-conforming product from the original production data; Based on the information of the non-conforming products, the production time range of the non-conforming products is determined. The production information of the qualified products within the production time range is extracted from the original production data. The non-conforming products and the qualified products are used as target products, and the information of the non-conforming products and the production information of the qualified products are used as target production information. The target production information is encoded according to a preset encoding rule to obtain the corresponding sample production data.

4. The production line fault location method as described in claim 3, characterized in that, Before the step of encoding the target production information according to a preset encoding rule to obtain the corresponding sample production data, the method further includes: Obtain the sub-material production information of the target product; The target production information and the sub-material production information are cascaded to synthesize corresponding product-related production process information, and the product-related production process information is used as new product production information.

5. The production line fault location method as described in claim 3, characterized in that, Before the step of encoding the target production information according to a preset encoding rule to obtain the corresponding sample production data, the method further includes: Based on the testing items of the target product, determine the relevant workstations and work positions corresponding to the testing items; Based on the relevant workstations and work positions, the target production information is filtered, and each production workstation corresponding to the relevant workstation and each production work position corresponding to the relevant work position in the target production information is retained to obtain new target production information.

6. The production line fault location method as described in claim 3, characterized in that, After the step of encoding the target production information according to a preset encoding rule to obtain the corresponding sample production data, the following steps are included: Obtain the first probability of the target product passing through each of the production stations; Remove the production stations whose first pass probability is lower than a preset threshold from the sample production data to obtain new sample production data.

7. The production line fault location method as described in claim 3, characterized in that, After the step of encoding the target production information according to a preset encoding rule to obtain the corresponding sample production data, the method further includes: Obtain the second probability of the target product passing through each of the workstations within each of the production stations; From the sample production data, retain any production station that the second probability characterization indicates must have been passed through, and obtain new sample production data.

8. The production line fault location method as described in claim 3, characterized in that, After the step of encoding the target production information according to a preset encoding rule to obtain the corresponding sample production data, the method further includes: Obtain the correlation coefficients between each pair of the aforementioned production stations; Remove either of the two production stations whose correlation coefficients indicate a corresponding relationship from the sample production data to obtain new sample production data.

9. The production line fault location method according to any one of claims 1 to 8, characterized in that, The step of generating corresponding risk workstation information and risk work position information based on the first risk probability and the second risk probability includes: Based on the first risk probability, each of the production stations is sorted to generate a corresponding risk station order, and the risk station order is used as risk station information. Based on the second risk probability, each production station is sorted to generate a corresponding risk station order, and the risk station order is used as risk station information.

10. A production line fault location device, characterized in that, The production line fault location device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the production line fault location method as described in any one of claims 1 to 9.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the production line fault location method as described in any one of claims 1 to 9.

Images