WO2023151523A1 - 基于数字孪生DaaS平台的深度学习编程方法及*** - Google Patents

基于数字孪生DaaS平台的深度学习编程方法及*** Download PDF

Info

Publication number
WO2023151523A1
WO2023151523A1 PCT/CN2023/074477 CN2023074477W WO2023151523A1 WO 2023151523 A1 WO2023151523 A1 WO 2023151523A1 CN 2023074477 W CN2023074477 W CN 2023074477W WO 2023151523 A1 WO2023151523 A1 WO 2023151523A1
Authority
WO
WIPO (PCT)
Prior art keywords
deep learning
sub
digital twin
operation data
equipment
Prior art date
Application number
PCT/CN2023/074477
Other languages
English (en)
French (fr)
Inventor
刘天琼
Original Assignee
深圳市爱云信息科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳市爱云信息科技有限公司 filed Critical 深圳市爱云信息科技有限公司
Publication of WO2023151523A1 publication Critical patent/WO2023151523A1/zh

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F8/00—Arrangements for software engineering
    • G06F8/60—Software deployment
    • G06F8/65—Updates
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F8/00—Arrangements for software engineering
    • G06F8/70—Software maintenance or management
    • G06F8/71—Version control; Configuration management
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Definitions

  • This application relates to the field of artificial intelligence Internet of Things, in particular to a deep learning programming method and system based on a digital twin DaaS platform.
  • the method of terminal equipment optimization and upgrading is usually based on the operation data of the terminal equipment, and the technicians manually upgrade the code, and manually optimize the terminal equipment.
  • the code requires high labor costs, which leads to low efficiency in the optimization and upgrading of terminal equipment, and affects the operation effect of terminal equipment.
  • the main purpose of this application is to provide a deep learning programming method based on the digital twin DaaS platform, which aims to solve the technical problems of low efficiency in business process and equipment optimization and upgrading and high labor costs in the prior art.
  • the deep learning programming method based on the digital twin DaaS platform includes:
  • equipment operation data of each sub-operation flow in the business flow of different terminal equipment wherein the equipment operation data includes the process number of each sub-operation flow
  • the sequence programming model is constructed based on pre-collected device operation data of each sub-operation process corresponding to the business process in different terminal devices.
  • the step of acquiring the equipment operation data of each sub-job process in the business process of different terminal devices includes:
  • the device operation data of each of the sub-job processes reported by the different terminal devices is acquired respectively.
  • the step of comparing and analyzing the equipment operation data of each of the sub-operation processes after data modeling is compared:
  • analysis results are obtained. analysis results.
  • the step of distributing and upgrading the target operating code to corresponding terminal devices includes:
  • the target operating code is distributed and upgraded to the sub-job flow of the corresponding terminal device through a preset device upgrade method.
  • the deep learning programming method based on the digital twin DaaS platform is also include:
  • the steps also include:
  • the algorithm engine is controlled to perform rollback processing to roll back to the operation code corresponding to the previous version number, and the abnormal operation code corresponding to the abnormal result is upgraded by the developer to The upgraded operating code is upgraded to the terminal device.
  • the preset communication protocol includes 2G, 3G, 4G, 5G, CAT1 network transmission, CAT4 network transmission, NB-IOT narrowband Internet of Things, LORA low-power remote wireless communication, MQTT message queue telemetry transmission, HTTP, One or more of TCP and UDP transport layer protocols.
  • the present application also provides a deep learning programming system based on the digital twin DaaS platform, the deep learning programming system based on the digital twin DaaS platform is a virtual system, and the deep learning programming system based on the digital twin DaaS platform is applied to artificial intelligence objects.
  • Networking platform the artificial intelligence Internet of Things platform is connected to at least one terminal device, and the deep learning programming system based on the digital twin DaaS platform includes:
  • the obtaining module is used to obtain the equipment operation data of each sub-operation process in the business process of different terminal devices, wherein the The equipment operation data includes the process number of each sub-operation flow;
  • the analysis module is used to perform data modeling and comparative analysis on the equipment operation data of the different terminal equipment, and judge whether the corresponding terminal equipment needs to be upgraded and optimized based on the analysis results;
  • the deep learning programming module is used to perform deep learning programming on the equipment operating data of each sub-job process based on the trained time sequence programming model, automatically generate the target operating code, and distribute the target operating code Upgrading to a corresponding terminal device, wherein the sequence programming model is constructed based on pre-collected device operation data of each sub-operation process corresponding to a business process in different terminal devices.
  • the present application also provides a deep learning programming device based on the digital twin DaaS platform, the deep learning programming device based on the digital twin DaaS platform is a physical device, and the deep learning programming device based on the digital twin DaaS platform includes: memory, processing The processor and the deep learning programming program based on the digital twin DaaS platform stored on the memory, the deep learning programming program based on the digital twin DaaS platform is executed by the processor to implement the above-mentioned deep learning based on the digital twin DaaS platform The steps of the programming method.
  • the present application also provides a storage medium, the storage medium is a computer-readable storage medium, and the deep learning programming program based on the digital twin DaaS platform is stored on the computer-readable storage medium, and the deep learning programming program based on the digital twin DaaS platform
  • the programming program is executed by the processor to realize the steps of the above-mentioned deep learning programming method based on the digital twin DaaS platform.
  • This application provides a deep learning programming method and system based on the digital twin DaaS platform. Compared with the technical means used in the prior art to manually upgrade and optimize equipment through developers, this application first obtains the business processes of different terminal equipment
  • the equipment operation data of each sub-operation process wherein the equipment operation data includes the process number of each sub-operation process, which realizes the collection of equipment operation data of the business process through the artificial intelligence Internet of Things platform, and the collected data is the corresponding terminal
  • the specific sub-job flow of the equipment so that after the subsequent deep learning automatic programming, the sub-job flow of the terminal equipment described in the automatic programming operation code can be determined based on the process number, and further, the equipment operation data of the different terminal equipment Perform comparative analysis after data modeling, and judge whether the corresponding terminal equipment needs to be upgraded and optimized based on the analysis results, and then, based on the trained timing programming model, perform deep learning programming on the equipment operation data of each sub-operation process , automatically generate the target running code, and distribute and upgrade the target running code to the corresponding terminal devices, where
  • the equipment operation data of each sub-operation process corresponding to the process is automatically programmed by deep learning, without manual programming, which greatly reduces labor costs and improves the efficiency of equipment optimization and upgrading. Further, it will automatically The generated operating code is distributed and upgraded to the corresponding terminal equipment, and the code is highly portable, thereby improving the intelligence of terminal equipment management.
  • Fig. 1 is a schematic flow diagram of the first embodiment of the deep learning programming method based on the digital twin DaaS platform of the present application;
  • FIG. 2 is a schematic flow diagram of the second embodiment of the deep learning programming method based on the digital twin DaaS platform of the present application;
  • FIG. 3 is a schematic structural diagram of a deep learning programming device based on the digital twin DaaS platform of the hardware operating environment involved in the embodiment of the present application;
  • FIG. 4 is a schematic diagram of the functional modules of the deep learning programming device of the present application.
  • the embodiment of the present application provides a deep learning programming method based on the digital twin DaaS (data as a service) platform.
  • the deep learning programming methods of the digital twin DaaS platform include:
  • Step S10 acquiring the equipment operation data of each sub-operation flow in the business flow of different terminal equipment, wherein the equipment operation data includes the process number of each sub-operation flow;
  • the deep learning programming is applied to the artificial intelligence Internet of Things (AIOT) platform
  • the artificial intelligence Internet of Things platform can be applied to fields such as intelligent transportation, intelligent security, intelligent medical treatment, and industrial Internet of Things.
  • the artificial intelligence Internet of Things platform includes a device layer, and the artificial intelligence Internet of Things platform is connected to at least one terminal device through the device layer, and the terminal device includes intelligent transportation equipment, industrial intelligent robots, network terminal equipment, human body intelligent wearable equipment, engineering intelligence Mechanical equipment, agricultural intelligent machinery and equipment, CNC machine tools, intelligent sensors, intelligent collectors, intelligent cameras, intelligent transmitters and embedded intelligent systems, etc.
  • each terminal device is configured with a corresponding device identifier
  • the business process includes a plurality of sub-operation processes
  • each of the sub-operation processes is configured with a corresponding process code. number, which includes multiple procedures or links in the entire business process, so that the sub-operation process of the corresponding terminal equipment can be determined through the equipment identification and process number.
  • the business process of terminal equipment It can be multiple sub-operation processes such as starting up, preheating, loading raw materials, polishing, and aging.
  • the inspection results of multiple parts of the body can be used as the sub-operation process.
  • Obtain the equipment operation data of each sub-operation process in the business process of different terminal devices specifically, after the terminal device is connected to the artificial intelligence Internet of Things platform, obtain the business processes corresponding to different terminal devices through the artificial intelligence Internet of Things platform , when the terminal device executes the corresponding business process, the equipment operation data of each sub-operation process in the business process is collected, and further, through the data communication protocol between the terminal device and the artificial intelligence Internet of Things platform, the various sub-operation processes The operating data of the equipment is reported to the artificial intelligence Internet of things platform.
  • the step of obtaining the device operation data of each sub-operation process in the business process of different terminal devices includes:
  • Step S11 based on a preset communication protocol, respectively acquire the device operation data of each of the sub-job processes reported by the different terminal devices.
  • the preset communication protocol includes at least one of 2G, 3G, 4G, 5G, CAT1, CAT4, NB-IOT, LORA, MQTT, HTTP, TCP, UDP and CoAP protocols or more.
  • the device operation data of each sub-job flow reported by the different terminal devices is acquired in real time.
  • Step S20 performing data modeling and comparative analysis on the equipment operation data of the different terminal equipment, and judging whether the corresponding terminal equipment needs to be upgraded and optimized based on the analysis results;
  • the device operation data of the different terminal devices is compared and analyzed after data modeling, and based on the analysis results, it is judged whether the corresponding terminal device needs to be upgraded and optimized, specifically, the collected device operation data is cleaned , filtering and other preprocessing, and combine the equipment operation data obtained by the preprocessing with 3D digital twin simulation technology for fusion analysis to obtain the analysis results.
  • the terminal device is connected to the artificial intelligence Internet of Things platform, and after receiving the access instruction corresponding to the terminal device's access to the artificial intelligence Internet of Things platform, collect each sub-job in the business process corresponding to different terminal devices
  • the equipment operation data of the process assuming that the business process corresponding to the terminal equipment includes 50 procedures, collect the equipment operation data corresponding to the 50 procedures in the business process, and conduct modeling analysis in conjunction with the 3D digital twin simulation model, so that the equipment of different processes can Operational data to determine the equipment operating status of different processes in the business process of terminal equipment.
  • AI viewing in smart medical care by connecting medical equipment to the artificial intelligence Internet of Things platform, users can collect The corresponding film and the film report stored in the test report database, so that when the user watches the film, the test report of different parts in the test report database is combined with the data 3D digital twin simulation technology for modeling analysis, so that the comprehensive analysis based on multiple parts is accurate Determine the user's detection results, and further, connect the AI camera to the artificial intelligence Internet of Things platform.
  • the image data captured by different AI cameras can be collected , and then in the AI camera monitoring of smart transportation, the captured images corresponding to different cameras in the same area can be combined with the data 3D digital twin simulation technology for data modeling and analysis, so as to accurately monitor the vehicle.
  • Step S21 based on the equipment operation data of each sub-operation process, combined with 3D digital twin simulation technology to perform fusion analysis to obtain analysis results.
  • the analysis results are obtained. Specifically, based on the equipment operation data of each of the sub-operation processes, through the 3D digital twin simulation technology conducts digital modeling analysis to obtain the analysis results.
  • Step S30 if yes, then based on the trained sequential programming model, perform deep learning programming on the equipment operation data of each sub-workflow, automatically generate the target operation code, and distribute and upgrade the target operation code to the corresponding The terminal device, wherein the sequence programming model is constructed based on pre-collected device operation data of each sub-operation process corresponding to the business process in different terminal devices.
  • the target operation code is automatically generated, and the target operation code is distributed and upgraded To the corresponding terminal device, wherein the sequence programming model is constructed based on the pre-collected equipment operation data of each sub-operation process corresponding to the business process in different terminal devices, specifically, if so, it proves that the terminal device needs to be upgraded optimization, and then based on the sequential programming model, perform deep learning programming on the equipment operation data of each of the sub-operation processes, thereby automatically generating the target operation code corresponding to each of the sub-operation processes, and passing the target operation code through the distribution
  • the form of the formula is intelligently upgraded to the corresponding terminal device, so that the terminal device can perform operation and debugging based on the target operation code, and improve the operation effect of the terminal device.
  • the business process corresponding to the terminal device includes 50 procedures (starting up, preheating, loading raw materials, polishing and aging, etc.), among which, assuming that in the traditional industrial Internet, the startup time is 8 seconds, the preheating time is 1 minute, and the raw material loading time is 5 minutes, based on the sub-operation process in the business process
  • the deep learning programming is carried out through the sequence programming model, thereby automatically optimizing the terminal equipment, which can make the terminal equipment assume that the boot time is 1 second, the warm-up time is 0 seconds, and the loading time of raw materials is 1 minute, thereby greatly improving
  • the operating efficiency of the equipment further, in the traditional terminal equipment control, professional technicians are required to control and upgrade and optimize the terminal equipment, and through this application, directly according to each business process in the terminal equipment After the equipment operation data corresponding to the process is modeled and analyzed, the upgrade package is generated through the sequential programming model, so that ordinary personnel can also operate and optimize the terminal equipment, greatly reducing the labor cost
  • the parameters of the terminal equipment are automatically optimized according to the test report database, so that the user can be accurately determined based on the comprehensive analysis of multiple parts.
  • the automatic programming is based on the test report database and the sequential programming model, so that medical experts do not need to analyze the medical film report, and ordinary doctors can also directly automatically according to the sequential programming model.
  • the medical terminal equipment optimized by deep learning programming can view the film report, which greatly reduces the labor cost and resources.
  • the AI camera monitoring of smart transportation it is necessary to obtain the image data captured by the smart camera for analysis.
  • the Different AI cameras are connected to the artificial intelligence Internet of Things platform to collect image data corresponding to different AI cameras.
  • the image data captured by different cameras in this area for the license plate is collected.
  • the image data of the camera is collected in an all-round way through the artificial intelligence Internet platform, and combined with the data 3D digital twin simulation technology for modeling and analysis, and then automatically programmed and optimized through the sequential programming model. Professional technicians monitor and analyze the image data of the camera, thereby greatly reducing labor costs.
  • the terminal equipment is connected to the artificial intelligence Internet platform in advance, so as to transmit the information of the terminal equipment based on the preset communication protocol.
  • Equipment operation data and automatically optimize the parameters of the terminal equipment based on the timing programming model, not only improves the operating efficiency of the terminal equipment, but also greatly reduces the cost of human resources.
  • the step of distributing the target operating code to the corresponding terminal device includes:
  • Step S31 classify the algorithm engines corresponding to the sub-operation processes of the business processes of the different terminal devices through the preset artificial intelligence algorithm platform;
  • Step S32 based on the engine classification result, distribute and upgrade the target running code to the sub-job flow of the corresponding terminal device through a preset device upgrade method.
  • the preset device upgrade method includes an OTA upgrade method, wherein the OTA (Over-the-AirTechnology) upgrade means that the terminal device downloads an upgrade package on a remote server through a wireless network, A technique for upgrading a system or application.
  • OTA Over-the-AirTechnology
  • different terminal devices are equipped with corresponding algorithm engines in the artificial intelligence algorithm center of the artificial intelligence Internet of Things platform, and the business process of a terminal device includes multiple sub-operation processes, therefore, it is necessary to combine different word operation processes
  • the corresponding algorithm engine is classified, so that based on the algorithm engine, the deep learning can be upgraded through the OTA upgrade method
  • the running code of learning programming is upgraded to the sub-job process of the corresponding terminal device.
  • Step A10 marking the version number corresponding to the target running code, and storing the target running code and the corresponding version number;
  • the programmed running code needs to be stored, and the version number of the running code should be marked.
  • the running code of the last deep learning programming will not be overwritten, but will be stored together on the artificial intelligence Internet of Things platform.
  • a preset number of times can be stored for deep learning
  • the running code of the programming for example, stores the running code 3 times before the current deep learning programming.
  • Step A20 obtaining the running result of debugging and optimizing the code based on the target running code through the terminal device;
  • the running result of debugging and optimizing the terminal device based on the target running code is obtained. Specifically, after the terminal device obtains the target running code, the terminal device performs pre-running based on the target running code, and obtains The running result is used to judge whether the terminal device is optimized and upgraded successfully.
  • Step A30 if the operation result is an abnormal result, control the algorithm engine to perform rollback processing to roll back to the operation code corresponding to the previous version number, and the abnormal operation code corresponding to the abnormal result is processed by the developer. and upgrading, so as to upgrade the upgraded running code to the terminal device.
  • the running result is an abnormal result
  • the algorithm engine corresponding to the abnormality is controlled to perform rollback processing, thereby rolling back to the previous version
  • the running code corresponding to the version number so that the terminal device runs according to the running code corresponding to the previous version number, avoiding the defect of terminal device running interruption, and then notifying the developer to manually upgrade the abnormal running code corresponding to the abnormal result, and
  • the upgraded operating code is upgraded to the terminal device, thereby completing the upgrade and optimization operation of the terminal device.
  • the embodiment of the present application provides a deep learning programming method based on the digital twin DaaS platform. Compared with the technical means of manually upgrading and optimizing the equipment adopted by developers in the prior art, the embodiment of the present application first obtains the business processes of different terminal devices The equipment operation data of each sub-operation process, wherein the equipment operation data includes the process number of each sub-operation process, which realizes the collection of equipment operation data of the business process through the artificial intelligence Internet of Things platform, and the collected data is detailed to the corresponding The specific sub-job flow of the terminal equipment, so that after the subsequent deep learning automatic programming, the sub-job flow of the terminal equipment described in the automatically programmed running code can be determined based on the process number, and further, for the different terminal equipment The equipment operation data of the terminal equipment is compared and analyzed after data modeling, and based on the analysis results, it is judged whether the corresponding terminal equipment needs to be upgraded and optimized.
  • the equipment operation data of the sub-job process is constructed, and the operation data of the specific job sub-process in different business processes can be learned through the model by splitting the business process into sub-job processes, so as to improve the accuracy of the timing programming model, and then Through the time series programming model, deep learning and automatic programming are performed on the equipment operation data of each sub-operation process corresponding to the business process, without manual programming, which greatly reduces labor costs and improves the efficiency of equipment optimization and upgrading. Further, the automatically generated operation code Distributed upgrade to the corresponding terminal equipment, the code portability is high, thereby improving the intelligence of terminal equipment management.
  • the equipment operation data of each of the sub-workflows is in-depth Learning programming, automatically generating target operating codes, and distributing and upgrading the target operating codes to corresponding terminal devices
  • the sequence programming model is based on pre-collected sub-job processes corresponding to business processes in different terminal devices
  • the deep learning programming method based on the digital twin DaaS platform also includes:
  • Step B10 obtaining business processes of different terminal devices, and collecting training device operation data of each sub-operation process in different business processes
  • Step B20 based on the training device operation data of each of the sub-workflows, train the initial model to be trained to obtain the sequence programming model.
  • the learning goal of training the sequential programming model is to allow the deep learning programming method based on the digital twin DaaS platform to imitate this behavior process for input data, and finally to form correct output data.
  • these conversion rules are encoded into the network parameters of the neural network.
  • the neural network performs feature extraction and representation on the input data to learn and predict based on the feature extraction results, thereby outputting the final code result. .
  • the training device operation data of each sub-operation process in the respective business processes of the same terminal device, and then extract the characteristic information of the training device operation data, so as to iteratively optimize the initial model to be trained based on the characteristic information network parameters, and then judge whether the optimized initial model to be trained satisfies the training end condition, wherein the training end condition includes conditions such as the loss function reaching convergence or the number of iterations reaching the preset number of iterations, etc., if satisfied, the timing sequence is obtained If the programming model is not satisfied, return to the execution step: obtain the business processes of different terminal devices, and collect the training device operation data of each sub-job process in different business processes.
  • the embodiment of the present application provides a deep learning programming method based on the digital twin DaaS platform, that is, to obtain the business processes of different terminal devices, and collect the training equipment operation data of each sub-operation process in different business processes, and then based on each
  • the training equipment operating data of the sub-operation process is used to train the initial model to be trained, and the sequence programming model is obtained, which realizes model training according to the equipment operation data of each sub-operation process in different business processes, so that the model can learn specific
  • the operation data of the job sub-process thereby improving the accuracy of the sequence programming model, and then based on the sequence programming model, the device operation data of different terminal devices can be automatically programmed by deep learning, and the efficiency of terminal device upgrade and optimization can be improved.
  • FIG. 3 is a schematic structural diagram of a deep learning programming device based on a digital twin DaaS platform of a hardware operating environment involved in the embodiment of the present application.
  • the deep learning programming device based on the digital twin DaaS platform may include: a processor 1001 , such as a CPU, a memory 1005 , and a communication bus 1002 .
  • the communication bus 1002 is used to realize connection and communication between the processor 1001 and the memory 1005 .
  • the memory 1005 can be a high-speed RAM memory, or a stable memory (non-volatile memory), such as a disk memory.
  • the memory 1005 may also be a storage device independent of the aforementioned processor 1001 .
  • the deep learning programming device based on the digital twin DaaS platform may also include a rectangular user interface, a network interface, a camera, an RF (Radio Frequency, radio frequency) circuit, a sensor, an audio circuit, a WiFi module, and the like.
  • the rectangular user interface may include a display screen (Display), an input sub-module such as a keyboard (Keyboard), and the optional rectangular user interface may also include a standard wired interface and a wireless interface.
  • the network interface may include a standard wired interface and a wireless interface (such as a WIFI interface).
  • the deep learning programming device structure based on the digital twin DaaS platform shown in Figure 3 does not constitute a limitation on the deep learning programming device based on the digital twin DaaS platform, and may include more or more Fewer components, or combinations of certain components, or different arrangements of components.
  • the memory 1005 as a computer storage medium may include an operating system, a network communication module, and a deep learning programming program based on the digital twin DaaS platform.
  • the operating system is a program that manages and controls the hardware and software resources of deep learning programming equipment based on the digital twin DaaS platform, and supports the operation of deep learning programming programs based on the digital twin DaaS platform and other software and/or programs.
  • the network communication module is used to realize the communication between various components inside the memory 1005, and communicate with other hardware and software in the deep learning programming method based on the digital twin DaaS platform.
  • the specific implementation of the deep learning programming device based on the digital twin DaaS platform of this application is basically the same as the above-mentioned embodiments of the deep learning programming method based on the digital twin DaaS platform, and will not be repeated here.
  • Figure 4 is a schematic diagram of the functional modules of the deep learning programming device of the present application, the present application also provides a deep learning programming system based on the digital twin DaaS platform, the deep learning programming system based on the digital twin DaaS platform include:
  • An acquisition module configured to acquire equipment operation data of each sub-operation flow in the business flow of different terminal equipment, wherein the equipment operation data includes the process number of each sub-operation flow;
  • the analysis module is used to perform data modeling and comparative analysis on the equipment operation data of the different terminal equipment, and judge whether the corresponding terminal equipment needs to be upgraded and optimized based on the analysis results;
  • the deep learning programming module is used to perform deep learning programming on the equipment operating data of each sub-job process based on the trained time sequence programming model, automatically generate the target operating code, and distribute the target operating code Upgrading to a corresponding terminal device, wherein the sequence programming model is constructed based on pre-collected device operation data of each sub-operation process corresponding to a business process in different terminal devices.
  • the acquisition module is also used for:
  • the device operation data of each of the sub-job processes reported by the different terminal devices is acquired respectively.
  • the analysis module is also used for:
  • the deep learning programming module is also used for:
  • the target operating code is distributed and upgraded to the sub-job flow of the corresponding terminal device through a preset device upgrade method.
  • the deep learning programming system based on the digital twin DaaS platform is also used for:
  • the deep learning programming system based on the digital twin DaaS platform is also used for:
  • the algorithm engine is controlled to perform rollback processing to roll back to the operation code corresponding to the previous version number, and the abnormal operation code corresponding to the abnormal result is upgraded by the developer to The upgraded operating code is upgraded to the terminal device.
  • the specific implementation of the deep learning programming system based on the digital twin DaaS platform of this application is basically the same as the above-mentioned embodiments of the deep learning programming system based on the digital twin DaaS platform, and will not be repeated here.
  • An embodiment of the present application provides a storage medium, the storage medium is a computer-readable storage medium, and the computer-readable storage medium stores one or more programs, and the one or more programs can also be used by one or more More than one processor is executed to implement the steps of the deep learning programming method based on the digital twin DaaS platform described in any one of the above.
  • the specific implementation of the computer-readable storage medium of the present application is basically the same as the above-mentioned embodiments of the deep learning programming method based on the digital twin DaaS platform, and will not be repeated here.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Security & Cryptography (AREA)
  • Biomedical Technology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Stored Programmes (AREA)

Abstract

本申请公开了基于数字孪生DaaS平台的深度学习编程方法及***，包括：获取不同终端设备的业务流程中各子作业流程的设备运行数据，其中，所述设备运行数据包括各子作业流程的工序编号，对所述不同终端设备中各子作业流程的设备运行数据进行数据建模后对比分析，并基于分析结果判断终端设备是否需要升级优化，则基于已训练好的时序编程模型，对各所述子作业流程的设备运行数据进行深度学习编程，自动生成目标运行代码，并将所述目标运行代码分布式升级至对应的终端设备，其中，所述时序编程模型为基于预先采集的不同终端设备中业务流程对应的各子作业流程的设备运行数据进行构建。本申请解决了业务流程和设备优化升级效率较低以及人力成本较高的问题。

Description

基于数字孪生DaaS平台的深度学习编程方法及*** 技术领域
本申请涉及人工智能物联网领域，尤其涉及一种基于数字孪生DaaS平台的深度学习编程方法及***。
背景技术
在终端设备运行过程中，往往需要检测终端设备是否需要升级优化，目前，终端设备优化升级的方法通常是基于终端设备的运行数据，通过技术人员手动进行代码的升级，而手动进行优化终端设备的代码，需要花费较高的人工成本，导致终端设备的优化升级的效率较低，影响终端设备的运行效果。
发明内容
本申请的主要目的在于提供一种基于数字孪生DaaS平台的深度学习编程方法，旨在解决现有技术中的业务流程和设备优化升级效率较低以及人力成本较高的技术问题。
为实现上述目的，本申请提供了一种基于数字孪生DaaS平台的深度学***台的深度学习编程方法包括：
获取不同终端设备的业务流程中各子作业流程的设备运行数据，其中，所述设备运行数据包括各所述子作业流程的工序编号；
对所述不同终端设备中各子作业流程的设备运行数据进行数据建模后对比分析，并基于分析结果判断对应的终端设备是否需要升级优化；
若是，则基于已训练好的时序编程模型，对各所述子作业流程的设备运行数据进行深度学习编程，自动生成目标运行代码，并将所述目标运行代码分布式升级至对应的终端设备，其中，所述时序编程模型为基于预先采集的不同终端设备中业务流程对应的各子作业流程的设备运行数据进行构建。
可选地，所述获取不同终端设备的业务流程中各子作业流程的设备运行数据的步骤包括：
基于预设通信协议，分别获取通过所述不同终端设备上报的各所述子作业流程的设备运行数据。
可选地，所述对各所述子作业流程的设备运行数据进行数据建模后对比分析的步骤：
基于各所述子作业流程的设备运行数据，结合3D数字孪生仿真技术进行融合分析，获得分 析结果。
可选地，所述将所述目标运行代码分布式升级至对应的终端设备的步骤包括：
通过预先设置的人工智能算法中台将所述不同终端设备的业务流程的子作业流程对应的算法引擎进行归类；
基于引擎归类结果，通过预先设置的设备升级方法将所述目标运行代码分布式升级至对应的终端设备的子作业流程中。
可选地，在所述若是，则基于时序编程模型，对各所述子作业流程的设备运行数据进行深度学***台的深度学习编程方法还包括：
获取不同终端设备的业务流程，并采集不同业务流程中各子作业流程的训练设备运行数据；基于各所述子作业流程的训练设备运行数据，对待训练初始模型进行训练，获得所述时序编程模型。
可选地，在所述基于时序编程模型，对各所述子作业流程的设备运行数据进行深度学习编程，自动生成目标运行代码，并将所述目标运行代码分布式升级至对应的终端设备的步骤之后，还包括：
标记所述目标运行代码对应的版本号，并将所述目标运行代码和对应的版本号进行存储；
获取通过终端设备基于所述目标运行代码进行调试和优化的运行结果；
若所述运行结果为异常结果，则控制所述算法引擎进行回滚处理，以回滚至上一版本号对应的运行代码，并通过开发人员对所述异常结果对应的异常运行代码进行升级，以将所述升级后的运行代码升级至所述终端设备。
可选地，所述预设通信协议包括2G、3G、4G、5G、CAT1网络传输、CAT4网络传输、NB-IOT窄带物联网、LORA低功耗远程无线通信、MQTT消息队列遥测传输、HTTP、TCP、UDP传输层协议中的一种或多种。
本申请还提供一种基于数字孪生DaaS平台的深度学***台的深度学***台的深度学***台，所述人工智能物联网平台连接至少一个终端设备，所述基于数字孪生DaaS平台的深度学习编程***包括：
获取模块，用于获取不同终端设备的业务流程中各子作业流程的设备运行数据，其中，所述 设备运行数据包括各所述子作业流程的工序编号；
分析模块，用于对所述不同终端设备的设备运行数据进行数据建模后对比分析，并基于分析结果判断对应的终端设备是否需要升级优化；
深度学习编程模块，用于若是，则基于已训练好的时序编程模型，对各所述子作业流程的设备运行数据进行深度学习编程，自动生成目标运行代码，并将所述目标运行代码分布式升级至对应的终端设备，其中，所述时序编程模型为基于预先采集的不同终端设备中业务流程对应的各子作业流程的设备运行数据进行构建。
本申请还提供一种基于数字孪生DaaS平台的深度学***台的深度学***台的深度学***台的深度学***台的深度学***台的深度学习编程方法的步骤。
本申请还提供一种存储介质，所述存储介质为计算机可读存储介质，所述计算机可读存储介质上存储基于数字孪生DaaS平台的深度学***台的深度学***台的深度学习编程方法的步骤。
本申请提供了一种基于数字孪生DaaS平台的深度学***台收集业务流程的设备运行数据，且采集的数据为对应终端设备的具体子作业流程，从而使得在后续深度学习自动编程后，可基于工序编号确定自动编程的运行代码所述的终端设备的子作业流程，进一步地，对所述不同终端设备的设备运行数据进行数据建模后对比分析，并基于分析结果判断对应的终端设备是否需要升级优化，进而若是，则基于已训练好的时序编程模型，对各所述子作业流程的设备运行数据进行深度学习编程，自动生成目标运行代码，并将所述目标运行代码分布式升级至对应的终端设备，其中，所述时序编程模型为基于预先采集的不同终端设备中业务流程对应的各子作业流程的设备运行数据进行构建，实现了通过将业务流程拆分为过个子作业流程，从而通过模型学习不同业务流程中的具体作业子流程的运行数据，提高时序编程模型的精准度，进而通过时序编程模型对业务流程对应的各子作业流程的设备运行数据进行深度学习自动编程，无需人工手动编程，大大降低了人力成本以及提高设备优化升级的效率，进一步地，将自动 生成的运行代码分布式升级至对应的终端设备中，代码移植性较高，从而提高对终端设备管理的智能性。
附图说明
此处的附图被并入说明书中并构成本说明书的一部分，示出了符合本申请的实施例，并与说明书一起用于解释本申请的原理。
为了更清楚地说明本申请实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，对于本领域默认技术人员而言，在不付出创造性劳动性的前提下，还可以根据这些附图获得其他的附图。
图1为本申请基于数字孪生DaaS平台的深度学习编程方法第一实施例的流程示意图；
图2为本申请基于数字孪生DaaS平台的深度学习编程方法第二实施例的流程示意图；
图3为本申请实施例方案涉及的硬件运行环境的基于数字孪生DaaS平台的深度学习编程设备结构示意图；
图4为本申请深度学习编程装置的功能模块示意图。
本发明目的的实现、功能特点及优点将结合实施例，参照附图做进一步说明。
具体实施方式
应当理解，此处所描述的具体实施例仅用以解释本申请，并不用于限定本申请。
本申请实施例提供一种基于数字孪生DaaS(数据即服务)平台的深度学***台的深度学***台的深度学习编程方法包括：
步骤S10，获取不同终端设备的业务流程中各子作业流程的设备运行数据，其中，所述设备运行数据包括各所述子作业流程的工序编号；
在实施例中，需要说明的是，所述深度学***台，人工智能物联网平台可应用于智慧交通，智慧安防，智慧医疗，工业物联网等领域，所述人工智能物联网平台包括设备层，所述人工智能物联网平台通过设备层连接至少一个终端设备，所述终端设备包括智能运输设备、工业智能机器人、网络终端设备、人体智能穿戴设备、工程智能机械设备、农业智能机械设备、数控机床、智能传感器、智能采集器、智能摄像头、智能变送器和嵌入式智能***等。
进一步需要说明的是，不同终端设备的工作流程不同，每个终端设备配置有对应的设备标识，所述业务流程包括多个子作业流程，各所述子作业流程配置各自对应的工序编 号，在整一个业务流程中包括多个工序或多个作业环节，从而可通过设备标识以及工序编号，确定对应的终端设备的子作业流程，例如，在工业互联网领域中，终端设备的业务流程可以为开机、预热、加载原材料、抛光和老化等多个子作业流程，在智慧医疗领域中，在用户进行身体检查时，身体多个部位的检查结果可作为所述子作业流程。
获取不同终端设备的业务流程中各子作业流程的设备运行数据，具体地，在终端设备接入所述人工智能物联网平台后，通过所述人工智能物联网平台获取不同终端设备对应的业务流程，当终端设备执行对应的业务流程时，收集业务流程中各个子作业流程的设备运行数据，进一步地，通过终端设备和人工智能物联网平台之间的数据通信协议，将所述各个子作业流程的设备运行数据上报至人工智能物联网平台。
其中，所述获取不同终端设备的业务流程中各子作业流程的设备运行数据的步骤包括：
步骤S11，基于预设通信协议，分别获取通过所述不同终端设备上报的各所述子作业流程的设备运行数据。
在本实施例中，需要说明的是，所述预设通信协议至少包括2G、3G、4G、5G、CAT1、CAT4、NB-IOT、LORA、MQTT、HTTP、TCP、UDP和CoAP协议的一种或多种。
具体地，在通过所述不同终端设备执行各自对应的业务流程中各子作业流程过程中，实时获取通过所述不同终端设备上报的各所述子作业流程的设备运行数据。
步骤S20，对所述不同终端设备的设备运行数据进行数据建模后对比分析，并基于分析结果判断对应的终端设备是否需要升级优化；
在本实施例中，对所述不同终端设备的设备运行数据进行数据建模后对比分析，并基于分析结果判断对应的终端设备是否需要升级优化，具体地，将采集到的设备运行数据进行清洗、过滤等预处理，并将预处理得到的设备运行数据，结合3D数字孪生仿真技术进行融合分析，获得分析结果，进一步地，基于所述分析结果，判断对应的终端设备是否需要升级优化，例如，在工业互联网中，将终端设备接入所述人工智能物联网平台，当接收到终端设备接入到人工智能物联网平台对应的接入指令后，收集不同终端设备对应业务流程中各个子作业流程的设备运行数据，假设终端设备对应的业务流程包括50道工序，收集该业务流程中50道工序分别对应的设备运行数据，结合3D数字孪生仿真模型进行建模分析，从而通过不同工序的设备运行数据，确定终端设备业务流程中不同工序的设备运行状态，另外地，在智慧医疗中的AI看片中，通过将医疗设备接入至人工智能物联网平台，从而可通过收集用户 对应的片子以及检测报告数据库中存储的片子报告，从而当用户进行看片时根据检测报告数据库中不同部位的检测报告结合数据3D数字孪生仿真技术进行建模分析，从而基于多部位的综合分析精准确定用户的检测结果，进一步，将AI摄像头接入人工智能物联网平台，当接收到AI摄像头接入到人工智能物联网平台对应的接入指令后，可收集不同AI摄像头所拍摄到的图像数据，进而在智慧交通的AI摄像头监控中，可将同一区域不同摄像头对应的拍摄图像结合数据3D数字孪生仿真技术进行数据建模后分析，从而精准对车辆进行监控。
其中，所述对各所述子作业流程的设备运行数据进行数据建模后对比分析的步骤：
步骤S21，基于各所述子作业流程的设备运行数据，结合3D数字孪生仿真技术进行融合分析，获得分析结果。
在本实施例中，基于各所述子作业流程的设备运行数据，结合3D数字孪生仿真技术进行融合分析，获得分析结果，具体地，基于各所述子作业流程的设备运行数据，通过所述3D数字孪生仿真技术进行数字化建模分析，获得所述分析结果。
步骤S30，若是，则基于已训练好的时序编程模型，对各所述子作业流程的设备运行数据进行深度学习编程，自动生成目标运行代码，并将所述目标运行代码分布式升级至对应的终端设备，其中，所述时序编程模型为基于预先采集的不同终端设备中业务流程对应的各子作业流程的设备运行数据进行构建。
在本实施例中，若是，则基于已训练好的时序编程模型，对各所述子作业流程的设备运行数据进行深度学习编程，自动生成目标运行代码，并将所述目标运行代码分布式升级至对应的终端设备，其中，所述时序编程模型为基于预先采集的不同终端设备中业务流程对应的各子作业流程的设备运行数据进行构建，具体地，若是，则证明该终端设备需要进行升级优化，进而基于所述时序编程模型，对各所述子作业流程的设备运行数据进行深度学习编程，从而自动生成各所述子作业流程对应的目标运行代码，并将所述目标运行代码通过分布式的形式智能化升级至对应的终端设备，从而使得终端设备可基于所述目标运行代码进行运行调式，提高终端设备的运行效果，例如，沿用上述步骤S20的例子，终端设备对应的业务流程包括50道工序(开机、预热、加载原材料、抛光和老化等)，其中，假设传统工业互联网中开机时间8秒、预热时间1分钟、加载原材料时间5分钟，基于业务流程中各子作业流程的设备运行数据，在本申请中，通过时序编程模型进行深度学习编程，从而自动优化终端设备，可以使得终端设备假设开机时间1秒、预热时间0秒、加载原材料时间1分钟，从而大大提高设备的运行效率，进一步地，在传统的终端设备控制中，需要专业的技术人员进行控制以及进行升级优化终端设备，而通过本申请中，直接根据终端设备中业务流程的各个 工序对应的设备运行数据进行建模分析后，通过时序编程模型生成升级包，从而使得普通人员也可以进行操作优化终端设备，大大降低企业的人工成本，另外地，在智慧医疗中的AI看片中，往往仅基于用户单个部位的检测报告进行检测，具有局限性，而在本申请中，通过时序编程模型，根据检测报告数据库自动优化终端设备的参数，从而基于多部位的综合分析精准确定用户的检测结果，提高AI看片的准确性，此外，根据检测报告数据库以及时序编程模型自动编程，从而使得医疗片子报告时无需一定要医疗专家进行分析，而普通医生也可直接根据时序编程模型自动深度学***台，从而收集得到不同AI摄像头对应的拍摄图像数据，可以理解地，当车辆在行驶至该区域时，收集该区域不同的摄像头对车牌拍摄的图像数据，即可获取360°全方位的图像数据，进一步地，在通过时序编程模型进行深度学***台全方位收集摄像头的图像数据，并结合结合数据3D数字孪生仿真技术进行建模分析后，进而通过时序编程模型进行自动编程优化，无需专业技术人员对摄像头的图像数据进行监控分析，从而大大降低人力成本，因此，在本申请中，通过预先将终端设备接入至人工智能互联网平台中，从而基于预设通信协议来传输终端设备的设备运行数据，并且基于时序编程模型来自动优化终端设备的参数，不仅提高了终端设备的运行效率，还能极大第降低人力资源的成本。
其中，所述将所述目标运行代码分布式升级至对应的终端设备的步骤包括：
步骤S31，通过预先设置的人工智能算法中台将所述不同终端设备的业务流程的子作业流程对应的算法引擎进行归类；
步骤S32，基于引擎归类结果，通过预先设置的设备升级方法将所述目标运行代码分布式升级至对应的终端设备的子作业流程中。
在本实施例中，需要说明的是，所述预先设置的设备升级方法包括OTA升级方法，其中，OTA(Over-the-AirTechnology)升级是指终端设备通过无线网络下载远程服务器上的升级包，对***或应用进行升级的技术。
进一步地，不同的终端设备在所述人工智能物联网平台中的人工智能算法中台配置有对应的算法引擎，而一个终端设备的业务流程包括多个子作业流程，因此，需要将不同字作业流程对应的算法引擎进行归类，从而使得基于算法引擎，通过OTA升级方法将深度学 习编程的运行代码升级至对应的终端设备的子作业流程中。
其中，在所述基于时序编程模型，对各所述子作业流程的设备运行数据进行深度学习编程，自动生成目标运行代码，并将所述目标运行代码分布式升级至对应的终端设备的步骤之后，还包括：
步骤A10，标记所述目标运行代码对应的版本号，并将所述目标运行代码和对应的版本号进行存储；
在本实施例中，需要说明的是，对于每一次深度学***台，在一种可实施方案中，为了缓解平台空间，可存储预设次数进行深度学习编程的运行代码，例如，存储距离当前深度学习编程前3次的运行代码。
步骤A20，获取通过终端设备基于所述目标运行代码进行调试和优化的运行结果；
在本实施例中，获取通过终端设备基于所述目标运行代码进行调试和优化的运行结果，具体地，终端设备在获取到目标运行代码之后，终端设备基于所述目标运行代码进行预运行，得到所述运行结果，以判断终端设备是否优化升级成功。
步骤A30，若所述运行结果为异常结果，则控制所述算法引擎进行回滚处理，以回滚至上一版本号对应的运行代码，并通过开发人员对所述异常结果对应的异常运行代码进行升级，以将所述升级后的运行代码升级至所述终端设备。
在本实施例中，具体地，若所述运行结果为异常结果，则证明深度学习编程得到的目标运行代码存在异常，进而控制存在异常对应的算法引擎进行回滚处理，从而回滚至上一版本号对应的运行代码，从而使得终端设备按照上一版本号对应的运行代码进行运行，避免终端设备运行中断的缺陷，进而通知开发人员对所述异常结果对应的异常运行代码进行手工升级，并将所述升级后的运行代码升级至所述终端设备，从而完成对终端设备的升级优化操作。
本申请实施例提供了基于数字孪生DaaS平台的深度学***台收集业务流程的设备运行数据，且采集的数据详细到对应终端设备的具体子作业流程，从而使得在后续深度学习自动编程后，可基于工序编号确定自动编程的运行代码所述的终端设备的子作业流程，进一步地，对所述不同终 端设备的设备运行数据进行数据建模后对比分析，并基于分析结果判断对应的终端设备是否需要升级优化，进而若是，则基于已训练好的时序编程模型，对各所述子作业流程的设备运行数据进行深度学习编程，自动生成目标运行代码，并将所述目标运行代码分布式升级至对应的终端设备，其中，所述时序编程模型为基于预先采集的不同终端设备中业务流程对应的各子作业流程的设备运行数据进行构建，实现了通过将业务流程拆分为过个子作业流程，从而通过模型学习不同业务流程中的具体作业子流程的运行数据，提高时序编程模型的精准度，进而通过时序编程模型对业务流程对应的各子作业流程的设备运行数据进行深度学习自动编程，无需人工手动编程，大大降低了人力成本以及提高设备优化升级的效率，进一步地，将自动生成的运行代码分布式升级至对应的终端设备中，代码移植性较高，从而提高对终端设备管理的智能性。
进一步地，参照图2，基于本申请中第一实施例，在本申请的另一实施例中，在所述若是，则基于时序编程模型，对各所述子作业流程的设备运行数据进行深度学***台的深度学习编程方法还包括：
步骤B10，获取不同终端设备的业务流程，并采集不同业务流程中各子作业流程的训练设备运行数据；
步骤B20，基于各所述子作业流程的训练设备运行数据，对待训练初始模型进行训练，获得所述时序编程模型。
在本实施例中，需要说明的是，训练时序编程模型的学***台的深度学习编程方法模仿这种针对输入数据的行为过程，最终能够形成正确地输出数据。当训练完毕后，这些转换规则就被编码到神经网络的网络参数中，当模型实际应用时，神经网络对输入数据进行特征提取与表示，以基于特征提取结果进行学习预测，从而输出最终代码结果。
具体地，首先获取采集同终端设备各自业务流程中的各子作业流程的训练设备运行数据，进而提取所述训练设备运行数据的特征信息，以基于所述特征信息迭代优化所述待训练初始模型的网络参数，进而判断优化后的待训练初始模型是否满足训练结束条件，其中，所述训练结束条件包括损失函数达到收敛或迭代次数达到预设迭代次数等条件，若满足，则获得所述时序编程模型，若不满足，则返回执行步骤：获取不同终端设备的业务流程，并采集不同业务流程中各子作业流程的训练设备运行数据。
本申请实施例提供了一种基于数字孪生DaaS平台的深度学习编程方法，也即，获取不同终端设备的业务流程，并采集不同业务流程中各子作业流程的训练设备运行数据，进而基于各所述子作业流程的训练设备运行数据，对待训练初始模型进行训练，获得所述时序编程模型，实现了根据不同业务流程中的各子作业流程的设备运行数据进行模型训练，从而使得模型能够学习具体作业子流程的运行数据，从而提高时序编程模型的精准度，进而即可基于所述时序编程模型对不同终端设备的设备运行数据进行深度学习自动编程，提高终端设备升级优化的效率。
参照图3，图3是本申请实施例方案涉及的硬件运行环境的基于数字孪生DaaS平台的深度学习编程设备结构示意图。
如图3所示，该基于数字孪生DaaS平台的深度学习编程设备可以包括：处理器1001，例如CPU，存储器1005，通信总线1002。其中，通信总线1002用于实现处理器1001和存储器1005之间的连接通信。存储器1005可以是高速RAM存储器，也可以是稳定的存储器(non-volatile memory)，例如磁盘存储器。存储器1005可选的还可以是独立于前述处理器1001的存储设备。
可选地，该基于数字孪生DaaS平台的深度学习编程设备还可以包括矩形用户接口、网络接口、相机、RF(Radio Frequency，射频)电路，传感器、音频电路、WiFi模块等等。矩形用户接口可以包括显示屏(Display)、输入子模块比如键盘(Keyboard)，可选矩形用户接口还可以包括标准的有线接口、无线接口。网络接口可选的可包括标准的有线接口、无线接口(如WIFI接口)。
本领域技术人员可以理解，图3中示出的基于数字孪生DaaS平台的深度学***台的深度学习编程设备的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件布置。
如图3所示，作为一种计算机存储介质的存储器1005中可以包括操作***、网络通信模块以及基于数字孪生DaaS平台的深度学***台的深度学***台的深度学***台的深度学习编程方法中其它硬件和软件之间通信。
在图3所示的基于数字孪生DaaS平台的深度学***台的深度学习编程程序，实现上述任一项所 述的一种基于数字孪生DaaS平台的深度学习编程方法的步骤。
本申请基于数字孪生DaaS平台的深度学***台的深度学习编程方法各实施例基本相同，在此不再赘述。
此外，请参照图4，图4是本申请深度学***台的深度学***台的深度学习编程***包括：
获取模块，用于获取不同终端设备的业务流程中各子作业流程的设备运行数据，其中，所述设备运行数据包括各所述子作业流程的工序编号；
分析模块，用于对所述不同终端设备的设备运行数据进行数据建模后对比分析，并基于分析结果判断对应的终端设备是否需要升级优化；
深度学习编程模块，用于若是，则基于已训练好的时序编程模型，对各所述子作业流程的设备运行数据进行深度学习编程，自动生成目标运行代码，并将所述目标运行代码分布式升级至对应的终端设备，其中，所述时序编程模型为基于预先采集的不同终端设备中业务流程对应的各子作业流程的设备运行数据进行构建。
可选地，所述获取模块还用于：
基于预设通信协议，分别获取通过所述不同终端设备上报的各所述子作业流程的设备运行数据。
可选地，所述分析模块还用于：
基于各所述子作业流程的设备运行数据，结合3D数字孪生仿真技术进行融合分析，获得分析结果。
可选地，所述深度学习编程模块还用于：
通过预先设置的人工智能算法中台将所述不同终端设备的业务流程的子作业流程对应的算法引擎进行归类；
基于引擎归类结果，通过预先设置的设备升级方法将所述目标运行代码分布式升级至对应的终端设备的子作业流程中。
可选地，所述基于数字孪生DaaS平台的深度学习编程***还用于：
获取不同终端设备的业务流程，并采集不同业务流程中各子作业流程的训练设备运行数据；基于各所述子作业流程的训练设备运行数据，对待训练初始模型进行训练，获得所述时序编程模型。
可选地，所述基于数字孪生DaaS平台的深度学习编程***还用于：
标记所述目标运行代码对应的版本号，并将所述目标运行代码和对应的版本号进行存储；
获取通过终端设备基于所述目标运行代码进行调试和优化的运行结果；
若所述运行结果为异常结果，则控制所述算法引擎进行回滚处理，以回滚至上一版本号对应的运行代码，并通过开发人员对所述异常结果对应的异常运行代码进行升级，以将所述升级后的运行代码升级至所述终端设备。
本申请基于数字孪生DaaS平台的深度学***台的深度学习编程***各实施例基本相同，在此不再赘述。
本申请实施例提供了一种存储介质，所述存储介质为计算机可读存储介质，且所述计算机可读存储介质存储有一个或者一个以上程序，所述一个或者一个以上程序还可被一个或者一个以上的处理器执行以用于实现上述任一项所述的基于数字孪生DaaS平台的深度学习编程方法的步骤。
本申请计算机可读存储介质具体实施方式与上述基于数字孪生DaaS平台的深度学习编程方法各实施例基本相同，在此不再赘述。
以上仅为本申请的优选实施例，并非因此限制本申请的专利范围，凡是利用本申请说明书及附图内容所作的等效结构或等效流程变换，或直接或间接运用在其他相关的技术领域，均同理包括在本申请的专利处理范围内。

Claims (10)

  1. 一种基于数字孪生DaaS平台的深度学***台的深度学***台，所述人工智能物联网平台连接至少一个终端设备，所述一种基于数字孪生DaaS平台的深度学习编程方法包括：
    获取不同终端设备的业务流程中各子作业流程的设备运行数据，其中，所述设备运行数据包括各所述子作业流程的工序编号；
    对所述不同终端设备中各子作业流程的设备运行数据进行数据建模后对比分析，并基于分析结果判断对应的终端设备是否需要升级优化；
    若是，则基于已训练好的时序编程模型，对各所述子作业流程的设备运行数据进行深度学习编程，自动生成目标运行代码，并将所述目标运行代码分布式升级至对应的终端设备，其中，所述时序编程模型为基于预先采集的不同终端设备中业务流程对应的各子作业流程的设备运行数据进行构建。
  2. 如权利要求1所述的基于数字孪生DaaS平台的深度学习编程方法，其特征在于，所述获取不同终端设备的业务流程中各子作业流程的设备运行数据的步骤包括：
    基于预设通信协议，分别获取通过所述不同终端设备上报的各所述子作业流程的设备运行数据。
  3. 如权利要求1所述的基于数字孪生DaaS平台的深度学习编程方法，其特征在于，所述对各所述子作业流程的设备运行数据进行数据建模后对比分析的步骤：
    基于各所述子作业流程的设备运行数据，结合3D数字孪生仿真技术进行融合分析，获得分析结果。
  4. 如权利要求1所述的基于数字孪生DaaS平台的深度学习编程方法，其特征在于，所述将所述目标运行代码分布式升级至对应的终端设备的步骤包括：
    通过预先设置的人工智能算法中台将所述不同终端设备的业务流程的子作业流程对应的算法引擎进行归类；
    基于引擎归类结果，通过预先设置的设备升级方法将所述目标运行代码分布式升级至对应的终端设备的子作业流程中。
  5. 如权利要求1所述的基于数字孪生DaaS平台的深度学***台的深度学习编程方法还包括：
    获取不同终端设备的业务流程，并采集不同业务流程中各子作业流程的训练设备运行数据；基于各所述子作业流程的训练设备运行数据，对待训练初始模型进行训练，获得所述时序编程模型。
  6. 如权利要求1所述的基于数字孪生DaaS平台的深度学习编程方法，其特征在于，在所述基于时序编程模型，对各所述子作业流程的设备运行数据进行深度学习编程，自动生成目标运行代码，并将所述目标运行代码分布式升级至对应的终端设备的步骤之后，还包括：
    标记所述目标运行代码对应的版本号，并将所述目标运行代码和对应的版本号进行标记和存储；
    获取通过终端设备基于所述目标运行代码进行调试和优化的运行结果；
    若所述运行结果为异常结果，则控制所述算法引擎进行回滚处理，以回滚至上一版本号对应的运行代码，并通过开发人员对所述异常结果对应的异常运行代码进行升级，将所述升级后的运行代码升级至所述终端设备。
  7. 如权利要求2所述的基于数字孪生DaaS平台的深度学习编程方法，其特征在于，所述预设通信协议包括3G、4G、5G、CAT1、CAT4网络传输、NB-IOT窄带物联网、LORA低功耗远程无线通信、MQTT消息队列遥测传输、HTTP、TCP传输层协议、UDP传输层协议中的一种或多种。
  8. 一种基于数字孪生DaaS平台的深度学***台的深度学***台，所述人工智能物联网平台连接至少一个终端设备，所述基于数字孪生DaaS平台的深度学习编程***包括：
    获取模块，用于获取不同终端设备的业务流程中各子作业流程的设备运行数据，其中，所述设备运行数据包括各所述子作业流程的工序编号；
    分析模块，用于对所述不同终端设备的设备运行数据进行数据建模后对比分析，并基于分析结果判断对应的终端设备是否需要升级优化；
    深度学习编程模块，用于若是，则基于已训练好的时序编程模型，对各所述子作业流程的设备运行数据进行深度学习编程，自动生成目标运行代码，并将所述目标运行代码分布式升级至对应的终端设备，其中，所述时序编程模型为基于预先采集的不同终端设备中业务流程对应的各子作业流程的设备运行数据进行构建。
  9. 一种基于数字孪生DaaS平台的深度学***台的深度学***台的深度学习编程程序，
    所述基于数字孪生DaaS平台的深度学***台的深度学习编程方法的步骤。
  10. 一种存储介质，所述存储介质为计算机可读存储介质，其特征在于，所述计算机可读存储介质上存储有基于数字孪生DaaS平台的深度学***台的深度学***台的深度学习编程方法的步骤。
PCT/CN2023/074477 2022-02-09 2023-02-03 基于数字孪生DaaS平台的深度学习编程方法及*** WO2023151523A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210120010.3A CN114153482B (zh) 2022-02-09 2022-02-09 基于数字孪生DaaS平台的深度学习编程方法及***
CN202210120010.3 2022-02-09

Publications (1)

Publication Number Publication Date
WO2023151523A1 true WO2023151523A1 (zh) 2023-08-17

Family

ID=80450027

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/074477 WO2023151523A1 (zh) 2022-02-09 2023-02-03 基于数字孪生DaaS平台的深度学习编程方法及***

Country Status (2)

Country Link
CN (1) CN114153482B (zh)
WO (1) WO2023151523A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117161892A (zh) * 2023-10-25 2023-12-05 苏州博宏源机械制造有限公司 一种基于抛光材料属性检测的抛光组件智能控制***
CN117255087A (zh) * 2023-11-20 2023-12-19 中国科学技术大学 一种支持海量异构设备接入的网络孪生实现方法与***

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114153482B (zh) * 2022-02-09 2022-05-17 深圳市爱云信息科技有限公司 基于数字孪生DaaS平台的深度学习编程方法及***
CN114693220B (zh) * 2022-05-30 2022-09-20 深圳市爱云信息科技有限公司 基于数字孪生DaaS平台的算法仓库管理方法及***
CN115562191B (zh) * 2022-09-26 2024-02-27 北京能科瑞元数字技术有限公司 基于工业数字孪生的生产力中台智能推测分析方法
CN115878131B (zh) * 2023-02-09 2023-05-09 天津汉云工业互联网有限公司 数字孪生应用的代码生成方法、装置和电子设备
CN116521150B (zh) * 2023-07-04 2023-10-13 北京凯思昊鹏软件工程技术有限公司 一种支持异构融合的分布式***及通信方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103235556A (zh) * 2013-03-27 2013-08-07 南京航空航天大学 基于特征的复杂零件数控加工制造方法
CN109799787A (zh) * 2019-01-10 2019-05-24 湖南科技大学 基于数控***的凸轮轴智能磨削工艺软件数据库***
US20190236489A1 (en) * 2018-01-30 2019-08-01 General Electric Company Method and system for industrial parts search, harmonization, and rationalization through digital twin technology
CN114153482A (zh) * 2022-02-09 2022-03-08 深圳市爱云信息科技有限公司 基于数字孪生DaaS平台的深度学习编程方法及***

Family Cites Families (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109445305A (zh) * 2018-10-26 2019-03-08 中国电子科技集团公司第三十八研究所 一种基于数字孪生的装配精度仿真分析方法与***
CN109934412A (zh) * 2019-03-18 2019-06-25 无锡雪浪数制科技有限公司 基于时序预测模型的实时设备异常检测装置及方法
CN109933035A (zh) * 2019-04-24 2019-06-25 中国科学院重庆绿色智能技术研究院 一种基于数字孪生的生产线控制***、方法及生产***
CN111027209B (zh) * 2019-12-10 2020-11-10 北京航空航天大学 一种数字孪生行为模型快速构建方法
CN111539583B (zh) * 2020-05-13 2023-10-24 中国电子科技集团公司第十四研究所 一种基于数字孪生的生产过程仿真优化方法
CN112069247B (zh) * 2020-09-09 2023-01-24 广东电网有限责任公司中山供电局 基于数字孪生技术的电力***运行数据可视化***及方法
CN112767058A (zh) * 2021-02-05 2021-05-07 深圳市爱云信息科技有限公司 AIOT DaaS数字孪生云平台
CN113110313A (zh) * 2021-03-26 2021-07-13 广东建设职业技术学院 一种基于数字孪生的施工过程管控方法
CN113157421B (zh) * 2021-04-29 2024-04-09 浙江工业大学 一种基于用户作业流程的分布式集群资源调度方法
CN113515848A (zh) * 2021-05-13 2021-10-19 三一重型装备有限公司 矿山设备的结构设计优化方法、***和可读存储介质
CN113421005B (zh) * 2021-07-01 2022-12-06 润联软件***(深圳)有限公司 工序数据和管理协同方法、装置、计算机设备及存储介质
CN113887016A (zh) * 2021-09-07 2022-01-04 中船智能科技(上海)有限公司 一种基于数字孪生的船舶数字化车间仿真方法与***
CN113836755A (zh) * 2021-11-29 2021-12-24 联想新视界（北京）科技有限公司 基于数字孪生模型的控制方法及装置

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103235556A (zh) * 2013-03-27 2013-08-07 南京航空航天大学 基于特征的复杂零件数控加工制造方法
US20190236489A1 (en) * 2018-01-30 2019-08-01 General Electric Company Method and system for industrial parts search, harmonization, and rationalization through digital twin technology
CN109799787A (zh) * 2019-01-10 2019-05-24 湖南科技大学 基于数控***的凸轮轴智能磨削工艺软件数据库***
CN114153482A (zh) * 2022-02-09 2022-03-08 深圳市爱云信息科技有限公司 基于数字孪生DaaS平台的深度学习编程方法及***

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
CHEN, CONG: "Research on Quality Control Method of Marine Diesel Engine Piston Processing Based on Digital Twin", MASTER'S THESIS, 12 June 2019 (2019-06-12), CN, pages 1 - 79, XP009548259, DOI: 10.27171/d.cnki.ghdcc.2019.000365 *
ZHAO, PENG: "Research on Evaluation Method of Machining Process Based on Digital Twin", DOCTORAL DISSERTATION, 5 June 2021 (2021-06-05), CN, pages 1 - 86, XP009548275, DOI: 10.27171/d.cnki.ghdcc.2021.000637 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117161892A (zh) * 2023-10-25 2023-12-05 苏州博宏源机械制造有限公司 一种基于抛光材料属性检测的抛光组件智能控制***
CN117161892B (zh) * 2023-10-25 2024-01-26 苏州博宏源机械制造有限公司 一种基于抛光材料属性检测的抛光组件智能控制***
CN117255087A (zh) * 2023-11-20 2023-12-19 中国科学技术大学 一种支持海量异构设备接入的网络孪生实现方法与***
CN117255087B (zh) * 2023-11-20 2024-02-23 中国科学技术大学 一种支持海量异构设备接入的网络孪生实现方法与***

Also Published As

Publication number Publication date
CN114153482A (zh) 2022-03-08
CN114153482B (zh) 2022-05-17

Similar Documents

Publication Publication Date Title
WO2023151523A1 (zh) 基于数字孪生DaaS平台的深度学习编程方法及***
Dang et al. Aiops: real-world challenges and research innovations
CN111835582B (zh) 物联网巡检设备的配置方法、装置、计算机设备
CN111797969A (zh) 神经网络模型的转换方法及相关装置
CN111835584A (zh) 物联网产品的巡检方法、装置、计算机设备和存储介质
CN111782635B (zh) 数据处理方法和装置、存储介质和电子装置
US20220366244A1 (en) Modeling Human Behavior in Work Environment Using Neural Networks
CN113282000A (zh) 数据中心的故障诊断方法、装置及动环监控***
US20220269595A1 (en) Testing iot-applications based on a fleet of iot-devices
CN113887749A (zh) 基于云边协同的电力物联网多维度监控处置方法、设备及平台
CN114138501B (zh) 用于现场安全监控的边缘智能服务的处理方法和装置
CN109299789A (zh) 设备点检方法及装置
CN110489093A (zh) 一种面向ai应用的持续构造方法、装置、终端及存储介质
CN115438768A (zh) 模型推理方法、装置、计算机设备及存储介质
KR102401137B1 (ko) 원격지원 장치 및 그 방법
CN111832758A (zh) 基于数字孪生的地铁站台门维保作业方法和装置以及设备
CN111859370B (zh) 识别服务的方法、装置、电子设备和计算机可读存储介质
CN113626022A (zh) 物模型创建方法、装置、存储介质及电子设备
US9128640B2 (en) Software product consistency assessment
Tan et al. Research on Deep Learning Programming Methods and Systems Based on Digital Mirror Platforms
CN103617044A (zh) 一种二次开发程序执行的控制方法
CN112448835B (zh) 一种物联网***
EP4270121A1 (en) Method and system for seamless transition of runtime system from controller device to digitalization platform
CN110874312B (zh) 一种适用于异质多智能主体的众智机及其实现方法
CN111782190A (zh) 异常定位方法和装置、存储介质和电子装置

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23752303

Country of ref document: EP

Kind code of ref document: A1