A machine learning method and system

By designing a machine learning platform that supports multiple models and multiple engines, the problem of insufficient business demand in vertical fields is solved, and a vertical field scenario-based solution for PHM scenarios is realized, which reduces development thresholds and costs and improves development efficiency.

CN112464569BActive Publication Date: 2024-11-15NANJING MING LUE TECH CO LTD
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
CN202011433273.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-12-10
Publication Date
2024-11-15
Estimated Expiration
2040-12-10

AI Technical Summary

Technical Problem

The existing machine learning platform has shortcomings in the business needs of vertical fields and cannot directly cope with specific needs in vertical fields, resulting in high development thresholds and low efficiency, especially in the failure prediction and health management (PHM) scenarios, which lacks targeted Solution.

Method used

A machine learning platform supporting multi-model and multi-engine is designed to provide a one-stop full-process machine learning solution from data management, algorithm management, task management to platform services, through the establishment of device object models, data storage, and algorithm upload , training task creation and prediction task creation, etc., provide vertical field scenario-based solutions for PHM scenarios.

Benefits of technology

Reduces the development threshold and development cost of machine learning models, improves user development efficiency, meets the vertical field needs of PHM scenarios, and provides a one-stop all-round range from data governance, algorithm management, model training, model prediction to model services. Serve.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN112464569B_ABST
    Figure CN112464569B_ABST
Patent Text Reader

Abstract

The present invention discloses a machine learning method and system. The machine learning method includes: a model building step: building a device object model; a data storage step: mysql stores the data source, and stores the mapping relationship between field information and the device object model, and the mapping relationship data in the MySql database; an algorithm management step: uploading the algorithm file and the configuration file in a package, and performing algorithm analysis according to the information of the configuration file; a training task creation step: creating a new training task and executing the training task according to the execution strategy; a prediction task creation step: creating a new prediction task and executing the prediction task according to the execution strategy. The present invention proposes a one-stop machine learning platform based on PHM data modeling, which is based on actual business, provides users with vertical field scenario solutions, and improves user development efficiency.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of data modeling, and in particular to a machine learning method and system. Background Art

[0002] Following the mobile Internet, artificial intelligence (AI) has become the hottest topic in the industry. Both giant manufacturers and newcomers are vigorously developing AI technology to enhance their technological competitiveness and empower their own businesses. Machine learning, as the core of AI technology, is the foundation of computer intelligence. At present, machine learning has brought many applications, such as face recognition and knowledge graphs. Various industries have high expectations for machine learning, but the contradiction is that the application threshold of machine learning technology is high. Each step of the process requires professional knowledge and a lot of time is needed to optimize each process. In order to solve the above problems, building a machine learning platform that integrates data processing, algorithm uploading, model management, task training / prediction, online prediction services and other functions to assist non-algorithm developers, help them get rid of tedious engineering development, and focus their limited energy on one-stop machine learning services has strong practical significance and a wide range of application scenarios. Through the machine learning platform, non-algorithm engineers can also build simple machine learning models based on actual business, complete data modeling, and solve the business needs of users in the vertical field. In addition, under the background of "Intelligent Manufacturing 2025", using information technology to solve the pain points in traditional manufacturing has become the current research focus. In order to solve the pain point of equipment failure and maintenance, the industry and information field have proposed Prognostics Health Management (PHM), whose key technology is fault diagnosis and prediction based on machine learning. In order to lower the threshold for use, the current machine learning platform has highly encapsulated machine learning algorithms, but the encapsulated algorithms only meet the needs of general fields and cannot directly meet the business needs of vertical fields.

[0003] Therefore, in view of the above situation, the present invention proposes a machine learning method and system. The present invention conducts a demand analysis based on the actual needs within the company, clarifies the problems to be solved, and designs the overall architecture of the platform by analyzing the needs of relevant business developers within the company. The main functions of several modules are introduced in detail in the form of a solution flowchart, showing the internal implementation of several modules and the collaborative relationship between several modules. Aiming at the PHM scenario, the platform establishes a machine learning platform that supports multiple models and multiple engines, providing users with a one-stop full-process machine learning solution from data management, algorithm management, task management to platform services. The platform is based on actual business, providing users with scenario-based solutions in vertical fields to improve user development efficiency. Summary of the invention

[0004] The embodiments of the present application provide a machine learning method and system to at least solve the problem of subjective factors in related technologies.

[0005] The present invention provides a machine learning method, comprising:

[0006] Model building steps: Establish device object model;

[0007] Data storage step: encapsulating a big data platform interface, calling the interface to store data source information in a database, configuring a mapping relationship between model information and field information of the device object model, and storing the mapping relationship and mapping data in the database;

[0008] Algorithm management step: uploading the algorithm, selecting the device object model, uploading the algorithm file and configuration file, parsing the algorithm parameters according to the information in the configuration file, and storing the algorithm parameters in the database;

[0009] Training task creation step: create a new training task, select the training task type and training algorithm parameters, locate the device object model through the training algorithm parameters, select instance data from the device object model according to the time range, configure the training task execution strategy, execute the device object model training, store the training verification results in the database, and then publish the training model;

[0010] Prediction task creation steps: create a new prediction task, add prediction task attributes, select the prediction task type and prediction algorithm parameters, locate the device object model and select the training model through the prediction algorithm parameters, publish the prediction task, and store the prediction results after executing the prediction task according to the strategy.

[0011] In the above-mentioned machine learning method, the model building step includes building a device object model, and adding device attributes and measurement point data information of the device object model.

[0012] In the above-mentioned machine learning method, the data storage step comprises:

[0013] Data source storage step: encapsulate the influxdb interface of the big data platform, call the influxdb interface to store the database name in the MySql data source table;

[0014] Data mapping storage step: configuring the mapping relationship between the attributes of the device object model, the measuring point information and the field information, and storing the mapping relationship and mapping data in the MySql database.

[0015] In the above-mentioned machine learning method, the algorithm management step comprises:

[0016] Algorithm upload step: upload the algorithm, add the attributes of the algorithm, select the device object model, and download the data of the device object model. If auxiliary files are required, select them and package and upload the algorithm file, configuration file and auxiliary files and save them to the server;

[0017] Algorithm parsing step: parsing the algorithm parameters according to the information in the configuration file, comparing the algorithm parameters with the measurement point data information of the device object model, and storing the algorithm parameters in the database.

[0018] In the above-mentioned machine learning method, the training task creation step includes: creating a new training task, adding a training task name, selecting a training task type and training algorithm parameters, locating the device object model through the training algorithm parameters, and selecting the instance data time range of the device object model, after selecting instance data from the device object model according to the instance data time range, the training task configures an execution strategy and saves the execution strategy, trains the device object model according to the execution strategy, and stores the training verification results in the database before publishing the training model.

[0019] In the above-mentioned machine learning method, the prediction task creation step includes: creating a new prediction task, adding prediction task attributes, selecting the prediction task type and prediction algorithm parameters, locating the device object model through the algorithm parameters, selecting the training model, if the prediction result data needs to be entered, the prediction task needs to be repeated, the previous Nth prediction result needs to be selected, the prediction task needs to be executed immediately, and a prediction result is selected. If the prediction result does not need to be entered, an online prediction service needs to be provided, the prediction task is executed, and an online API service is generated for the user to call. If the online prediction service is not needed, the prediction task is saved and then executed according to the strategy, and the device object model is predicted and the prediction result is stored.

[0020] The present invention also provides a machine learning system, characterized in that it is applicable to the machine learning method described above, including:

[0021] Model building unit: building device object model;

[0022] Data storage unit: encapsulates the big data platform interface, calls the interface to store data source information in the database, configures the mapping relationship between the model information and the field information of the device object model, and stores the mapping relationship and mapping data in the database;

[0023] Algorithm management unit: upload the algorithm, select the device object model, upload the algorithm file and configuration file, parse the algorithm parameters according to the information in the configuration file, and store the algorithm parameters in the database;

[0024] A training task creation unit: creates a new training task, selects a training task type and training algorithm parameters, locates the device object model through the training algorithm parameters, selects instance data from the device object model according to a time range, configures a training task execution strategy, executes the device object model training, and releases the training model after storing the training verification results in the database;

[0025] Prediction task creation unit: create a new prediction task, add prediction task attributes, select the prediction task type and prediction algorithm parameters, locate the device object model and select the training model through the prediction algorithm parameters, publish the prediction task, and store the prediction results after executing the prediction task according to the strategy.

[0026] In the above-mentioned machine learning system, the model building unit includes building a device object model and adding device attributes and measurement point data information of the device object model.

[0027] In the above-mentioned machine learning system, the data storage unit comprises:

[0028] Data source storage module: encapsulates the influxdb interface of the big data platform, and calls the influxdb interface to store the database name in the MySql data source table;

[0029] Data mapping storage module: configures the mapping relationship between the attributes of the device object model, the measuring point information and the field information, and stores the mapping relationship and mapping data in the MySql database.

[0030] In the above-mentioned machine learning system, the algorithm management unit includes:

[0031] Algorithm upload module: upload the algorithm, add the attributes of the algorithm, select the device object model, and download the data of the device object model. If auxiliary files are required, select them and package and upload the algorithm file, configuration file and auxiliary files and save them to the server;

[0032] Algorithm parsing module: parses the algorithm parameters according to the information in the configuration file, compares the algorithm parameters with the measurement point data information of the device object model, and stores the algorithm parameters in the database.

[0033] In the above-mentioned machine learning system, the training task creation unit includes creating a new training task, adding a training task name, selecting a training task type and training algorithm parameters, locating the device object model through the training algorithm parameters, and selecting the instance data time range of the device object model, after selecting instance data from the device object model according to the instance data time range, the training task configures an execution strategy and saves the execution strategy, trains the device object model according to the execution strategy, and stores the training verification results in the database before publishing the training model.

[0034] In the above-mentioned machine learning system, the prediction task creation unit includes creating a new prediction task, adding prediction task attributes, selecting the prediction task type and prediction algorithm parameters, locating the device object model through the algorithm parameters, selecting the training model, if the prediction result data needs to be entered, the prediction task needs to be repeated, the previous Nth prediction result needs to be selected, the prediction task needs to be executed immediately, and a prediction result is selected. If the prediction result does not need to be entered, an online prediction service needs to be provided, the prediction task is executed, and an online API service is generated for the user to call. If the online prediction service is not needed, the prediction task is saved and then executed according to the strategy, and the device object model is predicted and the prediction result is stored.

[0035] Compared with the related art, the present invention proposes a machine learning method and system. The present invention conducts demand analysis based on the actual needs within the company, clarifies the problems to be solved, and designs the overall architecture of the platform by analyzing the needs of relevant business developers within the company. The main functions of several modules are introduced in detail in the form of a solution flowchart, showing the internal implementation of several modules and the collaborative relationship between several modules. Aiming at the PHM scenario, the platform establishes a machine learning platform that supports multiple models and multiple engines, providing users with a one-stop full-process machine learning solution from data management, algorithm management, task management to platform services. The platform is based on actual business, providing users with scenario-based solutions in vertical fields to improve user development efficiency.

[0036] Details of one or more embodiments of the present application are set forth in the following drawings and description to make other features, objects, and advantages of the present application more readily apparent. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0038] Figure 1is an architecture diagram of a machine learning method according to an embodiment of the present application;

[0039] Figure 2 is a data management framework diagram according to an embodiment of the present application;

[0040] Figure 3 is an algorithm management framework diagram according to an embodiment of the present application;

[0041] Figure 4 is a task management framework diagram according to an embodiment of the present application;

[0042] Figure 5 is a flow chart of a machine learning method according to an embodiment of the present application;

[0043] Figure 6 is a flow chart of data storage steps according to an embodiment of the present application;

[0044] Figure 7 is a flowchart of algorithm management steps according to an embodiment of the present application;

[0045] Figure 8 A schematic diagram of the structure of a machine learning system of the present invention;

[0046] Fig. 9 is a framework diagram of an electronic device according to an embodiment of the present application.

[0047] Wherein, the accompanying drawings are marked as follows:

[0048] Model building unit: 41;

[0049] Data storage unit: 42;

[0050] Algorithm management unit: 43;

[0051] Training task creation unit: 44;

[0052] Prediction task creation unit: 45;

[0053] Data source storage module: 421;

[0054] Data mapping storage module: 422;

[0055] Algorithm upload module: 431;

[0056] Algorithm parsing module: 432;

[0057] 81: processor;

[0058] 82: memory;

[0059] 83: Communication interface;

[0060] 80: Bus. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is described and illustrated below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. Based on the embodiments provided in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application.

[0062] Obviously, the drawings described below are only some examples or embodiments of the present application. For ordinary technicians in this field, the present application can also be applied to other similar scenarios based on these drawings without creative work. In addition, it can also be understood that although the efforts made in this development process may be complicated and lengthy, for ordinary technicians in this field related to the content disclosed in this application, some changes in design, manufacturing or production based on the technical content disclosed in this application are just conventional technical means, and should not be understood as insufficient content disclosed in this application.

[0063] Reference to "embodiments" in this application 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 of ordinary skill in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0064] Unless otherwise defined, the technical terms or scientific terms involved in this application should be understood by people with ordinary skills in the technical field to which this application belongs. The words "one", "a", "a", "the" and the like involved in this application do not indicate a quantitative limitation, and may represent the singular or plural. The terms "include", "comprise", "have" and any of their variations involved in this application are intended to cover non-exclusive inclusions; for example, a process, method, system, product or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units that are not listed, or may also include other steps or units inherent to these processes, methods, products or devices. The words "connect", "connected", "coupled" and the like involved in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The "multiple" involved in this application refers to two or more. "And / or" describes the association relationship of associated objects, indicating that there may be three relationships, for example, "A and / or B" can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the objects before and after are in an "or" relationship. The terms "first", "second", "third", etc. involved in this application are only used to distinguish similar objects and do not represent a specific ordering of the objects.

[0065] The present invention is based on data modeling, which is briefly introduced below.

[0066] Machine learning platform is a concept that has emerged recently. It can highly encapsulate machine learning algorithms and provide the computing resources required for algorithm training, which reduces the threshold for using machine learning algorithms. At present, commercial machine learning platforms generally support functions such as data processing, visual modeling, model training, model management, and model publishing. These machine learning platforms allow users to directly operate in a "fool-style" manner without professional algorithm knowledge or attention to the subsequent operation and maintenance work of training. This brings some negative effects, that is, the encapsulated algorithms only meet the needs of general business, and are "stretched" to meet the needs of vertical fields. In addition, these products are basically highly customized, lack universal applicability, and cannot access more users. In response to the problems of current machine learning platforms, the present invention proposes a one-stop machine learning platform based on PHM data modeling, which aims to provide a one-stop full-process machine learning solution for PHM scenarios from data storage, data processing, algorithm management, model training, model publishing to model launch, reduce the development threshold and development cost of machine learning models, and professionally adapt to PHM scenarios to solve engineering problems such as training resources, model management, and service operation and maintenance for PHM engineers.

[0067] The one-stop machine learning platform based on PHM data modeling is divided into a data engine layer, a machine learning platform, and a business application layer from bottom to top. The data engine layer is managed by the company's big data platform team, including big data computing frameworks such as Spark, Flink, and Kafka, which provide data services for the machine learning platform; the machine learning platform is the core function of the present invention, including the entire process of machine learning, such as data storage, algorithm management, model management, task management, and model online services; the data storage provides a variety of database storage forms, the algorithm management provides commonly used algorithm modules for PHM scenarios, the model management is responsible for the release and version control of the training model, and the model online service can provide the trained model interface to the outside world; the business application layer is for the PHM scenario, and the present invention provides a one-stop full-process service from model training, model release, model prediction to API prediction. The present invention is aimed at the PHM field and proposes a one-stop machine learning platform based on PHM data modeling, which aims to provide PHM algorithm developers with one-stop comprehensive services from data governance, algorithm management, model training, model prediction to model services. The functions of the entire platform can cover the entire process of data storage, algorithm upload, model training, model publishing, model prediction and model evaluation in the algorithm development process. In addition, the model service provides API online use. Users can publish the model through simple configuration and apply it to their own business in the form of interface calls.

[0068] The present invention proposes a machine learning method and system. The present invention proposes a machine learning method and system. The present invention conducts a demand analysis based on the actual needs within the company, clarifies the problems to be solved, and designs the overall architecture of the platform by analyzing the needs of relevant business developers within the company. The main functions of several modules are introduced in detail in the form of a solution flowchart, showing the internal implementation of several modules and the collaborative relationship between several modules. Aiming at the PHM scenario, the platform establishes a machine learning platform that supports multiple models and multiple engines, providing users with a one-stop full-process machine learning solution from data management, algorithm management, task management to platform services. The platform is based on actual business, providing users with scenario-based solutions in vertical fields to improve user development efficiency.

[0069] The following will use data modeling as an example to illustrate the embodiments of the present application.

[0070] Embodiment 1

[0071] This embodiment provides a machine learning method. Figure 1-Figure 7 , Figure 1 is an architecture diagram of a machine learning method according to an embodiment of the present application; Figure 2 is a data management framework diagram according to an embodiment of the present application; Figure 3 is an algorithm management framework diagram according to an embodiment of the present application; Figure 4 is a task management framework diagram according to an embodiment of the present application; Figure 5 is a flow chart of a machine learning method according to an embodiment of the present application; Figure 6 is a flow chart of data storage steps according to an embodiment of the present application; Figure 7 It is a flowchart of the algorithm management steps according to an embodiment of the present application. As shown in the figure, a machine learning method includes the following steps:

[0072] Model building step S1: building a device object model;

[0073] Data storage step S2: encapsulating the big data platform interface, calling the interface to store the data source information in the database, configuring the mapping relationship between the model information and the field information of the device object model, and storing the mapping relationship and mapping data in the database;

[0074] Algorithm management step S3: uploading the algorithm, selecting the device object model, uploading the algorithm file and configuration file, parsing the algorithm parameters according to the information in the configuration file, and storing the algorithm parameters in the database;

[0075] Training task creation step S4: creating a new training task, selecting a training task type and training algorithm parameters, locating the device object model through the training algorithm parameters, selecting instance data from the device object model according to a time range, configuring a training task execution strategy, and executing the device object model training, storing the training verification results in the database, and then publishing the training model;

[0076] Prediction task creation step S5: create a new prediction task, add prediction task attributes, select the prediction task type and prediction algorithm parameters, locate the device object model and select the training model through the prediction algorithm parameters, publish the prediction task, and store the prediction results after executing the prediction task according to the strategy.

[0077] In the embodiment, the model building step S1 includes building a device object model, and adding device attributes and measurement point data information of the device object model.

[0078] Specifically, in the model building step S1, the device object model is first established, and its device attributes, such as device ID, device name, etc., are added, and then its measurement point data information, such as temperature, humidity, pressure, etc., are added. The subsequent steps are to perform training / prediction on the device.

[0079] In an embodiment, the data storage step S2 includes:

[0080] Data source storage step S21: encapsulate the influxdb interface of the big data platform, call the influxdb interface to store the database name in the MySql data source table;

[0081] Data mapping storage step S22: configuring the mapping relationship between the attributes of the device object model, the measuring point information and the field information, and storing the mapping relationship and mapping data in the MySql database.

[0082] Specifically, the data of the present invention comes from the IOT big data platform (company product, providing data services). The data storage bottom layer of the IOT big data platform includes multiple databases. Taking the influx database as an example, the technical solutions for the processes of data source storage, data mapping, and data storage are introduced below. First, the influxdb interface of the big data platform is encapsulated, and then the influxdb interface is called to store the database name in the MySql data source table, and update or create according to the table name. Requests and pymysql libraries are used in this process; the data source table can be directly queried through the pymysql library, and the database and the data table in the database can be obtained. In addition, all field information under the data table can be queried through the encapsulated influxdb interface and the data table name; according to the field information of the data table and the device object model information established in the previous step, the mapping relationship of the data field is configured and its data and mapping relationship are stored in the MySql database.

[0083] In the embodiment, the algorithm management step S3 includes:

[0084] Algorithm upload step S31: upload the algorithm, add the attributes of the algorithm, select the device object model, and download the data of the device object model. If auxiliary files are required, select and package the uploaded algorithm files, configuration files and auxiliary files and save them to the server;

[0085] Algorithm parsing step S32: parsing the algorithm parameters according to the information in the configuration file, comparing the algorithm parameters with the measurement point data information of the device object model, and storing the algorithm parameters in the database.

[0086] Specifically, in the algorithm management step S3, first add the algorithm name and algorithm description, select the device group, where the device group data comes from the data stored in the previous step; package the algorithm file and configuration file and upload it, where the configuration file needs to be modified according to the algorithm, and its role is to assist in completing the next step of algorithm parsing; if auxiliary files are needed, select them, and finally save the algorithm for algorithm parsing; save the uploaded algorithm file and configuration file to the server, where in order to ensure that large files do not use a large amount of system memory, the f.chunks() loop storage method is adopted; use the zipfile library to decompress the file to obtain the algorithm file and configuration file; parse the configuration file, obtain the main function file of the algorithm file and its function entry, use the importlib library to import the main function, and obtain the algorithm parameters through inspect.signature; if the algorithm parsing time exceeds 30s, it is considered that the parsing has failed, where time monitoring is implemented using multi-threading; put the algorithm parameters and the device group measurement points into a collection for comparison, and when the comparison is consistent, if there is an auxiliary file, it is stored in the same directory as the algorithm, otherwise an error is reported; classify the parsed algorithm parameters, store them in the database, and feed them back to the front-end page.

[0087] In an embodiment, the training task creation step S4 includes creating a new training task, adding a training task name, selecting a training task type and training algorithm parameters, locating the device object model through the training algorithm parameters, and selecting an instance data time range of the device object model, after selecting instance data from the device object model according to the instance data time range, the training task configures an execution strategy and saves the execution strategy, trains the device object model according to the execution strategy, and stores the training verification results in the database before publishing the training model.

[0088] Specifically, in the training task creation step S4, first add the task name and task type, select the training algorithm uploaded in step 2, and locate the device model through the algorithm; select certain device instances under the device model and device instance data in a certain time range as training data; configure the execution strategy, including scheduled execution and immediate execution; save the task, at which time the background downloads the device data to the server according to the time range and device instance ID. The data is the device data stored in MySql after executing step 1; install docker on the server, deploy the DS platform Note 2 into the container, and encapsulate its interface, call the DS platform for model training according to the execution strategy, and place the obtained model in the model folder; store the training verification results in the database and publish the model.

[0089] In the embodiment, the prediction task creation step S5 includes creating a new prediction task, adding prediction task attributes, selecting the prediction task type and prediction algorithm parameters, locating the device object model through the algorithm parameters, selecting the training model, if the prediction result data needs to be entered, the prediction task needs to be repeated, the previous Nth prediction result needs to be selected, the prediction task needs to be executed immediately, and a prediction result is selected. If the prediction result does not need to be entered, an online prediction service needs to be provided, the prediction task is executed, and an online API service is generated for the user to call. If the online prediction service is not needed, the prediction task is saved and then executed according to the strategy, and the device object model is predicted and the prediction result is stored.

[0090] Specifically, in the prediction task creation step S5, first add a prediction task, select the prediction algorithm uploaded in step 2, and select the published model, where the model supports keyword search; depending on the business, select new equipment or previous prediction results as the data input of the prediction algorithm; configure the execution strategy, including scheduled execution, immediate execution, and online service; save the task, download the prediction data in the background, and use the DS platform to perform model prediction, save the prediction results as a csv file, and provide a download interface; if it is an online prediction service, the background automatically deploys a Django service, uses Nginx for reverse proxy, and generates an API for users to call.

[0091] Therefore, the present invention proposes a machine learning method and system. The present invention conducts a demand analysis based on the actual needs within the company, clarifies the problems to be solved, and designs the overall architecture of the platform by analyzing the needs of relevant business developers within the company. The main functions of several modules are introduced in detail in the form of a solution flowchart, showing the internal implementation of several modules and the collaborative relationship between several modules. Aiming at the PHM scenario, the platform establishes a machine learning platform that supports multiple models and multiple engines, providing users with a one-stop full-process machine learning solution from data management, algorithm management, task management to platform services. The platform is based on actual business, providing users with scenario-based solutions in vertical fields to improve user development efficiency.

[0092] Embodiment 2

[0093] Please refer to Figure 8 , Figure 8 FIG. 1 is a schematic diagram of the structure of a machine learning system of the present invention. Figure 8 As shown, a machine learning system of the invention is applicable to the above-mentioned machine learning method, and the machine learning system includes:

[0094] Model building unit 41: building a device object model;

[0095] Data storage unit 42: encapsulates the big data platform interface, calls the interface to store data source information in the database, configures the mapping relationship between the model information and the field information of the device object model, and stores the mapping relationship and mapping data in the database;

[0096] Algorithm management unit: upload the algorithm, select the device object model, upload the algorithm file and configuration file, parse the algorithm parameters according to the information in the configuration file, and store the algorithm parameters in the database;

[0097] The training task creation unit 43 creates a new training task, selects a training task type and training algorithm parameters, locates the device object model through the training algorithm parameters, selects instance data from the device object model according to a time range, configures a training task execution strategy, executes the device object model training, and publishes the training model after storing the training verification result in the database;

[0098] Prediction task creation unit 45: creates a new prediction task, adds prediction task attributes, selects prediction task type and prediction algorithm parameters, locates the device object model and selects the training model through the prediction algorithm parameters, publishes the prediction task, and stores the prediction results after executing the prediction task according to the strategy.

[0099] In this embodiment, the model building unit 41 includes building a device object model and adding device attributes and measurement point data information of the device object model.

[0100] In this embodiment, the data storage unit 42 includes:

[0101] Data source storage module 421: encapsulates the influxdb interface of the big data platform, and calls the influxdb interface to store the database name in the MySql data source table;

[0102] Data mapping storage module 422: configures the mapping relationship between the attributes of the device object model, the measuring point information and the field information, and stores the mapping relationship and mapping data in the MySql database.

[0103] In this embodiment, the algorithm management unit 43 includes:

[0104] Algorithm upload module 431: upload the algorithm, add the attributes of the algorithm, select the device object model, and download the data of the device object model. If auxiliary files are required, select them and package and upload the algorithm file, configuration file and auxiliary files and save them to the server;

[0105] Algorithm parsing module 432: parses the algorithm parameters according to the information in the configuration file, compares the algorithm parameters with the measurement point data information of the device object model, and stores the algorithm parameters in the database.

[0106] In this embodiment, the training task creation unit 44 includes creating a new training task, adding a training task name, selecting a training task type and training algorithm parameters, locating the device object model through the training algorithm parameters, and selecting the instance data time range of the device object model. After selecting instance data from the device object model according to the instance data time range, the training task configures an execution strategy and saves the execution strategy, trains the device object model according to the execution strategy, and stores the training verification results in the database before publishing the training model.

[0107] In this embodiment, the prediction task creation unit 45 includes creating a new prediction task, adding prediction task attributes, selecting a prediction task type and prediction algorithm parameters, locating the device object model through the algorithm parameters, selecting the training model, if it is necessary to input prediction result data, repeating the prediction task, selecting the previous Nth prediction result, executing the prediction task immediately, selecting a prediction result, if it is not necessary to input the prediction result, providing an online prediction service, executing the prediction task, generating an online API service for users to call, if it is not necessary to provide the online prediction service, saving the prediction task and executing the prediction task according to the strategy, and performing a prediction of the device object model, and storing the prediction result.

[0108] Embodiment 3

[0109] Combination Fig. 9 As shown, this embodiment discloses a specific implementation of an electronic device. The electronic device may include a processor 81 and a memory 82 storing computer program instructions.

[0110] Specifically, the processor 81 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0111] Among them, the memory 82 may include a large-capacity memory for data or instructions. By way of example and not limitation, the memory 82 may include a hard disk drive (HDD), a floppy disk drive, a solid-state drive (SSD), a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory 82 may include a removable or non-removable (or fixed) medium. Where appropriate, the memory 82 may be inside or outside the data processing device. In a specific embodiment, the memory 82 is a non-volatile memory. In a specific embodiment, the memory 82 includes a read-only memory (ROM) and a random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (Programmable Read-Only Memory, PROM for short), an erasable PROM (Erasable ProgrammableRead-Only Memory, FPROM for short), an electrically erasable PROM (Electrically Erasable ProgrammableRead-Only Memory, EFPROM for short), an electrically alterable ROM (Electrically Alterable Read-Only Memory, EAROM for short) or a flash memory (FLASH) or a combination of two or more of these. Under appropriate circumstances, the RAM can be a static random access memory (SRAM) or a dynamic random access memory (DRAM), wherein the DRAM can be a fast page mode dynamic random access memory (FPMDRAM), an extended data output dynamic random access memory (EDODRAM), a synchronous dynamic random access memory (SDRAM), etc.

[0112] The memory 82 may be used to store or cache various data files required for processing and / or communication, as well as possible computer program instructions executed by the processor 81 .

[0113] The processor 81 implements any one of the file system capacity management optimization methods in the above embodiments by reading and executing computer program instructions stored in the memory 82 .

[0114] In some of the embodiments, the electronic device may further include a communication interface 83 and a bus 80. Fig. 9 As shown, the processor 81, the memory 82, and the communication interface 83 are connected via a bus 80 and communicate with each other.

[0115] The communication interface 83 is used to implement communication between the modules, devices, units and / or equipment in the embodiment of the present application. The communication port 83 can also implement data communication with other components such as: external devices, image / data acquisition equipment, databases, external storage, and image / data processing workstations.

[0116] The bus 80 includes hardware, software or both, and couples the components of the computer device to each other. The bus 80 includes but is not limited to at least one of the following: a data bus, an address bus, a control bus, an expansion bus, and a local bus. By way of example and not limitation, bus 80 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses or a combination of two or more of these. Where appropriate, bus 80 may include one or more buses. Although embodiments of the present application describe and illustrate a particular bus, the present application contemplates any suitable bus or interconnect.

[0117] Electronic devices can be connected to machine learning systems to achieve integration Figure 1-Figure 7 Describe the method.

[0118] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0119] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A machine learning method, characterized in that: Based on PHM data modeling, the machine learning method includes: Model building steps: Establish device object model; Data storage step: encapsulating a big data platform interface, calling the interface to store data source information in a database, configuring a mapping relationship between model information and field information of the device object model, and storing the mapping relationship and mapping data in the database; Algorithm management step: uploading the algorithm, selecting the device object model, uploading the algorithm file and configuration file, parsing the algorithm parameters according to the information in the configuration file, and storing the algorithm parameters in the database; Training task creation step: create a new training task, select the training task type and training algorithm parameters, locate the device object model through the training algorithm parameters, select instance data from the device object model according to the time range, configure the training task execution strategy, execute the device object model training, store the training verification results in the database, and then publish the training model; Prediction task creation step: create a new prediction task, add prediction task attributes, select the prediction task type and prediction algorithm parameters, locate the device object model and select the training model through the prediction algorithm parameters, publish the prediction task, and store the prediction results after executing the prediction task according to the strategy; The model building step includes building a device object model and adding device attributes and measurement point data information of the device object model, wherein the device attributes include device ID and device name, and the data information includes temperature, humidity, and pressure; Wherein, the algorithm management step includes: Algorithm upload step: upload the algorithm, add the attributes of the algorithm, select the device object model, and download the data of the device object model. If auxiliary files are required, select them and package and upload the algorithm file, configuration file and auxiliary files and save them to the server; Algorithm parsing step: parsing the algorithm parameters according to the information in the configuration file, comparing the algorithm parameters with the measurement point data information of the device object model, and storing the algorithm parameters in the database; Wherein, the data storage step comprises: Data source storage step: encapsulate the influxdb interface of the big data platform, call the influxdb interface to store the database name in the MySql data source table; Data mapping storage step: configuring the mapping relationship between the attributes of the device object model, the measuring point information and the field information, and storing the mapping relationship and mapping data in a MySql database.

2. A machine learning method according to claim 1, characterized in that: The training task creation step includes creating a new training task, adding a training task name, selecting a training task type and training algorithm parameters, locating the device object model through the training algorithm parameters, and selecting an instance data time range of the device object model. After selecting instance data from the device object model according to the instance data time range, the training task configures an execution strategy and saves the execution strategy, trains the device object model according to the execution strategy, and publishes the training model after storing the training verification results in the database.

3. A machine learning method according to claim 2, characterized in that: The prediction task creation step includes: creating a new prediction task, adding prediction task attributes, selecting the prediction task type and prediction algorithm parameters, locating the device object model through the algorithm parameters, selecting the training model, if it is necessary to input the prediction result data, repeating the prediction task, selecting the previous Nth prediction result, executing the prediction task immediately, selecting a prediction result, if it is not necessary to input the prediction result, providing an online prediction service, executing the prediction task, generating an online API service for users to call, if it is not necessary to provide the online prediction service, saving the prediction task and executing the prediction task according to the strategy, and performing the prediction of the device object model, and storing the prediction result.

4. A machine learning system, characterized in that: A machine learning method based on PHM data modeling, applicable to any one of claims 1 to 3 above, comprising: Model building unit: building device object model; Data storage unit: encapsulates the big data platform interface, calls the interface to store data source information in the database, configures the mapping relationship between the model information and the field information of the device object model, and stores the mapping relationship and mapping data in the database; Algorithm management unit: upload the algorithm, select the device object model, upload the algorithm file and configuration file, parse the algorithm parameters according to the information in the configuration file, and store the algorithm parameters in the database; A training task creation unit: creates a new training task, selects a training task type and training algorithm parameters, locates the device object model through the training algorithm parameters, selects instance data from the device object model according to a time range, configures a training task execution strategy, executes the device object model training, and releases the training model after storing the training verification results in the database; Prediction task creation unit: create a new prediction task, add prediction task attributes, select the prediction task type and prediction algorithm parameters, locate the device object model and select the training model through the prediction algorithm parameters, publish the prediction task, and store the prediction results after executing the prediction task according to the strategy; The model building unit includes building a device object model and adding device attributes and measurement point data information of the device object model, wherein the device attributes include device ID and device name, and the data information includes temperature, humidity, and pressure; Wherein, the algorithm management unit includes: Algorithm upload module: upload the algorithm, add the attributes of the algorithm, select the device object model, and download the data of the device object model. If auxiliary files are required, select them and package and upload the algorithm file, configuration file and auxiliary files and save them to the server; Algorithm parsing module: parsing the algorithm parameters according to the information in the configuration file, comparing the algorithm parameters with the measurement point data information of the device object model, and storing the algorithm parameters in the database; Wherein, the data storage unit comprises: Data source storage module: encapsulates the influxdb interface of the big data platform, and calls the influxdb interface to store the database name in the MySql data source table; Data mapping storage module: configures the mapping relationship between the attributes of the device object model, the measuring point information and the field information, and stores the mapping relationship and mapping data in a MySql database.

Citation Information

Patent Citations

  • Method and device for processing machine learning models

    CN108510082A

  • Apparatus and method for degradation analysis of a repairable system

    KR102092197B1

  • System and method for analysis and determination of relationships from a variety of data sources

    US20200133964A1

  • Distributed and redundant machine quality management

    US20200380417A1

  • System and method for cognitive engineering technology for automation and control of systems

    WO2018140365A1