Data Processing Method and Device for Multiple Industrial Models Based on Industrial Internet

Abstract

The present application discloses a data processing method and device for multiple industrial models based on the industrial Internet. The method includes: an application function server corresponding to an industrial model selects an application function adapted to the industry in a preset application function library according to a cloud server, a PaaS platform, and the industry corresponding to the industrial model; obtains personalized services based on the triggered application function; realizes model development and model testing in an online manner according to the operations performed by the model developer in the application function, so as to obtain a trained industrial model; packages the trained industrial model through a model packaging device and publishes it to the corresponding platform through the application function. Providing personalized application services according to the industry where the model developer is located can provide the required services for the model developer while reducing the appearance of redundant services, and can further improve the model development efficiency of the model developer.

Description

Technical Field

[0001] This application relates to the field of industrial Internet, and specifically to a data processing method and device for multiple industrial models based on the industrial Internet. Background Art

[0002] All aspects of an enterprise's R & D design, supply chain management, production manufacturing, quality control, operation management, warehousing and logistics, work safety, energy conservation and emission reduction, operation and maintenance services, etc. rely on the flow and application of data. Nowadays, simulation, mechanism inference, and data analysis are usually carried out in the form of industrial models.

[0003] Industrial models refer to various models applied in the industrial production environment, such as man-hour prediction models, mold production models, etc. They not only rely on computer information technologies such as architecture, language, database, computer graphics, etc., but also rely on the application of mathematical knowledge such as differential geometry, statistics, mechanics of materials, fluid mechanics, chemistry, molecular dynamics, thermodynamics, electrodynamics, etc. Finally, it also needs to go through the extraction of engineering experience, such as the practical tests of various processes, system, and process designs, before it can be actually applied.

[0004] Although the application of industrial models in various software has become mature, due to the closed nature of each software, no matter what industry the enterprise is in, the services it obtains are general services. General services usually only provide some basic services. For some industries, they cannot provide the services they need and will also provide redundant services that the enterprise does not need. This is not conducive to the enterprise's monitoring of production and cannot improve industrial production efficiency to a greater extent. Summary of the Invention

[0005] To solve the above problems, this application proposes a data processing method for multiple industrial models based on the industrial Internet, which is applied to an industrial model management system. The industrial model management system includes a cloud server for providing cloud infrastructure support, a PaaS platform for providing service PaaS support, one or more application function servers, and a model encapsulation device. The method includes: the application function server corresponding to the industrial model selects an application function adapted to the industry in a preset application function library according to the cloud server, the PaaS platform, and the industry corresponding to the industrial model; based on triggering the application function, personalized services are obtained; according to the operations performed by the model developer in the application function, model development and model testing are realized online to obtain the trained industrial model; the trained industrial model is encapsulated by the model encapsulation device and published to the corresponding platform through the application function.

[0006] In one example, to implement model testing in an online manner according to the operations performed by the model developer in the application function, it specifically includes: determining the data generalization level of the training samples uploaded by the model developer; if the data generalization level is lower than a first preset threshold, determining the data collection level corresponding to the industry of the model developer; if the data collection level is higher than a second preset threshold, generating virtual data based on the training samples uploaded by the model developer; replacing at least part of the data in the test set of the training samples with the virtual data, and implementing model testing in an online manner through the replaced test set.

[0007] In one example, the determining of the data generalization level of the training samples uploaded by the model developer specifically includes: for each training sample uploaded by the model developer, determining the values possessed in the key attributes of the training sample; determining the conceptual level at which the values are located in the space corresponding to the attributes; and determining the data generalization level of the training sample according to the proportion of the training sample in each conceptual level.

[0008] In one example, the generating of virtual data based on the training samples uploaded by the model developer specifically includes: determining the specified training samples whose conceptual levels are lower than the preset level in the training samples; determining the first values possessed in the key attributes of the specified training samples; generating second values with a higher conceptual level for the first values according to the industry status value of the industry, and replacing the first values with the second values; and replacing at least part of the non-key attributes of the training samples with other attributes that can be obtained based on the training samples; taking the replaced training samples as the generated virtual data.

[0009] In one example, to implement model testing in an online manner according to the operations performed by the model developer in the application function, it specifically includes:

[0010] Implementing model testing in an online manner respectively through the test set before replacement and the test set after replacement to obtain the first test result corresponding to before replacement and the second test result corresponding to after replacement; determining the replacement rate of the test set by the virtual data; evaluating the impact of the data generalization level on the model development according to the replacement rate and the difference between the first test result and the second test result; and presenting the evaluation result to the model developer.

[0011] In one example, before the trained industrial model is encapsulated by the model encapsulation device and published to the corresponding platform through the application function, the method further includes: obtaining the release target platform of the model developer; for each release form corresponding to the release target platform, determining the degree of fit between other models already released therein and the release target platform; according to the degree of fit, selecting at least one release form for the industrial model, so as to facilitate calling the model encapsulation device to encapsulate according to the selected release form.

[0012] In one example, the determining the degree of fit between other models already released therein and the release target platform for each release form corresponding to the release target platform specifically includes: for each release form corresponding to the release target platform, determining all the models already released therein; according to the application scenario and computing power of the industrial model, determining several models closest to the industrial model among all the models, and determining the degree of fit between the several models and the release target platform.

[0013] In one example, the cloud server can provide at least one of computing resources, storage resources, and network resources; the PaaS platform can provide at least one of Web service middleware, integrated development environment IDE, database, container engine, big data computing and storage components.

[0014] On the other hand, the present application also proposes an industrial model management 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, and the instructions are executed by the at least one processor so that the at least one processor can execute: the application function server selects an application function adapted to the industry in a preset application function library according to the cloud infrastructure support, the PaaS support, and the industry where the model developer is located; online displays the application function to the model developer, so that the model developer can obtain personalized services according to the application function; according to the operations performed by the model developer through the application function, online implements model development and model testing, so as to obtain a trained industrial model; calls the model encapsulation device, encapsulates the industrial model, and publishes it to the corresponding platform through the application function.

[0015] In one example, to implement model testing online according to the operations performed by the model developer in the application function, it specifically includes: determining the data generalization level of the training samples uploaded by the model developer; if the data generalization level is lower than the first preset threshold, determining the data collection level corresponding to the industry of the model developer; if the data collection level is higher than the second preset threshold, generating virtual data based on the training samples uploaded by the model developer; replacing at least part of the data in the test set of the training samples with the virtual data, and implementing model testing online through the replaced test set.

[0016] The data processing method for multiple industrial models based on the industrial Internet proposed by this application can bring the following beneficial effects:

[0017] Through the industrial model management system in this solution, it can provide functions such as online model development and online testing for model developers, and provide corresponding cloud resources and PaaS support, as well as externally encapsulated services, enabling model developers to focus on the development of industrial models, realizing the aggregation of developers, and also providing a supply and demand docking platform for model developers and the final user enterprises.

[0018] Moreover, by differentiating different industries and pre-setting an application function library for each industry unit, providing personalized application services according to the industry where the model developer is located, while being able to provide the required services for model developers, reducing the occurrence of redundant services, it can further improve the model development efficiency of model developers and is also more conducive to relevant model management work. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The drawings described herein are used to provide a further understanding of the present application and form a part of the present application. The schematic embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation to the present application. In the drawings:

[0020] Figure 1 is a schematic diagram of the architecture of the industrial model management system in the embodiment of the present application;

[0021] Figure 2 is a schematic flowchart of the data processing method for multiple industrial models based on the industrial Internet in the embodiment of the present application;

[0022] Figure 3 is a schematic diagram of the data processing device for multiple industrial models based on the industrial Internet in the embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments of this application and the corresponding drawings. Apparently, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the scope of protection of this application.

[0024] The following will, in conjunction with the drawings, elaborate on the technical solutions provided by each embodiment of this application.

[0025] As Figure 1 shown, the embodiment of this application provides a data processing method for multiple industrial models based on the industrial Internet, which is applied in an industrial model management system (hereinafter referred to as the management system). The management system mainly includes four layers, namely, a cloud server, a Platform as a Service (PaaS) layer, an application function server, and a model encapsulation device.

[0026] The cloud server is mainly used to provide cloud infrastructure support, including: computing resources (such as CPUs, GPUs, FPGAs), storage resources (such as cloud hard disks, object storage), and network resources (such as public network bandwidth and public network IPs).

[0027] The PaaS platform is mainly used to provide PaaS support, including: Web service middleware (such as frameworks like TomCat and Nginx for Web services), an integrated development environment IDE (supporting multi-language development such as Modelica, C, C++, Python, etc.), databases (such as time-series databases, relational databases, non-relational databases, and graph databases), container engines (such as Docker, K8S), big data computing and storage components (such as frameworks like Hadoop clusters, Kafka clusters, and Flink clusters for processing massive data in the industrial field), etc.

[0028] The application function server is mainly used to support the management and development of industrial models, and it includes multiple functions such as data management, model development, model management, model testing, and publishing services. The data management function is mainly used to support model developers in uploading and managing data sets, and it can publish personal data sets as public data sets. The model management function supports the viewing, uploading, and editing and modification of industrial models. The model development function allows model developers to perform online development and online debugging of industrial models, and it can directly call open-source algorithms. The model testing function allows model developers to test the input parameters and output results of models. Model developers can use the platform integration framework and open service interface (Restful API interface) set therein to call the model encapsulation device to encapsulate the model to implement the publishing service function.

[0029] The model encapsulation device is mainly used to encapsulate the model through the platform integration framework and the Restful API standard interface for external calling. The model developer can encapsulate the industrial model in industrial APPs, industrial software, applets, etc.

[0030] Such as Figure 2 shown, an embodiment of the present application provides a data processing method for multiple industrial models based on the industrial Internet, including:

[0031] S201: The application function server corresponding to the industrial model selects an application function adapted to the industry from a preset application function library according to the cloud server, the PaaS platform, and the industry corresponding to the industrial model.

[0032] As described above, the application function server can provide corresponding functions for the model developer according to the cloud infrastructure support and PaaS support. However, in this way, what the model developer gets are only conventional and general functions. Therefore, it is also possible to consider the industry where the model developer is located and select and provide adapted application functions for it.

[0033] Specifically, the application functions corresponding to different industries may be different. For example, in some industries, the data is relatively general, and the needs of the entire industry can be met with only a small number of industrial models. While in other industries, even for the same product, the data has relatively high specificity (for example, in the tire mold processing industry, even for the same type of vehicle, different industrial models are required for different tire patterns and structures). At this time, a larger number of industrial models are needed.

[0034] Based on this, an application function library is preset, and multiple application functions are set in the application function library. And for the same application function, corresponding personalized adjustments are set for different industries. For example, in the data management and model management functions, the function of "classification management" can be adaptively added or deleted to facilitate the model developer to accurately and quickly obtain the required data or models.

[0035] S202: Based on triggering the application function, obtain personalized services.

[0036] According to the different current application functions, the interface presented to the model developer can also change accordingly. This can be changes in the layout, content, etc. of the interface, so as to provide personalized services for the model developer through the services for different industries in the application function. At this time, the triggers for the model developer to the presented interface are also different. For example, as mentioned above, in the interface with the function of "classification management", corresponding buttons can be appropriately added for classification. Or, replace the background of the presented interface with relevant pictures of the industry where the model developer is located. In this way, a sense of familiarity can be brought to the model developer during the use process.

[0037] S203: According to the operations performed by the model developer in the application function, realize model development and model testing in an online manner, so as to obtain the trained industrial model.

[0038] The operations that the model developer can perform include creating tasks, selecting environments, writing code, conducting tests, etc. The management system calls the services required by the function through the interaction between the model developer and the server, supports uploading industrial model files in multiple formats, provides development, operation, and test environments in multiple languages, such as C language, C++, Python, Modelica, etc., and provides a variety of open-source scientific computing, simulation, and control toolkits to provide online development and test support for the model developer.

[0039] S204: Package the trained industrial model through the model packaging device and publish it to the corresponding platform through the application function.

[0040] The platform includes various forms, such as industrial software, industrial APPs, mini-programs, etc. By providing model publishing services and packaging interface functions, models in different languages and different formats can be conveniently unified and published as REST API services, which is convenient for the invocation between different models and applications, and can provide model interface invocation services externally.

[0041] In one embodiment, the model developer uses the management system in this article for model development and model testing. However, in some special cases, due to objective circumstances, for example, the industry where the model developer is located belongs to an emerging industry and the data that can be collected is less, or the data collected by the model developer is too similar. At this time, it is easy to cause the poor generalization ability of the finally trained industrial model, which is not conducive to the model developer to use or promote the industrial model.

[0042] Based on this, first determine the data generalization level of the training samples uploaded by the model developer. The data generalization level represents the generalization ability of the data. Generally speaking, the higher the similarity between the data, the lower the data generalization level, and the worse the generalization ability of the industrial model trained with it.

[0043] When the data generalization level is lower than the first preset threshold, determine the data collection level of the industry of the model developer. The data collection level indicates the difficulty of collecting various types of data in the industry. The higher the data collection level, the more difficult it is to collect training sample data in the industry. The data collection level can be evaluated according to the corresponding industry emergence time, the number of enterprises in the industry, enterprise scale, value brought, etc. When the data collection level is low, it means that it is easy to collect training sample data in the industry. However, its data generalization ability is low, and it is very likely that the model developer deliberately does so. At this time, no other additional processing is performed. When the data collection level is higher than the second preset threshold, it means that the data generalization level of the training samples provided by the user is low, which is very likely caused by the difficulty of collecting data. At this time, the generalization ability of the industrial model obtained by training is poor, which is very likely not what the user wants to see. Therefore, generate some virtual data according to the training samples, and replace some of the data in the test set with the virtual data, and perform model testing on the industrial model through the replaced test set. Among them, only the training samples in the test set are replaced, and the training set is not replaced, in order to ensure that the industrial model obtained by the user's training is strictly based on the data provided by itself. The virtual data generated by the management system is only used in the test set and will not affect the parameters of the industrial model itself, and is only used for its test evaluation, so that the model developer can have a more accurate understanding of the industrial model obtained by training.

[0044] Of course, during testing, the industrial model can be tested respectively through the test sets before and after replacement to obtain the corresponding first test result and second test result. And determine the replacement rate of the test set during the replacement process. The replacement rate indicates how many training samples in the test set are replaced.

[0045] Determine the difference between the first test result and the second test result. The difference calculation methods for different industrial models are also different. For example, if the output of the industrial model is an image recognition result, the difference between the recognition ranges of the two can be used as the difference. Or, if the output result of the industrial model is a numerical value, the difference can be obtained by taking the difference of the numerical values. According to the difference and the replacement rate, evaluate the impact of the low data generalization level on the industrial model, and display the evaluation result to the model developer. For example, if the replacement rate is very small but the difference is very large, it means that the model generalization level has a great impact on the industrial model; if the replacement rate is very large but the difference is very small, it means that the impact of the model generalization level on the industrial model is almost zero and can be ignored.

[0046] In one embodiment, as mentioned above, it is necessary to determine the data generalization level of the training samples uploaded by the model developer, and a detailed explanation is given here.

[0047] For each training sample uploaded by the model developer, determine the values possessed by the key attributes of the training sample. The key attributes indicate that they have a greater impact on the output result of the industrial model, and they can be preset in advance. For example, a certain training sample is: the user's name is A, the age is B, and the gender is C. Here, the extracted attributes include: name, age, and gender. If the key attribute is predefined as age, then the value it possesses is B.

[0048] Determine the conceptual level at which the value is located in the space corresponding to the attribute. Still taking the example in this embodiment for explanation, for age, there are multiple predefined conceptual levels, from high to low in sequence: population (for example, including the elderly, middle-aged, young, children, etc.), age group (for example, including 0 - 10 years old, 11 - 20 years old, 21 - 30 years old, etc.), specific age, etc. The higher the conceptual level, the less specific and more general the corresponding value is, while the lower the conceptual level, the more specific the corresponding value is.

[0049] According to the proportion of the training sample in each conceptual level, the data generalization level of the training sample can be finally obtained. When the proportion of the training sample in the lower conceptual level is larger, it indicates that the value of the training sample in the key attribute is too specific, and at this time, the data generalization level is lower. On the contrary, when the proportion of the training sample in the higher conceptual level is larger, it indicates that the data generalization level is higher.

[0050] In one embodiment, as mentioned above, virtual data is generated according to the training sample, and a detailed explanation is given here.

[0051] In the training sample, determine the specified training samples whose conceptual level is lower than the preset level. For this part of the training samples, their corresponding data generalization level is lower, and they need to be replaced to increase the overall data generalization level of the training sample.

[0052] Determine the first value of the specified training sample in the key attribute, and according to the industry status quo, generate a second value with a higher conceptual level for the first value, and replace the first value with the second value. Still taking the age mentioned above as an example, if the first value B of age is 20 years old, then generate a second value with a higher conceptual level, such as 11 - 20 years old, or a higher level of young. For the division of conceptual levels in different industries, it can be determined based on the current industry status quo.

[0053] In addition to replacing the key attributes, to further expand the difference between the training samples after replacement and those before replacement, at least some of the non-key attributes other than the key attributes are replaced with other attributes that can be obtained based on the training samples. Still taking the age in the above text as an example, its training sample is: the user name is A, the age is B, and the gender is C. After replacing the value of B, then replace its gender C with the height D. Since it is a non-key attribute, even if the replacement occurs, it will not have a great impact on the output result, and it can appropriately increase the generalization ability of the industrial model.

[0054] In one embodiment, before encapsulating the industrial model, first obtain the release target platform of the model developer. When the release target platform has multiple release forms (such as PC, Android, IOS, mini-program, etc.), for each release form, determine the compatibility between other models that have been released and the release target platform. The compatibility can be determined according to the view volume, download volume, favorable comment rate, etc. The higher the compatibility, the better the release effect of other models in this release form.

[0055] At this time, according to the compatibility, select at least one release form for the industrial model, and call the model encapsulation device to encapsulate and release the industrial model. Generally speaking, for different release forms, their target audiences may be correspondingly different. And pre-judging which release form is more suitable for the model developer for different release forms can improve the compatibility between the model and the release target platform after the model developer releases the industrial model, and improve the user experience.

[0056] Furthermore, when determining the compatibility in the release target platform, for each corresponding release form, all the models that have been released can be determined. Then, according to the application scenario of the industrial model itself (such as determined by the industry) and the computing power (such as determined by the accuracy of the output result), select several of the closest models among all the models, and determine the compatibility between these several models and the release target platform. In this way, selecting other models closer to the industrial model to calculate the compatibility can more accurately analyze the compatibility between the industrial model and the release target platform.

[0057] As Figure 3 shown, the embodiment of the present application also provides a data processing device for multiple industrial models based on the industrial Internet, including:

[0058] At least one processor; and,

[0059] A memory communicatively connected to the at least one processor; wherein,

[0060] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform:

[0061] The application function server corresponding to the industrial model selects, according to the cloud server, the PaaS platform, and the industry corresponding to the industrial model, an application function adapted to the industry from a preset application function library;

[0062] Based on triggering the application function, obtain personalized services;

[0063] According to the operations performed by the model developer in the application function, implement model development and model testing in an online manner, so as to obtain the trained industrial model;

[0064] Package the trained industrial model through the model packaging device and publish it to the corresponding platform through the application function.

[0065] In one embodiment, the implementing model testing in an online manner according to the operations performed by the model developer in the application function specifically includes:

[0066] Determine the data generalization level of the training samples uploaded by the model developer;

[0067] If the data generalization level is lower than a first preset threshold, determine the data collection level corresponding to the industry of the model developer;

[0068] If the data collection level is higher than a second preset threshold, generate virtual data according to the training samples uploaded by the model developer;

[0069] Replace at least part of the data in the test set of the training samples with the virtual data, and implement model testing in an online manner through the replaced test set.

[0070] The embodiment of the present application also provides a non-volatile computer storage medium storing computer-executable instructions, and the computer-executable instructions are set as:

[0071] The application function server corresponding to the industrial model selects, according to the cloud server, the PaaS platform, and the industry corresponding to the industrial model, an application function adapted to the industry from a preset application function library;

[0072] Based on triggering the application function, obtain personalized services;

[0073] According to the operations performed by the model developer in the application function, implement model development and model testing in an online manner, so as to obtain the trained industrial model;

[0074] The trained industrial model is encapsulated by the model encapsulation device and published to the corresponding platform through the application function.

[0075] Each embodiment in this application is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the device and medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiments.

[0076] The devices and media provided in the embodiments of this application correspond one by one to the methods. Therefore, the devices and media also have beneficial technical effects similar to those of the corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be elaborated here.

[0077] Those skilled in the art should understand that the embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0078] This application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of this application. It should be understood that each flow and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the flows and / or blocks in the flowchart and / or block diagram can also be implemented. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate for implementing in the process Figure 1 one process or multiple processes and / or blocks Figure 1 a device for the functions specified in one block or multiple blocks.

[0079] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device implements in the process Figure 1 one process or multiple processes and / or blocks Figure 1 a device for the functions specified in one block or multiple blocks.

[0080] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one process or multiple processes and / or one block or multiple blocks. Figure 1 One process or multiple processes and / or Figure 1 One block or multiple blocks.

[0081] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and memory.

[0082] The memory may include non-permanent memory in the form of computer-readable media, random access memory (RAM) and / or non-volatile memory such as read-only memory (ROM) or flash memory (flash RAM). The memory is an example of computer-readable media.

[0083] Computer-readable media includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette tapes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transitory media such as modulated data signals and carrier waves.

[0084] It should also be noted that the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, commodity or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or elements inherent to such process, method, commodity or device. Without further limitation, an element defined by the statement "comprising one..." does not exclude the presence of additional identical elements in the process, method, commodity or device comprising the element.

[0085] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.

Claims

1. A data processing method for multiple industrial models based on the industrial Internet, characterized in that, Applied to an industrial model management system, the industrial model management system includes a cloud server for providing cloud infrastructure support, a PaaS platform for providing service PaaS support, one or more application function servers, and a model encapsulation device; The method includes: The application function server corresponding to the industrial model selects an application function adapted to the industry from a preset application function library according to the cloud server, the PaaS platform, and the industry corresponding to the industrial model; Based on triggering the application function, personalized services are obtained; According to the operations performed by the model developer in the application function, model development and model testing are realized online, so as to obtain the trained industrial model; The trained industrial model is encapsulated by the model encapsulation device and published to the corresponding platform through the application function; The realizing model testing online according to the operations performed by the model developer in the application function specifically includes: Determining the data generalization level of the training samples uploaded by the model developer; If the data generalization level is lower than a first preset threshold, determining the data collection level corresponding to the industry of the model developer; If the data collection level is higher than a second preset threshold, generating virtual data according to the training samples uploaded by the model developer; Replacing at least part of the data in the test set of the training samples with the virtual data, and realizing model testing online through the replaced test set.

2. The method according to claim 1, wherein The determining the data generalization level of the training samples uploaded by the model developer specifically includes: For each training sample uploaded by the model developer, determining the values possessed in the key attributes of the training sample; Determining the conceptual level at which the value is located in the space corresponding to the attribute; Determining the data generalization level of the training sample according to the proportion of the training sample in each conceptual level.

3. The method according to claim 2, wherein The generating virtual data according to the training samples uploaded by the model developer specifically includes: Determining the specified training samples whose conceptual levels are lower than the preset level in the training samples; Determining the first values possessed in the key attributes of the specified training samples; According to the industry status value of the industry, generating a second value with a higher conceptual level for the first value, and replacing the first value with the second value; and Replacing at least part of the non-key attributes in the training samples with other attributes that can be obtained according to the training samples; Taking the replaced training samples as the generated virtual data.

4. The method according to claim 1, wherein The realizing model testing online according to the operations performed by the model developer in the application function specifically includes: Realizing model testing online through the test set before replacement and the test set after replacement respectively, and obtaining the first test result corresponding to before replacement and the second test result corresponding to after replacement; Determining the replacement rate of the test set by the virtual data; Evaluating the impact of the data generalization level on the model development according to the replacement rate and the difference between the first test result and the second test result; Show the evaluation results to the model developer.

5. The method according to claim 1, characterized in that, Before the trained industrial model is encapsulated by the model encapsulation device and published to the corresponding platform through the application function, the method further includes: Obtain the release target platform of the model developer; For each release form corresponding to the release target platform, determine the degree of fit between other models already released therein and the release target platform; According to the degree of fit, select at least one release form for the industrial model, so as to call the model encapsulation device to encapsulate according to the selected release form.

6. The method according to claim 5, characterized in that, The determining the degree of fit between other models already released therein and the release target platform for each release form corresponding to the release target platform specifically includes: For each release form corresponding to the release target platform, determine all the models already released therein; According to the application scenario and computing power of the industrial model, determine several models closest to the industrial model among all the models, and determine the degree of fit between the several models and the release target platform.

7. The method according to any one of claims 1-6, characterized in that, The cloud server can provide at least one of computing resources, storage resources, and network resources; the PaaS platform can provide at least one of Web service middleware, integrated development environment IDE, database, container engine, big data computing, and storage components.

8. An industrial model management device, characterized in that, Comprising: 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, and the instructions are executed by the at least one processor so that the at least one processor can execute the data processing method for multiple industrial models based on the industrial Internet according to claim 1.

Images