Machine learning method based on working conditions and related devices and medium program products

Abstract

The embodiment of the present application provides a machine learning method based on working conditions and related devices and medium program products. The method obtains material inspection data of the target working condition, production line working condition data of the target working condition and working condition coding rules, determines the target coding of the target working condition according to the material inspection data, production line working condition data and working condition coding rules, compares the target coding with the preset coding of the target working condition, and if the target coding is inconsistent with the preset coding, and the target working condition has not created a machine learning task, then establishes a first machine learning task for the target working condition, configures a first machine training parameter for the first machine learning task; inputs the first machine training parameter into the product parameter prediction model, performs machine training on the target working condition, obtains multiple first learning results, and determines the benchmark value from the multiple first learning results. In this way, an accurate product parameter prediction model is obtained through machine training, which can improve the efficiency of process optimization and obtain accurate product parameters.

Description

Technical Field

[0001] The present application belongs to the field of production data processing technology, and specifically relates to a machine learning method based on working conditions and related devices and medium program products. Background Art

[0002] At present, the production process optimization of traditional chemical plants and process manufacturing enterprises relies on the experience accumulation of front-line operators. This approach makes process improvement difficult and time-consuming, and excellent experience cannot be shared or solidified and promoted. At the same time, traditional experience lacks the support of big data and precise statistical analysis, and cannot meet the needs of current intelligent production. Summary of the invention

[0003] The embodiments of the present application provide a machine learning method based on working conditions and related devices and medium program products, in order to improve the efficiency of process optimization.

[0004] In a first aspect, an embodiment of the present application provides a machine learning method based on working conditions, which is applied to an electronic device, and the method includes:

[0005] Acquire material inspection data of a target operating condition, production line operating condition data of the target operating condition, and an operating condition coding rule, wherein the target operating condition is any production process of a target product in a production process, the material inspection data includes at least one first material type involved in the target operating condition, the production line operating condition data includes at least one first process parameter corresponding to a production process in a production process of the target operating condition, and the operating condition coding rule includes a first coding rule for the at least one first material type and a second coding rule for the at least one first process parameter;

[0006] Determine the target code of the target working condition according to the material inspection data, the production line working condition data and the working condition coding rule;

[0007] Comparing the target code with a preset code of the target working condition;

[0008] If the target code is inconsistent with the preset code, and the target working condition does not create a machine learning task, then establishing a first machine learning task for the target working condition;

[0009] configuring first machine training parameters for the first machine learning task;

[0010] The first machine training parameters are input into a product parameter prediction model, machine training is performed on the target working condition to obtain multiple first learning results, and a benchmark value is determined from the multiple first learning results. The benchmark value is provided as a product parameter standard for the target product in mass production testing.

[0011] In a second aspect, an embodiment of the present application provides a machine learning device based on working conditions, which is applied to an electronic device, and the device includes:

[0012] an acquisition unit, configured to acquire material inspection data of a target operating condition, production line operating condition data of the target operating condition, and an operating condition coding rule, wherein the target operating condition is any production process of a target product in a production process, the material inspection data includes at least one first material type involved in the target operating condition, the production line operating condition data includes at least one first process parameter corresponding to a production process in a production process of the target operating condition, and the operating condition coding rule includes a first coding rule for the at least one first material type and a second coding rule for the at least one first process parameter;

[0013] A coding unit, used to determine a target code of the target working condition according to the material inspection data, the production line working condition data and the working condition coding rule;

[0014] A comparison unit, used for comparing the target code with a preset code of the target working condition;

[0015] a creating unit, configured to establish a first machine learning task for the target operating condition if the target code is inconsistent with the preset code and the machine learning task is not created for the target operating condition;

[0016] a configuration unit, configured to configure a first machine training parameter for the first machine learning task,

[0017] A training unit is used to input the first machine training parameters into a product parameter prediction model, perform machine training on the target working condition, obtain multiple first learning results, and determine a benchmark value from the multiple first learning results. The benchmark value is provided as a product parameter standard for the target product in mass production testing.

[0018] In a third aspect, an embodiment of the present application provides an electronic device, comprising a processor, a memory, a communication interface, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the processor, and the program includes instructions for executing the steps of any one of the first and second aspects of the embodiments of the present application.

[0019] In a fourth aspect, an embodiment of the present application provides a computer storage medium, characterized in that it stores a computer program for electronic data exchange, wherein the computer program enables a computer to execute part or all of the steps described in any one of the first and second aspects of the present embodiment.

[0020] In a fifth aspect, an embodiment of the present application provides a computer program product, wherein the computer program product includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to execute some or all of the steps described in any one of the first and second aspects of the embodiment of the present application. The computer program product may be a software installation package.

[0021] It can be seen that in the embodiment of the present application, firstly, the material inspection data of the target working condition, the production line working condition data of the target working condition and the working condition coding rules are obtained, and then the target code of the target working condition is determined according to the material inspection data, the production line working condition data and the working condition coding rules, and then the target code is compared with the preset code of the target working condition. If the target code is inconsistent with the preset code, it is determined whether a machine learning task has been created for the target working condition. If a machine learning task has not been established for the target working condition, a first machine learning task is established for the target working condition; a first machine training parameter is configured for the first machine learning task, and the target working condition is machine trained to obtain a trained product parameter prediction model. In this way, an accurate product parameter prediction model is obtained through machine training, which can improve the efficiency of process optimization and obtain accurate product data. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0023] Figure 1 It is a schematic diagram of a system architecture provided by an embodiment of the present application;

[0024] Figure 2 It is a flowchart of a machine learning method based on working conditions provided in an embodiment of the present application;

[0025] Figure 3 It is a structural schematic diagram of a working condition-based machine learning device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0026] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0027] The terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, systems, products or devices.

[0028] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0029] The following is an introduction to the relevant terms involved in this application.

[0030] Machine learning: Machine learning evolved into a branch of artificial intelligence (AI) in the second half of the 20th century. It uses self-learning algorithms to obtain knowledge from data to make predictions. Machine learning does not require manual analysis of large amounts of data in advance, and then extracting rules and building models. Instead, it provides a more effective way to capture knowledge in the data and gradually improve the performance of the prediction model to complete data-driven decision-making.

[0031] Multiple working conditions: The definition of working conditions is the combination of the change range of the characteristic values ​​of the input factors (man-machine-material-method-environment-measurement) of the production process. In actual production, the combination of these factors will be divided and coded according to the characteristic range of the factors to form different working condition codes. Different working condition combinations will have a significant impact on the control parameters of the production process. Multiple working conditions means that the production process is classified and managed according to different working conditions.

[0032] At present, the production process optimization of traditional chemical plants and process manufacturing enterprises relies on the accumulation of experience of front-line operators. This method makes process improvement difficult and time-consuming, and excellent experience cannot be shared, solidified and promoted. At the same time, traditional experience lacks the support of big data and precise statistical analysis, and cannot adapt to the needs of current intelligent production. In addition, in the production process of traditional production enterprises, the operating boundary conditions of production equipment are constantly changing, including but not limited to: variable quality of incoming materials, variable climate, variable equipment, and variable load. This makes it difficult to apply excellent experience in equipment operation. At the same time, there is often only one large and comprehensive production process standard in the production process. The quality indicators of products produced according to this standard fluctuate greatly, and the quality is very unstable. Engineers need to adjust process parameters according to their own experience. There is a lack of a mechanism for multi-condition classification management of equipment operation with optimal parameters to ensure the achievement of product quality, cost, and efficiency indicators.

[0033] To solve the above problems, an embodiment of the present application provides a machine learning method based on working conditions. This method can be applied to scenarios where product parameters are predicted or optimized. The target code of the target working condition can be determined by obtaining the material inspection data of the target working condition, the production line working condition data and the working condition coding rules of the target working condition, and then the target code is determined according to the material inspection data, the production line working condition data and the working condition coding rules, and then the target code is compared with the preset code of the target working condition. If the target code is inconsistent with the preset code, it is determined whether a machine learning task has been created for the target working condition. If a machine learning task has not been established for the target working condition, a first machine learning task is established for the target working condition; a first machine training parameter is configured for the first machine learning task, and machine training is performed on the target working condition to obtain a trained product parameter prediction model. This solution can be applied to a variety of scenarios, including but not limited to the application scenarios mentioned above.

[0034] The following introduces the system architecture involved in the embodiments of the present application.

[0035] The present application also provides an electronic device 10, such as Figure 1 As shown, it includes at least one processor 11; a display screen 12; and a memory 13, and may also include a communications interface 15 and a bus 14. The processor 11, the display screen 12, the memory 13 and the communications interface 15 can communicate with each other through the bus 14. The display screen 12 is configured to display a preset user guide interface in the initial setting mode. The communications interface 15 can transmit information. The processor 11 can call the logic instructions in the memory 13 to execute the method in the above embodiment.

[0036] Optionally, the electronic device 10 may be a mobile electronic device, or an electronic device or other device, which is not limited here.

[0037] In addition, the logic instructions in the memory 13 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product.

[0038] The memory 13, as a computer-readable storage medium, can be configured to store software programs, computer executable programs, such as program instructions or modules corresponding to the methods in the embodiments of the present disclosure. The processor 11 executes functional applications and data processing by running the software programs, instructions or modules stored in the memory 13, that is, implementing the methods in the above embodiments.

[0039] The memory 13 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and at least one application required for a function; the data storage area may store data created according to the use of the electronic device 10, etc. In addition, the memory 13 may include a high-speed random access memory and may also include a non-volatile memory. For example, a variety of media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, may also be a transient storage medium.

[0040] The specific methods are introduced in detail below.

[0041] See also Figure 2 The present application also provides a machine learning method based on working conditions, which can be applied to the electronic device described above. The method includes:

[0042] Step 201: Obtain material inspection data of a target working condition, production line working condition data of the target working condition, and working condition coding rules.

[0043] Among them, the target operating condition is any production process of the target product in the production process, the material inspection data includes at least one first material type involved in the target operating condition, the production line operating condition data includes at least one first process parameter corresponding to the production process in the production process of the target operating condition, and the operating condition coding rule includes a first coding rule for the at least one first material type and a second coding rule for the at least one first process parameter.

[0044] Step 202: determine the target code of the target working condition according to the material inspection data, the production line working condition data and the working condition coding rule.

[0045] In a possible embodiment, determining the target code of the target operating condition based on the material inspection data, the production line operating condition data and the operating condition coding rule includes: determining at least one first material code corresponding to the at least one first material type from the first coding rule; determining at least one first process code corresponding to the at least one first process parameter from the second coding rule; and combining the at least one first material code and the at least one first process code according to the operating condition coding rule to obtain the target code.

[0046] In a specific implementation, the first coding rule includes the first material type and its corresponding first material code; for example, soybean free fatty acids, its first material code in the first coding rule is 1; similarly, the second coding rule includes the first process parameter and its corresponding first process code, such as ambient temperature, different ambient temperatures (i.e., first process parameters) have different first process codes, such as -10℃-10℃ is coded as 1, 10℃-20℃ is coded as 2, 20℃-30℃ is coded as 3, and greater than 30℃ is coded as 4. It can be seen that after determining the first material type, the first material code corresponding to the first material type can be queried in the first coding rule, and after determining the first process parameter corresponding to the first material type, the first process code corresponding to the first process parameter can be queried from the second coding rule. In addition, the operating condition coding rules also include the arrangement of the first material type and the first process parameter. For example, the first digit in the target code is used to indicate the first material type, and the second digit is used to indicate the process parameter. Therefore, taking the first material type as soybean free fatty acids and the first process parameter as the ambient temperature of 20°C-30°C as an example, after determining the first material code and the first process code, the target code obtained according to the arrangement in the operating condition coding rules is: 13.

[0047] It is understandable that there can be multiple first material types, and there can also be multiple first process parameters, which are not limited to ambient temperature, but can also be pressure, space size, humidity, usage, etc. The length of the target code can be adaptively adjusted according to the number of first material types and the number of first process parameters involved in the target working conditions, and no uniqueness limitation is made here.

[0048] It can be seen that in this embodiment, corresponding codes are set for different materials and processes, and then the codes are combined to obtain the target code, which realizes the classified management of materials and processes, improves the refinement of production, and further improves product quality.

[0049] Step 203: Compare the target code with the preset code of the target working condition.

[0050] In a specific implementation, the purpose of comparing the target code with the preset code of the target condition is to determine whether the target condition is an old condition for which a machine learning task has been established or a new condition for which a machine learning task has not been established. The following situations exist during the comparison process: (1) The preset code may not exist for the target condition, in which case the comparison result is inconsistent; (2) The preset code of the target condition is found, but the preset code is inconsistent with the target code, and the comparison result is inconsistent; (3) The preset code of the target condition is found, and the preset code is consistent with the target code, and the comparison result is consistent. When the comparison result is consistent, the target condition is determined to be the old condition; when the comparison result is inconsistent, the target condition is determined to be the new condition.

[0051] In a possible embodiment, after comparing the target code with the preset code of the target working condition, the method also includes: if the target code is consistent with the preset code and the target working condition has not started the machine learning task, determining whether the target working condition has the benchmark value; if the target working condition does not have the benchmark value, establishing a second machine learning task for the target working condition; configuring second machine training parameters for the second machine learning task, training the target working condition, and obtaining a benchmark value; or, setting the product parameters of the initial R&D version to the benchmark value of the target working condition; if the target working condition does not have the benchmark value, but there is a first learning result that meets the requirements, using the product parameters corresponding to the first learning result as the benchmark value.

[0052] In a specific implementation, when the target code is consistent with the preset code, it is confirmed that the target working condition is the old working condition for which the machine learning task has been established. If there is no benchmark value for the target working condition, it means that the machine learning task has not been executed, and then the product parameters of the initial R&D version determined in the R&D stage are queried as the benchmark value as the second machine training parameter of the second machine learning task, and the established machine learning task is executed to obtain the benchmark value. It can be understood that there is another situation in this embodiment, that is, the benchmark value does not exist for the target working condition, but there are learning results that meet the requirements. At this time, it means that the machine learning task has been executed, but the learning results that may be obtained do not meet the requirements, so that the benchmark value has not been determined. At this time, the product parameters corresponding to the learning results can be set to the benchmark value, and optimization processing can be performed after there is an improvement plan in the future.

[0053] It can be seen that in this embodiment, when the encoding comparison results are consistent, it is further determined whether the machine learning task is executed to ensure that each working condition can obtain the corresponding benchmark value.

[0054] In a possible embodiment, after setting the process parameters of the initial R&D version as the benchmark values ​​of the target working conditions, the method further includes: if the target code is consistent with the preset code, and the target working condition has started the machine learning task for machine training, querying the correspondence between the training set and the target code from the scorecard database according to the target code; querying the corresponding training set from the training set database according to the correspondence; and setting the training set as the machine training parameters of the machine learning task.

[0055] In a specific implementation, when the target code is consistent with the preset code, if it is detected that the target working condition has started the established machine learning task, it means that machine training is in progress, which may be one of multiple iterative trainings or a restarted machine training task. At this time, the correspondence between the training set and the target code is queried from the scorecard database according to the target code; the corresponding training set is queried from the training set database according to the correspondence; and the training set is set as the machine training parameter of the machine learning task.

[0056] It can be seen that in this embodiment, it is achieved that machine training parameters used for iteration are provided during the machine learning process to complete multiple machine trainings.

[0057] In a possible embodiment, after comparing the target code with the preset code of the target working condition, the method further includes: if the target code is inconsistent with the preset code, and the target working condition has established the machine learning task, querying the correspondence between the training set and the target code from the scorecard database according to the target code; querying the corresponding training set from the training set database according to the correspondence; and setting the training set as the first machine training parameter of the machine learning task.

[0058] In a specific implementation, when the target code is inconsistent with the preset code, if it is detected that the target working condition has established the machine learning task, it means that the working condition is a newly created idle working condition, but the corresponding machine training data is not configured. Therefore, it is only necessary to query the corresponding relationship between the training set and the target code from the scorecard database according to the target code; query the corresponding training set from the training set database according to the corresponding relationship; and set the training set as the first machine training parameter of the machine learning task. After completing the parameter setting, the machine training and subsequent steps of obtaining the benchmark value are performed.

[0059] It can be seen that in this embodiment, the purpose of detecting the idle condition to restart the idle condition is achieved.

[0060] Step 204: If the target code is inconsistent with the preset code, and the target working condition does not create a machine learning task, then establish a first machine learning task for the target working condition.

[0061] In a specific implementation, when the target code is inconsistent with the preset code, it indicates that the target working condition currently being processed has not been machine trained or the parameters (material type, process parameters, etc.) in the target working condition have changed. Therefore, it is first determined whether a machine learning task has been created for the target working condition before, so as to facilitate subsequent processing. At this time, the above two situations are determined by determining whether a machine learning task has been established for the target working condition. If a machine learning task has been established, it means that a parameter change has occurred; if a machine learning task has not been established, it means that machine training has not been performed; after determining the situation, the desired identifier can be added to mark which situation this processing is in, such as adding a first identification code to indicate that a parameter change has occurred in the target working condition, and adding a second identification code to indicate that the target working condition has not been machine trained. If it is determined that a machine learning task has not been established for the target working condition, a first learning task is established for the target working condition.

[0062] In one possible embodiment, if a machine learning task has been established for the target working condition, the first material type and the first process parameter are used as new machine training parameters and the target working condition replaces the original machine training parameters of the machine learning task for the target working condition.

[0063] It can be seen that in this embodiment, when the parameters in the target working condition change, corresponding adjustments can be made to re-perform machine learning to obtain new benchmark values.

[0064] Step 205: configuring first machine training parameters for the first machine learning task;

[0065] In a possible embodiment, configuring the first machine training parameter for the first machine learning task includes: determining whether there is a first operating condition close to the target operating condition in the operating condition database; if the first operating condition exists, setting the first parameter to be the first machine training parameter of the first machine learning task, the first parameter being the first material inspection data and the first process parameter corresponding to the first benchmark value of the first operating condition, the first benchmark value being the optimal parameter value among multiple first product parameters obtained by the product parameter prediction model for the first operating condition.

[0066] For example, the operating condition database stores operating condition data of multiple operating conditions, and the operating condition data includes the above-mentioned material inspection data, production line operating condition data, benchmark values, and historical data of these parameters and other parameters, and the other parameters include but are not limited to preset codes.

[0067] In a specific implementation, since the above comparison result is inconsistent, the target working condition is a new working condition. In order to speed up machine learning, the new working condition queries a first working condition close to the target working condition in the database, and sets the first parameter corresponding to the benchmark value of the first working condition as the first machine training parameter of the first machine learning task, so that only fewer parameters need to be adjusted during the machine training process.

[0068] It can be seen that in this embodiment, the machine training of the target working condition is performed using parameters of similar working conditions, which speeds up the machine learning of the target working condition.

[0069] In a possible embodiment, the target code includes multiple sub-codes, each of the multiple sub-codes is the first material code or the first process code; the determination of whether there is a first operating condition similar to the target operating condition in the operating condition database includes: determining the impact level of the material or production process corresponding to each sub-code in the target code on the product, and the impact level is used to indicate the degree of impact of the material or production process on the product; querying at least one first code corresponding to at least one second operating condition with the highest similarity to the target code from the operating condition database; determining each second code in the at least one first code that is different from the target code, and obtaining at least one second code; determining a third code from the at least one second code, and the third code refers to the second code corresponding to the sub-code with the lowest impact level among the multiple sub-codes; and determining the second operating condition corresponding to the third code as the first operating condition.

[0070] For example, the target code and the first code have the same number of bits, and the target code and the code at the same position in the first code can be compared in sequence to determine whether there is a difference.

[0071] In a specific implementation, the operating condition with the smallest difference (i.e., the highest similarity) in the operating condition database is taken as the second operating condition. In the best case, there is only one second code that is different from the target code in the first code corresponding to each second operating condition. After determining the at least one second code, the second code corresponding to the subcode with the lowest impact level is determined according to the impact level of at least one subcode corresponding to the at least one second code, which is the third code. Then the second operating condition corresponding to the third code is determined as the first operating condition.

[0072] Specifically, there may be a situation where the difference between the second code and the target code is two-digit code, three-digit code or more-digit code. Since the large difference will increase the difficulty of machine learning, a first threshold is set. If the number of difference codes is greater than the first threshold, a prompt message is output to prompt the user to manually add the first machine training parameters.

[0073] It is understandable that the impact of materials on products is not necessarily greater or less than the impact of production processes on products. For example, some materials can be replaced. After changing to replaceable materials, the impact on products may be smaller. However, production processes such as temperature and humidity may also have a serious impact on products, resulting in inferior products, and vice versa.

[0074] It can be seen that in this embodiment, the first operating condition closest to the target operating condition can be queried based on the operating condition database, and then the target operating condition can be machine trained using the parameters of the similar operating condition, thereby speeding up the machine learning of the target operating condition.

[0075] 206. Input the first machine training parameter into a product parameter prediction model, perform machine training on the target working condition, obtain multiple learning results, and determine a benchmark value from the multiple learning results.

[0076] The benchmark value is provided as a product parameter standard for the target product in mass production testing.

[0077] In a possible embodiment, the step of inputting the first machine training parameter into a product parameter prediction model, performing machine training on the target working condition, obtaining a plurality of learning results, and determining a benchmark value from the plurality of learning results includes: inputting the first machine training parameter into the product parameter prediction model for multiple iterative training to obtain a plurality of first learning results; scoring the plurality of first learning results according to the target parameters corresponding to the target working condition, the target parameters being the expected parameters of the product corresponding to the target working condition; if there is at least one second learning result among the plurality of first learning results whose score is greater than a first preset value, determining that the training is completed; setting the product parameter corresponding to the first learning result with the highest score among the at least one second learning result as the benchmark value, and setting the first machine training parameter corresponding to the benchmark value as the benchmark process parameter.

[0078] In a specific implementation, the target parameter is set for the first machine learning task according to the product corresponding to the target working condition. After each training is completed and a learning result is obtained, the first learning result obtained by executing the first machine learning task is scored according to the target parameter. Scoring methods include: (1) scoring according to the error range with the target parameter; (2) setting multiple levels of target parameters, and giving the score of the corresponding level according to which level of target parameter the first learning result meets; (3) using the first method and the second method at the same time. After completing this scoring step, it is determined whether the score of the first learning result is greater than the first preset value. If so, it is determined as the second learning result. After obtaining the first second learning result, multiple iterations can be performed. If the scores of the obtained multiple second learning results are significantly different from those of the first second learning result, the iteration is continued. If the scores of the obtained multiple second learning results are different from those of the first second learning result within the first preset range, the machine learning is completed, and the second learning result with the highest score is used as the benchmark value, and the first machine training parameter corresponding to the benchmark value is set as the benchmark process parameter.

[0079] Furthermore, if the score of the first learning result is less than the first preset value but greater than the second preset value, it means that the first product parameter corresponding to the first learning result does not meet the standard. At this time, the electronic device sets the first product parameter corresponding to the first learning result to a non-mass production verification version. The first product parameter of the non-mass production verification version does not participate in the subsequent actual mass production verification phase process. The user can choose whether to directly use the first product parameter as the benchmark value. It is understandable that the first product parameter of the non-mass production verification version can be modified by the electronic device receiving instructions input by the user, and can be modified to a mass production verification version. At this time, the modified first product parameter can participate in the subsequent actual mass production verification phase process.

[0080] Furthermore, if the score of the first learning result is less than the third preset value, it means that the first product parameter corresponding to the first learning result not only fails to meet the standard, but also differs greatly from the expected parameter. At this time, the training parameter correction is performed based on the first product parameter, and then machine learning continues.

[0081] It can be seen that in this embodiment, multiple iterations of training are performed and the learning results are scored to determine the product parameters that best meet expectations, thereby ensuring the effectiveness of machine learning.

[0082] In a possible embodiment, after determining the benchmark value from the multiple first learning results, the method further includes: obtaining M first product parameters corresponding to N first products obtained through N actual productions based on the benchmark value, wherein N is a positive integer greater than zero, and M is a positive integer greater than or equal to N; calculating the product index of the product parameter corresponding to each of the N first products to obtain N product index values; if K of the M product index values ​​are greater than a first preset index value, determining that the benchmark value can be applied to actual production, wherein K is a positive integer less than or equal to M; if K of the M product index values ​​are within the range of the first preset index value and the second preset index value, determining that the benchmark value can be applied to actual production, wherein the second preset index value is less than the first preset index value; if K of the M product index values ​​are less than the second preset index value, determining that the benchmark value cannot be applied to actual production.

[0083] In a specific implementation, after completing machine training and determining the benchmark value, it is necessary to conduct an actual test on the benchmark value to ensure that the benchmark value can meet the requirements of the actual product. In the actual test, a small amount of trial production will be carried out on the production line. For example, N first products are obtained by N productions, and the obtained products are statistically analyzed to obtain the corresponding M first product data. The electronic device obtains the M first product data and obtains the corresponding product index value by calculation. The product index value can be a CPK value. Since the optimal CPK value is 1.33 and the CPK value required for a qualified product is 1.0, the first preset index value can be set to 1.33, and the second index value of the next level can be set to 1.0. If the CPK value is less than 1.0, it is a failed product. Furthermore, if more than K product index values ​​among the N CPK values ​​are greater than the first preset index value, it means that the first product is very good. Therefore, it is determined that the benchmark value can be applied to actual production. If more than K product index values ​​among the N CPK values ​​are within the range of the first preset index value and the second preset index value, it means that the yield of the first product is within the preset range, so it is determined that the benchmark value can be applied to actual production. If K product index values ​​among the N product index values ​​are less than the second preset index value, it means that the product yield is too low and there are many defective products, so it is determined that the benchmark value cannot be applied to actual production.

[0084] It is understandable that in the above three cases, the value of K may be the same or different, and is not limited to uniqueness.

[0085] It can be seen that in this embodiment, the actual production data is verified to ensure that the benchmark value obtained through training is a usable value, thereby avoiding significant losses in large-scale production.

[0086] In summary, in the embodiment of the present application, firstly, the material inspection data of the target working condition, the production line working condition data and the working condition coding rules of the target working condition are obtained, and then the target code of the target working condition is determined according to the material inspection data, the production line working condition data and the working condition coding rules, and then the target code is compared with the preset code of the target working condition. If the target code is inconsistent with the preset code, it is determined whether a machine learning task has been created for the target working condition. If a machine learning task has not been established for the target working condition, a first machine learning task is established for the target working condition; a first machine training parameter is configured for the first machine learning task, and the target working condition is machine trained to obtain a trained product parameter prediction model. In this way, an accurate product parameter prediction model is obtained through machine training, which can improve the efficiency of process optimization and obtain accurate product data.

[0087] The above mainly introduces the scheme of the embodiment of the present application from the perspective of the execution process on the method side. It is understandable that, in order to realize the above functions, the electronic device includes a hardware structure and / or software module corresponding to the execution of each function. It should be easily appreciated by those skilled in the art that, in combination with the units and algorithm steps of each example described in the embodiments provided herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present application.

[0088] The embodiment of the present application can divide the electronic device into functional units according to the above method example. For example, each functional unit can be divided according to each function, or two or more functions can be integrated into one processing unit. The above integrated unit can be implemented in the form of hardware or in the form of software functional units. It should be noted that the division of units in the embodiment of the present application is schematic and is only a logical function division. There may be other division methods in actual implementation.

[0089] See also Figure 3 , Figure 3 : is a structural diagram of a machine learning device based on working conditions provided in an embodiment of the present application. The machine learning device 40 based on working conditions is applied to an electronic device, and the device includes:

[0090] An acquisition unit 401 is used to acquire material inspection data of a target operating condition, production line operating condition data of the target operating condition, and an operating condition coding rule, wherein the target operating condition is any production process of a target product in a production process, the material inspection data includes at least one first material type involved in the target operating condition, the production line operating condition data includes at least one first process parameter corresponding to a production process in a production process of the target operating condition, and the operating condition coding rule includes a first coding rule for the at least one first material type and a second coding rule for the at least one first process parameter;

[0091] An encoding unit 402 is used to determine a target code of the target working condition according to the material inspection data, the production line working condition data and the working condition coding rule;

[0092] A comparison unit 403, configured to compare the target code with a preset code of the target working condition;

[0093] A creating unit 404, configured to create a first machine learning task for the target working condition if the target code is inconsistent with the preset code and the machine learning task is not created for the target working condition;

[0094] A configuration unit 405, configured to configure a first machine training parameter for the first machine learning task;

[0095] The training unit 406 is used to input the first machine training parameters into the product parameter prediction model, perform machine training on the target working condition, obtain multiple learning results, and determine a benchmark value from the multiple learning results. The benchmark value is provided as a product parameter standard for the target product in mass production testing.

[0096] It can be seen that in this embodiment, firstly, the material inspection data of the target working condition, the production line working condition data of the target working condition and the working condition coding rules are obtained, and then the target code of the target working condition is determined according to the material inspection data, the production line working condition data and the working condition coding rules, and then the target code is compared with the preset code of the target working condition. If the target code is inconsistent with the preset code, it is determined whether a machine learning task has been created for the target working condition. If a machine learning task has not been established for the target working condition, a first machine learning task is established for the target working condition; a first machine training parameter is configured for the first machine learning task, and the target working condition is machine trained to obtain a trained product parameter prediction model. In this way, an accurate product parameter prediction model is obtained through machine training, which can improve the efficiency of process optimization and obtain accurate product data.

[0097] In a possible embodiment, in terms of determining the target code of the target working condition based on the material inspection data, the production line working condition data and the working condition coding rule, the coding unit 402 is specifically used to: determine at least one first material code corresponding to the at least one first material type from the first coding rule; determine at least one first process code corresponding to the at least one first process parameter from the second coding rule; and combine the at least one first material code and the at least one first process code according to the working condition coding rule to obtain the target code.

[0098] In a possible embodiment, after comparing the target code with the preset code of the target working condition, the creation unit 404 is also used to: if the target code is consistent with the preset code and the target working condition has not started the machine learning task, determine whether the benchmark value exists for the target working condition; if the benchmark value does not exist for the target working condition, establish a second machine learning task for the target working condition; the configuration unit 405 is also used to: configure second machine training parameters for the second machine learning task; the training unit 406 is also used to: train the target working condition to obtain a benchmark value; or, set the product parameters of the initial R&D version to the benchmark value of the target working condition; if the benchmark value does not exist for the target working condition, but there is a first learning result that meets the requirements, the product parameters corresponding to the first learning result are used as the benchmark value.

[0099] In a possible embodiment, after setting the process parameters of the initial R&D version as the benchmark values ​​of the target working condition, the configuration unit 405 is also used to: if the target code is consistent with the preset code, and the target working condition has started the machine learning task for machine training, then query the correspondence between the training set and the target code from the scorecard database according to the target code; query the corresponding training set from the training set database according to the correspondence; and set the training set as the machine training parameter of the machine learning task.

[0100] In a possible embodiment, after setting the process parameters of the initial R&D version as the benchmark values ​​of the target working condition, the configuration unit 405 is also used to: if the target code is inconsistent with the preset code, and the target working condition has established the machine learning task, query the correspondence between the training set and the target code from the scorecard database according to the target code; query the corresponding training set from the training set database according to the correspondence; and set the training set as the first machine training parameter of the machine learning task.

[0101] In a possible embodiment, in terms of configuring the first machine training parameters for the first machine learning task, the configuration unit 405 is specifically used to: determine whether there is a first operating condition close to the target operating condition in the operating condition database; if the first operating condition exists, set the first parameter to be the first machine training parameter of the first machine learning task, the first parameter being the first material inspection data and the first process parameter corresponding to the first benchmark value of the first operating condition, and the first benchmark value being the optimal parameter value among multiple first product parameters obtained by the product parameter prediction model for the first operating condition.

[0102] In a possible embodiment, the target code includes multiple sub-codes, each of the multiple sub-codes is the first material code or the first process code; in determining whether there are aspects of the first operating condition similar to the target operating condition in the operating condition database, the configuration unit 405 is specifically used to: determine the impact level of the material or production process corresponding to each sub-code in the target code on the product, and the impact level is used to indicate the degree of impact of the material or production process on the product; query at least one first code corresponding to at least one second operating condition with the highest similarity to the target code from the operating condition database; determine each second code in the at least one first code that is different from the target code, and obtain at least one second code; determine a third code from the at least one second code, and the third code refers to the second code corresponding to the sub-code with the lowest impact level among the multiple sub-codes; determine the second operating condition corresponding to the third code as the first operating condition.

[0103] In a possible embodiment, in terms of inputting the first machine training parameter into a product parameter prediction model, performing machine training on the target working condition, obtaining multiple learning results, and determining a benchmark value from the multiple learning results, the training unit 406 is specifically used to: input the first machine training parameter into the product parameter prediction model for multiple iterative training to obtain multiple first learning results; score the multiple first learning results according to the target parameters corresponding to the target working condition, the target parameters being the expected parameters of the product corresponding to the target working condition; if there is at least one second learning result among the multiple first learning results whose score is greater than a first preset value, it is determined that the training is completed; and the product parameter corresponding to the first learning result with the highest score among the at least one second learning result is set as the benchmark value, and the first machine training parameter corresponding to the benchmark value is set as the benchmark process parameter.

[0104] In a possible embodiment, after determining the benchmark value from the multiple first learning results, the device further includes: an acquisition unit, used to acquire M first product parameters corresponding to N first products obtained by performing N actual productions based on the benchmark value, wherein N is a positive integer greater than zero, and M is a positive integer greater than or equal to N; a calculation unit, used to calculate the product index of the product parameter corresponding to each of the N first products to obtain N product index values; a determination unit, used to determine that the benchmark value can be applied to actual production when K product index values ​​among the N product index values ​​are greater than a first preset index value, wherein K is a positive integer less than or equal to N; when K product index values ​​among the N product index values ​​are within the range of the first preset index value and the second preset index value, it is determined that the benchmark value can be applied to actual production, wherein the second preset index value is less than the first preset index value; when K product index values ​​among the N product index values ​​are less than the second preset index value, it is determined that the benchmark value cannot be applied to actual production.

[0105] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired or wireless means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0106] An embodiment of the present application also provides a computer storage medium, wherein the computer storage medium stores a computer program for electronic data exchange, wherein the computer program enables a computer to execute part or all of the steps of any method described in the above method embodiments, and the above computer includes an electronic device.

[0107] The embodiment of the present application also provides a computer program product, the computer program product includes a non-transitory computer-readable storage medium storing a computer program, the computer program is operable to cause a computer to execute some or all of the steps of any method described in the method embodiment. The computer program product may be a software installation package, and the computer includes an electronic device.

[0108] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0109] In the several embodiments provided in the present application, it should be understood that the disclosed methods, devices and systems can be implemented in other ways. For example, the device embodiments described above are merely schematic; for example, the division of the units is only a logical function division, and there may be other division methods in actual implementation; for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0110] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0111] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may be physically included separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0112] The above-mentioned integrated unit implemented in the form of a software functional unit can be stored in a computer-readable storage medium. The above-mentioned software functional unit is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform some steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a volatile memory or a non-volatile memory. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM) or a flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM) and direct RAM bus RAM (DR RAM). Various media that can store program code are available.

[0113] Although the present invention is disclosed as above, the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions without departing from the spirit and scope of the present invention, and can make various changes and modifications, including the combination of the above-mentioned different functions and implementation steps, including software and hardware implementation methods, all of which are within the scope of protection of the present invention.

Claims

1. A machine learning method based on working conditions, characterized in that: The method comprises: Acquire material inspection data of a target operating condition, production line operating condition data of the target operating condition, and an operating condition coding rule, wherein the target operating condition is any production process of a target product in a production process, the material inspection data includes at least one first material type involved in the target operating condition, the production line operating condition data includes at least one first process parameter corresponding to a production process in a production process of the target operating condition, and the operating condition coding rule includes a first coding rule for the at least one first material type and a second coding rule for the at least one first process parameter; Determine the target code of the target working condition according to the material inspection data, the production line working condition data and the working condition coding rule; Comparing the target code with a preset code of the target working condition; If the target code is inconsistent with the preset code, and the target working condition does not create a machine learning task, then establishing a first machine learning task for the target working condition; configuring first machine training parameters for the first machine learning task; Inputting the first machine training parameter into a product parameter prediction model, performing machine training on the target working condition to obtain a plurality of first learning results, and determining a benchmark value from the plurality of first learning results, wherein the benchmark value is provided as a product parameter standard for the target product in mass production testing; The configuring of the first machine training parameter for the first machine learning task also includes: determining whether there is a first operating condition similar to the target operating condition in the operating condition database; the target code includes multiple sub-codes, each of the multiple sub-codes is the first material code or the first process code; the determining whether there is a first operating condition similar to the target operating condition in the operating condition database includes: determining the impact level of the material or production process corresponding to each sub-code in the target code on the product, the impact level is used to indicate the degree of impact of the material or production process on the product; querying at least one first code corresponding to at least one second operating condition with the highest similarity to the target code from the operating condition database; determining a second code in which each first code in the at least one first code is different from the target code, and obtaining at least one second code; determining a third code from the at least one second code, the third code being the second code corresponding to the sub-code with the lowest impact level among the multiple sub-codes; and determining the second operating condition corresponding to the third code as the first operating condition; If the number of the third codes is greater than the first threshold, a prompt message is output to prompt the user to manually add the first machine training parameter.

2. The method according to claim 1, characterized in that The determining the target code of the target working condition according to the material inspection data, the production line working condition data and the working condition coding rule comprises: Determining at least one first material code corresponding to the at least one first material type from the first coding rule; Determining at least one first process code corresponding to the at least one first process parameter from the second coding rule; The at least one first material code and the at least one first process code are combined according to the working condition coding rule to obtain the target code.

3. The method according to claim 1 or 2, characterized in that: After comparing the target code with the preset code of the target working condition, the method further includes: If the target code is consistent with the preset code, and the target operating condition does not initiate the machine learning task, determining whether the target operating condition has the benchmark value; If the benchmark value does not exist for the target operating condition, establishing a second machine learning task for the target operating condition; configuring a second machine training parameter for the second machine learning task, training the target working condition, and obtaining a benchmark value; or, Setting the product parameters of the initial R&D version to the benchmark values ​​of the target working conditions; If the benchmark value does not exist for the target operating condition, but there is a first learning result that meets the requirements, the product parameter corresponding to the first learning result is used as the benchmark value.

4. The method according to any one of claims 1 to 3, characterized in that: After comparing the target code with the preset code of the target working condition, the method further includes: If the target code is inconsistent with the preset code, and the target working condition has established the machine learning task, querying the corresponding relationship between the training set and the target code from the scorecard database according to the target code; According to the corresponding relationship, a corresponding training set is searched from a training set database; The training set is set as a first machine training parameter of the machine learning task.

5. The method according to any one of claims 1 to 4, characterized in that: The configuring the first machine training parameter for the first machine learning task also includes: If the first working condition exists, the first parameter is set to the first machine training parameter of the first machine learning task, the first parameter is the first product parameter corresponding to the first benchmark value of the first working condition, and the first benchmark value is the optimal parameter value among the multiple first product parameters obtained by the product parameter prediction model for the first working condition.

6. The method according to any one of claims 1 to 5, characterized in that: The step of inputting the first machine training parameter into a product parameter prediction model, performing machine training on the target working condition, obtaining a plurality of learning results, and determining a benchmark value from the plurality of learning results includes: Inputting the first machine training parameter into the product parameter prediction model for multiple iterative training to obtain multiple first learning results; Scoring the plurality of first learning results according to a target parameter corresponding to the target operating condition, wherein the target parameter is an expected parameter of the product corresponding to the target operating condition; If there is at least one second learning result among the plurality of first learning results whose score is greater than a first preset value, then determining that the training is completed; The product parameter corresponding to the first learning result with the highest score among the at least one second learning result is set as the benchmark value, and the first machine training parameter corresponding to the benchmark value is set as the benchmark process parameter.

7. A machine learning device based on working conditions, characterized in that: The device comprises: an acquisition unit, configured to acquire material inspection data of a target operating condition, production line operating condition data of the target operating condition, and an operating condition coding rule, wherein the target operating condition is any production process of a target product in a production process, the material inspection data includes at least one first material type involved in the target operating condition, the production line operating condition data includes at least one first process parameter corresponding to a production process in a production process of the target operating condition, and the operating condition coding rule includes a first coding rule for the at least one first material type and a second coding rule for the at least one first process parameter; A coding unit, used to determine a target code of the target working condition according to the material inspection data, the production line working condition data and the working condition coding rule; A comparison unit, used for comparing the target code with a preset code of the target working condition; a creating unit, configured to establish a first machine learning task for the target operating condition if the target code is inconsistent with the preset code and the machine learning task is not created for the target operating condition; a configuration unit, configured to configure a first machine training parameter for the first machine learning task, A training unit is used to input the first machine training parameters into a product parameter prediction model, perform machine training on the target working condition, obtain multiple first learning results, and determine a benchmark value from the multiple first learning results, wherein the benchmark value is provided as a product parameter standard for the target product in mass production testing; the configuration of the first machine training parameters for the first machine learning task also includes: determining whether there is a first working condition similar to the target working condition in the working condition database; the target code includes multiple sub-codes, each of the multiple sub-codes is the first material code or the first process code; the determination of whether there is a first working condition similar to the target working condition in the working condition database includes: determining the material corresponding to each sub-code in the target code or the impact level of the material or production process on the product, the impact level is used to indicate the degree of impact of the material or production process on the product; query at least one first code corresponding to at least one second working condition with the highest similarity to the target code from the working condition database; determine the second code in which each first code in the at least one first code is different from the target code, and obtain at least one second code; determine a third code from the at least one second code, the third code refers to the second code corresponding to the subcode with the lowest impact level among the multiple subcodes; determine the second working condition corresponding to the third code as the first working condition; if the number of the third codes is greater than the first threshold, output a prompt message to prompt the user to manually add the first machine training parameter.

8. An electronic device, characterized in that: The method comprises a processor, a memory, a communication interface, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the processor, wherein the programs include instructions for executing the steps in the method according to any one of claims 1 to 6, or the programs include instructions for the steps of the device according to claim 7.

9. A computer-readable storage medium or a computer program product, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program enables a computer to execute instructions for the steps in the method as described in any one of claims 1 to 6, or instructions for the steps in the device as described in claim 7; the computer program product enables a computer to execute instructions for the steps in the method as described in any one of claims 1 to 6, or instructions for the steps in the device as described in claim 7.

Images