Abnormality detection method and device, electronic equipment and storage medium

Abstract

The invention provides an anomaly detection method and device, electronic equipment and a storage medium. The method comprises the following steps: receiving product parameters of a product; under the condition that the state of the product parameter is abnormal, based on an abnormal reason estimation model, an estimated abnormal reason corresponding to the product parameter is determined, the product parameter and the estimated abnormal reason are presented, and the state comprises normal and abnormal states; and under the condition that an actual abnormal reason indicated by a user for the product parameter is received, performing incremental training on the abnormal reason estimation model based on the product parameter and the actual abnormal reason. According to the embodiment of the invention, the problems of low efficiency and relatively high cost of a manual method are solved, and the defects of high deployment cost, poor adaptability to different technological processes, difficulty in maintenance and the like of a computer anomaly monitoring system are also overcome, so that the automatic anomaly detection method can be widely applied to various industrial scenes, and the working efficiency of technicians is greatly improved.

Description

Technical Field

[0001] The present disclosure relates to the technical field of industrial anomaly detection, and in particular to anomaly detection methods, devices, electronic equipment and storage media. Background Art

[0002] In industrial production, it is often necessary to inspect the products produced in order to promptly discover abnormalities in production equipment and avoid long-term existence of industrial production abnormalities that may lead to major economic losses or even operational accidents.

[0003] At present, the classification of production processes is becoming more and more detailed, and there are many optional options. In addition, machine tool manufacturers are often different and the actual production environment is often different. Therefore, the current mainstream industrial anomaly detection method is still manual inspection, such as manual observation of camera images or manual sample sampling. This method is inefficient and costly. Summary of the invention

[0004] To overcome the problems existing in the related art, the present disclosure provides an abnormality detection method, device, electronic device and storage medium.

[0005] A first aspect of the present disclosure provides an anomaly detection method, the method comprising:

[0006] Receive product parameters of the product;

[0007] In the case where the state of the product parameter is abnormal, based on the abnormal cause estimation model, determining the estimated abnormal cause corresponding to the product parameter, and presenting the product parameter and the estimated abnormal cause, wherein the state includes normal and abnormal;

[0008] When an actual abnormality cause indicated by a user for the product parameter is received, incremental training is performed on the abnormality cause estimation model based on the product parameter and the actual abnormality cause.

[0009] Optionally, the method further includes:

[0010] If the product parameter is within a threshold range, the state of the product parameter is determined to be abnormal, wherein the abnormal cause estimation model is used to determine the threshold range.

[0011] Optionally, the method further includes:

[0012] After the abnormality cause estimation model is incrementally trained each time based on the product parameters and the actual abnormality cause, the current threshold range is updated based on the output result of the abnormality cause estimation model after the incremental training.

[0013] Optionally, the incrementally training the abnormality cause estimation model based on the product parameters and the actual abnormality cause includes:

[0014] The state and the actual abnormal cause indicated by the user for the product parameter are received, and the state and the actual abnormal cause are used as labels corresponding to the product parameter and input into the abnormal cause estimation model to perform incremental training on the abnormal cause estimation model.

[0015] Optionally, the method further includes:

[0016] Based on the estimated abnormal cause corresponding to the product parameter and the preset knowledge base corresponding to the product, a solution to the estimated abnormal cause is presented.

[0017] Optionally, the abnormal cause estimation model includes a large language model;

[0018] The presenting of a solution to the estimated abnormal cause according to the estimated abnormal cause corresponding to the product parameter and the preset knowledge base corresponding to the product includes:

[0019] The text output by the abnormal cause estimation model is input into a decision tree model so that the decision tree model presents a solution to the estimated abnormal cause according to a preset knowledge base corresponding to the product, wherein the text is used to indicate the estimated abnormal cause.

[0020] Optionally, presenting the product parameters and the estimated abnormality reasons includes:

[0021] Determining the process stage corresponding to the estimated abnormal cause;

[0022] The abnormal parameters and the estimated abnormal causes are presented to the account or terminal corresponding to the process stage.

[0023] Optionally, the method further includes:

[0024] Record the process stage corresponding to the estimated abnormal cause;

[0025] When receiving the actual abnormal reason indicated by the user for the product parameter, updating the recorded process stage according to the process stage corresponding to the actual abnormal reason;

[0026] According to the correspondence between the recorded abnormal causes and the process stages, the abnormality rate of each process stage in the process corresponding to the product is presented.

[0027] A second aspect of the present disclosure provides an abnormality detection device, the device comprising:

[0028] A parameter receiving module, used for receiving product parameters of a product;

[0029] a cause estimation module, configured to determine, when the state of the product parameter is abnormal, an estimated abnormal cause corresponding to the product parameter based on an abnormal cause estimation model, and present the product parameter and the estimated abnormal cause, wherein the state includes normal and abnormal;

[0030] The incremental training module is used to perform incremental training on the abnormality cause estimation model based on the product parameters and the actual abnormality cause when receiving the actual abnormality cause indicated by the user for the product parameters.

[0031] Optionally, the device further includes an abnormality judgment module, configured to perform the following steps:

[0032] If the product parameter is within a threshold range, the state of the product parameter is determined to be abnormal, wherein the abnormal cause estimation model is used to determine the threshold range.

[0033] Optionally, the device further comprises:

[0034] The threshold updating module is used to update the current threshold range based on the output result of the abnormal cause estimation model after incremental training each time based on the product parameters and the actual abnormal cause.

[0035] Optionally, when the incremental training module is used to perform incremental training on the abnormal cause estimation model based on the product parameters and the actual abnormal cause, it is specifically used to:

[0036] The state and the actual abnormal cause indicated by the user for the product parameter are received, and the state and the actual abnormal cause are used as labels corresponding to the product parameter and input into the abnormal cause estimation model to perform incremental training on the abnormal cause estimation model.

[0037] Optionally, the device further comprises:

[0038] The solution estimation module is used to present a solution to the estimated abnormal cause based on the estimated abnormal cause corresponding to the product parameter and the preset knowledge base corresponding to the product.

[0039] Optionally, the abnormal cause estimation model includes a large language model;

[0040] The solution estimation module is used to present a solution to the estimated abnormal cause according to the estimated abnormal cause corresponding to the product parameter and the preset knowledge base corresponding to the product, and is specifically used to:

[0041] The text output by the abnormal cause estimation model is input into a decision tree model so that the decision tree model presents a solution to the estimated abnormal cause according to a preset knowledge base corresponding to the product, wherein the text is used to indicate the estimated abnormal cause.

[0042] Optionally, when presenting the product parameters and the estimated abnormal causes, the cause estimation module is specifically used to:

[0043] Determining the process stage corresponding to the estimated abnormal cause;

[0044] The abnormal parameters and the estimated abnormal causes are presented to the account or terminal corresponding to the process stage.

[0045] Optionally, the device further includes an abnormality rate analysis module, configured to perform the following steps:

[0046] Record the process stage corresponding to the estimated abnormal cause;

[0047] When receiving the actual abnormal reason indicated by the user for the product parameter, updating the recorded process stage according to the process stage corresponding to the actual abnormal reason;

[0048] According to the correspondence between the recorded abnormal causes and the process stages, the abnormality rate of each process stage in the process corresponding to the product is presented.

[0049] A third aspect of the present disclosure provides a computer program product, including a computer program / instruction, which implements the method described in the first aspect when executed by a processor.

[0050] A fourth aspect of the present disclosure provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in the first aspect when executing the computer program.

[0051] A fifth aspect of the present disclosure provides a computer-readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the method described in the first aspect.

[0052] The technical solution provided by the embodiments of the present disclosure may have the following beneficial effects:

[0053] The method architecture provided by the embodiments of the present disclosure can be easily embedded in various abnormality monitoring systems of computers, and after one deployment, it can be adapted to different processes, as well as different process stages or different production batches of the same process flow. It not only overcomes the problems of low efficiency and high cost of manual methods, but also overcomes the shortcomings of high deployment cost, poor adaptability to different process flows, and difficult maintenance of computer abnormality monitoring systems, so that the automatic abnormality detection method can be widely used in various industrial scenarios, greatly improving the work efficiency of technicians.

[0054] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The accompanying drawings, which are incorporated in and constitute a part of this disclosure, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0056] Figure 1 The following is a flow chart of an anomaly detection method shown in some exemplary embodiments.

[0057] Figure 2 is a schematic diagram of a process stage shown in some exemplary embodiments.

[0058] Figure 3 is a selection branch diagram of an anomaly detection method shown in some exemplary embodiments.

[0059] Figure 4 is a state transition diagram of an anomaly detection method shown in some exemplary embodiments.

[0060] Figure 5 is a block diagram of an abnormality detection device showing some exemplary embodiments.

[0061] Figure 6 It is a hardware structure diagram of an electronic device shown in some exemplary embodiments. DETAILED DESCRIPTION

[0062] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0063] As mentioned in the background technology, due to the large number of process classifications, coupled with the fact that machine tool manufacturers are often different and the actual production environment is often different, therefore, automated anomaly detection can usually only be effectively applied in large-scale enterprises or factories with good management order. For most manufacturers, the current mainstream industrial anomaly detection method is still manual inspection, such as manual observation of camera images or manual sample inspection, which is inefficient and costly.

[0064] In view of this, the present disclosure provides an abnormality detection method, device, electronic device and storage medium. Next, the embodiments of the present disclosure are described in detail.

[0065] The first aspect of the present disclosure provides an anomaly detection method. Figure 1 , which may include steps S101 to S103.

[0066] Step S101, receiving product parameters of a product.

[0067] The product can be a product obtained at any stage of any process, for example, see Figure 2 , the raw material undergoes process stage 1 to obtain product 1 (intermediate product), and product 1 undergoes process stage 2 to obtain product 2 (final product), and both product 1 and product 2 can be used as the product in step S101. For glass products, product parameters may include glass thickness, light transmittance, compressive strength and other parameters, which can be collected based on acquisition equipment such as spectrometers and scanning electron microscopes. Similarly, for other types of products, it is only necessary to collect preset parameters based on corresponding acquisition equipment. In addition, for the parameter collection of product 1 and product 2, a variety of different types of acquisition equipment can be used to perform them respectively, and the specific types of parameters collected and the corresponding abnormal threshold ranges can also be different.

[0068] After collecting product parameters (that is, Figure 2 After obtaining the indicator data in the dataset, data cleaning can be performed first. Data cleaning can be performed based on a deep learning model or a preset algorithm. For example, empty data output by the acquisition device and obvious abnormal data (such as data that does not conform to physical rules or data greater than a certain threshold) can be removed.

[0069] Next, the status of the product parameters can be determined. Among them, the status includes normal and abnormal, that is, the status of the product parameters can characterize whether the product corresponding to the product parameters is normal or abnormal (these abnormalities may be caused by abnormal raw material ratios, machine tool failures, incorrect mechanical parameter settings, equipment aging, or excessive operating errors). Of course, the abnormality can also be subdivided into multiple abnormality levels, but the idea remains unchanged, so it will not be repeated here; in addition, it should be understood that the abnormal state of the product obtained in each process stage is not necessarily only affected by the process stage, but may also include the influence of certain process stages before the process stage.

[0070] Exemplarily, the anomaly can be completed based on an anomaly cause estimation model or other deep learning models, that is, the method can also include inputting the product parameters into the corresponding deep learning model to obtain the status of the product parameters.

[0071] In another exemplary embodiment, the state judgment step of the product parameter can be performed independently of the deep learning model, but at the same time maintain the influence of the deep learning model on the step, thereby combining the advantages of high adaptability and high accuracy of the deep learning model with the advantages of high efficiency and low overhead of threshold judgment. For example, the method may also include: if the product parameter is within a threshold range, then judging the state of the product parameter as abnormal, wherein the abnormal cause estimation model is used to determine the threshold range.

[0072] See also Figure 3 The product parameters of the process product can be executed by a lightweight judge, which can compare the product parameters with the abnormal threshold to determine whether the product parameters fall within the abnormal range indicated by the threshold. If so, the product parameters are judged to be abnormal, and steps S102 to S103 are executed; if not, the product parameters can be considered normal, and this abnormality detection ends. The above abnormal threshold range can be maintained by a deep learning model, which can be an abnormal cause estimation model (to be introduced later) or other independent models.

[0073] Of course, it should be understood that when the status of the product parameters is normal, logs can also be recorded, and the product parameters can be used to train any of the above models. In addition, the above product parameters and abnormal thresholds can have multiple sub-items, and the product parameters can be judged to be abnormal when any product parameter sub-item falls into the threshold range sub-item corresponding to the sub-item, or the product parameters can be considered abnormal when a preset number of product parameter sub-items (such as all sub-items) fall into the abnormal range sub-item corresponding to each product parameter sub-item.

[0074] Based on this, each status judgment of the product parameters no longer needs to go through the abnormal cause estimation model, but can be directly determined by a simple threshold comparison. At the same time, the threshold is determined by the abnormal cause estimation model, so the change of the threshold can also be integrated into the method provided by the present invention and updated according to user instructions, thereby forming an adaptive judgment of the status (that is, automatically adapting to different process stages of the product, or different batches of the same process stage).

[0075] Step S102, when the state of the product parameter is abnormal, based on the abnormal cause estimation model, determine the estimated abnormal cause corresponding to the product parameter, and present the product parameter and the estimated abnormal cause.

[0076] The industrial parameter model can be a classification model, such as Support Vector Machine (SVM), Random Forest, etc.; it can also be a large language model, such as Generative Pre-trained Transformer (GPT); or Bidirectional Encoder Representations from Transformers (BERT), etc.; for example, when the industrial parameter model is a large language model (of course, the industrial parameter model is not limited to a single model, it can also be a combination of multiple models), the text output by the large language model can include abnormal threshold range, abnormal state, estimated abnormal cause, etc. (these contents can be selected and matched according to actual conditions), and can also include estimated solutions (that is, it can partially or completely act as a decision tree model to realize its function).

[0077] See also Figure 3 , when the product parameters are judged to be abnormal (that is, the state is abnormal; the optional judgment method is described later), the abnormal parameters can be input into the abnormal cause estimation model to obtain the estimated abnormal cause output by the abnormal cause estimation model (the model can be pre-trained according to the pre-divided process categories, and then the accuracy of the estimated abnormal cause output by the model can be further improved when the user applies it to the specific production application process). The abnormal parameters and the corresponding estimated abnormal causes can be presented to the user to provide a reference for the user. The estimated abnormal cause can have various optional granularities, such as being located on a certain production component (such as a milling cutter, a valve), or located on one or some production equipment, or located on one or some process stages.

[0078] Of course, in addition to this, the method can also be connected with other automatic exception handling devices or exception handling models, so that other models or devices can automatically handle corresponding exceptions according to the estimated causes of the exceptions, so that the process can return to normal.

[0079] Step S103: upon receiving the actual abnormal cause indicated by the user for the product parameter, incremental training is performed on the abnormal cause estimation model based on the product parameter and the actual abnormal cause.

[0080] You can continue to see Figure 3 When the user thinks that the error of the estimated abnormal cause is large, the actual abnormal cause can be entered. When the user thinks that the abnormal parameter is not actually abnormal, the actual state (such as "normal") can be entered. Then, the content entered by the user (that is, the state indicated by the user, which can also be called the actual state) can be used together with the corresponding abnormal parameters as training data to update the abnormal cause estimation model.

[0081] Among them, incremental training can gradually update the model without having to retrain the entire model every time new data is received. For example, when new product parameters and actual abnormal causes are received, they are added to the original data set or directly applied to the model. By using these new data to train the abnormal cause estimation model, the model gradually learns and adapts to new experiences while retaining the original knowledge, so that the model can continuously improve its prediction ability, improve adaptability, and more accurately identify and predict possible abnormalities in the future. It can effectively utilize new data and avoid the high time cost and resource consumption that may be incurred when training traditional batch models.

[0082] For example, optionally, the incremental training of the abnormal cause estimation model based on the product parameters and the actual abnormal cause may include: receiving the status and actual abnormal cause indicated by the user for the product parameters, and inputting the status (i.e., the status indicated by the user, which can also be considered as the actual status of the product parameter) and the actual abnormal cause as labels corresponding to the product parameters into the abnormal cause estimation model, so as to perform incremental training on the abnormal cause estimation model.

[0083] That is, "Is the product parameter really abnormal" and "What is the abnormal reason for the product parameter (if abnormal)" can be used as the first label and the second label of the product parameter (sample), respectively, and then the sample is input into the abnormal reason estimation model to perform incremental training on it. In this way, the abnormal reason estimation model (and the operating architecture provided in the present disclosure) can adapt to new product batches, new process stages and even new processes after only a small amount of incremental training guided by the user, so as to more accurately identify and predict the possible conditions in the future production process, and provide effective estimated abnormal reasons and even estimated solutions.

[0084] In addition, the method may further include: after incremental training of the abnormal cause estimation model based on the product parameters and the actual abnormal cause each time, updating the current threshold range based on the output result of the abnormal cause estimation model after the incremental training.

[0085] In other words, after the abnormal cause estimation model is updated, a new threshold corresponding to the process can be output based on the updated abnormal cause estimation model (the new threshold can cover part or all of the sub-items of the threshold range, or a new sub-item can be introduced into the threshold range), and then the threshold range (the corresponding sub-item) can be updated based on these new thresholds. Taking the aforementioned glass products as an example, the original threshold range may only include one sub-item of transmittance. When the transmittance of the process product is within a certain range (for example, below a certain threshold), the product is considered abnormal; and the above-mentioned update process can be to let the abnormal cause estimation model output a new transmittance range, or to introduce another detection dimension for process product detection, such as the compressive strength of glass. Optionally, this optional step can be triggered after the user enters the actual state (when the user only enters the actual abnormal cause and does not specify the abnormal state, it can be considered that the current abnormal threshold does not need to be updated), thereby further improving the training efficiency.

[0086] It is worth noting that if the user only provides one of the first label and the second label, the steps performed by the model can be adaptively adjusted according to the specific content provided by the user. For example, the threshold range is used to predetermine whether the product parameters are abnormal. Then the model can update the threshold range when the user provides "whether the product parameters are indeed abnormal", but when the user only provides "the actual abnormal reason", it will not be executed, or it will be executed but the threshold range will be updated with a lower weight.

[0087] In addition, if the threshold adjustment model responsible for outputting the abnormal threshold range and the abnormal cause estimation model responsible for outputting the estimated abnormal cause in the method are different models, the actual state indicated by the user can be used as a label of the product parameters and input into the threshold adjustment model, and the actual abnormal cause indicated by the user can be used as a label of the product parameters and input into the abnormal cause estimation model. The two models can be trained and run separately, and the present disclosure does not limit this.

[0088] In summary, the disclosed embodiment can estimate the abnormal cause corresponding to the product parameter based on the abnormal cause estimation model when the product parameter of the product is received and the state of the product parameter is abnormal, and present the product parameter and the estimated abnormal cause to the user; if the cause is wrong, the user can indicate the actual abnormal cause of the product parameter, and the actual abnormal cause will be used to perform incremental training on the abnormal cause estimation model. The above method architecture can be easily embedded in various abnormal monitoring systems of the computer, and after one deployment, it can be adapted to different processes, different process stages or different production batches of the same process flow. This is because the product parameter characteristics obtained by different processes and different process stages or different production batches of the same process flow are likely to fluctuate, and due to factors such as product updates and iterations, replacement of equipment and machine tools, and introduction of new production technologies, in the above cases, even if the product parameter value collected by the acquisition device has not changed, the parameter value may correspond to different abnormal states, and in the process of industrial updates and iterations, new process stages or new products may also be introduced. In the method provided by the present disclosure, such changes can be easily adapted based on a small amount of user guidance, so that the above method can be directly deployed in an existing industrial production environment without affecting the update and iteration of the industrial production environment.

[0089] Therefore, the embodiments of the present disclosure not only overcome the problems of low efficiency and high cost of manual methods, but also overcome the shortcomings of high deployment cost, poor adaptability to different process flows, and difficult maintenance of computer anomaly monitoring systems, so that automatic anomaly detection methods can be widely used in various industrial scenarios, greatly improving the work efficiency of technicians.

[0090] In some embodiments, the method may further include: presenting a solution to the estimated abnormal cause based on the estimated abnormal cause corresponding to the product parameter and a preset knowledge base corresponding to the product.

[0091] The preset knowledge base can be a pre-prepared information base about the process of the above-mentioned product, such as "if a certain error occurs in a certain component, it may cause certain parameters of the product to change in a certain direction", or "the concentration of the input raw materials in a certain process stage will directly affect which characteristics of the product". Exemplarily, after the abnormal cause estimation model obtains the estimated abnormal cause based on the product parameter analysis, the estimated abnormal cause can be matched with the solution in the preset knowledge base. For example, for the thickness parameter abnormality of product 1, the possible estimated abnormal cause is the wrong raw material ratio through analysis, and then the corresponding solution is searched and matched in the preset knowledge base, such as adjusting the raw material ratio.

[0092] In another exemplary embodiment, the abnormal cause estimation model includes a large language model. In this case, the output of the model can be directly presented to the user after being trimmed, and the output of the model can also be made highly versatile, for example, it can be directly used as the input of other models. In addition, presenting a solution to the estimated abnormal cause according to the estimated abnormal cause corresponding to the product parameter and the preset knowledge base corresponding to the product includes: inputting the text output by the abnormal cause estimation model into a decision tree model, so that the decision tree model presents a solution to the estimated abnormal cause according to the preset knowledge base corresponding to the product, wherein the text is used to indicate the estimated abnormal cause.

[0093] That is to say, in addition to the abnormal cause estimation model, a decision tree model for determining solutions can also be deployed. The decision tree model can be a classification prediction model based on a tree structure, which can divide data into different categories or prediction results through a series of decision rules. Specifically, the decision tree model includes a root node, an internal node, and a leaf node, each node represents a decision point or a judgment condition, and the leaf node represents the final classification result or prediction value. The preset knowledge base can be used as a training sample or reference data for the decision tree model. By inputting actual production data and known solutions into the decision tree model, the model can be trained and its decision rules can be optimized. For example, for the thickness parameter abnormality of the product, the historical data of different states and the corresponding solutions can be recorded in the knowledge base, and then these data can be used to train the decision tree model so that it can quickly provide corresponding solutions when encountering similar abnormalities in the future.

[0094] In some embodiments, presenting the product parameters and the estimated abnormal cause may include: determining the process stage corresponding to the estimated abnormal cause; and presenting the abnormal parameters and the estimated abnormal cause to the account or terminal corresponding to the process stage.

[0095] The account can be an account issued by the platform that executes the method, or an email account or phone number of a certain person; for example, if an abnormality occurs in workshop A at a certain process stage, the relevant parameters and reasons will be sent to the corresponding staff of workshop A; or, if an abnormality occurs in workshop B, the relevant parameters and reasons will be sent to the staff of workshop B. In this way, the pertinence of human-computer interaction can be enhanced.

[0096] In some embodiments, the method may further include: recording the process stage corresponding to the estimated abnormal cause; upon receiving the actual abnormal cause indicated by the user for the product parameters, updating the recorded process stage according to the process stage corresponding to the actual abnormal cause; and presenting the abnormality rate of each process stage in the process corresponding to the product based on the correspondence between the recorded abnormal cause and the process stage.

[0097] That is to say, the estimated abnormal cause can be obtained by analyzing the product parameters, and the abnormal cause can be associated with the corresponding process stage and recorded in the database. Then, after the user provides the actual abnormal cause, the actual abnormal cause can be matched with the corresponding process stage, and the previously recorded process stage information can be updated. Then, when the user needs to view the statistical results, or at the end of a statistical cycle, the abnormal rate of the product in each process stage can be calculated and presented based on the correspondence between the recorded abnormal cause and the process stage. For example, the status of different process stages such as raw material ratio, melting, and molding can be recorded separately, and the abnormal rate of each stage can be calculated, so that users can monitor and manage the quality control of each link in the production process.

[0098] The above-mentioned embodiments can be integrated on the same software platform, thereby providing users with a functionally rich and highly versatile (i.e., ready-to-use) industrial anomaly detection platform. As an example, the process shown in the embodiment of the present disclosure (e.g. Figure 3 The process shown in the figure can be executed by a software development kit (Software Development Kit, usually referred to as SDK) or an application programming interface (Application Programming Interface, usually referred to as API) (for example, their corresponding software devices are encapsulated in the SDK), so that it can be more conveniently deployed on a third-party platform or each manufacturer's own software platform, making it easier to deploy and maintain.

[0099] for example, Figure 3In the figure, the "process product" and "user" shown by the dotted rectangle can be two different dimensional interfaces provided by the SDK. The "process product" interface is used to connect with the collection equipment or sampling personnel, and the "user" is used to connect with the front end (the front end can interact with the corresponding user based on various user interfaces such as web page controls). The above mainly introduces the execution logic inside the SDK, and the following will combine Figure 4 , the entire application scenario is briefly introduced.

[0100] First, the SDK can receive product parameters from the acquisition device, and then compare the product parameters with the threshold range. If the product parameters fall within the corresponding threshold range, the estimated abnormal cause of the abnormal parameter is determined based on the abnormal cause estimation model. Then, the product parameters and the estimated abnormal cause are presented to the user as a task (that is, Figure 4 The "abnormal parameters, abnormal reasons" shown in the dotted rectangle in the middle) can also provide the responsible person of the process stage corresponding to the abnormal reason through email, text message, phone call, etc. At this time, the task is in the waiting state.

[0101] In industrial applications, the staff can be roughly divided into analysts responsible for analyzing the causes of abnormalities, troubleshooters responsible for actual troubleshooting and error repair in the working environment, and acceptance personnel responsible for checking whether the errors have been repaired. When the task is established and is in the waiting state, the analyst can take the analysis task and perform the analysis. If there are obvious problems with the estimated abnormal cause, or the analyst has learned that this batch of products will have deviations in certain parameters, the actual abnormal cause can be indicated, thereby replacing the estimated abnormal cause generated by the model in the task (at the same time, the actual abnormal cause can be returned to the model to perform incremental training on the model).

[0102] Then, the task can be converted to the pending troubleshooting state; the troubleshooter can receive the troubleshooting task and conduct troubleshooting to determine whether the analyst's analysis is incorrect. If it is incorrect, the task can be returned to the analyst (at this time, the analyst can see that the task has been returned in his own user interface), and then the analyst re-provides the actual abnormal reason. The re-provided actual abnormal reason can replace the original actual abnormal reason and return to the model again to perform incremental training on the model, and so on. If the troubleshooter believes that the analysis is correct, the error can be fixed. If it can be fixed, then after the repair, the task will be converted to the pending verification state. The task in the pending verification state can be received and accepted by the acceptance personnel to determine whether the task has indeed been repaired (if the acceptance fails, the task can be converted back to the pending troubleshooting state or the pending analysis state). If the investigation finds that the task is temporarily unrepairable, the anomaly can be monitored (the task is switched to the monitored state) to prevent the anomaly from causing serious economic losses or operational accidents. Alternatively, the task can be returned to the analyst to determine whether there are other causes of the anomaly. Alternatively, the task can be re-investigated to determine whether the analyst's analysis is correct and whether the previous investigation is correct.

[0103] It can be seen that the implementation method provided by the present disclosure can be adapted to different industrial production lines and production batches, and can also be adapted to different production departments, so as to adapt to the current industrial production model to the greatest extent.

[0104] Corresponding to the embodiments of the aforementioned method, the present disclosure also provides embodiments of a device and a terminal to which the device is applied.

[0105] A second aspect of the present disclosure provides an anomaly detection device (for example, which can be deployed in a software development kit), see Figure 5 , the device comprises:

[0106] The parameter receiving module 501 is used to receive product parameters of the product;

[0107] A cause estimation module 502 is used to determine, when the state of the product parameter is abnormal, an estimated abnormal cause corresponding to the product parameter based on an abnormal cause estimation model, and present the product parameter and the estimated abnormal cause, wherein the state includes normal and abnormal;

[0108] The incremental training module 503 is used to perform incremental training on the abnormality cause estimation model based on the product parameters and the actual abnormality cause when receiving the actual abnormality cause indicated by the user for the product parameters.

[0109] Optionally, the device further includes an abnormality judgment module, configured to perform the following steps:

[0110] If the product parameter is within a threshold range, the state of the product parameter is determined to be abnormal, wherein the abnormal cause estimation model is used to determine the threshold range.

[0111] Optionally, the device further comprises:

[0112] The threshold updating module is used to update the current threshold range based on the output result of the abnormal cause estimation model after incremental training each time based on the product parameters and the actual abnormal cause.

[0113] Optionally, when the incremental training module is used to perform incremental training on the abnormal cause estimation model based on the product parameters and the actual abnormal cause, it is specifically used to:

[0114] The state and the actual abnormal cause indicated by the user for the product parameter are received, and the state and the actual abnormal cause are used as labels corresponding to the product parameter and input into the abnormal cause estimation model to perform incremental training on the abnormal cause estimation model.

[0115] Optionally, the device further comprises:

[0116] The solution estimation module is used to present a solution to the estimated abnormal cause based on the estimated abnormal cause corresponding to the product parameter and the preset knowledge base corresponding to the product.

[0117] Optionally, the abnormal cause estimation model includes a large language model;

[0118] The solution estimation module is used to present a solution to the estimated abnormal cause according to the estimated abnormal cause corresponding to the product parameter and the preset knowledge base corresponding to the product, and is specifically used to:

[0119] The text output by the abnormal cause estimation model is input into a decision tree model so that the decision tree model presents a solution to the estimated abnormal cause according to a preset knowledge base corresponding to the product, wherein the text is used to indicate the estimated abnormal cause.

[0120] Optionally, when presenting the product parameters and the estimated abnormality causes, the cause estimation module is specifically used to:

[0121] Determining the process stage corresponding to the estimated abnormal cause;

[0122] The abnormal parameters and the estimated abnormal causes are presented to the account or terminal corresponding to the process stage.

[0123] Optionally, the device further includes an abnormality rate analysis module, configured to perform the following steps:

[0124] Record the process stage corresponding to the estimated abnormal cause;

[0125] When receiving the actual abnormal reason indicated by the user for the product parameter, updating the recorded process stage according to the process stage corresponding to the actual abnormal reason;

[0126] According to the correspondence between the recorded abnormal causes and the process stages, the abnormality rate of each process stage in the process corresponding to the product is presented.

[0127] The implementation process of the functions and effects of each module in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, which will not be repeated here.

[0128] A third aspect of the present disclosure provides a computer program product, including a computer program / instruction, which implements the method described in the first aspect when executed by a processor.

[0129] For the device embodiments and computer program product embodiments, since they basically correspond to the method embodiments, the relevant parts can refer to the partial description of the method embodiments. In addition, the device embodiments described above are merely schematic, wherein the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, that is, they may be located in one place, or they may be distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the disclosed solution. Those of ordinary skill in the art can understand and implement it without paying creative labor.

[0130] In a fourth aspect, the embodiments of the abnormality detection device provided by the present disclosure can be applied to electronic devices. Figure 6 , which exemplarily shows a hardware schematic diagram of an electronic device. For example, the device 600 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, etc.

[0131] Device 600 may include one or more of the following components: a processing component 601 , a memory 602 , a power component 603 , a multimedia component 604 , an audio component 605 , an input / output (I / O) interface 606 , a sensor component 607 , and a communication component 608 .

[0132] The processing component 601 generally controls the overall operation of the device 600, such as operations associated with display, phone calls, data communications, camera operations, and recording operations. The processing component 601 may include one or more processors 609 to execute instructions to complete all or part of the steps of the above-mentioned method. In addition, the processing component 601 may include one or more modules to facilitate the interaction between the processing component 601 and other components. For example, the processing component 601 may include a multimedia module to facilitate the interaction between the multimedia component 604 and the processing component 601.

[0133] The memory 602 is configured to store various types of data to support operations on the device 600. Examples of such data include instructions for any application or method operating on the device 600, contact data, phone book data, messages, pictures, videos, etc. The memory 602 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0134] The power supply component 603 provides power to the various components of the device 600. The power supply component 603 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device 600.

[0135] The multimedia component 604 includes a screen providing an output interface between the device 600 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundaries of the touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 604 includes a front camera and / or a rear camera. When the device 600 is in an operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera may receive external multimedia data. Each front camera and the rear camera may be a fixed optical lens system or have a focal length and optical zoom capability.

[0136] The audio component 605 is configured to output and / or input audio signals. For example, the audio component 605 includes a microphone (MIC), and when the device 600 is in an operating mode, such as a call mode, a recording mode, and a speech recognition mode, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in the memory 602 or sent via the communication component 608. In some embodiments, the audio component 605 also includes a speaker for outputting audio signals.

[0137] The input / output (I / O) interface 606 provides an interface between the processing component 601 and the peripheral interface modules, which may be keyboards, click wheels, buttons, etc. These buttons may include but are not limited to: a home button, a volume button, a start button, and a lock button.

[0138] The sensor assembly 607 includes one or more sensors for providing various aspects of status assessment for the device 600. For example, the sensor assembly 607 can detect the open / closed state of the device 600, the relative positioning of components, such as the display and keypad of the device 600, and the sensor assembly 607 can also detect the position change of the device 600 or a component of the device 600, the presence or absence of user contact with the device 600, the orientation or acceleration / deceleration of the device 600, and the temperature change of the device 600. The sensor assembly 607 can also include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 607 can also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 607 can also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0139] The communication component 608 is configured to facilitate wired or wireless communication between the device 600 and other devices. The device 600 can access a wireless network based on a communication standard, such as WiFi, 2G or 3G, 4G or 5G or a combination thereof. In an exemplary embodiment, the communication component 608 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 608 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0140] In an exemplary embodiment, the device 600 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-mentioned anomaly detection method for electronic devices.

[0141] In a fifth aspect, the present disclosure, in an exemplary embodiment, further provides a non-transitory computer-readable storage medium including instructions, such as a memory 602 including instructions, and the instructions can be executed by a processor 609 of a device 600 to complete the above-mentioned abnormality detection method of an electronic device. For example, the non-transitory computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like.

[0142] The above describes specific embodiments of the present disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0143] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the inventions claimed herein. The present disclosure is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not claimed in the present disclosure. The description and examples are to be considered exemplary only, and the true scope and spirit of the present disclosure are indicated by the following claims.

[0144] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

[0145] The above description is only a preferred embodiment of the present disclosure and is not intended to limit the present disclosure. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present disclosure shall be included in the scope of protection of the present disclosure.

Claims

1. An anomaly detection method, characterized in that: The method comprises: Receive product parameters of the product; In the case where the state of the product parameter is abnormal, based on the abnormal cause estimation model, determining the estimated abnormal cause corresponding to the product parameter, and presenting the product parameter and the estimated abnormal cause, wherein the state includes normal and abnormal; When an actual abnormality cause indicated by a user for the product parameter is received, incremental training is performed on the abnormality cause estimation model based on the product parameter and the actual abnormality cause.

2. The anomaly detection method according to claim 1, characterized in that: The method further comprises: If the product parameter is within a threshold range, the state of the product parameter is determined to be abnormal, wherein the abnormal cause estimation model is used to determine the threshold range.

3. The anomaly detection method according to claim 2, characterized in that: The method further comprises: After the abnormality cause estimation model is incrementally trained based on the product parameters and the actual abnormality cause each time, the threshold range is updated based on the output result of the abnormality cause estimation model after the incremental training.

4. The anomaly detection method according to claim 1, characterized in that: The incrementally training the abnormality cause estimation model based on the product parameters and the actual abnormality cause comprises: The state and the actual abnormal cause indicated by the user for the product parameter are received, and the state and the actual abnormal cause are used as labels corresponding to the product parameter and input into the abnormal cause estimation model to perform incremental training on the abnormal cause estimation model.

5. The anomaly detection method according to claim 1, characterized in that: The method further comprises: Based on the estimated abnormal cause corresponding to the product parameter and the preset knowledge base corresponding to the product, a solution to the estimated abnormal cause is presented.

6. The anomaly detection method according to claim 5, characterized in that: The abnormal cause estimation model includes a large language model; The presenting of a solution to the estimated abnormal cause according to the estimated abnormal cause corresponding to the product parameter and the preset knowledge base corresponding to the product includes: The text output by the abnormal cause estimation model is input into a decision tree model so that the decision tree model presents a solution to the estimated abnormal cause according to a preset knowledge base corresponding to the product, wherein the text is used to indicate the estimated abnormal cause.

7. The abnormality detection method according to claim 1, characterized in that: The presenting of the product parameters and the estimated abnormality reasons includes: Determining the process stage corresponding to the estimated abnormal cause; The abnormal parameters and the estimated abnormal causes are presented to the account or terminal corresponding to the process stage.

8. The anomaly detection method according to claim 1, characterized in that: The method further comprises: Record the process stage corresponding to the estimated abnormal cause; When receiving the actual abnormal reason indicated by the user for the product parameter, updating the recorded process stage according to the process stage corresponding to the actual abnormal reason; According to the correspondence between the recorded abnormal causes and the process stages, the abnormality rate of each process stage in the process corresponding to the product is presented.

9. An abnormality detection device, characterized in that: The device comprises: A parameter receiving module, used for receiving product parameters of a product; a cause estimation module, configured to determine, when the state of the product parameter is abnormal, an estimated abnormal cause corresponding to the product parameter based on an abnormal cause estimation model, and present the product parameter and the estimated abnormal cause, wherein the state includes normal and abnormal; The incremental training module is used to perform incremental training on the abnormality cause estimation model based on the product parameters and the actual abnormality cause when receiving the actual abnormality cause indicated by the user for the product parameters.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the method according to any one of claims 1 to 8 is implemented.

11. An electronic device, characterized in that: The method comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method according to any one of claims 1 to 8 is implemented.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images