Material rejection rate prediction method and device, equipment and storage medium
By extracting material processing history features and aggregating data, and combining trend, cycle, and residual scrap rate prediction models, the problems of low efficiency and insufficient accuracy in material scrap rate prediction are solved, achieving more accurate and efficient prediction.
Patent Information
- Application Number
- CN202510969727.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-14
- Publication Date
- 2025-11-14
AI Technical Summary
Existing methods for predicting material scrap rates during material production are inefficient and have low accuracy. They are influenced by many dynamic factors, leading to large prediction errors.
By acquiring the processing history of materials, extracting processing characteristic values, performing data aggregation processing, identifying trends, cycles, and residual scrap rates, and using predictive models to fuse and process the data to obtain the predicted scrap rate.
It improved the accuracy and efficiency of material scrap rate prediction and reduced labor costs.
Smart Images
Figure CN120952218A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of machine learning technology, and in particular to a method, apparatus, device, and storage medium for predicting material scrap rates. Background Technology
[0002] In the material production process, when it is necessary to understand the scrap rate of materials in the future production process, it is usually done by manually calculating the historical defect rate of materials in each workshop. Then, the product of the historical defect rate and the future allocation is used as the scrap rate prediction. This prediction method is inefficient, and there are many factors that affect the material defect rate, such as the processing parameters of production equipment, defective cutting tools, processing time, etc. These factors are not static, and some factors change dynamically. Therefore, the above scrap rate prediction method has a large error and low accuracy. Summary of the Invention
[0003] In view of this, this application provides a method, apparatus, equipment and storage medium for predicting material scrap rate, in order to solve the above-mentioned technical problems.
[0004] In a first aspect, embodiments of this application provide a method for predicting material scrap rate, the method comprising: Obtain the processing history of each piece of material within a historical time period; The processing history of each piece of material is processed by feature extraction to obtain the processing feature value of each piece of material; The processing characteristic values of each piece of material are aggregated to obtain the target characteristic value corresponding to the historical time period; Based on the target feature values, the trend scrap rate, periodic scrap rate, and residual scrap rate are obtained; The trend scrap rate, the periodic scrap rate, and the residual scrap rate are fused together to obtain the predicted scrap rate.
[0005] In this embodiment of the application, historical data is used to predict the future scrap rate. By identifying long-term trend data, recurring patterns or periodic data, and unpredictable random fluctuations in the scrap data, the future scrap rate is predicted in combination, making the predicted scrap rate more accurate and improving the efficiency of obtaining the predicted scrap rate, thus saving labor costs.
[0006] In some possible embodiments, the step of performing feature extraction processing on the processing history of each piece of material to obtain the processing feature value of each piece of material includes: Perform the following for each piece of material: Based on the material's coding information, obtain the material's processing station information and processing process information; The processing station information and the processing process information are processed to obtain the processing feature value corresponding to the material.
[0007] In some possible embodiments, the processing feature value includes at least one of the following: Workstation transfer interval, process transfer interval, dwell time at each workstation, dwell time at each process, dwell percentage, transfer time percentage, number of process processing steps, number of workstation processing steps, process inspection results, and yield.
[0008] In some possible embodiments, the data aggregation processing of the processing feature values of each piece of material to obtain the target feature value corresponding to the historical time period includes: Perform any one or a combination of the following steps for each type of processing feature value: The average value of the processing characteristic value of each piece of material is obtained by averaging the values. The variance of the processing characteristic value of each piece of material is processed to obtain the variance characteristic value; The processing characteristic value of each piece of material is processed by quantiles to obtain quantile characteristic values.
[0009] In some possible embodiments, obtaining the trend scrap rate, periodic scrap rate, and residual scrap rate based on the target feature value includes: Determine a first target feature value corresponding to the trend scrap rate from the target feature values; Determine a second target feature value corresponding to the cycle scrap rate from the target feature values; A third target feature value corresponding to the residual scrap rate is determined from the target feature values; Input the first target feature value into the first prediction model to obtain the trend scrap rate output by the first prediction model; Input the second target feature value into the second prediction model to obtain the cycle scrap rate output by the second prediction model; The third target feature value is input into the third prediction model to obtain the residual scrap rate output by the third prediction model.
[0010] In some possible embodiments, the fusion processing of the trend scrap rate, the periodic scrap rate, and the residual scrap rate to obtain the predicted scrap rate includes: Obtain the end date corresponding to the historical time period; At least one target date is obtained based on the preset duration window length of the end date; Obtain the target trend scrap rate, target period scrap rate, and target residual scrap rate for each target date; Dependency parameters are obtained based on the target trend scrap rate, target periodic scrap rate, and target residual scrap rate corresponding to each target date; wherein, the dependency parameters characterize the dependency relationship between the trend scrap rate, the periodic scrap rate, and the residual scrap rate; By inputting the dependency parameters into the fully connected network model, we can obtain the first weight corresponding to the trend scrap rate, the second weight corresponding to the period scrap rate, and the third weight corresponding to the residual scrap rate. The predicted scrap rate is obtained by weighting and summing the first weight, the second weight, the third weight, the trend scrap rate, the periodic scrap rate, and the residual scrap rate.
[0011] In some possible embodiments, the method further includes: Determine the time period corresponding to the historical time period; Based on the time period, the residual scrap rate of multiple time periods corresponding to the first time period is obtained; wherein, the first time period is before the historical time period; The residual mean and standard deviation are obtained based on the residual scrap rate corresponding to the historical time period and the residual scrap rate corresponding to the first time period; The residual threshold is obtained based on the residual mean and the standard deviation. If the absolute value of the residual scrap rate is greater than the residual threshold, the predicted scrap rate is corrected.
[0012] In some possible embodiments, the correction process for the predicted scrap rate includes: Obtain the processing feature values corresponding to specific dates within the historical time period; The processing characteristic value, the trend scrap rate, the periodic scrap rate, and the residual scrap rate are input into the activation function to obtain the activation correction value. The trend scrap rate is corrected based on the activation correction value to obtain the corrected trend scrap rate; and the periodic scrap rate is corrected based on the activation correction value to obtain the corrected periodic scrap rate. The predicted scrap rate is corrected based on the corrected trend scrap rate and the corrected periodic scrap rate to obtain the corrected predicted scrap rate.
[0013] Secondly, embodiments of this application also provide a material scrap rate prediction device, the device comprising: The acquisition module is used to acquire the processing history of each piece of material within a historical time period; The extraction module is used to perform feature extraction processing on the processing history of each piece of material to obtain the processing feature value of each piece of material; The aggregation module is used to perform data aggregation processing on the processing feature values of each piece of material to obtain the target feature value corresponding to the historical time period; The prediction module is used to obtain the trend scrap rate, periodic scrap rate, and residual scrap rate based on the target feature value; The fusion module is used to fuse the trend scrap rate, the periodic scrap rate, and the residual scrap rate to obtain the predicted scrap rate.
[0014] In some possible embodiments, the extraction module is specifically used for: Perform the following for each piece of material: Based on the material's coding information, obtain the material's processing station information and processing process information; The processing station information and the processing process information are processed to obtain the processing feature value corresponding to the material.
[0015] In some possible embodiments, the processing feature value includes at least one of the following: Workstation transfer interval, process transfer interval, dwell time at each workstation, dwell time at each process, dwell percentage, transfer time percentage, number of process processing steps, number of workstation processing steps, process inspection results, and yield.
[0016] In some possible embodiments, the aggregation module is specifically used for: Perform any one or a combination of the following steps for each type of processing feature value: The average value of the processing characteristic value of each piece of material is obtained by averaging the values. The variance of the processing characteristic value of each piece of material is processed to obtain the variance characteristic value; The processing characteristic value of each piece of material is processed by quantiles to obtain quantile characteristic values.
[0017] In some possible embodiments, the prediction module is specifically used for: Determine a first target feature value corresponding to the trend scrap rate from the target feature values; Determine a second target feature value corresponding to the cycle scrap rate from the target feature values; A third target feature value corresponding to the residual scrap rate is determined from the target feature values; Input the first target feature value into the first prediction model to obtain the trend scrap rate output by the first prediction model; Input the second target feature value into the second prediction model to obtain the cycle scrap rate output by the second prediction model; The third target feature value is input into the third prediction model to obtain the residual scrap rate output by the third prediction model.
[0018] In some possible embodiments, the fusion module is specifically used for: Obtain the end date corresponding to the historical time period; At least one target date is obtained based on the preset duration window length of the end date; Obtain the target trend scrap rate, target period scrap rate, and target residual scrap rate for each target date; Dependency parameters are obtained based on the target trend scrap rate, target periodic scrap rate, and target residual scrap rate corresponding to each target date; wherein, the dependency parameters characterize the dependency relationship between the trend scrap rate, the periodic scrap rate, and the residual scrap rate; By inputting the dependency parameters into the fully connected network model, we can obtain the first weight corresponding to the trend scrap rate, the second weight corresponding to the period scrap rate, and the third weight corresponding to the residual scrap rate. The predicted scrap rate is obtained by weighting and summing the first weight, the second weight, the third weight, the trend scrap rate, the periodic scrap rate, and the residual scrap rate.
[0019] In some possible embodiments, the fusion module is further configured to: determine the time period corresponding to the historical time period; Based on the time period, the residual scrap rate of multiple time periods corresponding to the first time period is obtained; wherein, the first time period is before the historical time period; The residual mean and standard deviation are obtained based on the residual scrap rate corresponding to the historical time period and the residual scrap rate corresponding to the first time period; The residual threshold is obtained based on the residual mean and the standard deviation. If the absolute value of the residual scrap rate is greater than the residual threshold, the predicted scrap rate is corrected.
[0020] In some possible embodiments, the fusion module is specifically used for: Obtain the processing feature values corresponding to specific dates within the historical time period; The processing characteristic value, the trend scrap rate, the periodic scrap rate, and the residual scrap rate are input into the activation function to obtain the activation correction value. The trend scrap rate is corrected based on the activation correction value to obtain the corrected trend scrap rate; and the periodic scrap rate is corrected based on the activation correction value to obtain the corrected periodic scrap rate. The predicted scrap rate is corrected based on the corrected trend scrap rate and the corrected periodic scrap rate to obtain the corrected predicted scrap rate.
[0021] Thirdly, another embodiment of this application also provides an electronic device, including at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform any of the methods provided in the first aspect embodiment of this application.
[0022] Fourthly, another embodiment of this application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program for causing a computer to perform any of the methods provided in the first aspect of this application.
[0023] Other features and advantages of this application will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings. Attached Figure Description
[0024] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0025] Figure 1 This is a schematic diagram illustrating an application scenario of a material scrap rate prediction method provided in an embodiment of this application. Figure 2 This is a schematic diagram of the overall process of a material scrap rate prediction method provided in an embodiment of this application; Figure 3 This is a flowchart illustrating the process of extracting features from the processing history of a material to obtain its processing feature values, as provided in an embodiment of this application, for a method to predict the scrap rate of a material. Figure 4 A schematic diagram of a workstation and process for a material scrap rate prediction method provided in this application embodiment; Figure 5 A schematic diagram illustrating the process of data aggregation of processing characteristic values in a material scrap rate prediction method provided in this application embodiment; Figure 6A schematic diagram of the processing time curve of process 1 for a material scrap rate prediction method provided in this application embodiment; Figure 7 A schematic diagram of the aggregated curve of a material scrap rate prediction method provided in this application embodiment; Figure 8 This is a flowchart illustrating a material scrap rate prediction method based on target feature values to obtain trend scrap rate, periodic scrap rate, and residual scrap rate, as provided in an embodiment of this application. Figure 9 A schematic diagram illustrating a material scrap rate prediction method based on target feature values, yielding trend scrap rate, periodic scrap rate, and residual scrap rate, as provided in this application embodiment. Figure 10 A schematic diagram illustrating the process of fusing trend scrap rate, periodic scrap rate, and residual scrap rate in a material scrap rate prediction method provided in this application embodiment; Figure 11 A schematic diagram of the trend scrap rate curve, the periodic scrap rate curve, the residual scrap rate curve, and the predicted scrap rate curve provided in this application embodiment for a material scrap rate prediction method; Figure 12 A flowchart illustrating the determination of the accuracy of the predicted scrap rate using a material scrap rate prediction method provided in this application embodiment; Figure 13 A schematic diagram illustrating the process of correcting the predicted scrap rate in a material scrap rate prediction method provided in this application embodiment; Figure 14 A schematic diagram of an apparatus for predicting material scrap rate provided in an embodiment of this application; Figure 15 This is a schematic diagram of an electronic device for a material scrap rate prediction method provided in an embodiment of this application. Detailed Implementation
[0026] To better understand the technical solution of this application, the embodiments of this application will be described in detail below with reference to the accompanying drawings.
[0027] It should be understood that the described embodiments are merely some, not all, of the embodiments in this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.
[0028] The terminology used in the embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. The singular forms “a,” “the,” and “the” used in the embodiments of this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.
[0029] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.
[0030] This application provides a method, apparatus, device, and storage medium for predicting material scrap rates, addressing the problems mentioned in the background art. The inventive concept of this application can be summarized as follows: acquiring the processing history of each piece of material within a historical time period; performing feature extraction processing on the processing history of each piece of material to obtain processing feature values for each piece of material; performing data aggregation processing on the processing feature values of each piece of material to obtain target feature values corresponding to the historical time period; obtaining trend scrap rate, periodic scrap rate, and residual scrap rate based on the target feature values; and fusing the trend scrap rate, periodic scrap rate, and residual scrap rate to obtain the predicted scrap rate.
[0031] In this embodiment of the application, historical data is used to predict the future scrap rate. Furthermore, by identifying long-term trend data, recurring patterns or periodic data, and unpredictable random fluctuations in the scrap data, the future scrap rate is predicted in a more accurate manner.
[0032] For ease of understanding, the following detailed description, in conjunction with the accompanying drawings, illustrates a method for predicting material scrap rates according to an embodiment of this application: like Figure 1 The diagram shown illustrates an application scenario of a material scrap rate prediction method according to an embodiment of this application. The diagram includes: a machine, a material scrap rate prediction device, and materials; wherein: The material scrap rate prediction device acquires the processing history of each piece of material on the machine within a historical time period; performs feature extraction processing on the processing history of each piece of material to obtain the processing feature value of each piece of material; performs data aggregation processing on the processing feature value of each piece of material to obtain the target feature value corresponding to the historical time period; based on the target feature value, it obtains the trend scrap rate, periodic scrap rate, and residual scrap rate; and performs fusion processing on the trend scrap rate, periodic scrap rate, and residual scrap rate to obtain the predicted scrap rate.
[0033] The machine tool can be one or more pieces of equipment used in the production process, such as a processing machine tool, an inspection machine tool, a cleaning machine tool, etc., or it can be auxiliary equipment, such as a clamping machine tool. The machine tool, material scrap rate prediction device, and material can be one or more. Correspondingly, there can be one material scrap rate prediction device, but multiple material scrap rate prediction devices can also be set according to actual usage needs. The material scrap rate prediction method of this application can be applied to one or more of the following equipment: machine tool, computer, material scrap rate prediction device, etc. For ease of understanding, the following description focuses on the application of the material scrap rate prediction method to a material scrap rate prediction device. The description in this application only details multiple machine tools, material scrap rate prediction devices, and multiple materials. However, those skilled in the art should understand that the machine tools, material scrap rate prediction devices, and materials shown are intended to represent the operation of the machine tools, material scrap rate prediction devices, and materials involved in the technical solution of this application, and do not limit the quantity, type, or location of the machine tools, material scrap rate prediction devices, and materials. It should be noted that adding additional modules to or removing individual modules from the illustrated environment will not change the underlying concept of the example embodiments of this application.
[0034] It should be noted that the material scrap rate prediction method proposed in this application is not only applicable to... Figure 1 The application scenarios shown can also be applied to any device that needs to predict material scrap rates.
[0035] like Figure 2 The diagram shown is a schematic representation of the overall process of a material scrap rate prediction method provided in an embodiment of this application, wherein: In step 201: Obtain the processing history of each piece of material within the historical time period.
[0036] In this embodiment, the processing history of each piece of material can be acquired by the machine and sent to the server, or it can be acquired by other devices and sent to the server. Alternatively, the processing history of each piece of material can be generated from initial data sent by the machine and processed by the server. For ease of understanding, in this embodiment, the processing history of the material is generated by the machine.
[0037] Optionally, the historical time period can be the week before the current time or the first half of the current time period. The time cycle corresponding to the historical time period can be set according to the needs. That is, this application does not limit the specific time cycle of the historical time period.
[0038] In step 202: the processing history of each piece of material is processed by feature extraction to obtain the processing feature value of each piece of material.
[0039] In this embodiment of the application, the material processing history contains all the information of the material from the start to the end of processing. In order to avoid wasting computing resources by processing useless data, feature extraction is performed on the material processing history to process only the data related to the scrap rate of the material, thus saving computing resources.
[0040] In some possible embodiments, for each piece of material, feature extraction processing is performed on the processing history of that material to obtain the processing feature values of the material. Specifically, this can be implemented as follows: Figure 3 The steps shown are as follows: In step 301: Based on the material's coding information, obtain the material's processing station information and processing process information.
[0041] In this embodiment, the processing station information includes, but is not limited to: the station number, the dwell time at each station, the transfer time between adjacent stations, and the number of processing operations at each station; the processing process information includes, but is not limited to: the material dwell time at each process, the number of processing operations at each process, the inspection results at each process, the material yield, the transfer time between adjacent processes, and the inspection results at each process. The process includes one or more of CNC machining, laser welding, anodizing, physical vapor deposition coating, grinding, and polishing, and each station in each process can be equipped with one or more machines.
[0042] In step 302: the processing station information and processing process information are processed to obtain the processing feature value corresponding to the material.
[0043] In the embodiments of this application, the processing characteristic values include at least one or a combination of the following: workstation transfer interval, process transfer interval, dwell time at each workstation, dwell time at each process, dwell percentage, transfer time percentage, number of process processing times, number of workstation processing times, process inspection results, and yield.
[0044] To facilitate understanding, the above processing feature values are explained below with reference to the accompanying drawings: For example, the station transfer interval characterizes the length of time that materials stay between adjacent stations.
[0045] For example: Figure 4As shown, the processing of material A includes three processes: process 1, process 2, and process 3. Process 1 includes workstations 1, 2, and 3; process 2 includes workstations 4 and 5; and process 3 includes workstations 6, 7, and 8. The transfer intervals for material A at each workstation include: the time interval between the end time of processing material A at workstation 1 and the start time of processing material A at workstation 2 (first transfer interval); the time interval between the end time of processing material A at workstation 2 and the start time of processing material A at workstation 3 (second transfer interval); and the time interval between the end time of processing material A at workstation 3 and the start time of processing material A at workstation 3. The time interval between the end time of material A at station 4 and the start time of processing at station 5 (third transfer interval); the time interval between the end time of material A at station 4 and the start time of processing at station 5 (fourth transfer interval); the time interval between the end time of material A at station 5 and the start time of processing at station 6 (fifth transfer interval); the time interval between the end time of material A at station 6 and the start time of processing at station 7 (sixth transfer interval); and the time interval between the end time of material A at station 7 and the start time of processing at station 8 (seventh transfer interval).
[0046] For example, process transfer intervals characterize the length of time materials remain between adjacent processes.
[0047] Continue with Figure 4 Taking an example, the process transfer intervals corresponding to material A include: the time interval between the processing end time of material A in process 1 and the processing start time in process 2 (first process interval), and the time interval between the processing end time of material A in process 2 and the processing start time in process 3 (second process interval).
[0048] For example, the dwell time percentage represents the proportion of material transfer intervals at workstations in the total processing time.
[0049] Continue with Figure 4 Taking an example: The workstation transfer intervals corresponding to material A include: the first transfer interval, the second transfer interval, the third transfer interval, the fourth transfer interval, the fifth transfer interval, the sixth transfer interval, and the seventh transfer interval; the total processing time is denoted as T, then the dwell percentages corresponding to material A include: the first transfer interval / T = the first dwell percentage, the second transfer interval / T = the second dwell percentage, the third transfer interval / T = the third dwell percentage, the fourth transfer interval / T = the fourth dwell percentage, the fifth transfer interval / T = the fifth dwell percentage, the sixth transfer interval / T = the sixth dwell percentage, and the seventh transfer interval / T = the seventh dwell percentage.
[0050] For example, the percentage of material transfer time represents the proportion of the material transfer interval in the total processing time.
[0051] Continue with Figure 4 Taking an example: The process transfer intervals corresponding to material A include: the first process interval and the second process interval. The total processing time is denoted as T. Then, the transfer time percentages corresponding to material A include: the first process interval / T = the percentage of the first transfer time, and the second process interval / T = the percentage of the second transfer time.
[0052] For example, the number of processing steps represents the total number of times the material is processed at each station in the process; where the number of processing steps at a station represents the number of times the material is processed at that station.
[0053] Continue with Figure 4 Taking an example, the number of times material A is processed at each station includes: the number of times it is processed at station 1, the number of times it is processed at station 2, the number of times it is processed at station 3, the number of times it is processed at station 4, the number of times it is processed at station 5, the number of times it is processed at station 6, the number of times it is processed at station 7, and the number of times it is processed at station 8.
[0054] It is understandable that the number of processing steps in the characteristic values corresponding to each piece of material is the same throughout the historical period.
[0055] Continue with Figure 4 For example, the number of times material A is processed in process 1 is the sum of the number of times it is processed in stations 1, 2, and 3. Assuming that material A is processed 2 times in station 1, 1 time in station 2, and 1 time in station 3, then material A is processed 4 times in process 1. Similarly, the number of times material A is processed in process 2 is the sum of the number of times it is processed in stations 4 and 5, and the number of times material A is processed in process 3 is the sum of the number of times it is processed in stations 6, 7, and 8.
[0056] For example, process inspection results characterize the inspection results of materials after processing in the last process. Continue with Figure 4 Taking this as an example, the last process of material A is process 3. The test result after process 3 is 1 (1 indicates that the test result of the material is qualified, and 0 indicates that the test result of the material is unqualified). Therefore, the process test result of material A is 1 (qualified).
[0057] For example, yield characterizes the yield of each process corresponding to a material.
[0058] Each piece of material undergoes a qualification test after each process is completed to determine the corresponding test result. Each process determines its corresponding yield based on the test result of each piece of material.
[0059] Continue with Figure 4Taking this example, 1 represents a qualified test result for the material, and 0 represents a failed test result. The yield corresponding to material A includes the yield of process 1 and the yield of process 2. The yield corresponding to each process is obtained based on the test result of each piece of material processed in that process. For example: the test result of material A in process 1 is 1, and the test result in process 2 is 1; the test result of material B in process 1 is 1, and the test result in process 2 is 0; the test result of material C in process 1 is 0, and the test result in process 2 is 0; the test result of material D in process 1 is 1, and the test result in process 2 is 1; the test result of material E in process 1 is 1, and the test result in process 2 is 1. Therefore, the yield corresponding to process 1 is 80%, and the yield corresponding to process 2 is 60%. Thus, the yield corresponding to material A includes: a yield of 80% in process 1 and a yield of 60% in process 2.
[0060] In step 203: The processing characteristic values of each piece of material are aggregated to obtain the target characteristic values corresponding to the historical time period.
[0061] In this embodiment of the application, by performing data aggregation processing on processing feature values of the same category, the data corresponding to multiple pieces of material are aggregated into one for subsequent processing, thereby reducing the amount of computation.
[0062] In some possible embodiments, data aggregation is performed on processing feature values of the same category to obtain target feature values corresponding to historical time periods. For each category of processing feature values, the specific implementation can be as follows: Figure 5 Any one or combination of the steps shown, wherein: In step 501: the processing characteristic values of each piece of material are averaged to obtain the average characteristic value.
[0063] In this embodiment of the application, the processing characteristic value of each piece of material is averaged, that is, the average processing characteristic value of each piece of material of the same category is calculated.
[0064] In step 502: Variance processing is performed on the processing characteristic values of each piece of material to obtain variance characteristic values.
[0065] In the embodiments of this application, the variance of the processing characteristic value of each piece of material is processed, that is, the standard deviation of the processing characteristic value corresponding to each piece of material of the same category is calculated.
[0066] In step 503: the processing characteristic value of each piece of material is processed by quantiles to obtain quantile characteristic values.
[0067] In the embodiments of this application, each piece of material is processed by quantiles, including but not limited to: 25th quantile, 50th quantile, and 75th quantile.
[0068] It should be noted that, in addition to data aggregation processing of processing feature values of the same category, Figure 5 In addition to the methods mentioned above, it may also include: the sum of each type of processing feature value, the minimum value of each type of processing feature value, the maximum value of each type of processing feature value, etc.; that is, the specific implementation method of the aggregation process in this application embodiment is not limited, and can be set according to the needs in specific implementation.
[0069] To facilitate understanding, the following example illustrates the process of data aggregation for processing feature values: For example: Figure 6 As shown, the processing time of process 1 for each piece of material is used as an example for explanation. The vertical axis of the process 1 processing time curve represents the processing time for each piece of material. The unit of processing time can be milliseconds, seconds, minutes, hours, days, etc. This application uses hours as an example for processing time. The horizontal axis represents the quantity of materials. That is, each point (x, y) on the process 1 processing time curve means that there are x pieces of material with a processing time of y in process 1. Figure 7 As shown, for Figure 6 The process 1 processing time curve shown is processed by data aggregation to obtain aggregated data. The aggregated target feature values corresponding to the process 1 processing time are: the sum of the process 1 processing time for each piece of material, the mean of the process 1 processing time for each piece of material, the standard deviation of the process 1 processing time for each piece of material, the minimum process 1 processing time for each piece of material, the 25th percentile of the process 1 processing time for each piece of material, the 50th percentile of the process 1 processing time for each piece of material, the 75th percentile of the process 1 processing time for each piece of material, and the maximum process 1 processing time for each piece of material.
[0070] Understandable, Figure 7 The specific values in the figures are merely illustrative examples for ease of understanding and are not intended to limit the interpretation of these figures.
[0071] This method allows us to obtain the target feature value corresponding to each type of processing feature value within a historical time period.
[0072] In step 204: Based on the target feature value, the trend scrap rate, periodic scrap rate and residual scrap rate are obtained.
[0073] In this embodiment of the application, in order to improve the accuracy of predicting the scrap rate of materials, the scrap rate of materials is predicted by combining the trend scrap rate, the periodic scrap rate, and the residual scrap rate.
[0074] Among these, the trend scrap rate represents the direction of long-term changes in the scrap rate. In the early stages of production, the trend scrap rate is higher, gradually decreasing as mass production begins. Seasonality refers to recurring patterns or changes in data within a specific time period. These changes are usually related to cyclical factors such as weather, holidays, and workdays. Specific times each month exhibit recurring patterns; for example, the scrap rate might be higher in the last week of each month, or it might show a pattern of one week low, one week high, or scrap being cleared only every two weeks. Irregular or residual refers to unpredictable, irregular fluctuations in the data, usually considered noise. The scrap rate can increase due to abnormalities in production equipment or human factors.
[0075] In some possible embodiments, based on the target feature values, the trend scrap rate, periodic scrap rate, and residual scrap rate are obtained, which can be specifically implemented as follows: Figure 8 The steps shown are as follows: In step 801: Determine the first target feature value corresponding to the trend scrap rate from the target feature values.
[0076] In this embodiment of the application, a feature value that is associated with the long-term change direction of the scrap rate is determined from the target feature values as the first target feature value corresponding to the trend scrap rate.
[0077] In step 802: Determine the second target feature value corresponding to the periodic scrap rate from the target feature values.
[0078] In this embodiment of the application, the feature value that affects the scrap rate within a specific time period is determined from the target feature value as the second target feature value corresponding to the periodic scrap rate.
[0079] In step 803: Determine the third target feature value corresponding to the residual scrap rate from the target feature values.
[0080] In this embodiment of the application, irregular feature values that affect the scrap rate are determined from the target feature values as the third target feature value corresponding to the residual scrap rate.
[0081] For example, such as Figure 9 As shown, a filter can be used to perform convolution operations on the target feature values to filter out the first target feature value corresponding to the trend scrap rate; the first target feature value is removed from the target feature values to obtain the fluctuation data; the fluctuation data is processed according to the data aggregation period to obtain the second target feature value corresponding to the period scrap rate; the second target feature value is removed from the fluctuation data, and the remaining feature value is the third target feature value corresponding to the residual scrap rate.
[0082] It is understood that the method given above is only a method for determining the first target feature value, the second target feature value, and the third target feature value from the target feature values, and is not a limitation on the method for determining the first target feature value, the second target feature value, and the third target feature value from the target feature values. In specific implementation, the method for determining the first target feature value, the second target feature value, and the third target feature value from the target feature values can be set according to the needs.
[0083] In step 804: the first target feature value is input into the first prediction model to obtain the trend scrap rate output by the first prediction model.
[0084] In step 805: the second target feature value is input into the second prediction model to obtain the periodic scrap rate output by the second prediction model.
[0085] In step 806: Input the third target feature value into the third prediction model to obtain the residual scrap rate output by the third prediction model.
[0086] It is understandable that the training processes of the first prediction model, the second prediction model, and the third prediction model can be the same or similar, or of course, different. Optionally, the training processes of the first prediction model, the second prediction model, and the third prediction model can be the same or similar to the training processes of neural network models in related technologies, or of course, different, which will not be elaborated here.
[0087] It should be noted that, Figure 8 The order of implementation is not limited to steps 801, 802, and 803. In specific implementation, step 801 can be completed before step 804, step 802 before step 805, and step 803 before step 806. Furthermore, this application does not limit the implementation order of the above steps 804, 805, and 806. They can be implemented simultaneously or the implementation order of steps 804, 805, and 806 can be set according to requirements.
[0088] In step 205: The trend scrap rate, periodic scrap rate, and residual scrap rate are fused to obtain the predicted scrap rate.
[0089] In this embodiment of the application, the scrap rate is predicted by identifying long-term trend data, recurring patterns or periodic data, and unpredictable random fluctuations in the scrap data, making the obtained scrap rate more accurate.
[0090] In some possible embodiments, the trend scrap rate, periodic scrap rate, and residual scrap rate are fused to obtain the predicted scrap rate. Specifically, this can be implemented as follows: Figure 10 The steps shown are as follows: In step 1001: Obtain the end date corresponding to the historical time period.
[0091] In this embodiment of the application, the end date is recorded as the date of the last day of the historical time period. For example, if the historical time period is 2025 / 4 / 1-2025 / 4 / 30, then the end date is 2025 / 4 / 30.
[0092] In step 1002: At least one target date is obtained based on the preset duration window length of the end date.
[0093] In this embodiment of the application, the preset duration window length can be denoted as: For example, the end date is 2025 / 4 / 30. Therefore, the target dates can be determined as: 2025 / 4 / 30, 2025 / 4 / 29, 2025 / 4 / 28, 2025 / 4 / 27, 2025 / 4 / 26.
[0094] In step 1003: obtain the target trend scrap rate, target period scrap rate, and target residual scrap rate for each target date.
[0095] In the embodiments of this application, the target trend scrap rate, target cycle scrap rate, and target residual scrap rate for each date are obtained based on the processing history of the materials for each date.
[0096] In step 1004: Dependency parameters are obtained based on the target trend scrap rate, target periodic scrap rate, and target residual scrap rate corresponding to each target date; wherein, the dependency parameters characterize the dependency relationship between the trend scrap rate, periodic scrap rate, and residual scrap rate.
[0097] In some possible embodiments, the method for obtaining the dependency parameters can be specifically implemented as Equation 1, wherein: , (Formula 1) in, This is a dependency relationship; For the first The target trend scrap rate corresponding to each target date For the first Each target date corresponds to a target cycle scrap rate. For the first The target residual scrap rate corresponding to each target date; The preset duration window length, The end date of the historical period. It is the dimension scaling factor, and These are pre-set preset values.
[0098] In step 1005: Input the dependent parameters into the fully connected network model to obtain the first weight corresponding to the trend scrap rate, the second weight corresponding to the periodic scrap rate, and the third weight corresponding to the residual scrap rate.
[0099] In some possible embodiments, the weights corresponding to each scrap rate can be specifically implemented as formula 2, where: , (Formula 2) in, The first weight corresponding to the trend scrap rate, The second weight corresponding to the periodic scrap rate, The third weight corresponding to the residual scrap rate. It is a fully connected network model. It is 3× D A dimensional trainable matrix is used to specify the degree of influence of different features on the weights, that is, to... and adaptation vector The concatenation result is linearly mapped to the first weight, the second weight, and the third weight. This is the adaptation vector corresponding to the fully connected network model. This is a three-dimensional vector in a fully connected network model, used to adjust the baseline allocation ratio of each weight.
[0100] For example: it can be done The document specifies the degree of influence of different characteristics on different weights (for example, the characteristic of promotional events will increase the seasonality weight). The default weighting for trend is 60%.
[0101] In step 1006: the first weight, the second weight, the third weight, the trend scrap rate, the periodic scrap rate, and the residual scrap rate are weighted and summed to obtain the predicted scrap rate.
[0102] In some possible embodiments, the weighted summation process can be specifically implemented as Equation 3, where: , (Formula 3) in, To predict the scrap rate, As the first weight, As the second weight, As the third weight, For trend scrap rate, For the periodic scrap rate, This refers to the residual scrap rate.
[0103] For example: Figure 11 As shown, the curve corresponding to the predicted scrap rate is obtained based on the curves corresponding to the trend scrap rate, the periodic scrap rate, and the residual scrap rate.
[0104] In some possible embodiments, in order to further improve the accuracy of predicting scrap rates, therefore according to... Figure 12 The steps shown are to determine the accuracy of the predicted scrap rate, and to correct the predicted scrap rate when the accuracy is low. In step 1201: Determine the time period corresponding to the historical time period.
[0105] In this embodiment of the application, the time period corresponding to the historical time period is the duration of the historical time period. For example, if the historical time period is from 2024 / 3 / 1 to 2024 / 3-7, then the time period can be determined to be one week, or 7 days.
[0106] In step 1202: Based on the time period, obtain the residual scrap rate of multiple time periods corresponding to the first time period; wherein, the first time period is before the historical time period.
[0107] In this embodiment of the application, the length of the first time period is not limited, nor is the number of time periods limited.
[0108] For example: If the historical time period is from 2024 / 5 / 24 to 2024 / 5 / 30, the time cycle is one week, and the number of time cycles is 5, then the 5 time cycles are: 2024 / 5 / 17-2024 / 5 / 23, 2024 / 5 / 10-2024 / 5 / 16, 2024 / 5 / 03-2024 / 5 / 09, 2024 / 4 / 26-2024 / 5 / 02, and 2024 / 4 / 19-2024 / 4 / 25.
[0109] In step 1203: the residual mean and standard deviation are obtained based on the residual scrap rate corresponding to the historical time period and the residual scrap rate corresponding to the first time period.
[0110] For example, given a historical time period from May 24, 2024 to May 30, 2024, the residual rejection rates for the first time period are as follows: May 17, 2024 to May 23, 2024; May 10, 2024 to May 16, 2024; May 3, 2024 to May 9, 2024; April 26, 2024 to May 2, 2024; and April 19, 2024 to April 25, 2024. The mean of the residuals is obtained by averaging these six residual rejection rates. The standard deviation is obtained by calculating the standard deviation of these six residual rejection rates.
[0111] In step 1204: the residual threshold is obtained based on the residual mean and standard deviation.
[0112] In some possible embodiments, the residual threshold obtained based on the residual mean and standard deviation can be specifically implemented as Equation 4, where: , (Formula 4) in, The residual threshold, The mean of the residuals, Standard deviation, The sensitivity coefficient is set in advance.
[0113] In step 1205: If the absolute value of the residual scrap rate is greater than the residual threshold, the predicted scrap rate is corrected.
[0114] In this embodiment of the application, after obtaining the residual scrap rate, the absolute value of the residual scrap rate is compared with the residual threshold. If the absolute value of the residual scrap rate is greater than the residual threshold, it indicates that factors such as equipment or human factors have a significant impact on the predicted scrap rate, and therefore the predicted scrap rate needs to be corrected.
[0115] In some possible embodiments, the predicted scrap rate is corrected, specifically as follows: Figure 13 The steps shown are as follows: In step 1301: Obtain the processing feature values corresponding to special dates within the historical time period.
[0116] In this embodiment of the application, considering that holidays and other special dates have a significant impact on the scrap rate, it is necessary to determine whether there are special dates within the historical time period. If there are special dates within the historical time period, it is necessary to obtain the processing feature value corresponding to the special date and use the processing feature value corresponding to the special date to process the predicted scrap rate.
[0117] In step 1302: The processing feature value, trend scrap rate, periodic scrap rate and residual scrap rate are input into the activation function to obtain the activation correction value.
[0118] In some possible embodiments, the process of obtaining the activation correction value can be specifically implemented as shown in Formula 5, wherein: , (Formula 5) in, To activate the correction value, For the Sigmoid function, For trend scrap rate, For the periodic scrap rate, The residual scrap rate, Processing characteristic value corresponding to a special date and Here, are the function parameters corresponding to the activation function. It is a dimension The trainable row vectors, where 1 corresponds to the activation correction value. The output dimension, The total dimension corresponding to the input feature vector is determined by The splicing length is determined by; Each element in the table represents the activation correction value for the corresponding input feature pair. Influence weights: where, The first element in ( (This refers to the trend scrap rate) The weight represents the effect of the current trend scrap rate on the activation correction value. The influence; the second element ( ( ) represents the periodic scrap rate The weight represents the effect of the periodic scrap rate on the activation correction value. The influence; the third element ( ) represents the residual scrap rate component The weights represent the effect of residual scrap rate on activation correction values. The influence of the 4th element ( () represents the processing characteristic value corresponding to a specific date. The weight represents the processing feature value corresponding to a specific date. Activation correction value The impact; The trainable bias parameter in the activation function is used to adjust the baseline offset of the activation correction value; that is, when the features in the input activation parameters are zero, the activation correction value... ;like Negative values (e.g.) = ),default A value close to 0 indicates no significant need for correction; if Positive values (e.g.) =0.5), default A value close to 0.62 indicates the strength of the baseline correction.
[0119] In step 1303: the trend scrap rate is corrected based on the activation correction value to obtain the corrected trend scrap rate; and the cycle scrap rate is corrected based on the activation correction value to obtain the corrected cycle scrap rate.
[0120] In some possible embodiments, the corrected trend scrap rate can be specifically implemented as Equation 6, where: , (Formula 6) in, This is the corrected trend scrap rate. The original trend scrap rate. To activate the correction value, The residual scrap rate, For processing feature values corresponding to specific dates, For a dimension of 1× k Trainable vectors, Each element in corresponds to The direction and magnitude of the correction to the trend scrap rate, i.e. Will As a moderating factor, it controls the direction (positive / negative) and magnitude (-1 to 1 times the residual) of the correction to the trend scrap rate.
[0121] In some possible embodiments, the corrected cycle failure rate can still be obtained as Equation 7, where: , (Formula 7) in, This is the corrected cycle scrap rate. The original cycle scrap rate. To activate the correction value, The residual scrap rate, It is a rectified unit function, used to ensure that only positive corrections are performed. It is a trainable scalar parameter with a dimension of 1×1, used to characterize the degree of influence of residual scrap rate on periodic scrap rate, and used to set the ratio of residual scrap rate to periodic scrap rate correction gain.
[0122] In step 1304: The predicted scrap rate is corrected based on the corrected trend scrap rate and the corrected periodic scrap rate to obtain the corrected predicted scrap rate.
[0123] In some possible embodiments, the corrected predicted scrap rate can be specifically implemented as Equation 8, wherein: , (Formula 8) in, This is the revised predicted scrap rate. The original predicted scrap rate. For the pre-set attenuation factor, This is the corrected trend scrap rate. This is the corrected periodic scrap rate.
[0124] In summary, the material scrap rate prediction method provided in this application uses historical data to predict future scrap rates. Furthermore, it identifies long-term trends, recurring patterns, or periodic data within the scrap data, as well as unpredictable random fluctuations, to jointly predict future scrap rates, resulting in more accurate predictions. Secondly, accurate scrap rate prediction reduces raw material waste caused by defective products during production. The prediction results help in rationally planning inventory levels, avoiding over-purchasing or insufficient inventory, and reducing inventory holding costs. A reduced scrap rate means fewer products requiring reprocessing or repair, thus saving on the manpower and material costs associated with rework. Finally, the prediction results can be used to identify weak points in the production process, allowing for timely improvements and enhancing the overall efficiency of the production line. Scrap prediction also helps quality management departments identify potential quality problems in advance, take preventative measures, and ensure stable product quality.
[0125] Based on the same inventive concept, after introducing a material scrap rate prediction method provided by the embodiments of this application, as follows: Figure 14 As shown, the following describes a material scrap rate prediction device 1400 provided in an embodiment of this application. The device includes: The acquisition module 14001 is used to acquire the processing history of each piece of material within a historical time period; The extraction module 14002 is used to perform feature extraction processing on the processing history of each piece of material to obtain the processing feature value of each piece of material; The aggregation module 14003 is used to perform data aggregation processing on the processing characteristic values of each piece of material to obtain the target characteristic value corresponding to the historical time period; The prediction module 14004 is used to obtain the trend scrap rate, periodic scrap rate and residual scrap rate based on the target feature value; The fusion module 14005 is used to fuse the trend scrap rate, the periodic scrap rate and the residual scrap rate to obtain the predicted scrap rate.
[0126] In some possible embodiments, the extraction module 14002 is specifically used for: Perform the following for each piece of material: Based on the material's coding information, obtain the material's processing station information and processing process information; The processing station information and the processing process information are processed to obtain the processing feature value corresponding to the material.
[0127] In some possible embodiments, the processing feature value includes at least one of the following: Workstation transfer interval, process transfer interval, dwell time at each workstation, dwell time at each process, dwell percentage, transfer time percentage, number of process processing steps, number of workstation processing steps, process inspection results, and yield.
[0128] In some possible embodiments, the aggregation module 14003 is specifically used for: Perform any one or a combination of the following steps for each type of processing feature value: The average value of the processing characteristic value of each piece of material is obtained by averaging the values. The variance of the processing characteristic value of each piece of material is processed to obtain the variance characteristic value; The processing characteristic value of each piece of material is processed by quantiles to obtain quantile characteristic values.
[0129] In some possible embodiments, the prediction module 14004 is specifically used for: Determine a first target feature value corresponding to the trend scrap rate from the target feature values; Determine a second target feature value corresponding to the cycle scrap rate from the target feature values; A third target feature value corresponding to the residual scrap rate is determined from the target feature values; Input the first target feature value into the first prediction model to obtain the trend scrap rate output by the first prediction model; Input the second target feature value into the second prediction model to obtain the cycle scrap rate output by the second prediction model; The third target feature value is input into the third prediction model to obtain the residual scrap rate output by the third prediction model.
[0130] In some possible embodiments, the fusion module 14005 is specifically used for: Obtain the end date corresponding to the historical time period; At least one target date is obtained based on the preset duration window length of the end date; Obtain the target trend scrap rate, target period scrap rate, and target residual scrap rate for each target date; Dependency parameters are obtained based on the target trend scrap rate, target periodic scrap rate, and target residual scrap rate corresponding to each target date; wherein, the dependency parameters characterize the dependency relationship between the trend scrap rate, the periodic scrap rate, and the residual scrap rate; By inputting the dependency parameters into the fully connected network model, we can obtain the first weight corresponding to the trend scrap rate, the second weight corresponding to the period scrap rate, and the third weight corresponding to the residual scrap rate. The predicted scrap rate is obtained by weighting and summing the first weight, the second weight, the third weight, the trend scrap rate, the periodic scrap rate, and the residual scrap rate.
[0131] In some possible embodiments, the fusion module 14005 is further configured to: determine the time period corresponding to the historical time period; Based on the time period, the residual scrap rate of multiple time periods corresponding to the first time period is obtained; wherein, the first time period is before the historical time period; The residual mean and standard deviation are obtained based on the residual scrap rate corresponding to the historical time period and the residual scrap rate corresponding to the first time period; The residual threshold is obtained based on the residual mean and the standard deviation. If the absolute value of the residual scrap rate is greater than the residual threshold, the predicted scrap rate is corrected.
[0132] In some possible embodiments, the fusion module 14005 is specifically used for: Obtain the processing feature values corresponding to specific dates within the historical time period; The processing characteristic value, the trend scrap rate, the periodic scrap rate, and the residual scrap rate are input into the activation function to obtain the activation correction value. The trend scrap rate is corrected based on the activation correction value to obtain the corrected trend scrap rate; and the periodic scrap rate is corrected based on the activation correction value to obtain the corrected periodic scrap rate. The predicted scrap rate is corrected based on the corrected trend scrap rate and the corrected periodic scrap rate to obtain the corrected predicted scrap rate.
[0133] Corresponding to the above embodiments, this application also provides an electronic device. Figure 15 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. The electronic device 1500 may include a processor 1501, a memory 1502, and a communication unit 1503. These components communicate through one or more buses. Those skilled in the art will understand that the structure of the electronic device shown in the figure does not constitute a limitation on the embodiment of the present invention. It may be a bus topology or a star topology, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.
[0134] The communication unit 1503 is used to establish a communication channel, enabling the electronic device to communicate with other devices. It receives user data from other devices or sends user data to other devices.
[0135] The processor 1501 serves as the control center of the electronic device, connecting various parts of the device via interfaces and lines. It executes software programs and / or modules stored in the memory 1502 and retrieves data stored in the memory to perform various functions and / or process data. The processor may be composed of integrated circuits (ICs), such as a single packaged IC or multiple packaged ICs with the same or different functions connected together. For example, the processor 1501 may consist only of a central processing unit (CPU). In this embodiment, the CPU may have a single processing core or include multiple processing cores.
[0136] The memory 1502 is used to store the execution instructions of the processor 1501. The memory 1502 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.
[0137] When the execution instructions in memory 1502 are executed by processor 1501, the electronic device 1500 is able to perform operations. Figure 2 Some or all of the steps in the illustrated embodiments.
[0138] In a specific implementation, the present invention also provides a computer storage medium, wherein the computer storage medium may store a program, which, when executed, may include some or all of the steps in the various embodiments of the material scrap rate prediction method provided by the present invention. The storage medium may be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.
[0139] Those skilled in the art will clearly understand that the techniques in the embodiments of the present invention can be implemented using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solutions in the embodiments of the present invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or certain parts of the embodiments of the present invention.
[0140] The same or similar parts between the various embodiments in this specification can be referred to mutually. In particular, the device embodiments and terminal embodiments are basically similar to the method embodiments, so the description is relatively simple, and the relevant parts can be referred to the description in the method embodiments.
Claims
1. A method for predicting material scrap rate, characterized in that, The method includes: Obtain the processing history of each piece of material within a historical time period; The processing history of each piece of material is processed by feature extraction to obtain the processing feature value of each piece of material; The processing characteristic values of each piece of material are aggregated to obtain the target characteristic value corresponding to the historical time period; Based on the target feature values, the trend scrap rate, periodic scrap rate, and residual scrap rate are obtained; The trend scrap rate, the periodic scrap rate, and the residual scrap rate are fused together to obtain the predicted scrap rate.
2. The method according to claim 1, characterized in that, The process of extracting features from the processing history of each piece of material to obtain the processing feature values of each piece of material includes: Perform the following for each piece of material: Based on the material's coding information, obtain the material's processing station information and processing process information; The processing station information and the processing process information are processed to obtain the processing feature value corresponding to the material.
3. The method according to claim 2, characterized in that, The processing feature value includes at least one of the following: Workstation transfer interval, process transfer interval, dwell time at each workstation, dwell time at each process, dwell percentage, transfer time percentage, number of process processing steps, number of workstation processing steps, process inspection results, and yield.
4. The method according to claim 1, characterized in that, The process of aggregating the processing feature values of each piece of material to obtain the target feature value corresponding to the historical time period includes: Perform any one or a combination of the following steps for each type of processing feature value: The average value of the processing characteristic value of each piece of material is obtained by averaging the values. The variance of the processing characteristic value of each piece of material is processed to obtain the variance characteristic value; The processing characteristic value of each piece of material is processed by quantiles to obtain quantile characteristic values.
5. The method according to claim 1, characterized in that, The process of obtaining the trend scrap rate, periodic scrap rate, and residual scrap rate based on the target feature value includes: Determine a first target feature value corresponding to the trend scrap rate from the target feature values; Determine a second target feature value corresponding to the cycle scrap rate from the target feature values; A third target feature value corresponding to the residual scrap rate is determined from the target feature values; Input the first target feature value into the first prediction model to obtain the trend scrap rate output by the first prediction model; Input the second target feature value into the second prediction model to obtain the cycle scrap rate output by the second prediction model; The third target feature value is input into the third prediction model to obtain the residual scrap rate output by the third prediction model.
6. The method according to claim 1, characterized in that, The process of fusing the trend scrap rate, the periodic scrap rate, and the residual scrap rate to obtain the predicted scrap rate includes: Obtain the end date corresponding to the historical time period; At least one target date is obtained based on the preset duration window length of the end date; Obtain the target trend scrap rate, target period scrap rate, and target residual scrap rate for each target date; Dependency parameters are obtained based on the target trend scrap rate, target periodic scrap rate, and target residual scrap rate corresponding to each target date; wherein, the dependency parameters characterize the dependency relationship between the trend scrap rate, the periodic scrap rate, and the residual scrap rate; By inputting the dependency parameters into the fully connected network model, we can obtain the first weight corresponding to the trend scrap rate, the second weight corresponding to the period scrap rate, and the third weight corresponding to the residual scrap rate. The predicted scrap rate is obtained by weighting and summing the first weight, the second weight, the third weight, the trend scrap rate, the periodic scrap rate, and the residual scrap rate.
7. The method according to claim 6, characterized in that, The method further includes: Determine the time period corresponding to the historical time period; Based on the time period, the residual scrap rate of multiple time periods corresponding to the first time period is obtained; wherein, the first time period is before the historical time period; The residual mean and standard deviation are obtained based on the residual scrap rate corresponding to the historical time period and the residual scrap rate corresponding to the first time period; The residual threshold is obtained based on the residual mean and the standard deviation. If the absolute value of the residual scrap rate is greater than the residual threshold, the predicted scrap rate is corrected.
8. The method according to claim 7, characterized in that, The correction process for the predicted scrap rate includes: Obtain the processing feature values corresponding to specific dates within the historical time period; The processing characteristic value, the trend scrap rate, the periodic scrap rate, and the residual scrap rate are input into the activation function to obtain the activation correction value. The trend scrap rate is corrected based on the activation correction value to obtain the corrected trend scrap rate; and the periodic scrap rate is corrected based on the activation correction value to obtain the corrected periodic scrap rate. The predicted scrap rate is corrected based on the corrected trend scrap rate and the corrected periodic scrap rate to obtain the corrected predicted scrap rate.
9. An electronic device, characterized in that, It includes a memory for storing computer program instructions and a processor for executing the program instructions, wherein when the computer program instructions are executed by the processor, the electronic device is triggered to perform the method of any one of claims 1-8.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein, when the program is executed, it controls the device on which the computer-readable storage medium is located to perform the method according to any one of claims 1-8.