Jig life prediction method, life classification model training method and related equipment

Abstract

The invention relates to the technical field of intelligent manufacturing, and provides a jig service life prediction method, a service life classification model training method and related equipment, and the method comprises the steps: obtaining the service life information of a jig; if the service life information does not comprise the service life classification identifier, acquiring a first processing parameter when the jig currently passes through all processing equipment; and predicting a life classification identifier of the jig based on the first processing parameter and the life classification model. The service life classification identification of the jig is predicted through the service life classification model, and the service life of the jig in normal use and the service life of the jig in abnormal use are classified, so that the accuracy of subsequent service life prediction is improved. Besides, the classification result of the life classification model can reflect whether the jig is used normally or not, manual abnormality judgment is not needed, and therefore the overall life prediction time can be shortened, and manpower can be saved.

Description

Technical Field

[0001] This application relates to the field of intelligent manufacturing technology, and particularly to a method for predicting the life of a fixture, a method for training a life classification model, and related devices. Background Art

[0002] In the manufacturing process based on the low-injection-pressure overmolding process, liquid silicone rubber (LSR) in the LSR fixture is usually used to clamp products and perform glue injection processing.

[0003] However, during multiple uses of the LSR in the LSR fixture, it may be damaged or deformed due to various reasons (such as wear, collision, etc.). These damages or deformations may affect the life of the LSR, thereby affecting the subsequent clamping effect of the product, and even causing damage to the product itself. Therefore, after using the LSR fixture to clamp the product and inject glue, it is usually necessary to predict the life of the LSR fixture in a timely manner to discover potential performance degradation or failure risks, so as to take preventive measures to avoid product defects or production interruptions caused by fixture problems during the production process.

[0004] In the related art, after processing with the LSR fixture, an automatic optical inspection (AOI) device is used to perform an appearance inspection on the LSR in the LSR fixture. If the AOI device determines that the LSR fixture is abnormal, it prompts an operator to rejudge the LSR fixture to determine the remaining life of the LSR fixture. If the manual rejudgment is abnormal, the LSR fixture is replaced. This process requires high experience of the rejudgment personnel, consumes a large amount of manpower, takes a long time, and has low accuracy. Summary of the Invention

[0005] In view of this, this application provides a method for predicting the life of a fixture, a method for training a life classification model, and related devices, so as to solve the problems in the related art of consuming a large amount of manpower, taking a long time, and having low accuracy.

[0006] In the first aspect of the embodiments of this application, a method for predicting the life of a fixture is provided. The fixture circulates through at least one processing device. The method for predicting the life of the fixture parts includes: obtaining the life information of the fixture; if the life information does not include a life classification identifier, obtaining the first processing parameters when the fixture currently circulates through all the processing devices; and predicting the life classification identifier of the fixture based on the first processing parameters and the life classification model.

[0007] In some embodiments, after predicting the life classification identifier of the fixture based on the first processing parameter and the life classification model, the method further includes: if the type of the life classification identifier is an abnormal life identifier, inputting the first processing parameter into a life prediction model to output a first remaining life.

[0008] In some embodiments, the life information includes the target identity identifier of the fixture, and the method further includes: obtaining the current remaining life of the fixture according to the preset mapping relationship between the target identity identifier and the remaining life, where the preset mapping relationship is used to represent the mapping relationship between the identity identifier and the remaining life; updating the current remaining life according to the first remaining life.

[0009] In some embodiments, the life information includes the target identity identifier of the fixture. After predicting the life classification identifier of the fixture based on the first processing parameter and the life classification model, the method further includes: if the type of the life classification identifier is a normal life identifier, obtaining the current remaining life of the fixture according to the preset mapping relationship between the target identity identifier and the remaining life, where the preset mapping relationship is used to represent the mapping relationship between the identity identifier and the remaining life; updating the current remaining life.

[0010] In some embodiments, after predicting the life classification identifier of the fixture based on the first processing parameter and the life classification model, the method further includes: updating the life information according to the predicted life classification identifier.

[0011] In some embodiments, the life information includes the target identity identifier of the fixture, and the method further includes: if the life information includes a life classification identifier and the type of the life classification identifier is an abnormal life identifier, obtaining the current remaining life of the fixture according to the preset mapping relationship between the target identity identifier and the remaining life, where the preset mapping relationship is used to represent the mapping relationship between the identity identifier and the remaining life; updating the current remaining life to obtain an updated remaining life.

[0012] In some embodiments, the method further includes: if the updated remaining life is less than or equal to a preset deactivation number, outputting a prompt message to prompt replacement of the fixture.

[0013] In some embodiments, the lifespan information includes the target identity identifier of the fixture, and the method further includes: if the lifespan information includes a lifespan classification identifier and the type of the lifespan classification identifier is a normal lifespan identifier, obtaining second processing parameters of the fixture currently flowing through all the processing devices; inputting the second processing parameters into the lifespan classification model, and using the lifespan classification model to predict a new lifespan classification identifier of the fixture; if the type of the new lifespan classification identifier is a normal lifespan identifier, obtaining the current remaining lifespan of the fixture according to the target identity identifier and a preset mapping relationship, where the preset mapping relationship is used to represent the mapping relationship between the identity identifier and the remaining lifespan; and updating the current remaining lifespan.

[0014] In some embodiments, the method further includes: if the type of the new lifespan classification identifier is an abnormal lifespan identifier, inputting the second processing parameters into a lifespan prediction model to obtain a second remaining lifespan; updating the current remaining lifespan of the fixture according to the second remaining lifespan; and updating the lifespan classification identifier of the fixture according to the abnormal lifespan identifier.

[0015] In some embodiments, the method further includes: if the second remaining lifespan is less than or equal to a preset deactivation number of times, outputting a prompt message to prompt replacement of the fixture.

[0016] A second aspect of the embodiments of the present application provides a method for training a lifespan classification model of a fixture, and the method includes: obtaining a training data set, where the training data set includes a plurality of sample data, and each sample data includes sample processing parameters corresponding to at least one processing device through which a sample fixture flows during a single processing and a lifespan classification label corresponding to the single processing; performing feature processing on the sample processing parameters and the lifespan classification label in each sample data to obtain a sample parameter feature set and a label feature corresponding to each sample data; and training an initial classification model based on all the sample parameter feature sets and the label features corresponding to each sample parameter feature set to obtain the lifespan classification model.

[0017] In some embodiments, the feature processing of the sample processing parameters and the life classification labels in each sample data to obtain the sample parameter feature set and the label feature corresponding to each sample data includes: performing feature encoding on the life classification labels in each sample data to obtain corresponding label features; performing feature encoding on the sample processing parameters in each sample data to obtain initial parameter features corresponding to each sample processing parameter; obtaining all the initial parameter features corresponding to each type of sample processing parameter; calculating the feature average value and the feature standard deviation corresponding to each type of sample processing parameter based on the feature values of all the initial parameter features; performing parameter normalization processing on the initial parameter features corresponding to each type of sample processing parameter based on the feature average value and the feature standard deviation to obtain the sample parameter features corresponding to each type of sample processing parameter; and obtaining the sample parameter feature set corresponding to each sample data according to the sample parameter features corresponding to each type of sample processing parameter.

[0018] A third aspect of the embodiments of the present application provides an electronic device, including a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor. When the processor executes the computer-readable instructions, the life prediction method of the above-mentioned jig or the life classification model training method of the above-mentioned jig is implemented.

[0019] A fourth aspect of the embodiments of the present application provides a computer-readable storage medium storing computer-readable instructions, and when the computer-readable instructions are executed by a processor, the life prediction method of the above-mentioned jig parts or the life classification model training method of the above-mentioned jig is implemented.

[0020] The embodiments of the present application provide a life prediction method for a jig. Considering that the life of the jig during normal use may be different from the life of the jig during abnormal use due to faults or other reasons, the present application predicts the life classification identifier of the jig through a life classification model, realizes the classification of the life of the jig during normal use and the life of the jig during abnormal use, thereby helping to improve the accuracy of subsequent life prediction. In addition, the classification result of the life classification model can reflect whether the jig is used normally, without manual judgment of abnormalities, thus saving the overall life prediction time and manpower. Description of the Drawings

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for the description of the embodiments of the present application will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0022] Figure 1 It is a device diagram of the fixture life prediction method and the life classification model training method provided by the embodiments of the present application.

[0023] Figure 2 It is a flowchart of the implementation of the fixture life prediction method provided by the first embodiment of the present application.

[0024] Figure 3 It is a flowchart of the implementation of the fixture life prediction method provided by the second embodiment of the present application.

[0025] Figure 4 It is a flowchart of the implementation of the fixture life prediction method provided by the third embodiment of the present application.

[0026] Figure 5 It is a flowchart of the implementation of the fixture life classification model training method provided by the embodiments of the present application.

[0027] Figure 6 It is an algorithm model example diagram of the life classification model provided by the embodiments of the present application.

[0028] Figure 7 It is a model update example diagram of the life classification model provided by the embodiments of the present application.

[0029] Figure 8 It is another model update example diagram of the life classification model provided by the embodiments of the present application.

[0030] Figure 9 It is yet another model update example diagram of the life classification model provided by the embodiments of the present application.

[0031] Figure 10 It is a schematic structural diagram of the fixture life prediction device provided by the embodiments of the present application.

[0032] Figure 11 It is a schematic structural diagram of the fixture life classification model training device provided by the embodiments of the present application. Detailed implementation manners

[0033] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0034] Hereinafter, the terms "first" and "second" are for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the embodiments of the present application, words such as "exemplary" or "for example" are used to mean for example, illustration or explanation. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be construed as being more preferred or having more advantages than other embodiments or designs. Rather, the use of words such as "exemplary" or "for example" is intended to present relevant concepts in a specific manner.

[0035] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which this application belongs. The terms used in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application. It should be understood that unless otherwise specified in this application, " / " means "or". For example, A / B may mean A or B. The "and / or" in this application is merely a description of the relationship between associated objects, indicating that three relationships may exist. For example, A and / or B may mean: A exists alone, A and B exist simultaneously, and B exists alone, these three cases. "At least one" means one or more. "Multiple" means two or more than two. For example, at least one of a, b, or c may mean: a, b, c, a and b, a and c, b and c, a, b, and c, these seven cases.

[0036] During the manufacturing process based on the low-injection-pressure overmolding process, the LSR fixture is designed to accurately position and fix the products to be processed (such as the screen and the metal frame) to facilitate the uniform injection of glue between the screen and the metal frame. When the LSR fixture is applied in the processing of the low-injection-pressure overmolding process, during one processing cycle of the LSR fixture, the workstations through which the LSR fixture flows include, but are not limited to, the dimension inspection workstation, the appearance inspection workstation, the clamping workstation, the glue filling workstation, the visible light curing workstation, and the demolding workstation, etc. Specifically, the LSR fixture flows to the dimension inspection workstation, and the dimension inspection workstation performs dimension inspections (such as length, width, height, depth, etc.) on the upper mold, the lower mold, the LSR, and the molding cavity in the LSR of the LSR fixture to ensure the accuracy of the LSR fixture. After the dimension inspection of the LSR fixture, the appearance inspection is carried out on the LSR fixture (such as inspecting whether the LSR fixture is deformed, caked, soiled, worn, etc.). If the LSR in the LSR fixture is deformed or worn, it will affect the service life of the LSR fixture, as well as the clamping effect of the subsequent products, and even cause damage to the products to be processed themselves. At the clamping workstation, the products to be processed are clamped to the LSR fixture. At the glue filling workstation, glue is poured along the LSR fixture through the glue filling equipment. At the visible light curing workstation, the glue in the LSR fixture is cured by irradiating with an ultraviolet (UV) light source. After the glue is cured, the upper mold and the lower mold of the LSR fixture are opened, and the processed products are pulled out from the LSR fixture through the demolding equipment at the demolding workstation. Thus, the LSR fixture completes one processing process. Before the LSR fixture is used again for the next round, the appearance inspection workstation performs an appearance inspection on the LSR fixture to determine whether the LSR fixture can continue to be used.

[0037] In the related art, an automatic optical inspection (AOI) device is used to perform an appearance inspection on the LSR fixture. During the appearance inspection using the AOI device, there are usually a large number of LSR fixtures that are over-killed. To avoid excessive waste of LSR fixtures, if the AOI device determines that the LSR fixture is abnormal, it prompts the operator to perform a re-judgment on the LSR fixture to determine the remaining service life of the LSR fixture. If the manual re-judgment is abnormal, the LSR fixture is replaced. However, this process requires high experience of the re-judgment personnel, consumes a large amount of manpower, takes a long time, and has low accuracy.

[0038] To solve the above problems, an embodiment of the present application provides a method for predicting the life of a jig. By using a life classification model to predict the life classification identifier of the jig and determine the life state of the jig, it is possible to more accurately determine whether there are abnormal conditions in the jig, such as wear and deformation that cause damage to the life, and then facilitate a more accurate prediction of the remaining life of the jig. In addition, using this method can avoid over-killing a large number of LSR jigs, thus eliminating the need for manual judgment of abnormalities, saving the overall life prediction time and manpower.

[0039] Please refer to Figure 1 as shown, which is a device diagram corresponding to a method for predicting the life of a jig and a method for training a life classification model provided by an embodiment of the present application. As Figure 1 shown, the electronic device 100 includes a memory 101, at least one processor 102, at least one communication bus 103, at least one network interface 104, and other user interfaces 105. Among them, the processor 102 is used to implement the method for predicting the life of a jig and the method for training a life classification model when executing the computer program stored in the memory 101. At least one communication bus 103 is set to realize the connection and communication between the memory 101 and at least one processor 102, etc. The network interface 104 may include a Wireless Fidelity (Wi-Fi) interface, a 5G interface, and other communication interfaces, and the user interface 105 may include a USB interface and other standard interfaces.

[0040] In some embodiments, a computer program is stored in the memory 101. When the computer program is executed by at least one processor 102, all or part of the steps in the method for predicting the life of a jig and the method for training a life classification model as described above are implemented. Exemplarily, the computer program may be divided into one or more modules / units. The one or more modules / units are stored in the memory 101 and executed by the processor 102 to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, and these instruction segments are used to describe the execution process of the computer program in the electronic device.

[0041] In some embodiments, the memory 101 includes a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electrically-erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM), or other optical disc memories, magnetic disk memories, tape memories, or any other computer-readable medium that can be used to carry or store data.

[0042] In some embodiments, at least one processor 102 is the control core (Control Unit) of the electronic device 100, connecting various components of the entire electronic device 100 through various interfaces and lines. By running or executing programs or modules stored in the memory 101, and by invoking data stored in the memory 101, it performs various functions of the electronic device 100 and processes data. For example, when at least one processor 102 executes the computer program stored in the memory, it implements all or part of the steps of the fixture life prediction method and the life classification model training method in the embodiments of the present application; or implements all or part of the functions of the alarm index processing device. At least one processor 102 can be composed of integrated circuits. For example, it can be composed of a single packaged integrated circuit, or can be composed of multiple integrated circuits with the same or different functions packaged together, including a combination of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips, etc.

[0043] It can be understood that the structure illustrated in the embodiments of the present application does not constitute a specific limitation on the electronic device 100. In other embodiments of the present application, the electronic device 100 may include more or fewer components than shown in the figure, or combine certain components, or split certain components, or have different component arrangements. The illustrated components can be implemented in hardware, software, or a combination of software and hardware.

[0044] In this embodiment, the electronic device includes, but is not limited to, one or more of devices such as a computer, a server, and a Programmable Logic Controller (PLC). Other existing or future electronic products that can be adapted to this application should also be included within the protection scope of this application and are hereby incorporated by reference.

[0045] In some embodiments, the electronic device can be communicatively connected to the processing devices passed through by the LSR fixture in each workstation to control the processing devices and obtain the processing parameters of each workstation.

[0046] In this embodiment, the fixture can be an LSR fixture, and its form includes, but is not limited to, an upper mold, a lower mold, LSR, etc. The LSR can be in a frame shape, a strip shape, an arc shape, etc. It has a forming groove, and the forming groove cooperates with the product to form a forming cavity. The glue injection needle is inserted into the forming cavity for glue injection so that glue matching the forming cavity is formed on the product. The specific form of the LSR fixture in the embodiments of this application is not limited. Correspondingly, in the embodiments of this application, the life prediction of the LSR fixture by the electronic device specifically includes predicting the life of the LSR, and the replacement of the LSR fixture by the electronic device specifically includes replacing the LSR. For the convenience of description, the fixture is used hereinafter to take the LSR fixture as an example. Of course, in other embodiments, it can also be a fixture without LSR, such as a CNC fixture, an assembly fixture, etc.

[0047] In this embodiment, the processing device refers to the corresponding processing device of each processing workstation when the fixture passes through the processing workstations (such as a clamping workstation, a glue injection workstation, a visible light curing workstation, a demolding workstation, etc.) for product processing. In some embodiments, when using the LSR fixture to complete one processing, the workstations passed through by the LSR fixture include, but are not limited to, one or more of a dimension inspection workstation, a clamping workstation, a glue injection workstation, a visible light curing workstation, and a demolding workstation. For example, the LSR fixture usually undergoes a dimension inspection once after molding, and then the dimension of the LSR fixture is usually not inspected during the processing using the LSR fixture. In this case, during the first use of the LSR fixture for processing, the workstations passed through by the LSR fixture include a dimension inspection workstation, a clamping workstation, a glue injection workstation, a visible light curing workstation, and a demolding workstation. During the non - first use of the LSR fixture for processing, the workstations passed through by the LSR fixture include a clamping workstation, a glue injection workstation, a visible light curing workstation, and a demolding workstation, reducing the complex procedures of the dimension inspection workstation.

[0048] In this embodiment, after each use of the LSR fixture to complete one processing and before using the LSR fixture to enter the next round of processing, the life prediction method of the fixture provided by the embodiments of this application is used to predict the life of the LSR fixture.

[0049] Please refer to Figure 2 As shown, it is a flowchart of the implementation of the fixture life prediction method provided by the first example of this application. This method is applied to an electronic device. In the embodiments of this application, this method is described by taking the electronic device 100 in Figure 1 as an example. This method includes the following steps. S11: Obtain the life information of the fixture.

[0050] In some embodiments, the life information may include, but is not limited to, one or more of the identity identifier of the fixture, the used life (i.e., the number of times used), the remaining life, the total life, the life classification identifier, etc.

[0051] In some embodiments, the identity identifier of the fixture is the unique identifier of the fixture. The identity identifier can be set customarily. For example, the identity identifier of the fixture can be set according to the model or number of the fixture. Or, the identity identifier of the fixture can be set using a barcode or a QR code. Or, the identity identifier of the fixture can be set using a text description, a graphic identifier, etc. Or, the glass code of the glass plate adhered to the fixture can also be used as the identity identifier of the fixture. The embodiments of this application do not limit the specific setting form of the identity identifier.

[0052] In some embodiments, the life of the fixture can be represented by the number of times the fixture is used. The used life represents the number of times the fixture has been used. For example, if the fixture has been used 5 times, then the used life of the fixture is 5 times. The total life of the fixture represents the maximum number of times the fixture can be used. For example, if the maximum number of times the fixture can be used is 35 times, then the life of the fixture is 35 times. The remaining life of the fixture represents the remaining number of times the fixture can be used. For example, if the life of the fixture is 35 times, it has been used 30 times, and the remaining number of times it can be used is 5 times, then the remaining life of the fixture is 5 times.

[0053] In some embodiments, each time the fixture is used to complete a processing process, the used number of the fixture is incremented by 1, and the electronic device decrements the remaining service life of the fixture by 1. When the remaining service life of the fixture is less than or equal to the preset deactivation number of times, the electronic device can prompt to replace the fixture. Among them, the preset deactivation number of times can be set customarily. For example, the preset deactivation number of times can be set to 0, 1, 2, etc.

[0054] In some embodiments, after the fixture is used beyond the preset number of times, it may be impossible to effectively determine whether there are any abnormalities in the fixture (such as deformation, wear, etc.). In other words, after the fixture is used beyond the preset number of times, whether there are any abnormalities in the fixture may not be detected through visual inspection, and problems may occur during processing, which may lead to damage to the fixture. In this case, if no abnormalities are found in the fixture through visual inspection and the fixture is continued to be used for processing, the probability of defective products processed by the fixture is likely to increase. To solve the above problems, the present application sets the upper limit of the service life of the fixture to the preset number of times.

[0055] In some embodiments, the preset number of times can be custom-set according to experience. For example, the preset number of times can be set to 35 times. The specific setting of the preset number of times in the embodiments of the present application is not limited. For example, the preset number of times can also be set to 10 times, 20 times, 50 times, 60 times, 100 times, etc., which is mainly determined by factors such as the LSR material and the potting pressure during the processing and the pulling force of the product during demolding.

[0056] In some embodiments, the total life of the fixture can be divided into two categories, namely the total life equal to the preset number of times and the total life less than the preset number of times. The total life equal to the preset number of times means that during the normal use of the fixture throughout its life cycle, the loss of the fixture is within the preset error range, and the influence of the processing environment, etc. on the fixture is small, thus not affecting the total life of the fixture. The preset error range can be custom-set. The total life less than the preset number of times means that there are abnormalities during the entire life cycle of the fixture, the loss of the fixture is outside the preset error range, the fixture has large losses and suffers from wear, deformation, etc., or the processing environment has a certain or large impact on the fixture, resulting in a reduction in the total life of the fixture.

[0057] In some embodiments, the life classification identifier is used to indicate the type of the total life of the fixture. The types included in the life classification identifier can include the normal life classification identifier and the abnormal life classification identifier. When the life classification identifier of the fixture is the normal life classification identifier, it means that the total life of the fixture is equal to the preset number of times. When the life classification identifier of the fixture is the abnormal life classification identifier, it means that the total life of the fixture is less than the preset number of times.

[0058] In some embodiments, during the process of obtaining the life information of the fixture, the electronic device will request to obtain the life classification identifier of the fixture, so as to determine the type of the total life of the fixture according to the life classification identifier, and thus more accurately predict the remaining life of the fixture according to the type of the total life of the fixture.

[0059] However, since in the case of first using the fixture for processing, the electronic device cannot determine the type of the total life of the fixture, the electronic device cannot obtain the life classification identifier of the fixture. In this case, there is also no life classification identifier in the life information of the fixture.

[0060] In the case of processing with the fixture for non-first time, the electronic device can predict the type of the total life of the fixture based on the processing parameters of the previous processing of the fixture, so as to determine the life classification identifier of the fixture. In this case, during the process of obtaining the fixture life information, the electronic device can match the corresponding life classification identifier according to the identity identifier of the fixture, so that the life classification identifier is included in the fixture life information.

[0061] In some embodiments, after the electronic device determines the life information such as the used life, remaining life, total life, and life classification identifier of the fixture, it can store the identity identifier of the fixture corresponding to the used life, remaining life, total life, and life classification identifier of the fixture. Therefore, when obtaining the life information of the fixture, according to the identity identifier of the fixture, the corresponding life information is matched and obtained.

[0062] S12: Determine whether the life information includes a life classification identifier.

[0063] S13: If the life information does not include a life classification identifier, obtain the first processing parameters when the fixture currently passes through all processing devices.

[0064] Specifically, if the life information does not include a life classification identifier, it indicates that the fixture is circulated through all processing devices for the first time and the life classification identifier has not been predicted by the life classification model.

[0065] In some embodiments, the first processing parameters can represent all the processing parameters when the fixture currently passes through all processing devices.

[0066] In some embodiments, after each use of the fixture to complete a processing and before using the fixture to enter the next round of processing, the electronic device will collect the processing parameters of this processing process and use the collected processing parameters to predict the life of the fixture.

[0067] In this embodiment, the first processing parameters when the fixture currently passes through all processing devices represent the processing parameters when the fixture passed through all processing devices during the last processing completed by using the fixture.

[0068] In some embodiments, all processing devices can include the processing devices in the corresponding workstations when the fixture passes through one or more workstations such as the size inspection workstation, clamping workstation, potting workstation, visible light curing workstation, and demolding workstation.

[0069] In some embodiments, the processing parameters when the fixture passes through the processing equipment at the dimensional inspection station may include dimensional parameters such as the length, width, height, angle, inner diameter, outer diameter, depth, etc. of key points or predefined points in the fixture. The processing parameters when the fixture passes through the processing equipment at the clamping station may include the assembly pressure, the number of times the fixture has been used, the identity identifier, the three-axis (X-axis, Y-axis, Z-axis) offset of the assembly, and the height detection value after assembly. The processing parameters when the fixture passes through the processing equipment at the glue filling station may include the glue filling vacuum pressure, the glue filling pressure, the glue filling time-sharing pressure (for example, the pressure of injecting glue at multiple time points), and the maximum glue filling pressure. The processing parameters when the fixture passes through the processing equipment at the demolding station may include the demolding force, the top blowing time, the top blowing pressure, the bottom blowing pressure, etc. The embodiments of the present application do not limit the processing parameters when the fixture passes through the processing equipment at any station.

[0070] S14: Based on the first processing parameter and the life classification model, predict the life classification identifier of the fixture.

[0071] In some embodiments, the life classification model may be a deep learning model, such as the XGBOOST model.

[0072] In some embodiments, after the electronic device obtains the first processing parameter, it may input the first processing parameter into the life classification model and use the life classification model to predict the life classification identifier of the fixture. Among them, the training method of the life classification model can refer to Figure 5 and the corresponding embodiment content.

[0073] Please refer to Figure 3 As shown, it is a flowchart of the implementation of the life prediction method for the fixture provided by the second embodiment of the present application. This method is applied to an electronic device. In the embodiments of the present application, this method is described by taking the electronic device 100 in Figure 1 as an example. This method includes the following steps. S21: Obtain the life information of the fixture.

[0074] S22: Determine whether the life information includes a life classification identifier.

[0075] S23: If the life information does not include a life classification identifier, obtain the first processing parameter when the fixture currently passes through all processing equipment.

[0076] S24: Based on the first processing parameter and the life classification model, predict the life classification identifier of the fixture.

[0077] In some embodiments, the specific implementation manners of steps S21 to S24 may refer to the specific embodiment content of steps S11 to S14.

[0078] S25: If the type of the life classification identifier is the normal life identifier, obtain the current remaining life of the jig according to the target identity identifier and the preset mapping relationship.

[0079] S26: Update the current remaining life.

[0080] In some embodiments, the preset mapping relationship is used to represent the mapping relationship between the identity identifier and the remaining life. Among them, the remaining life recorded in the preset mapping relationship represents the latest updated remaining life of the corresponding jig.

[0081] In the case where the life information does not include the life classification identifier, the electronic device predicts the life classification identifier of the jig based on the first processing parameter and the life classification model. If the type of the predicted life classification identifier is the normal life identifier, obtain the current remaining life of the jig according to the target identity identifier and the preset mapping relationship, and update the current remaining life.

[0082] In some embodiments, each update of the current remaining life includes decreasing the current remaining life by a preset decreasing quantity. The preset decreasing quantity can be 1, 2, etc. For example, the preset decreasing quantity is 1, and the current remaining life is 24. Updating the current remaining life is 24 - 1 = 23. That is, the updated current remaining life is 23, which means the remaining usage times of the jig is 23, and update this to the preset mapping relationship so that when the jig passes through the processing equipment in the next cycle, if it is a normal life identifier, it will be decreased based on 23.

[0083] In some embodiments of the present application, the life prediction method of the jig further includes the following steps. S27: If the type of the life classification identifier is the abnormal life identifier, input the first processing parameter into the life prediction model to output the first remaining life.

[0084] In some embodiments, the life prediction model can be a neural network model, such as a Long Short-Term Memory (LSTM).

[0085] In other embodiments, other models can also be selected for the life prediction model, and the type of the life prediction model is not limited in the embodiments of the present application.

[0086] In the case where the life information does not include a life classification identifier, the electronic device predicts the life classification identifier of the jig based on the first processing parameter and the life classification model. In this case, if the type of the predicted life classification identifier is an abnormal life identifier, it means that the total life of the jig is less than the preset number of times. However, since the life classification identifier of the jig cannot be obtained before predicting the life classification identifier of the jig based on the first processing parameter and the life classification model, the type of the total life of the jig cannot be determined either. In this case, if the life is directly decreased based on the current remaining life of the jig, it may lead to inaccurate calculation of the remaining life, and even may lead to processing with a damaged jig, thus affecting the product processing quality and processing efficiency.

[0087] For example, the electronic device presets the initial value of the remaining life of the jig as the preset number of times. When using the jig for the first processing, since the life information of the jig does not include a life classification identifier, the electronic device obtains the first processing parameter when the jig passes through all processing devices during the first processing, and predicts the life classification identifier of the jig based on the first processing parameter and the life classification model. At this time, in the case where the predicted life classification identifier of the jig is an abnormal life classification identifier each time, if the electronic device obtains the current remaining life, that is, the preset number of times, and obtains the remaining life after the first processing by decreasing the preset number of times by the preset decreasing quantity each time, then, in this way, the total life is still equal to the preset number of times. This is contrary to the conclusion that the predicted life classification identifier is an abnormal life classification identifier.

[0088] To solve the above problems, in the case where the type of the life classification identifier of the jig predicted based on the first processing parameter and the life classification model is an abnormal life identifier, the electronic device inputs the first processing parameter into the life prediction model, and uses the life prediction model to predict the remaining life of the jig. The electronic device defines the remaining life predicted based on the first processing parameter as the first remaining life. The first remaining life predicted in this way is more accurate.

[0089] In some embodiments of the present application, the life information includes the target identity identifier of the jig. The life prediction method of the jig further includes the following steps.

[0090] S28: Obtain the current remaining life of the jig according to the target identity identifier and the preset mapping relationship.

[0091] Specifically, the target identity identifier represents the identity identifier of the current jig. The preset mapping relationship is used to represent the mapping relationship between the identity identifier and the remaining life. Among them, the remaining life recorded in the preset mapping relationship represents the remaining life of the jig updated most recently.

[0092] If the lifespan information does not include a lifespan classification identifier, it indicates that the fixture is in its first circulation and use. At this time, the current remaining lifespan of the fixture in the preset mapping relationship is the initial value, a null value, or a preset number of times.

[0093] S29: Update the current remaining lifespan according to the first remaining lifespan.

[0094] Specifically, update the current remaining lifespan in the preset mapping relationship to the first remaining lifespan. For example, if the current remaining lifespan is a null value, directly replace the null value with the first remaining lifespan. In some embodiments of the present application, after predicting the lifespan classification identifier of the fixture based on the first processing parameter and the lifespan classification model, the lifespan prediction method of the fixture further includes the step of updating the lifespan information of the fixture according to the predicted lifespan classification identifier.

[0095] In some embodiments, if the electronic device predicts the lifespan classification identifier of the fixture, it can update the predicted lifespan classification identifier of the fixture to the lifespan information of the fixture according to the corresponding relationship between the target identity identifier of the fixture and the predicted lifespan classification identifier of the fixture.

[0096] Please refer to Figure 4 As shown, it is a flowchart of the implementation of the lifespan prediction method of the fixture provided by the third embodiment of the present application. This method is applied to an electronic device. In the embodiments of the present application, this method is described by taking the electronic device 100 in Figure 1 as an example. The method includes the following steps. S31: Obtain the lifespan information of the fixture.

[0097] S32: Determine whether the lifespan information includes a lifespan classification identifier.

[0098] S33: If the lifespan information does not include a lifespan classification identifier, obtain the first processing parameter when the fixture currently passes through all processing devices during circulation.

[0099] S34: Predict the lifespan classification identifier of the fixture based on the first processing parameter and the lifespan classification model.

[0100] In some embodiments, the specific implementation manners of steps S31 to S34 can refer to the specific embodiment contents of steps S11 to S14.

[0101] S35: If the type of the lifespan classification identifier is a normal lifespan identifier, obtain the current remaining lifespan of the fixture according to the target identity identifier and the preset mapping relationship.

[0102] S36: Update the current remaining lifespan.

[0103] S37: If the type of the lifespan classification identifier is an abnormal lifespan identifier, predict the lifespan of the fixture based on the first processing parameter using the lifespan prediction model to obtain the first remaining lifespan.

[0104] In some embodiments, the specific implementation manners of steps S35 to S37 may refer to the specific embodiment content of steps S25 to S27. S38: If the lifespan information includes a lifespan classification identifier, identify the type of the lifespan classification identifier.

[0105] In some embodiments, the types of the lifespan classification identifier include a normal lifespan identifier and an abnormal lifespan identifier. S381: If the lifespan information includes a lifespan classification identifier and the type of the lifespan classification identifier is an abnormal lifespan identifier, obtain the current remaining lifespan of the jig according to the target identity identifier and the preset mapping relationship.

[0106] S382: Update the current remaining lifespan to obtain the updated remaining lifespan.

[0107] In some embodiments, the preset mapping relationship is used to represent the mapping relationship between the identity identifier and the remaining lifespan. Among them, the remaining lifespan recorded in the preset mapping relationship represents the latest updated remaining lifespan of the corresponding jig.

[0108] In some embodiments, when it is determined that the type of the lifespan classification identifier of the jig is an abnormal lifespan identifier, or in other words, when the total lifespan of the jig is less than the preset number of times, if the jig is used for processing again, the electronic device can obtain the current remaining lifespan of the jig according to the target identity identifier and the preset mapping relationship, and update the current remaining lifespan to obtain the updated remaining lifespan.

[0109] In some embodiments of the present application, the method for predicting the lifespan of the jig further includes the step: If the updated remaining lifespan is less than or equal to the preset number of deactivation times, output a prompt message to prompt to replace the jig.

[0110] In some embodiments, the preset number of deactivation times can be set customarily. For example, it can be set to 5 times, 3 times, 2 times, 0 times, etc. The present application embodiment does not limit the specific setting of the preset number of deactivation times.

[0111] In some embodiments, that the updated remaining lifespan is less than or equal to the preset number of deactivation times means that the corresponding jig has approached or reached the limit of its service life, or has approached or completed the maximum number of times it can be used. This means that the performance, accuracy and reliability of the jig may drop significantly and cannot meet the normal use requirements. In this case, the electronic device can output a prompt message to prompt to prepare to replace the jig or replace the jig.

[0112] In some embodiments, the electronic device prompt information includes, but is not limited to, one or more of the following ways: text prompt, voice prompt, visual reminder, tactile reminder, etc. Optionally, the visual reminder includes a visual reminder display or a visual replacement reminder display. The visual reminder display includes a display when the remaining life is 1-5 times, so that the staff can make preparations for the fixture replacement according to the prompt; the visual replacement reminder display includes a display when the remaining life is 0 times, so that the staff can replace the fixture according to the prompt.

[0113] In some embodiments, when replacing the fixture, the information related to the fixture to be disassembled, such as the life information, can be deleted.

[0114] S383: If the life information includes a life classification identifier and the type of the life classification identifier is a normal life identifier, obtain the second processing parameter of the fixture currently flowing through all processing devices.

[0115] Specifically, the second processing parameter is the above-mentioned first processing parameter, that is, it represents the processing parameters when the fixture flowed through all processing devices during the last processing process completed using the fixture.

[0116] S384: Input the second processing parameter into the life classification model, and use the life classification model to predict the new life classification identifier of the fixture. In some embodiments, in some embodiments, when it has been determined that the type of the life classification identifier of the fixture is a normal life identifier, or in other words, when the total life of the fixture is equal to the preset number of times, if the fixture is used for processing again, the electronic device needs to re-predict the life classification identifier of the fixture based on the processing parameters of the re-processing using the life classification model to obtain a new life classification identifier. This is because the fixture may be damaged during the re-processing, resulting in the total life of the fixture being less than the preset number of times.

[0117] In some embodiments of the present application, the method for predicting the life of the fixture may further include the following steps.

[0118] S3841: If the type of the new life classification identifier is a normal life identifier, obtain the current remaining life of the fixture according to the target identity identifier and the preset mapping relationship.

[0119] S3842: Update the current remaining life.

[0120] In some embodiments, if the type of the new life classification identifier is still a normal life identifier, the electronic device can obtain the current remaining life of the fixture according to the target identity identifier and the preset mapping relationship and update the current remaining life.

[0121] In some embodiments, the preset mapping relationship is used to represent the mapping relationship between the identity identifier and the remaining life. Among them, the remaining life recorded in the preset mapping relationship represents the remaining life of the corresponding fixture updated most recently.

[0122] In some embodiments, updating the current remaining life includes updating the current remaining life according to preset conditions, and the preset conditions may be to decrease by 1 based on the current remaining life according to a preset decreasing quantity.

[0123] In some embodiments of the present application, the method for predicting the life of a fixture may further include the following steps.

[0124] S3843: If the type of the new life classification identifier is an abnormal life identifier, based on the second processing parameter, use the life prediction model to predict the life of the fixture and obtain the second remaining life.

[0125] S3844: Update the current remaining life of the fixture according to the second remaining life, and update the life classification identifier of the fixture according to the abnormal life identifier.

[0126] In some embodiments, if the type of the new life classification identifier is an abnormal life identifier, the electronic device may, based on the second processing parameter, use the life prediction model to predict the life of the fixture and obtain the second remaining life. The electronic device may update the current remaining life of the fixture to the second remaining life, and update the life classification identifier of the fixture to an abnormal life identifier.

[0127] In some embodiments of the present application, the method for predicting the life of a fixture may further include the following steps.

[0128] S3845: If the second remaining life is less than or equal to the preset deactivation times, output a prompt message to prompt to replace the fixture.

[0129] In some embodiments, the preset deactivation times can be custom-set. For example, it can be set to 3 times, 2 times, 0 times, etc. The specific setting of the preset deactivation times in the embodiments of the present application is not limited.

[0130] In some embodiments, the second remaining life being less than or equal to the preset deactivation times means that the fixture corresponding to the second remaining life has approached or reached the limit of its service life, or has approached or completed the maximum number of times it can be used. This means that the performance, accuracy, and reliability of the fixture may decrease significantly and cannot meet the normal usage requirements. In this case, the electronic device can output a prompt message to prompt to replace the fixture.

[0131] In some embodiments, the ways for the electronic device to prompt to replace the fixture include but are not limited to one or more of the ways such as text prompt, voice prompt, visual reminder, tactile reminder, etc.

[0132] In some embodiments, when replacing the fixture, information related to the fixture to be disassembled, such as lifespan information, can be deleted.

[0133] The embodiment of the present application provides a method for predicting the lifespan of a fixture. Considering that the lifespan of the fixture during normal use may be different from the lifespan of the fixture during abnormal use due to reasons such as malfunctions, the present application predicts the lifespan classification identifier of the fixture through a lifespan classification model, so as to classify the lifespan of the fixture during normal use and the lifespan of the fixture during abnormal use, thereby helping to improve the accuracy of subsequent lifespan prediction. In addition, the classification result of the lifespan classification model can reflect whether the fixture is used normally, without manual judgment of abnormalities, thus saving the overall lifespan prediction time and labor.

[0134] Please refer to Figure 5 As shown, it is a flowchart for implementing the method for training the lifespan classification model of the fixture provided by the embodiment of the present application. This method is applied to an electronic device. The embodiment of the present application takes the method applied to the Figure 1 electronic device 100 as an example for illustration. This method includes the following steps. S41: Obtain a training data set. The training data set includes a plurality of sample data. Each sample data includes the sample processing parameters corresponding to at least one processing device through which the sample fixture flows during a single processing process and the corresponding lifespan classification label for the single processing process.

[0135] In some embodiments, the lifespan classification label represents the corresponding lifespan classification identifier when the sample fixture completes a single processing process.

[0136] In some embodiments, the processing device may include the process of the sample fixture flowing through one or more of the size inspection station, clamping station, potting station, visible light curing station, and demolding station, and flowing through the processing devices in the corresponding stations.

[0137] In some embodiments, the sample processing parameters when the sample fixture flows through the processing equipment at the dimension inspection station may include dimension parameters such as the length, width, height, angle, inner diameter, outer diameter, depth, etc. of the key points or predefined points in the sample fixture. The sample processing parameters when the sample fixture flows through the processing equipment at the clamping station may include the assembly pressure, the number of times the sample fixture has been used, the identity identifier, the three-axis (X-axis, Y-axis, Z-axis) offset after assembly, and the height detection value after assembly. The sample processing parameters when the sample fixture flows through the processing equipment at the potting station may include the potting vacuum pressure, the potting pressure, the potting time-sharing pressure, and the maximum potting pressure. The sample processing parameters when the sample fixture flows through the processing equipment at the demolding station may include the demolding force, the top blowing time, the top blowing pressure, the bottom blowing pressure, etc. The embodiments of the present application do not limit the sample processing parameters when the sample fixture flows through the processing equipment at any station.

[0138] In some embodiments, if a life classification model is used to predict the life classification label of any type of fixture, then in the process of training the life classification model, the same type of sample fixture should also be used.

[0139] In some embodiments, the sample fixtures corresponding to different sample data may be the same sample fixture or different sample fixtures.

[0140] In some embodiments, the electronic device can obtain the training data set from a database (such as a relational database (PostgreSQL)) or a real-time data stream (such as kafka).

[0141] In some embodiments, if the total life of the sample fixture is equal to the preset number of times, the electronic device can label the life classification labels of the processing processes corresponding to the preset number of times of the sample fixture as normal classification identifiers. If the total life of the sample fixture is less than the preset number of times, the electronic device can label the processing process corresponding to the specified number of times of the sample fixture as an abnormal life identifier, and label the other processing processes except the processing process corresponding to the specified number of times as normal life identifiers.

[0142] Taking the preset number of times as 35 times and the specified number of times as the last 10 times as an example, the total life of the sample jig A is equal to 35 times, the total life of the sample jig B is 25 times, and the total life of the sample jig C is 15 times. The electronic device obtains the sample processing parameters of the 35 processing processes corresponding to the sample jig A, and labels the life classification label corresponding to each processing process as a normal life identifier. The electronic device obtains the sample processing parameters of the 25 processing processes corresponding to the sample jig B, labels the life classification label of the first 15 processing processes corresponding to the sample jig B as a normal life identifier, and labels the life classification label of the last 10 processing processes corresponding to the sample jig B as an abnormal life identifier. The electronic device obtains the sample processing parameters of the 25 processing processes corresponding to the sample jig C, labels the life classification label of the first 5 processing processes corresponding to the sample jig C as a normal life identifier, and labels the life classification label of the last 10 processing processes corresponding to the sample jig C as an abnormal life identifier.

[0143] In some embodiments, the sample data corresponding to the sample jig with a total life of 35 times can be used as the first type of sample data, and the sample data corresponding to the sample jig with a total life less than 35 times can be used as the second type of sample data. To ensure the prediction accuracy of the life classification model, the electronic device can obtain the first type of sample data and the second type of sample data simultaneously during the process of selecting sample data. However, since the number of the first type of sample data is usually much larger than that of the second type of sample data, it may cause the uneven distribution of the sample data. In this case, it may lead to the life classification model tending to learn the first type of sample data, with a more accurate prediction result for the first type of sample data and a distorted prediction result for the second type of sample data.

[0144] To solve the above problems, the electronic device can use the undersampling technology to balance the number of the two types of sample data.

[0145] S42: Perform feature processing on the sample processing parameters and life classification labels in each sample data to obtain the sample parameter feature set and label feature corresponding to each sample data.

[0146] In some embodiments, the ways of feature processing include but are not limited to feature encoding and parameter regularization processing.

[0147] In some embodiments of the present application, performing feature processing on the sample processing parameters and life classification labels in each sample data to obtain the sample parameter feature set and label feature corresponding to each sample data includes the following steps.

[0148] S421: Perform feature encoding on the life classification labels in each sample data to obtain the corresponding label features.

[0149] S422: Perform feature encoding on the sample processing parameters in each sample data to obtain the initial parameter features corresponding to each sample processing parameter.

[0150] S423: Obtain all the initial parameter features corresponding to each type of sample processing parameter.

[0151] S424: Based on the feature values of all the initial parameter features, calculate the feature average value and the feature standard deviation corresponding to each type of sample processing parameter.

[0152] S425: Based on the feature average value and the feature standard deviation, perform parameter regularization processing on the initial parameter features corresponding to each type of sample processing parameter to obtain the sample parameter features corresponding to each type of sample processing parameter.

[0153] S426: According to the sample parameter features corresponding to each type of sample processing parameter, obtain the sample parameter feature set corresponding to each sample data.

[0154] In some embodiments, the electronic device performs feature encoding on the life classification label and the sample processing parameters in each sample data, and converts the life classification label and the sample processing parameters into the format required by the initial classification model or the life classification model, such as a numerical type.

[0155] In some embodiments, the label feature and the initial parameter feature can be numerical features. The electronic device can convert the life classification label and the sample processing parameters into numerical features through feature encoding, which is convenient for mathematical calculations and comparisons during the training process of the life classification model.

[0156] In some embodiments, the ways of feature encoding include but are not limited to any one or more of encoding ways such as binary encoding, label encoding, target encoding, one-hot encoding, etc. Taking the electronic device using binary encoding to perform feature encoding on the life classification label as an example, if the life classification label is a normal life identifier, the corresponding label feature can be set to 0 after feature encoding. If the life classification label is an abnormal life identifier, the corresponding label feature can be set to 1 after feature encoding.

[0157] In some embodiments, if the scale differences of the initial parameter features corresponding to the sample processing parameters are very large (for example, the value range of a numerical feature is between 0 and 1, while the value range of another initial parameter feature is between 1000 and 10000), then during the training process of the life classification model, the model convergence speed may become slow because different initial parameter features contribute differently to the gradient update, and some initial parameter features may dominate the gradient update due to their large numerical ranges, thus masking the influence of other parameter features on the model, resulting in a decrease in the model accuracy and the model stability.

[0158] To solve the above problems, after the initial parameter features corresponding to each sample processing parameter, the electronic device can perform parameter normalization processing on the initial parameter features to perform feature scaling, scale the feature values of the initial parameter features to between 0 and 1, so that the feature values are scaled to a distribution with unit variance and zero mean.

[0159] In the process of implementing parameter normalization processing on the initial parameter features, the electronic device can obtain all the initial parameter features corresponding to each type of sample processing parameter, and calculate the feature average value and feature standard deviation corresponding to each type of sample processing parameter based on the feature values of all the initial parameter features. The electronic device can perform parameter normalization processing on the initial parameter features corresponding to each type of sample processing parameter based on the feature average value and feature standard deviation, and obtain the sample parameter features corresponding to each type of sample processing parameter.

[0160] In some embodiments, the electronic device can calculate the feature average value according to the following formula.

[0161] 。

[0162] In the formula, represents the feature average value; represents the i-th initial parameter feature corresponding to any type of sample processing parameter, and i takes an integer greater than 0; represents the sum of all the initial parameter features corresponding to any type of sample processing parameter; represents the number of all the initial parameter features corresponding to any type of sample processing parameter.

[0163] In some embodiments, the electronic device can calculate the feature standard deviation according to the following formula.

[0164] 。

[0165] In the formula, represents the feature standard deviation; represents the number of all the initial parameter features corresponding to any type of sample processing parameter; X represents the initial parameter feature corresponding to any type of sample processing parameter.

[0166] In some embodiments, after obtaining the feature average value and feature standard deviation corresponding to any type of sample processing parameter, the electronic device can perform parameter normalization processing on the initial parameter features corresponding to any type of sample processing parameter according to the following formula to obtain the sample parameter features corresponding to the any type of sample processing parameter.

[0167] 。

[0168] In the formula, represents the initial parameter feature corresponding to any type of sample processing parameter; represents the average value of the features corresponding to any type of sample processing parameter; represents the standard deviation of the features corresponding to any type of sample processing parameter; represents the sample parameter feature corresponding to any type of sample processing parameter.

[0169] In some embodiments, the electronic device can use the StandardScaler tool to calculate the average value of the features and the standard deviation of the features corresponding to each type of sample processing parameter, and perform parameter regularization processing on the initial parameter features corresponding to each type of sample processing parameter based on the average value of the features and the standard deviation of the features, so that the feature values of the initial parameter features are scaled between 0 and 1 to obtain the corresponding sample parameter features.

[0170] In some embodiments, the electronic device can regularly obtain new training data and update the average value of the features and the standard deviation of the features corresponding to the sample processing parameters by using the new training data. In some embodiments, if the sample processing parameters in the sample data are themselves numerical features and these numerical features are already at an appropriate scale, then generally there is really no need to perform additional feature processing on these numerical features.

[0171] S43: Train the initial classification model based on all sample parameter feature sets and the label features corresponding to each sample parameter feature set to obtain a life classification model.

[0172] In some embodiments, the initial classification model is a deep learning model or a machine learning model. In the embodiments of the present application, it is described by taking the initial classification model as an XGBOOST model.

[0173] In some embodiments, the XGBOOST model is a machine learning model based on the gradient boosting framework, which can construct a series of decision trees by optimizing its objective function, so as to achieve accurate prediction of data. As Figure 6 shown, the XGBOOST model constructs a strong classifier by integrating multiple weak classifiers (usually decision trees 1 to n). The generation of the latter decision tree will consider the prediction results of the previous decision tree, so as to achieve accurate prediction of data, and finally determine the comprehensive prediction results of decision trees 1 to n and output. The XGBOOST model will continuously optimize the weights and structures of each weak classifier during the training process to minimize the prediction error.

[0174] In the XGBoost model, each weak classifier is usually a decision tree. The structure of a decision tree includes a root node, internal nodes (split nodes), and leaf nodes. The root node is the starting point of the entire tree. Internal nodes perform splits based on feature values, and leaf nodes contain the final predicted values or classes.

[0175] In some embodiments, the information gain before and after a split node can be derived from the objective function of the XGBoost model. By calculating the information gain before and after the split, the XGBoost model can determine whether a certain node should be split and the optimal way to split. This helps to construct a more compact and efficient decision tree structure.

[0176] During the construction of a decision tree, the contribution degrees of different input features to the prediction result are different. The XGBoost model can evaluate the importance of input features by calculating the information gain of each feature at the split node. This enables the XGBoost model to screen out the input features that have the greatest impact on the prediction result and thus assign higher weights to them in subsequent model training.

[0177] In addition, the weights to be assigned to each leaf node can also be derived from the objective function of the XGBoost model. These weights represent the contribution degrees of different leaf nodes to the prediction result. During the process of making a prediction through a decision tree, the XGBoost model can distribute the weights to the corresponding leaf nodes according to the input features and calculate the final prediction result based on the weight of that leaf node.

[0178] During the construction of a decision tree, the XGBoost model can use a greedy algorithm to determine the partitioning of each node of the decision tree. The XGBoost model constructs the final prediction model by calculating the weights of each node. These weights reflect the contribution degrees of different nodes to the prediction result.

[0179] During the model training process, an electronic device can pre-create an empty initial classification model and set the initial model parameters of the initial classification model, such as model parameters like the learning rate, the number of decision trees, and the depth of the decision tree. The electronic device iteratively trains the initial classification model based on all sample parameter feature sets and the label features corresponding to each sample parameter feature set to obtain a life classification model.

[0180] Specifically, during the first iterative training process, the electronic device can use the initial model parameters to predict the sample parameter feature set to obtain an initial predicted value. The electronic device can calculate the first-order gradient and second-order derivative of the loss function based on the true label corresponding to the sample parameter feature set and the initial predicted value. The electronic device can use a greedy algorithm to construct a decision tree and update the model parameters (such as the weights of leaf nodes) according to the constructed decision tree.

[0181] During subsequent iterative training, the electronic device can use the updated model parameters to predict the sample parameter feature set and obtain new predicted values. The electronic device can calculate the value of the loss function based on the new predicted values and the corresponding label features. If the value of the loss function or other metrics (such as the number of iterations, etc.) meet the conditions for stopping training (such as reaching a preset number of iterations, the value of the loss function no longer decreasing significantly, etc.), the training loop ends. If the value of the loss function or other metrics (such as the number of iterations, etc.) do not meet the stopping conditions, the electronic device will continue to construct a new decision tree and update the model parameters. After training is completed, a life classification model is obtained. The life classification model can predict the life classification identifier of the jig based on the processing parameters when the jig passes through at least one processing device during fixture circulation. In some embodiments, after the model is trained, the electronic device can perform cross-validation on the life classification model to avoid overtraining and improve the performance of the model.

[0182] In some embodiments of the present application, the electronic device can train to obtain a life prediction model in the following manner: S51: Obtain a training sample set. The training sample set includes at least one training sample. Each training sample includes the sample processing parameters of a predetermined number of processing processes corresponding to the sample jig and the life classification label corresponding to each processing process.

[0183] S52: Input the training sample set into the initial prediction model and train the initial prediction model to obtain a life prediction model.

[0184] In some embodiments, the sample processing parameters may include dimensional parameters such as the length, width, height, angle, inner diameter, and outer diameter of key points or predefined points in the sample jig. The sample processing parameters when the sample jig passes through the processing device at the clamping workstation may include assembly pressure, the number of times the sample jig has been used, the identity identifier, the offset of the three axes (X-axis, Y-axis, Z-axis) of the assembly, and the height detection value after assembly. The sample processing parameters when the sample jig passes through the processing device at the glue filling workstation may include glue filling vacuum pressure, glue filling pressure, glue filling time-sharing pressure (such as dividing into 10 times and the glue filling pressure at each time point), and the maximum glue filling pressure. The sample processing parameters when the sample jig passes through the demolding workstation's processing device may include demolding force, top blowing time, top blowing pressure, bottom blowing pressure, etc. The present application does not limit the sample processing parameters when the sample jig passes through the processing device at any workstation.

[0185] During the process of training the initial prediction model to obtain a life prediction model, the electronic device adjusts the model parameters of the initial prediction model. Among them, the model parameters of the initial prediction model include but are not limited to the input dimension, output dimension, number of hidden layers, and batch training times.

[0186] In some embodiments, the life prediction model can be a neural network model. For example, the life prediction model can be a Long Short-term Memory (LSTM) model.

[0187] In some embodiments, as the processing requirements change over different periods, it may be necessary to dynamically adjust the processing parameters and retrain the life classification model and the life prediction model to improve their accuracy. However, there is a gap between the data distribution of the training dataset formed based on the adjusted processing parameters and the data distribution during inference. For example, when the life classification model and the life prediction model are trained in the first period, one of the sample processing parameters, the glue injection pressure, has a range value of 1-3 after feature processing. When further learning needs to be updated in the second period, the range value of the glue injection pressure after feature processing falls within 4-9. Since the processing parameters are dynamically adjusted at different times during the processing, in order to reduce the impact of data drift and improve the inference accuracy of the model, the electronic device of the present application collects new processing parameters in real time or regularly to form a training dataset for incremental learning of the life classification model and the life prediction model.

[0188] In some embodiments, the incremental learning of the electronic device for the life classification model includes parameter regularization and model update. The incremental learning of the electronic device for the life prediction model includes model update.

[0189] In some embodiments, model update only uses new training data to update the model, rather than adding the new training data to the existing original training data and inputting it into the model for retraining. Suppose there are 1000 training data now. During the process of model training by model update, the first version of the model can be trained with the first piece of data (e.g., 300 training data) in the training data. Next, the second piece of data (another 700 training data) in the training data is used to train the first version of the model to obtain the second version of the model. The difference between training by model update and directly using 1000 training data for model training is that when using the second piece of data to update the model, the model learns the features of the second piece of data and does not learn the features of the first piece of data. Therefore, compared with the first version of the model, the second version of the model will be more fitted to the second piece of data. In other words, compared with the original training data (e.g., the first piece of data), the new training data (e.g., the second piece of data) has a greater impact on the model.

[0190] Taking the electronic device's model training for the life classification model (such as the XGBOOST model) as an example, as Figures 7 to 9 shown, Figure 7 represents the initially trained XGBOOST model, Figure 8 and Figure 9It represents two cases of model updates for the initially trained XGBoost model. Figure 8 It represents changing the weights of the leaf nodes (such as sample parameter features) in the XGBoost model without changing the leaf nodes. Figure 9 It represents changing the weights of the leaf nodes in the XGBoost model and changing the leaf nodes.

[0191] By periodically performing model updates on the life classification model and the life prediction model, the life classification model and the life prediction model can be made to better fit the new processing parameters, thereby improving the model prediction accuracy.

[0192] It should be understood that the magnitudes of the sequence numbers of the steps in the above embodiments do not imply the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0193] Please refer to Figure 10 As shown, it is the structural diagram of the life prediction device for the fixture provided by the embodiment of the present application, which can implement the details of the life prediction method for the fixture parts in the above embodiment and achieve the same effect. As Figure 10 As shown, the life prediction device 10 for the fixture can be applied to an electronic device with data processing functions. The life prediction device 10 for the fixture includes: an acquisition module 11 for acquiring the life information of the fixture; a judgment module 12 for, if the life information does not include a life classification identifier, acquiring the first processing parameters when the fixture currently passes through all processing devices; a prediction module 13 for predicting the life classification identifier of the fixture based on the first processing parameters and the life classification model.

[0194] In some embodiments, after predicting the life classification identifier of the fixture based on the first processing parameters and the life classification model, the prediction module 13 is further configured to: if the type of the life classification identifier is an abnormal life identifier, input the first processing parameters into the life prediction model and output the first remaining life.

[0195] In some embodiments, the life prediction device 10 further includes an analysis and update module. The life information includes the target identity identifier of the fixture. The analysis and update module is configured to: obtain the current remaining life of the fixture according to the target identity identifier and the preset mapping relationship, where the preset mapping relationship is used to represent the mapping relationship between the identity identifier and the remaining life; update the current remaining life according to the first remaining life.

[0196] In some embodiments, the life information includes the target identity identifier of the fixture. The analysis and update module is further configured to: if the type of the life classification identifier is a normal life identifier, obtain the current remaining life of the fixture according to the target identity identifier and the preset mapping relationship, where the preset mapping relationship is used to represent the mapping relationship between the identity identifier and the remaining life; update the current remaining life.

[0197] In some embodiments, the analysis and update module is further configured to: update the lifespan information according to the predicted lifespan classification identifier.

[0198] In some embodiments, the lifespan information includes the target identity identifier of the jig. The analysis and update module is further configured to: if the lifespan information includes a lifespan classification identifier and the type of the lifespan classification identifier is an abnormal lifespan identifier, obtain the current remaining lifespan of the jig according to the target identity identifier and a preset mapping relationship, where the preset mapping relationship is used to represent the mapping relationship between the identity identifier and the remaining lifespan; update the current remaining lifespan to obtain the updated remaining lifespan.

[0199] In some embodiments, the analysis and update module is further configured to: if the updated remaining lifespan is less than or equal to the preset deactivation times, output a prompt message to prompt replacement of the jig.

[0200] In some embodiments, the analysis and update module is further configured to: if the lifespan information includes a lifespan classification identifier and the type of the lifespan classification identifier is a normal lifespan identifier, obtain the second processing parameter of the jig currently flowing through all processing devices; input the second processing parameter into the lifespan classification model, and use the lifespan classification model to predict the new lifespan classification identifier of the jig; if the type of the new lifespan classification identifier is a normal lifespan identifier, obtain the current remaining lifespan of the jig according to the target identity identifier and a preset mapping relationship, where the preset mapping relationship is used to represent the mapping relationship between the identity identifier and the remaining lifespan; update the current remaining lifespan.

[0201] In some embodiments, the analysis and update module is further configured to: if the type of the new lifespan classification identifier is an abnormal lifespan identifier, input the second processing parameter into the lifespan prediction model to obtain the second remaining lifespan; update the current remaining lifespan of the jig according to the second remaining lifespan; and update the lifespan classification identifier of the jig according to the abnormal lifespan identifier.

[0202] In some embodiments, the analysis and update module is further configured to: if the second remaining lifespan is less than or equal to the preset deactivation times, output a prompt message to prompt replacement of the jig.

[0203] For the specific limitations of the lifespan prediction device 10 of the jig, reference may be made to the limitations of the lifespan prediction method of the jig parts in the above text, which will not be elaborated here. Each module in the above lifespan prediction device 10 of the jig can be implemented in whole or in part by software, hardware, and their combination. The above modules can be embedded in the processor of the electronic device in hardware form or independent of it, or stored in the memory of the electronic device in software form, so that the processor can call and execute the operations corresponding to the above modules.

[0204] Please refer to Figure 11, The following is a structural diagram of the fixture life classification model training device provided by the embodiments of the present application, which can implement the details of the fixture life classification model training method in the above embodiments and achieve the same effect. As Figure 11 As shown, the fixture life classification model training device 20 can be applied to an electronic device with data processing capabilities. The fixture life classification model training device 20 includes: an acquisition module 21, configured to acquire a training data set. The training data set includes a plurality of sample data, and each sample data includes sample processing parameters corresponding to at least one processing device through which a sample fixture flows during a single processing and a life classification label corresponding to the single processing; a feature processing module 22, configured to perform feature processing on the sample processing parameters and the life classification label in each sample data to obtain a sample parameter feature set and a label feature corresponding to each sample data; a training module 23, configured to train an initial classification model based on all the sample parameter feature sets and the label features corresponding to each sample parameter feature set to obtain a life classification model.

[0205] In some embodiments, the feature processing module 22 is further configured to perform feature encoding on the life classification label in each sample data to obtain a corresponding label feature; perform feature encoding on the sample processing parameters in each sample data to obtain an initial parameter feature corresponding to each sample processing parameter; acquire all the initial parameter features corresponding to each type of sample processing parameter; calculate a feature average value and a feature standard deviation corresponding to each type of sample processing parameter based on the feature values of all the initial parameter features; perform parameter regularization processing on the initial parameter features corresponding to each type of sample processing parameter based on the feature average value and the feature standard deviation to obtain a sample parameter feature corresponding to each type of sample processing parameter; and obtain a sample parameter feature set corresponding to each sample data according to the sample parameter features corresponding to each type of sample processing parameter.

[0206] For the specific limitations of the fixture life classification model training device 20, reference can be made to the limitations on the fixture part life prediction method in the above text, which will not be elaborated here. Each module in the above fixture life classification model training device 20 can be implemented in whole or in part by software, hardware, and their combination. The above modules can be embedded in or independent of a processor in an electronic device in the form of hardware, or stored in a memory in the electronic device in the form of software, so as to facilitate the processor to call and execute the operations corresponding to the above modules.

[0207] The embodiments of the present application further provide a computer-readable storage medium, on which a computer program is stored. The computer program includes program instructions, and the method implemented when the program instructions are executed can refer to the fixture life prediction method or the fixture life classification model training method in the above embodiments of the present application. Among them, the computer-readable storage medium may be the internal memory of the electronic device described in the above embodiments, such as the hard disk or memory of the electronic device. The computer-readable storage medium may also be an external storage device of the electronic device, such as a plug-in hard disk equipped on the electronic device, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. Furthermore, the computer-readable storage medium may mainly include a storage program area and a storage data area. Among them, the storage program area may store an operating system, application programs required for at least one function, etc.; the storage data area may store data created according to the use of the electronic device, etc.

[0208] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit them. Although the present application has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present application can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A method for predicting the life of a fixture, wherein the fixture circulates in at least one processing equipment, characterized in that: The life prediction method of the fixture includes: Obtaining life information of the fixture; If the life information does not include a life classification identifier, obtaining a first processing parameter of the fixture when it is currently circulating through all the processing equipment; Based on the first processing parameter and the life classification model, a life classification identifier of the fixture is predicted.

2. The life prediction method of a fixture according to claim 1, characterized in that: After predicting the life classification identifier of the fixture based on the first processing parameter and the life classification model, the method further includes: If the type of the life classification mark is an abnormal life mark, the first processing parameter is input into a life prediction model, and a first remaining life is output.

3. The life prediction method of a fixture as claimed in claim 2, characterized in that: The life information includes a target identity of the fixture, and the method further includes: According to the target identity and a preset mapping relationship, the current remaining life of the fixture is obtained, wherein the preset mapping relationship is used to represent the mapping relationship between the identity and the remaining life; The current remaining life is updated according to the first remaining life.

4. The life prediction method of a fixture according to claim 1, characterized in that: The life information includes a target identity of the fixture. After predicting the life classification identity of the fixture based on the first processing parameter and the life classification model, the method further includes: If the type of the life classification identifier is a normal life identifier, the current remaining life of the fixture is obtained according to the target identity identifier and a preset mapping relationship, where the preset mapping relationship is used to represent the mapping relationship between the identity identifier and the remaining life; The current remaining lifespan is updated.

5. The life prediction method of a fixture according to claim 1, characterized in that: After predicting the life classification identifier of the fixture based on the first processing parameter and the life classification model, the method further includes: The life information is updated according to the predicted life classification identifier.

6. The life prediction method of a fixture according to claim 1, characterized in that: The life information includes a target identity of the fixture, and the method further includes: If the life information includes a life classification identifier, and the type of the life classification identifier is an abnormal life identifier, the current remaining life of the fixture is obtained according to the target identity identifier and a preset mapping relationship, and the preset mapping relationship is used to characterize the mapping relationship between the identity identifier and the remaining life; The current remaining life is updated to obtain an updated remaining life.

7. The life prediction method of a fixture according to claim 6, characterized in that: The method further comprises: If the remaining life after the update is less than or equal to the preset number of deactivations, a prompt message is output to prompt the user to replace the fixture.

8. The life prediction method of a fixture as claimed in claim 1, characterized in that: The life information includes a target identity of the fixture, and the method further includes: If the life information includes a life classification identifier, and the type of the life classification identifier is a normal life identifier, obtaining a second processing parameter of all the processing equipment through which the fixture currently circulates; Inputting the second processing parameter into the life classification model, and using the life classification model to predict a new life classification identifier of the fixture; If the type of the new life classification identifier is a normal life identifier, the current remaining life of the fixture is obtained according to the target identity identifier and a preset mapping relationship, where the preset mapping relationship is used to represent the mapping relationship between the identity identifier and the remaining life; The current remaining lifespan is updated.

9. The life prediction method of a fixture as claimed in claim 8, characterized in that: The method further comprises: If the type of the new life classification mark is an abnormal life mark, inputting the second processing parameter into a life prediction model to obtain a second remaining life; updating the current remaining life of the fixture according to the second remaining life; and According to the abnormal life mark, the life classification mark of the fixture is updated.

10. The life prediction method of a fixture according to claim 9, characterized in that: The method further comprises: If the second remaining life is less than or equal to the preset number of outages, a prompt message is output to prompt the user to replace the fixture.

11. A method for training a life classification model of a fixture, characterized in that: The method comprises: Acquire a training data set, the training data set comprising a plurality of sample data, each sample data comprising a sample processing parameter corresponding to a sample fixture flowing through at least one processing device during a single processing process and a life classification label corresponding to the single processing process; Performing feature processing on the sample processing parameters and life classification labels in each sample data to obtain a sample parameter feature set and label features corresponding to each sample data; Based on all the sample parameter feature sets and the label features corresponding to each sample parameter feature set, the initial classification model is trained to obtain the life classification model.

12. The method for training a life classification model of a fixture according to claim 11, characterized in that: The feature processing of the sample processing parameters and life classification labels in each sample data to obtain the sample parameter feature set and label features corresponding to each sample data includes: Performing feature encoding on the lifespan classification label in each sample data to obtain corresponding label features; Performing feature encoding on the sample processing parameters in each sample data to obtain initial parameter features corresponding to each sample processing parameter; Obtain all initial parameter features corresponding to each type of sample processing parameters; Based on the characteristic values ​​of all the initial parameter characteristics, calculating the characteristic average value and characteristic standard deviation corresponding to each type of sample processing parameter; Based on the feature average value and the feature standard deviation, parameter normalization processing is performed on the initial parameter features corresponding to each type of sample processing parameters to obtain sample parameter features corresponding to each type of sample processing parameters; According to the sample parameter features corresponding to each type of sample processing parameters, a sample parameter feature set corresponding to each sample data is obtained.

13. An electronic device, characterized in that: The invention comprises a memory, a processor and computer-readable instructions stored in the memory and executable on the processor, wherein the computer-readable instructions, when executed by the processor, implement the life prediction method of the fixture as described in any one of claims 1 to 10 or implement the life classification model training method of the fixture as described in claim 11 or 12.

14. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the life prediction method of a fixture as described in any one of claims 1 to 10 or the life classification model training method of a fixture as described in claim 11 or 12.

Images