WO2024011601A1 - 用于设备功能下降型故障预警的工业物联网、方法及介质 - Google Patents

用于设备功能下降型故障预警的工业物联网、方法及介质 Download PDF

Info

Publication number
WO2024011601A1
WO2024011601A1 PCT/CN2022/106031 CN2022106031W WO2024011601A1 WO 2024011601 A1 WO2024011601 A1 WO 2024011601A1 CN 2022106031 W CN2022106031 W CN 2022106031W WO 2024011601 A1 WO2024011601 A1 WO 2024011601A1
Authority
WO
WIPO (PCT)
Prior art keywords
platform
data
time
sub
equipment
Prior art date
Application number
PCT/CN2022/106031
Other languages
English (en)
French (fr)
Inventor
邵泽华
权亚强
吴岳飞
陈于浩
周莙焱
Original Assignee
成都秦川物联网科技股份有限公司
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 成都秦川物联网科技股份有限公司 filed Critical 成都秦川物联网科技股份有限公司
Priority to PCT/CN2022/106031 priority Critical patent/WO2024011601A1/zh
Priority to CN202280007664.3A priority patent/CN116940956A/zh
Priority to US18/329,679 priority patent/US20230315062A1/en
Publication of WO2024011601A1 publication Critical patent/WO2024011601A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B19/00—Programme-control systems
    • G05B19/02—Programme-control systems electric
    • G05B19/418—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM]
    • G05B19/4184—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM] characterised by fault tolerance, reliability of production system

Definitions

  • This specification relates to the technical field of the Internet of Things, and in particular to an industrial Internet of Things, method and medium for early warning of equipment function degradation faults.
  • production efficiency directly determines production costs
  • production efficiency has always been the focus of enterprises.
  • intelligent manufacturing a production line is set up and run by multiple intelligent manufacturing equipment according to the process sequence or manufacturing execution time sequence. If each equipment can ensure and improve its own production efficiency, then the overall production efficiency of the production line will be improved, thereby reducing Cost of production.
  • obstacles may occur that reduce the function of the equipment, leading to a decrease in production efficiency.
  • the Industrial Internet of Things includes a management platform, and the management platform is configured to perform the following operations: obtain the performance of the device in a preset time period The execution status of at least one production task; based on the execution status of the at least one production task, determine whether the equipment has a functional degradation fault; in response to the existence of the functional degradation fault in the equipment, issue an early warning and The equipment is repaired accordingly.
  • One embodiment of this specification provides a control method for the industrial Internet of Things for early warning of equipment function degradation faults, which is implemented based on the management platform of the Industrial Internet of Things for early warning of equipment function degradation faults; the control method includes: obtaining The execution status of at least one production task executed by the equipment within a preset time period; based on the execution status of the at least one production task, determining whether the equipment has a functional degradation fault; responding to the existence of the functional degradation fault in the equipment , issue an early warning and repair the equipment accordingly.
  • One embodiment of this specification provides a computer-readable storage medium that stores computer instructions. When the computer instructions are executed by a processor, the aforementioned control method is implemented.
  • Figure 1 is a schematic diagram of an application scenario of the Industrial Internet of Things for early warning of device function degradation faults according to some embodiments of this specification;
  • Figure 2 is a schematic diagram of the structure of the Industrial Internet of Things for early warning of device function degradation faults according to some embodiments of this specification;
  • Figure 3 is an exemplary flow chart of a control method of the Industrial Internet of Things for equipment function degradation type fault early warning according to some embodiments of this specification;
  • Figure 4 is an exemplary flow chart of a control method of the Industrial Internet of Things for equipment function degradation type fault warning according to some embodiments of this specification;
  • Figure 5 is an exemplary flowchart of a process for determining equipment function degradation failure based on qualifications according to some embodiments of this specification
  • Figure 6 is an exemplary schematic diagram of the application of a failure probability prediction model according to some embodiments of this specification.
  • system means of distinguishing between different components, elements, parts, portions or assemblies at different levels.
  • said words may be replaced by other expressions if they serve the same purpose.
  • Figure 1 is a schematic diagram of an application scenario of the Industrial Internet of Things for early warning of device function degradation faults according to some embodiments of this specification.
  • the application scenario 100 of the Industrial Internet of Things for equipment function degradation fault warning may include a server 110 , a network 120 , a terminal device 130 , a production line 140 and a storage device 150 .
  • the application scenario 100 can realize early warning and repair of equipment by implementing the industrial Internet of Things and control methods for early warning of equipment function degradation faults disclosed in this specification.
  • the execution status of at least one production task executed in a preset time period can be obtained and sent to the server 110; Based on the execution of the at least one production task, the server 110 determines whether the equipment in the production line 140 has a functional degradation fault; in response to the equipment in the production line 140 having the functional degradation fault, the server 110 issues an early warning and The equipment in the production line 140 is repaired accordingly to ensure that the production equipment can operate normally and ensure production efficiency.
  • server 110 may be used to process information and/or data related to application scenario 100. For example, the server 110 can determine whether the equipment in the production line 140 has a functional degradation obstacle or the like based on at least one production task situation.
  • server 110 may be a single server or a group of servers. The server group can be centralized or distributed, dedicated or simultaneously served by other devices or systems.
  • server 110 may be local or remote. In some embodiments, server 110 may be implemented on a cloud platform or provided in a virtual manner.
  • server 110 may include processing equipment.
  • Processing equipment may process data and/or information obtained from other equipment or system components.
  • the processing device may execute program instructions based on these data, information, and/or processing results to perform one or more functions described in this application.
  • network 120 may facilitate the exchange of information and/or data.
  • one or more components of the application scenario 100 eg, server 110, network 120, terminal device 130, production line 140, and storage device 150
  • server 110 may obtain information and/or data over network 120.
  • the terminal device 130 may be used as an electronic device for implementing data processing and data communication.
  • mobile phone 130-1, tablet computer 130-2, laptop computer 130-3 or other electronic devices are not too limited here.
  • the data and/or information obtained from the server 110 can be analyzed by the laptop 130-3 and displayed on the screen of the laptop 130-3.
  • the production line 140 can implement assembly line production of products.
  • the production line 140 may include various smart manufacturing equipment for product manufacturing. For example, cylinder block processing equipment, cylinder block positioning and turning equipment, cam assembly installation equipment, bolt assembly installation equipment, etc. in the automobile engine assembly production line.
  • storage device 150 is used to store data and/or instructions.
  • the storage device 150 may store relevant information of the production line 140 . For example, data such as the time when the equipment in the production line 140 performs at least one production task.
  • storage device 150 may store data and/or instructions used by server 110 to perform the exemplary methods described herein.
  • storage device 150 may be implemented on a cloud platform.
  • the storage device 150 may be connected to the network 120 to communicate with one or more components of the application scenario 100 (eg, the server 110, the network 120, the terminal device 130, and the production line 140). One or more components of the application scenario 100 may access data or instructions stored in the storage device 150 via the network 120 . In some embodiments, storage device 150 may directly connect or communicate with one or more components of application scenario 100. In some embodiments, storage device 150 may be part of server 110 .
  • application scenarios are provided for illustrative purposes only and are not intended to limit the scope of this specification.
  • application scenarios can also include databases.
  • application scenarios can be implemented on other devices to achieve similar or different functions.
  • changes and modifications may not depart from the scope of this specification.
  • the Internet of Things system is an information processing system that includes some or all of the user platform, service platform, management platform, sensor network platform, and object platform.
  • the user platform is the leader of the entire Internet of Things operating system and can be used for Obtain user needs.
  • User needs are the basis and premise for the formation of the Internet of Things operating system.
  • the connections between the various platforms of the Internet of Things system are to meet the needs of users.
  • FIG. 2 is a schematic structural diagram of an industrial Internet of Things system for equipment function degradation type fault warning according to some embodiments of this specification.
  • the industrial Internet of Things system 200 for early warning of equipment function degradation faults can be implemented based on the Internet of Things system.
  • the system 200 includes a user platform 210, a service platform 220, a management platform 230, a sensor network platform 240 and an object platform 250 that interact sequentially from top to bottom.
  • User platform 210 may be used for interactive users.
  • the user platform 210 may be configured as a terminal device, such as a mobile phone, a computer and other smart devices.
  • the service platform 220 can be used to receive instructions from the user platform 210 and send them to the management platform 230, and extract information needed to process the user platform 210 from the management platform 230.
  • the service platform 220 may be configured as the first server.
  • the service platform 220 adopts a front split platform arrangement.
  • the front-sub-platform arrangement means that the corresponding platform is equipped with a main platform and multiple sub-platforms.
  • the multiple sub-platforms store and process data of different types or different receiving objects sent by the lower platform.
  • a main platform handles multiple sub-platforms.
  • the platform data is aggregated, stored and processed, and the data is transmitted to the upper platform.
  • the service platform 220 includes a general platform of a service platform and sub-platforms of multiple service platforms.
  • the sub-platforms of multiple service platforms respectively store and process data of different types or different receiving objects sent by the management platform 230.
  • the main platform of the service platform summarizes, stores and processes the data of the sub-platforms of multiple service platforms, and transmits it. Data to user platform 210.
  • the management platform 230 can be used to control the operation of the object platform 250 and receive feedback data from the object platform 250 .
  • management platform 230 may be configured as a second server.
  • the management platform 230 may be arranged in a front-part platform arrangement.
  • the management platform 230 may include a general management platform and sub-platforms of multiple management platforms.
  • the sub-platforms of multiple management platforms respectively store and process data of different types or different receiving objects sent by the sensor network platform 240.
  • the main platform of the management platform summarizes and stores and processes the data of the sub-platforms of multiple management platforms. and transmit the data to the service platform 220.
  • the first server and the second server may be a single server or a server cluster, without too many limitations here.
  • the sensor network platform 240 may be used for interaction between the object platform 250 and the management platform 230.
  • sensor network platform 240 may be configured as a communications network and gateway.
  • the sensor network platform 240 can use multiple groups of gateway servers or multiple groups of smart routers, without too many limitations here.
  • the sensor network platform 240 adopts a stand-alone arrangement.
  • the independent arrangement means that the sensor network platform 240 uses different sub-platforms for data storage, data processing and/or data transmission of data from different object platforms 250 .
  • the sensor network platform 240 includes multiple sensor network sub-platforms.
  • Object platform 250 may be a monitoring object.
  • object platform 250 may be configured as a smart manufacturing device in a production line that performs manufacturing.
  • object platform 250 may include multiple different object platforms, such as object platform 1, object platform 2, object platform 3, and so on.
  • the industrial Internet of Things for early warning of equipment function degradation faults is built based on a five-platform structure.
  • the service platform and the management platform are arranged in front sub-platforms.
  • the main platform of the service platform or management platform is based on the data of the upper platform. Unified reception, analysis and processing facilitates the processing and classification of data from the upper platform and each sub-platform.
  • Each sub-platform corresponding to the main platform operates independently of each other and can be divided into several independent data processing channels based on needs, thereby processing different data.
  • Data uses different channels for data processing and transmission, which can share the data processing pressure of the corresponding main platform, reduce the data processing capacity requirements of each sub-platform, and also ensure the independence of each data, ensuring data classification transmission, traceability and instruction Classification and processing make the structure and data processing of the Internet of Things clear and controllable, which facilitates the management, control and data processing of the Internet of Things.
  • the independent arrangement of the sensor network platform can independently transmit the data of different object platforms, ensuring that different The uplink or downlink data of the object platform are independent and do not interfere with each other during data transmission, which also facilitates effective data classification and data source identification.
  • management platform and service platform can be integrated into one component.
  • each component may share a storage device, or each component may have its own storage device. Such deformations are within the scope of this manual.
  • FIG. 3 is an exemplary flow chart of a control method of the Industrial Internet of Things for early warning of device function degradation faults according to some embodiments of this specification.
  • the process 300 may include the following steps: In some embodiments, the process 300 may be executed by the management platform 230 .
  • Step 310 Obtain the execution status of at least one production task executed by the device within a preset time period.
  • the preset time period may be a certain period of time set in advance and stored in the server. For example, any time period within a day such as 00:00-02:00, 02:00-04:00, etc. There can be only one preset time period, or there can be multiple; multiple time periods can overlap or not; they can be set at a fixed time interval or at any interval, without too many limitations here.
  • Production tasks can be tasks that equipment in a production line needs to perform.
  • the cylinder positioning and flipping equipment needs to perform the task of flipping the cylinder
  • the cam installation assembly needs to perform the task of installing the cam assembly, etc.
  • the execution situation can be the situation that occurs when the equipment performs production tasks. For example, due to lubrication failure and increased friction of the cylinder flipping arm in the cylinder positioning and flipping equipment, the flipping positioning accuracy is reduced and the positioning correction time is increased.
  • the execution status can be obtained through a variety of functional sensors, such as light sensors, speed sensors, displacement sensors, etc.
  • the execution status can also be obtained by analyzing the difference in single-piece manufacturing time. In some embodiments, when the equipment is being manufactured, if the difference between adjacent single-piece manufacturing times continuously appears negative, it means that the single-piece manufacturing time of the equipment continues to increase, which means that there is a problem with the equipment.
  • Step 320 Based on the execution of at least one production task, determine whether the equipment has a functional degradation fault.
  • Functional degradation failure refers to the fact that during the manufacturing process of the equipment, due to mutual friction, external force, stress and other physical or chemical reactions of the internal structure of the equipment, some parts will suffer from wear, corrosion, fracture, vibration, disengagement, etc., resulting in Equipment production accuracy, efficiency, stability, safety, etc. are reduced.
  • the wear and tear of the component grabbing mechanism of the cam component installation equipment the component cannot be grasped well or the component is repeatedly grasped, thereby reducing the grabbing efficiency.
  • Functional degradation faults can be determined by manual inspection, sensor monitoring and other methods. For example, through regular manual inspection of equipment, if it is found that the equipment has problems such as wear and lubrication, resulting in low production accuracy and production efficiency, it can be said that the equipment has a functional degradation failure.
  • the management platform can obtain the actual time taken by the equipment to execute each production task, and determine multiple time-consuming differences based on the actual time-consuming performed by each production task and the corresponding standard time-consuming. Based on the multiple time-consuming differences, Determine whether the equipment has a functional degradation fault.
  • Actual time consumption refers to the time it actually takes to complete the production task. For example, during the manufacturing process, it takes 20 seconds for the grabbing mechanism in the cam component installation equipment to grab a component, and these 20 seconds are the actual time taken.
  • the actual time consumption can be recorded manually or through other methods such as obtaining it from the server.
  • Standard time consumption refers to the time it takes to complete production tasks under normal operation of the equipment.
  • the grabbing mechanism in the cam component installation equipment normally only takes 10 seconds to grab a component, and these 10 seconds are the standard time consumption.
  • Standard elapsed time can be determined in several ways. For example, the average time to complete a production task within a period of time can be used as the standard time consumption, or the time specified by the user to complete a production task can be used as the standard time consumption, etc.
  • the time-consuming difference can be the absolute value of the actual time-consuming and the standard time-consuming. For example, if the actual time taken by the grabbing mechanism to grab a component is 20 seconds, and the standard time taken is 10 seconds, then the time difference between the two is 10 seconds.
  • the time difference can be determined in a number of ways. For example, using processing equipment calculations in the server, manual statistics, etc.
  • a degradation fault may be determined based on multiple elapsed time differences. For example, when the time difference exceeds the preset difference threshold more than the preset times threshold, the device is considered to have a functional degradation fault.
  • Step 330 In response to the presence of a functional degradation fault in the equipment, an early warning is issued and the equipment is repaired accordingly.
  • Early warning can be warning information when an abnormal situation occurs.
  • warning There are many ways of warning, such as light warning, sound warning, etc.
  • Restoration can be the repair of a device to restore it to normal functionality. There are many ways to repair it. For example, faulty equipment can be repaired manually. Another example is that equipment can self-check and repair itself. For example, the self-healing instructions in the maintenance instructions in the device server are used to perform self-healing operations on the equipment.
  • Some embodiments of this specification determine whether the equipment has a functional degradation fault based on the execution of at least one production task, and provide early warning and timely processing for equipment with functional degradation faults, which can ensure that the equipment is discovered and processed at the best time or in advance. Solve the problem of equipment function degradation, thereby ensuring the safety and stability of the equipment, and ensuring the service life and production efficiency of the equipment.
  • FIG. 4 is a flowchart of an exemplary process 400 of an industrial Internet of Things control method for device function degradation type fault early warning according to some embodiments of this specification.
  • an industrial Internet of Things control method for early warning of equipment function degradation faults includes the following steps:
  • Step 410 When the manufacturing equipment performs manufacturing, upload the single-piece manufacturing parameter data to the corresponding sub-platform of the sensor network platform; the single-piece manufacturing parameter data at least includes the total time-consuming data of the manufacturing equipment during single-piece manufacturing.
  • Sub-platforms of the sensor network platform can use gateway servers, etc. For more information on the sub-platforms of the sensor network platform, see Figure 2 and its description.
  • Step 420 The sub-platform of the sensor network platform converts the single-piece manufacturing parameter data into a data file identifiable by the management platform and sends it to the corresponding sub-platform of the management platform.
  • the sub-platform of the management platform can be set as a sub-server of the second server, etc.
  • the sub-platforms of the management platform see Figure 2 and its description.
  • Step 430 The sub-platform of the management platform receives the data file and extracts the total time-consuming data. Based on the total time-consuming data, the time difference between two adjacent total time-consuming data is calculated in sequence according to the single-piece manufacturing time, and all time differences are calculated according to the single-piece manufacturing time.
  • the time difference data set is formed by sorting in chronological order, and the data file and time difference data set are stored and sent to the main platform of the management platform.
  • Step 440 After the main platform of the management platform receives the time difference data set, it performs equipment function degradation fault analysis based on the time difference data set, and executes step 450 or step 460 based on the analysis results.
  • the main platform of the management platform after the main platform of the management platform receives the time difference data set, it performs equipment function degradation fault analysis based on the time difference data set, including: selecting to calculate the absolute values of two adjacent time differences in sequence according to the manufacturing time of a single piece. Difference value; when the difference value appears negative, and the number of consecutive negative values in the difference value is greater than the threshold set by the total platform of the management platform, it is determined that the equipment function degradation fault exists, and the analysis result is determined to be abnormal; otherwise, the equipment function is determined to be degraded Type fault does not exist, and the analysis result is confirmed to be normal.
  • the main platform of the management platform after receiving the time difference data set, performs equipment function degradation fault analysis based on the time difference data set, which further includes: the main platform of the management platform stores a threshold table, and each sub-platform of the management platform stores a threshold table.
  • the main platform of the management platform analyzes the time difference data set of the sub-platforms of each management platform, and retrieves the corresponding thresholds and differences in the threshold table that appear continuously The difference is calculated based on the number of negative values; based on the difference calculation results: if the number of consecutive negative values in the difference is less than the threshold, the device is judged to be normal; if the number of consecutive negative values in the difference is greater than the threshold, the device is judged to be faulty. Failure analysis can then be performed.
  • Step 450 In response to the analysis result being normal, the main platform of the management platform deletes the time difference data set and waits for analysis of the re-uploaded time difference data set.
  • Step 460 In response to the analysis result being abnormal, combine the data file, abnormal result data and time difference data set into analysis data and upload it to the corresponding sub-platform of the service platform.
  • Step 470 The sub-platform of the service platform receives the analysis data and obtains the abnormal time node based on the data file, and sends the abnormal time node, abnormal result data and time difference data set as packaged data to the main platform of the service platform.
  • the sub-platform of the service platform receives the analysis data and obtains the abnormal time nodes based on the data file, and sends the abnormal time nodes, abnormal result data and time difference data set as packaged data to the main platform of the service platform, specifically as follows:
  • the single-piece manufacturing parameter data also includes the manufacturing start time of the manufacturing equipment during single-piece manufacturing; when the manufacturing equipment uploads the single-piece manufacturing parameter data, the manufacturing start time of the single-piece manufacturing is associated with the total time-consuming data and uploaded;
  • the sub-platform of the service platform extracts the data files after receiving the analysis data, extracts the time differences corresponding to the negative values in the time difference data set, and obtains multiple corresponding total time-consuming data based on the time differences; based on multiple total time-consuming data , obtain multiple manufacturing start times corresponding to multiple total time-consuming data; sequence the multiple manufacturing start times according to the single-piece manufacturing time sequence to form the abnormal time node.
  • Abnormal time nodes can help determine the moment and time period when manufacturing equipment is abnormal, and can then assist in determining the cause of the abnormality based on the abnormal time nodes.
  • the abnormal time nodes can be extracted to determine the time when the abnormality occurred. and reasons so that the exception can be handled accordingly.
  • Step 480 The main platform of the service platform receives and stores the packaged data, performs fault classification based on the abnormal result data in the packaged data, and sends the corresponding level information after classification to the user platform.
  • Fault classification refers to classifying the severity of equipment failures. For example, general equipment failures in the mechanical field can be classified from low to high into minor level E, ordinary level D, general level C, severe level B and major level A.
  • a fault classification method is determined for the use of smart manufacturing equipment, specifically: among the corresponding differences with consecutive negative values, the difference with the largest absolute value is set to T1, and the difference with the smallest absolute value is set to T1. The difference is set to T2, the specific number of consecutive negative values is set to N, and the grading benchmark is F, then the grading benchmark F satisfies:
  • Each smart manufacturing equipment will have different faults based on its operating time, service life, and corresponding processing parameters and conditions. When faults occur, they are usually reflected in product manufacturing.
  • the manufacturing equipment can assist in troubleshooting based on the product manufacturing situation. Faults are classified.
  • this embodiment calculates a grading benchmark in the actual manufacturing process by obtaining the corresponding differences in consecutive negative values, and then obtains the corresponding differences in different consecutive negative values allowed by different manufacturing equipment.
  • the allowed grading benchmark is considered to be the maximum grading benchmark of the manufacturing equipment, that is, the maximum allowable fault level.
  • the actual standard and the allowable standard are then determined by a numerical value between the actual standard and the allowable standard. Class fault levels can be accurately graded for different manufacturing equipment.
  • the main platform of the service platform will send the corresponding level information after classification to the user platform for execution at the same time: the main platform of the service platform will issue an early warning instruction to the corresponding service based on the fault classification.
  • the sub-platform of the sensor network platform, the early warning instruction data packet is stored in the main platform of the management platform; the sub-platform of the sensor network platform converts the early warning instruction and the early warning instruction data packet into a configuration file identifiable by the object platform and sends it to The corresponding object platform; the object platform performs early warning operations based on the configuration file.
  • the main platform of the service platform can issue early warning instructions of the corresponding level to the user platform in advance and perform corresponding early warning operations, such as alarms, remote warnings, etc. , so that corresponding early warning processing can be carried out in advance, and fault processing can be carried out in advance or the lag of fault processing can be reduced.
  • the user platform issues a maintenance instruction to the main platform of the service platform.
  • the maintenance instruction includes at least one self-repair sub-instruction; the main platform of the service platform receives Maintenance instructions are analyzed based on the maintenance instructions, at least one self-repair sub-instruction is obtained, and at least one self-repair sub-instruction is sent to the sub-platform of the corresponding service platform; the sub-platform of the service platform receives the self-repair sub-instruction and sends the self-repair sub-instruction The command is sent to the main platform of the management platform; the main platform of the management platform receives the self-repair sub-instruction, obtains the corresponding instruction code data packet, associates the instruction code data packet with the self-repair sub-instruction and sends them to the sub-platform of the management platform.
  • the instruction code data packet is pre-stored in the main platform of the management platform; the sub-platform of the management platform receives the instruction code data packet and the self-repair sub-command and sends it to the sub-platform of the corresponding sensor network platform; the sub-platform of the sensor network platform will The instruction code data package and the self-repair sub-instruction are converted into a configuration file that can be recognized by the object platform and then sent to the corresponding object platform.
  • the object platform retrieves the instruction code data in the instruction code data package based on the self-repair sub-instruction to perform self-repair. .
  • the user platform after the user platform obtains the level information, it can issue some instructions in advance according to the fault level to eliminate or assist in eliminating exceptions and faults.
  • the manufacturing equipment may have precision issues in grabbing parts, positioning parts, and assembling them, requiring multiple actions to be completed, resulting in an increase in the manufacturing time of a single piece.
  • Such failures Generally, it is a low-level fault and can be automatically handled through equipment self-checking, self-repair, etc., without the need for further troubleshooting.
  • the main platform of the service platform when the maintenance instructions correspond to different execution times, writes the execution time into the corresponding self-repair sub-instructions after parsing; when the object platform retrieves the instruction code data package based on the self-repair sub-instructions After the instruction code data in the system, self-repair is performed at the corresponding execution time to reduce the impact of manufacturing equipment tasks on product manufacturing.
  • fault handling can be performed in advance to avoid the severity of the fault.
  • self-repair instructions are issued in advance to enable the equipment to proactively self-repair, which can reduce the complexity and time of fault processing and reduce the impact of fault processing on production efficiency.
  • FIG. 5 is an exemplary flowchart of a process 500 for determining a device function degradation type fault based on qualifications according to some embodiments of this specification.
  • process 500 may include the following steps.
  • the process 500 can be applied to various scenarios of equipment function degradation type fault warning.
  • the process 500 can be applied to automobile parts production lines, for example, automobile engine parts production lines, automobile brake system parts production lines, electrical instrument parts production lines, etc.
  • the management platform may determine the actual pass rate of each production task based on the ratio of qualified parts of the auto parts produced in at least one production task of the auto parts production equipment.
  • the management platform further conducts a comparative analysis based on the actual qualification rate of multiple production tasks and the preset standard qualification rate, determines the multiple qualifications of the corresponding auto parts production equipment, and determines whether there is functional degradation of the production equipment based on the multiple qualifications. type failure.
  • the process 500 will be described in detail below using a food production scenario as an example.
  • Step 510 Determine the actual qualification rate of each production task in at least one production task based on the ratio of qualified food in the output food of at least one production task.
  • Qualified food can refer to food produced in production tasks that meets various preset indicators.
  • the food produced meets the requirements of length, shape, volume, color, taste and other indicators.
  • the preset indicators may include parameters such as puffing degree and shape of the product.
  • the qualification rate can refer to the proportion of the quantity of qualified food among the total quantity of food produced by a certain production task of the equipment. For example, the pass rate is 0.95, etc.
  • the management platform can determine the actual pass rate of the corresponding production task based on the pass rate of at least one production task within a certain period of time (such as a preset production cycle). For example, the pass rate of each production task within a production cycle is regarded as the actual pass rate of the production task.
  • the management platform can determine whether the output food is qualified food through the corresponding relationship between various indicators of the output food and standard food. For example, various standard food data tables are preset, and the management platform can determine whether the output food is qualified food by comparing and analyzing the indicators of the actual output food with the indicators in the data table.
  • whether the output food is qualified or not is related to the puffing degree compliance and shape compliance of the output food.
  • the management platform can perform image recognition on the output food, and determine the puffing degree compliance and shape compliance of the output food based on the images of the output food and the standard food. For example, the management platform can compare and analyze the real-time image of the output food obtained by the camera device with the preset standard image of the output food to determine whether the expansion degree and shape of the output food meet the standard requirements to further determine the output. Whether the food is qualified food.
  • the degree of puffing compliance can refer to the degree to which the ratio of the volume of the output food to the volume of the standard food meets the requirements. When the puffing degree compliance degree is closer to 1, it means that the volume of the output food is more consistent with the volume of the standard food.
  • a range of puffing degree compliance can be preset, such as the [0.9, 1.2] interval. When the puffing degree compliance of the output food is within this range, such as 0.95, it means that the puffing degree of the output food conforms to meets the preset requirements.
  • Shape compliance can refer to the degree to which the shape of the output food is consistent with the shape of the standard output food.
  • the standard shape of snow rice cakes should be round.
  • the shape of the actual produced snow rice cakes is rectangular, it can be considered that its shape conformity is low.
  • the management platform can determine the puffing degree compliance and shape compliance of the output food based on the first image recognition model.
  • the first image recognition model may refer to a model used to determine the puffing degree compliance and shape compliance of the output food and thereby determine whether the output food is qualified.
  • the first image recognition model can be a trained machine learning model. For example, any one or combination of convolutional neural networks, deep neural network models or other customized model structures.
  • the first image recognition model may include a feature extraction layer and a judgment layer.
  • the feature extraction layer may include two convolutional neural network models for respectively extracting food features of real-time images of produced food and food features in standard images, where the two convolutional neural network models have the same initial parameters, and Parameter sharing.
  • the judgment layer can be a deep neural network model, which is used to determine the puffing degree compliance and shape compliance of the output food, and then determine whether the output food is qualified.
  • the management platform can obtain real-time images of the output food in the production task and preset standard images of the output food through the camera device, and input them to the first image recognition model, based on the two feature extraction layers.
  • the convolutional neural network model processes real-time images and standard images respectively, and outputs output food feature vectors and standard food feature vectors respectively.
  • Food feature vectors can be used to represent the expansion degree (such as length, width, height features, etc.) and shape (such as outline, area) features of food.
  • the management platform inputs the output food feature vector and the standard food feature vector to the judgment layer, and processes the above two food feature vectors through the judgment layer.
  • the judgment layer can calculate the food feature vectors of the output food and standard food.
  • the vector distance (such as Euclidean distance), when the vector distance is less than the preset threshold, it means that the similarity between the output food and the standard food meets the preset requirements, thereby determining that the expansion degree and shape of the output food meet the requirements of the standard, thereby determining that the product The food produced is qualified food.
  • the judgment layer can also compare the puffing degree and shape characteristics of the output food with the puffing degree and shape characteristics of the standard food respectively to output the puffing degree compliance and shape compliance of the output food, and then based on the output Whether the puffing degree compliance and shape compliance of the food meet the preset conditions are then judged whether the output food is qualified food.
  • the first image recognition model can be obtained through joint training of a feature extraction layer and a judgment layer.
  • the training samples can be multiple sets of images of food produced by historical production tasks, as well as corresponding standard images of food.
  • the label of the training sample can be set based on the output of the first image recognition model. If the output is required to be the puffing degree compliance and shape compliance of the output food, then the label of the training sample can be the output food of the historical production task. Actual puffing degree compliance and shape compliance. For another example, if the model is required to directly output the judgment result of whether the product is qualified food, the label of the training sample can be the result of whether the output food is qualified corresponding to the historical output image. Labels can be based on manual annotation.
  • the label can be 1 or 0, with 1 indicating qualified and 0 indicating unqualified.
  • the management platform can establish a loss function based on the labels of the training samples and the results output by the judgment layer, and update the parameters of the model.
  • the two convolutional neural network models of the feature extraction layer can be updated simultaneously.
  • the management platform iteratively updates the parameters of the first image recognition model based on the loss function until the training is completed when the preset conditions are met, and the trained first image recognition model is obtained.
  • the preset condition can be that the loss function is less than the threshold, converges, or the training cycle reaches the threshold.
  • the first image recognition model is used to process the real-time image of the output food and the image of the standard food, which helps to quickly confirm whether the puffing degree compliance and shape compliance of the output food meet the preset requirements. , thereby improving the efficiency of judging whether the output food is qualified food.
  • whether the output food is qualified is also related to the uniformity of the feeding per unit area of the output food.
  • the management platform can determine the uniformity of the output food based on the conditions of different areas of the output food (such as appearance, color, etc.), and then determine whether the output food is qualified.
  • Feeding can refer to the operation of adding a certain amount of food accessories during the production of a certain type of food. For example, to produce French fries, a preset amount of tomato powder and/or other auxiliary ingredients need to be added to the surface of the French fries.
  • Uniformity can refer to the uniformity of auxiliary materials per unit area of food during the production process. For example, when the degree of uniformity is low, that is, when the amount of ingredients in different areas of the food differs greatly, the color of the area with a large amount of ingredients may be concentrated and close to the color of the auxiliary materials, and the color of the area with a small amount of ingredients may be relatively small. Close to the color of the output product itself.
  • the degree of uniformity can be a value in the range [0-1], such as 0.8. The larger the value, the more uniform it is. You can preset the uniformity threshold, such as 0.7. When the uniformity of the output food is greater than the threshold, it means that the output food meets the uniformity requirements.
  • the management platform can determine the uniformity of the output food based on the second image recognition model, and further determine whether the output food is qualified based on the uniformity of the output food.
  • the second image recognition model may refer to a model used to determine the uniformity of the output food and thereby determine whether the output food is qualified.
  • the second image recognition model can be a trained machine learning model.
  • the second image recognition model may include a convolutional neural network model. It is used to extract the color distribution characteristics of different areas of the produced food in the real-time image to determine the uniformity of the produced food.
  • the management platform can obtain real-time images of the food produced during the production task through the camera device and input it to the second image recognition model. The real-time image is processed through the second image recognition model and the uniformity of the food produced is output.
  • the second image recognition model can be obtained through training.
  • the training samples can be multiple sets of images of food produced in historical production tasks.
  • the label of the training sample can be the uniformity of the output food corresponding to the historical output image, such as 0.5, etc.
  • the training label can be manually labeled.
  • the management platform can establish a loss function based on the labels of the training samples and the output of the second image recognition model, update the parameters of the model, and iteratively update the parameters of the second image recognition model based on the loss function until the preset conditions are satisfied.
  • the preset condition can be that the loss function is less than the threshold, converges, or the training cycle reaches the threshold.
  • the real-time image of the output food is processed through the second image recognition model, which helps to quickly and real-time confirm whether the uniformity of the output food meets the preset requirements, thereby improving the accuracy of the output food. Improve the efficiency of judgment of qualified food and reduce the consumption of time and energy caused by manual analysis.
  • Step 520 Determine multiple qualifications of the equipment based on the actual qualification rate of each production task and the corresponding standard qualification rate.
  • the standard pass rate may refer to a preset pass rate.
  • the standard pass rate can be determined based on production experience such as the type of food and production difficulty. For example, 0.95 can be preset as the standard pass rate.
  • Qualification can refer to the degree to which the qualification rate of production tasks meets the qualification requirements. For example, the standard pass rate is 0.95. If the actual pass rate of a certain production task is 0.95 or above, the pass rate is considered to meet the requirements, and the corresponding pass rate of the production task can be 1; if the actual pass rate of the production task is low If it is less than 0.95, it is considered that the qualification rate does not meet the requirements, and the qualification of the production task will also be reduced accordingly.
  • Compliance can be determined by various calculation methods.
  • Step 530 Based on multiple qualifications, determine whether the device has a functional degradation fault.
  • the management platform can obtain the qualifications of multiple production tasks of the equipment, and then obtain multiple qualifications of the equipment, and determine whether there is a functional degradation fault in the equipment based on the multiple qualifications.
  • the qualification threshold can be preset, and the management platform can calculate the average value of multiple qualifications. When the average value is less than the preset qualification threshold, it is determined that the device has a functional degradation fault.
  • the management platform can determine whether the device has a probability of functional degradation failure through a failure probability prediction model. See Figure 6 and its description for further explanation.
  • Figure 6 is an exemplary schematic diagram of a failure probability prediction model according to some embodiments of this specification.
  • the management platform can predict the failure probability through the failure probability prediction model 620, which is a machine learning model.
  • the failure probability prediction model 620 is a machine learning model.
  • the management platform can predict the failure probability through the failure probability prediction model 620, which is a machine learning model.
  • the failure probability prediction model 620 is a machine learning model.
  • the failure probability prediction model 620 may refer to a model used to predict the probability of a device having a functional degradation failure.
  • failure probability prediction model 620 includes parameter augmentation layer 630 and probability prediction layer 670.
  • Parameter enrichment layer 630 may refer to a processing layer for processing device-related features.
  • the parameter expansion layer 630 can be used to process the qualification characteristics of the food produced by the equipment, the time-consuming characteristics of the equipment performing production tasks, etc.
  • parameter augmentation layer 630 may include first feature layer 640.
  • the first feature layer 640 may include a first timing model 641 and a first embedding layer 643.
  • the first sequential model 641 may be a long short-term memory network model
  • the first embedding layer 643 may be a deep neural network model.
  • the management platform can input the qualification sequence 611 to the first feature layer 640, process the qualification sequence 611 through the first feature layer 640, and output the first feature vector 661.
  • the qualification sequence 611 may refer to a sequence composed of the qualifications of multiple historical production tasks of the equipment as of the current time, arranged in chronological order. For example, (0.9, 0.9, 0.6) represents the sequence composed of the qualifications of three historical moments (such as T1, T2, T3) up to the current time, which can be represented by a vector.
  • the qualification sequence 611 can be obtained based on the management platform through the historical execution of multiple production tasks of the equipment.
  • the first feature vector 661 may refer to a vector representation of the qualification feature.
  • the first feature vector 661 includes the qualifications at multiple times in history and the predicted qualifications at multiple times in the future.
  • the first feature vector 661 composed of the qualifications of three historical moments T1, T2, and T3 and two future moments T4 and T5 is (0.9, 0.9, 0.6, 0.94, 0.92).
  • the first time series model 641 may refer to a model used to predict qualifications at multiple times in the future.
  • the first time series model can process the qualification sequence 611 input by the management platform to the first feature layer 640 and output the qualifications 642 of multiple future moments. For example, based on the qualification sequence of three historical moments T1, T2, and T3, the qualification degrees of two future moments T4 and T5 are output, 0.94 and 0.92. It should be noted that the qualifications 642 of multiple future moments may also be in the form of a sequence in chronological order.
  • the first embedding layer 643 may refer to a processing layer for processing the qualification sequence 611 and the qualifications 642 of multiple future moments. In some embodiments, the first embedding layer 643 inputs the qualification sequence 611 and the qualifications 642 of multiple future moments, and outputs the first feature vector 661.
  • the predicted qualification as the input of the subsequent probabilistic prediction layer to predict the probability of functional degradation failure
  • the number of actual tests of the equipment can be reduced, the test cost can be reduced, and the prediction accuracy of the probabilistic prediction layer can be increased.
  • the probability prediction layer 670 may refer to a processing layer used to predict the probability of a device having a functional degradation type failure.
  • Probabilistic prediction layer 670 may be a deep neural network model.
  • the management platform inputs the first feature vector 661 and the device information 663 of the device to the probability prediction layer 670, processes the first feature vector 661 and the device information 663 through the probability prediction layer 670, and outputs the information that the device is faulty. Probability 680.
  • the management platform can input the qualification sequence 611 to the failure probability prediction model 620, process the qualification sequence 611 through the first feature layer 640 of the parameter expansion layer 630, and output the first feature vector 661.
  • the first feature vector 661 serves as the input of the probability prediction layer 670.
  • the first feature vector 661 and the device information 663 are processed by the probability prediction layer 670, and a probability 680 of a device failure is output.
  • a probability threshold may be preset, such as 0.7. In response to the probability 680 of equipment failure output by the failure probability prediction model 620 being greater than the preset probability threshold, it is determined that the corresponding equipment has a functional degradation fault.
  • the failure probability prediction model 620 can be obtained through training.
  • the training samples include multiple sets of historical qualification sequences. Multiple sets of training samples can be obtained from historical production data.
  • the training sample can be a qualification sequence composed of the qualifications of several consecutive production tasks of the equipment in the past year.
  • the label of the training sample can be the equipment failure situation corresponding to each group of samples. Labels can be manually annotated or annotated in other feasible ways. For example, 0 indicates that the device has failed, and 1 indicates that the device has not failed.
  • the management platform can input multiple combinations of grid degree sequences 611 in the training sample to the fault probability prediction model 620, construct a loss function based on the output and labels of the fault probability prediction model 620, and simultaneously and iteratively update the initial first time series model 641 based on the loss function. , the parameters of the first embedding layer 643 and the probability prediction layer 670 until the preset conditions are met and the training is completed, and the trained fault probability prediction model 620 is obtained.
  • the preset condition can be that the loss function is less than the threshold, converges, or the training cycle reaches the threshold.
  • the failure probability prediction model 620 may also include a second feature layer 650.
  • the second feature layer 620 may include a second timing model 651 and a second embedding layer 654.
  • the second timing model 651 may be a long short-term memory network model
  • the second embedding layer 654 may be a deep neural network model.
  • the management platform can input the time-consuming difference sequence 612 to the second feature layer 650, process the time-consuming difference sequence 612 through the second feature layer 650, and output it as a second feature vector 662.
  • the time-consuming difference sequence 612 may refer to a sequence constructed according to the time sequence of the time-consuming differences of multiple historical production tasks up to the current time. For example, (0s, 0s, 0s, 1s, 0s) represents the time-consuming difference sequence of five historical production tasks up to the current moment, which can be represented by a vector.
  • the time-consuming difference sequence can be obtained based on the management platform through the historical time-consuming conditions of multiple production tasks of the equipment. For an explanation of the time difference, see the relevant content in Figure 3.
  • the second feature vector 662 may refer to a vector representation of the time difference feature.
  • the second feature vector includes the time-consuming difference of multiple historical production tasks and the predicted time-consuming difference of multiple future production tasks.
  • the second feature vector composed of the time difference between five historical production tasks and three future production tasks is (10s, 1s, 0s, 0s, 0s, 2s, 3s, 0s).
  • the first five time-consuming differences are the time-consuming differences of five historical production tasks.
  • the second timing model 651 may refer to a model used to predict the time-consuming difference of production tasks at multiple times in the future.
  • the second timing model 651 can process the time-consuming difference sequence 612 input by the management platform to the second feature layer 650 and output the time-consuming differences 652 of multiple future moments. For example, based on the above sequence of time-consuming differences of the five historical moments T1, T2, T3, T4, and T5 (10s, 1s, 0s, 0s, 0s), output the time-consuming differences of the three future moments T6, T7, and T8 (2s, 3s, 0s).
  • the second embedding layer 654 may refer to a processing layer for processing the time-consuming difference sequence 612 and the time-consuming differences 652 of multiple future moments.
  • the management platform can input the time-consuming difference sequence 612 and the time-consuming differences 652 of multiple future moments to the second embedding layer 654, and output the second feature vector 662 through processing by the second embedding layer 654.
  • the input to the probability prediction layer 670 further includes the second feature vector 662 output by the second feature layer 650 .
  • the probability prediction layer 670 can process the first feature vector 661, the second feature vector 662, and the device-related information 663, and output a probability 680 of a device failure.
  • the accuracy of prediction can be improved by introducing the historical time difference of production tasks and the predicted future time difference data to predict the probability of equipment function degradation failure.
  • the output of the first feature layer 640 further includes a first confidence level 664 and the output of the second feature layer 650 further includes a second confidence level 665 .
  • the input of the probability prediction layer 670 also includes a first confidence level 664 and a second confidence level 665.
  • the first confidence level may refer to a degree of confidence in the predicted qualifications 642 of multiple future moments.
  • the first confidence level can be expressed as a value within the interval of (0, 1), such as 0.8. The larger the value, the higher the degree of confidence. It is understandable that when the predicted future time is further away from the current time, the confidence level is lower.
  • the first feature layer 640 may also include a first confidence calculation module 644.
  • the first confidence level 664 may be obtained through calculation by the first confidence level calculation module 644 .
  • the calculation method may be determined based on the relationship between the qualification sequence 611 and the predicted number of qualifications at a future time.
  • each production task has a corresponding production cycle.
  • the number of qualifications predicted at the future moment is greater, it means that the span of the predicted future time is larger, and the corresponding first confidence level will be lower.
  • the number of historical qualifications used is greater, it means that the data is more fully supported, and the corresponding first confidence level will be higher.
  • the second confidence level may refer to the degree of credibility of the predicted time-consuming differences at multiple future moments.
  • the second confidence level can be expressed as a value within the interval of (0, 1), such as 0.8. The larger the value, the higher the degree of confidence. It is understandable that when the predicted future time is further away from the current time, the confidence level is lower.
  • the second feature layer 650 may also include a second confidence calculation module 653.
  • the second confidence level 665 may be obtained based on the second confidence level calculation module 653 .
  • the calculation method of the first confidence is the same and will not be repeated here.
  • the accuracy of the prediction can be further improved by considering the influence of the confidence level of the qualification data and the time-consuming difference data when predicting the probability of equipment function degradation failure.
  • the failure probability prediction model 620 may be obtained through joint training of the parameter expansion layer 630 and the probability prediction layer 670 .
  • the training samples include the qualification sequence 611 of multiple groups of historical production tasks, and the time-consuming difference sequence 612 of the production tasks corresponding to each set of qualification sequences 611.
  • the labels of the training samples can be manually labeled based on the corresponding fault conditions of the equipment.
  • the management platform can input multiple combinations of grid degree sequences 611 and time-consuming difference sequences 612 in the training samples to the parameter expansion layer 630.
  • the qualification sequence 611 is input to the first feature layer 640, and at the same time, the time difference sequence 612 is input to the second feature layer 650.
  • the qualification sequence 611 is input to the first temporal model 641, and the qualifications 642 of multiple future moments are output; the qualifications 642 of multiple future moments and the qualification sequence 611 are used as the first embedding
  • the input of layer 643 is processed by the first embedding layer 643 to obtain the first feature vector 661.
  • the first confidence calculation module 644 obtains the first feature vector 661 based on the qualification 642 and the qualification sequence 611 of multiple future moments through a preset calculation formula. First confidence level 664.
  • the time-consuming difference sequence 612 is input to the second time series model 651, and the time-consuming difference 652 of multiple future moments is output; the time-consuming difference 652 of multiple future moments and the time-consuming difference sequence are 612 is used as the input of the second embedding layer 653, and the second feature vector 662 is obtained through the processing of the first embedding layer 643.
  • the second confidence calculation module 653 is based on the time-consuming difference 652 and the time-consuming difference sequence 612 of multiple future moments through the predetermined Assume the calculation formula to obtain the second confidence level 665.
  • the management platform uses multiple sets of first feature vectors 661 and first confidence levels 664 output by the first feature layer 640 and multiple sets of second feature vectors 662 and second confidence levels 665 output by the second feature layer as probability prediction layers. 670 training sample data. At the same time, the training samples of the probability prediction layer 670 also include device related information 663.
  • the management platform inputs the above-mentioned multiple sets of training samples of the probability prediction layer 670 to the probability prediction layer 670, constructs a loss function based on the output and labels of the probability prediction layer 670, and simultaneously iteratively updates the initial first feature layer 640 and the initial feature layer 640 based on the loss function.
  • Parameters of the second feature layer 650 and the probability prediction layer 670 The parameter update of the first feature layer 640 includes updating the parameters of the first timing model 641 and the first embedding layer 643; the parameter update of the second feature layer 650 includes updating the parameters of the second timing model 651 and the second embedding layer 653.
  • the termination condition of the parameter iterative update may be that the preset conditions are met, then the training is completed, the trained first feature layer 640, the second feature layer 650 and the probability prediction layer 670 are obtained, and finally the failure probability prediction model 620 is obtained.
  • the preset condition can be that the loss function is less than the threshold, converges, or the training cycle reaches the threshold.
  • the failure probability prediction model 620 is obtained through joint training of the parameter expansion layer 630 and the probability prediction layer 670, which helps to reduce the complexity of obtaining training samples and improve the efficiency of training.
  • Some embodiments of this specification use a failure probability prediction model to predict equipment's functional degradation faults, which helps to provide early warning and early prevention for equipment that may have functional degradation faults.
  • This manual provides an industrial Internet of Things and its control method for early warning of equipment function degradation faults. It builds an Industrial Internet of Things based on a five-platform structure and adopts a platform layout method that combines front-platform layout and independent layout. , which can ensure the independence of data transmission and facilitate data classification and processing. By grading faults that degrade equipment function, and based on the fault classification, the main service platform issues fault processing commands on its own, which can provide early warning of faults to the equipment to ensure that the production line can operate normally, thus achieving the purpose of ensuring production efficiency.
  • numbers are used to describe the quantities of components and properties. It should be understood that such numbers used to describe the embodiments are modified by the modifiers "about”, “approximately” or “substantially” in some examples. Grooming. Unless otherwise stated, “about,” “approximately,” or “substantially” means that the stated number is allowed to vary by ⁇ 20%. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximations that may vary depending on the desired features of the individual embodiment. In some embodiments, numerical parameters should account for the specified number of significant digits and use general digit preservation methods. Although the numerical ranges and parameters used to identify the breadth of ranges in some embodiments of this specification are approximations, in specific embodiments, such numerical values are set as accurately as is feasible.

Landscapes

  • Engineering & Computer Science (AREA)
  • General Engineering & Computer Science (AREA)
  • Manufacturing & Machinery (AREA)
  • Quality & Reliability (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

一种用于设备功能下降型故障预警的工业物联网，工业物联网包括管理平台(230)，管理平台(230)被配置为执行以下操作：获取设备在预设时间段执行的至少一个生产任务的执行情况(310)；基于至少一个生产任务的执行情况，判断设备是否存在功能下降型故障(320)；响应于设备存在功能下降型故障，发出预警并对设备进行相应修复(330)。

Description

用于设备功能下降型故障预警的工业物联网、方法及介质 技术领域
本说明书涉及物联网技术领域，特别涉及一种用于设备功能下降型故障预警的工业物联网、方法及介质。
背景技术
由于生产效率直接决定着生产成本，所以生产效率一直是企业关注的重点。在智能制造中，生产线是由多个智能制造设备按照工序顺序或制造执行时间顺序设置并运行，如果每个设备能够保证并提高其自身的生产效率，那么生产线整体生产效率将得到提升，进而降低生产成本。但由于设备内部结构的相互摩擦、应力等原因，可能会出现设备功能下降型障碍，导致生产效率下降。
因此，希望提出一种用于设备功能下降型故障预警的工业物联网及控制方法，以提前获取设备故障情况并及时解决故障，确保生产线设备的正常使用、保障生产效率。
发明内容
本说明书实施例之一提供一种用于设备功能下降型故障预警的工业物联网，所述工业物联网包括管理平台，所述管理平台被配置为执行以下操作：获取设备在预设时间段执行的至少一个生产任务的执行情况；基于所述至少一个生产任务的执行情况，判断所述设备是否存在功能下降型故障；响应于所述设备存在所述功能下降型故障，发出预警并对所述设备进行相应修复。
本说明书实施例之一提供一种用于设备功能下降型故障预警的工业物联网的控制方法，基于用于设备功能下降型故障预警的工业物联网的管理平台实现；所述控制方法包括：获取设备在预设时间段执行的至少一个生产任务的执行情况；基于所述至少一个生产任务的执行情况，判断所述设备是否存在功能下降型故障；响应于所述设备存在所述功能下降型故障，发出预警并对所述设备进行相应修复。
本说明书实施例之一提供一种计算机可读存储介质，所述存储介质存储计算机指令，当所述计算机指令被处理器执行时实现前述的控制方法。
附图说明
本说明书将以示例性实施例的方式进一步说明，这些示例性实施例将通过附图进行详细描述。这些实施例并非限制性的，在这些实施例中，相同的编号表示相同的结构，其中：
图1是根据本说明书一些实施例所示的用于设备功能下降型故障预警的工业物联网的应用场景示意图；
图2是根据本说明书一些实施例所示的用于设备功能下降型故障预警的工业物联网的结构示例性示意图；
图3是根据本说明书一些实施例所示的用于设备功能下降型故障预警的工业物联网的控制方法的示例性流程图；
图4是根据本说明书一些实施例所示的用于设备功能下降型故障预警的工业物联网的控制方法的示例性流程图；
图5是根据本说明书一些实施例所示的基于合格度确定设备功能下降型故障流程的示例性流程图；
图6是根据本说明书一些实施例所示的故障概率预测模型应用的示例性示意图。
具体实施方式
为了更清楚地说明本说明书实施例的技术方案，下面将对实施例描述中所需要使用的附图作简单的介绍。显而易见地，下面描述中的附图仅仅是本说明书的一些示例或实施例，对于本领域的普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图将本说明书应用于其它类似情景。除非从语言环境中显而易见或另做说明，图中相同标号代表相同结构或操作。
应当理解，本文使用的“系统”、“装置”、“单元”和/或“模块”是用于区分不同级别的不同组件、元件、部件、部分或装配的一种方法。然而，如果其他词语可实现相同的目的，则可通过其他表达来替换所述词语。
如本说明书和权利要求书中所示，除非上下文明确提示例外情形，“一”、“一个”、“一种” 和/或“该”等词并非特指单数，也可包括复数。一般说来，术语“包括”与“包含”仅提示包括已明确标识的步骤和元素，而这些步骤和元素不构成一个排它性的罗列，方法或者设备也可能包含其它的步骤或元素。
本说明书中使用了流程图用来说明根据本说明书的实施例的系统所执行的操作。应当理解的是，前面或后面操作不一定按照顺序来精确地执行。相反，可以按照倒序或同时处理各个步骤。同时，也可以将其他操作添加到这些过程中，或从这些过程移除某一步或数步操作。
图1是根据本说明书一些实施例所示的用于设备功能下降型故障预警的工业物联网的应用场景示意图。
如图1所示，用于设备功能下降型故障预警的工业物联网的应用场景100可以包括服务器110、网络120、终端设备130、生产线140和存储设备150。
在一些实施例中，应用场景100可以通过实施本说明书中披露的用于设备功能下降型故障预警的工业物联网及控制方法来实现设备的提前预警与修复。例如，在一个典型的应用场景中，当需要实现生产线140中的设备的提前预警与修复时，可以先通过将获取在预设时间段执行的至少一个生产任务的执行情况发送至服务器110；由服务器110基于所述至少一个生产任务的执行情况，判断所述生产线140中的设备是否存在功能下降型故障；服务器110响应于所述生产线140中的设备存在所述功能下降型故障，发出预警并对所述生产线140中的设备进行相应修复，以确保生产设备可以正常运行，保证生产效率。
在一些实施例中，服务器110可以用于处理与应用场景100有关的信息和/或数据。例如，服务器110可以基于至少一个生产任务情况，判断所述生产线140中的设备是否存在功能下降型障碍等。在一些实施例中，服务器110可以是单一服务器或服务器组。该服务器组可以是集中式或分布式的，可以是专用的也可以由其他设备或系统同时提供服务。
在一些实施例中，服务器110可以是区域的或者远程的。在一些实施例中，服务器110可以在云平台上实施，或者以虚拟方式提供。
在一些实施例中，服务器110可以包括处理设备。处理设备可以处理从其他设备或系统组成部分中获得的数据和/或信息。处理设备可以基于这些数据、信息和/或处理结果执行程序指令，以执行本申请中描述的一个或多个功能。
在一些实施例中，网络120可以促进信息和/或数据的交换。在一些实施例中，应用场景100的一个或以上组件(例如，服务器110、网络120、终端设备130、生产线140和存储设备150)可以经由网络120将信息和/或数据发送至应用场景100的其他组件。例如，服务器110可以通过网络120获取信息和/或数据。
在一些实施例中，终端设备130可以用于实现数据处理及数据通信的电子设备。例如，手机130-1、平板电脑130-2、笔记本电脑130-3或其他电子设备，在此不作过多限定。在一些实施例中，可以通过笔记本电脑130-3将从服务器110中获取的数据和/或信息进行分析，并显示在笔记本电脑130-3屏幕上。
在一些实施例中，生产线140可以实现产品的流水线生产。在一些实施例中，生产线140可以包括各种用于产品制造的智能制造设备。例如，汽车发动机装配生产线中的缸体处理设备、缸体定位翻转设备、凸轮组件安装设备、螺栓组件安装设备等。
在一些实施例中，存储设备150用于存储数据和/或指令。在一些实施例中，存储设备150可以存储生产线140的相关信息。例如，生产线140中的设备执行至少一个生产任务的时间等数据。在一些实施例中，存储设备150可以储存服务器110用来执行完成本申请中描述的示例性方法的数据及/或指令。在一些实施例中，存储设备150可在云平台上实现。
在一些实施例中，存储设备150可连接到网络120，以与应用场景100的一个或以上组件(例如，服务器110、网络120、终端设备130和生产线140)通信。应用场景100的一个或以上组件可以经由网络120访问存储设备150中存储的数据或指令。在一些实施例中，存储设备150可以与应用场景100的一个或以上组件直接连接或者进行通信。在一些实施例中，存储设备150可以是服务器110的一部分。
应当注意应用场景仅仅是为了说明的目的而提供，并不意图限制本说明书的范围。对于本领域的普通技术人员来说，可以根据本说明书的描述，做出多种修改或变化。例如，应用场景还可以包括数据库。又例如，应用场景可以在其他设备上实现以实现类似或不同的功能。然而，变化和修改不会背离本说明书的范围。
物联网系统是一种包括用户平台、服务平台、管理平台、传感网络平台、对象平台中部分或全部平台的信息处理系统，其中，用户平台是整个物联网运行体系的主导者，可以用于获取用户需求。用户需求是物联网运行体系形成的基础和前提，物联网系统的各平台之间的联系均是为了满足用户的需求。
图2是根据本说明书一些实施例所示的用于设备功能下降型故障预警的工业物联网系统的结构示例性示意图。
如图2所示，用于设备功能下降型故障预警的工业物联网系统200可以基于物联网系统实现。系统200包括从上到下依次交互的用户平台210、服务平台220、管理平台230、传感网络平台240以及对象平台250。
用户平台210可以用于交互用户。在一些实施例中，用户平台210可以被配置为终端设备，例如手机、电脑等智能设备。
服务平台220可以用于接收用户平台210的指令并发送至管理平台230，并从管理平台230中提取处理用户平台210需要的信息。在一些实施例中，服务平台220可以被配置为第一服务器。在一些实施例中，服务平台220采用前分平台式布置。其中，前分平台式布置是指对应平台设置有一个总平台和多个分平台，多个分平台分别存储和处理下层平台发送的不同类型或不同接收对象的数据，一个总平台对多个分平台的数据进行汇总后存储和处理，并传输数据至上层平台。
在一些实施例中，服务平台220包括一个服务平台的总平台、多个服务平台的分平台。多个服务平台的分平台分别储存和处理管理平台230发送的不同类型或不同接收对象的数据，服务平台的总平台分别对多个服务平台的分平台的数据进行汇总后储存和处理，并传输数据至用户平台210。
管理平台230可以用于控制对象平台250运行，并接收对象平台250的反馈数据。在一些实施例中，管理平台230可以被配置为第二服务器。在一些实施例中，管理平台230可采用前分平台式布置。在一些实施例中，管理平台230可以包括一个管理平台的总平台、多个管理平台的分平台。多个管理平台的分平台分别储存和处理传感网络平台240发送的不同类型或不同接收对象的数据，管理平台的总平台分别对多个管理平台的分平台的数据进行汇总后储存和处理，并传输数据至服务平台220。
在一些实施例中，第一服务器与第二服务器可以采用单一服务器，也可以采用服务器集群，在此不作过多限定。
传感网络平台240可以用于对象平台250和管理平台230之间的交互。在一些实施例中，传感网络平台240可以被配置为通信网络和网关。传感网络平台240可以采用多组网关服务器，或者多组智能路由器，在此不作过多限定。
在一些实施例中，传感网络平台240采用独立式布置。其中，独立式布置是指传感网络平台240对不同对象平台250的数据采用不同的分平台进行数据存储、数据处理和/或数据传输。在一些实施例中，传感网络平台240包括多个传感网络分平台。
对象平台250可以是监控对象。在一些实施例中，对象平台250可以被配置为执行制造的生产线中的智能制造设备。在一些实施例中，对象平台250可以包括多个不同对象平台，例如对象平台1、对象平台2、对象平台3等。
基于五个平台结构搭建的用于设备功能下降型故障预警的工业物联网，其中，服务平台和管理平台均采用前分平台式布置，服务平台或管理平台的总平台通过对上层平台的数据进行统一接收、分析及处理，便于对上层平台和各个分平台的数据进行处理以及分类，而总平台对应的每个分平台相互独立运行，可基于需求分成若干独立的数据处理通路，进而对不同的数据采用不同通路进行数据处理和传输，从而可以分担对应总平台的数据处理压力，降低各个分平台的数据处理能力需求，并且也能保证各个数据的独立性，确保数据分类传输、溯源以及指令的分类下达和处理，使得物联网结构和数据处理清晰可控，方便了物联网的管控和数据处理，而传感网络平台采用独立式布置则能将不同的对象平台的数据进行独立传输，确保不同对象平台的上行或下行数据进行数据传输时的独立和互不干扰，也便于基于数据进行有效分类和数据来源识别。
需要注意的是，以上对于系统及其组成部分的描述，仅为描述方便，并不能把本说明书限制在所举实施例范围之内。可以理解，对于本领域的技术人员来说，在了解该系统的原理后，可能在不背离这一原理的情况下，对各个组成部分进行任意组合，或者构成子系统与其他组成部分连接。例如，管理平台和服务平台可以整合在一个组成部分中。又例如，各个组成部分可以共用一个存储设备，各个组成部分也可以分别具有各自的存储设备。诸如此类的变形，均在本说明书的保护范围之内。
图3是根据本说明书一些实施例所示的用于设备功能下降型故障预警的工业物联网的控制方法的示例性流程图。如图3所示，流程300可以包括以下步骤：在一些实施例中，流程300可以由管理平台230执行。
步骤310，获取设备在预设时间段执行的至少一个生产任务的执行情况。
预设时间段可以是提前设置并存储在服务器中的某一段时间。例如，00：00-02:00、02:00-04:00等一天内的任意一个时间段。预设时间段可以只有一个，也可以有多个；多个时间段可以重叠，也可以不重叠；可以是固定时间间隔设置，也可以是任意间隔设置，在此不作过多限定。
生产任务可以是生产线中的设备所需执行的任务。例如，缸体定位翻转设备需要执行的翻转缸体的任务、凸轮安装组件需要执行的安装凸轮组件的任务等。
执行情况可以是设备在执行生产任务出现的情况。例如，缸体定位翻转设备中的缸体翻转臂由于润滑失效、摩擦增大，导致翻转定位精度降低、定位修正时间增加等情况。执行情况可以通过多种功能传感器获取，例如光传感器、速度传感器、位移传感器等。执行情况也可以通过分析单件制造时间的差值等方式获取。在一些实施例中，当设备进行生产制造，相邻单件制造时间的差值若连续出现负值时，说明设备的单件制造时间持续增加，即说明设备存在问题。
步骤320，基于至少一个生产任务的执行情况，判断设备是否存在功能下降型故障。
功能下降型故障是指设备在制造过程中，由于设备内结构的相互摩擦、外力、应力及其它物理反应或化学反应，导致一些零部件出现磨损、腐蚀、断裂、振动、脱离配合等，从而导致设备生产精度、效率、稳定性、安全性等降低。例如，凸轮组件安装设备由于组件抓取机构出现磨损，导致无法很好抓取组件或反复抓取组件，而降低抓取的效率等。
功能下降型故障可以利用人工检查、传感器监测等多种方式去判断。例如，通过人工定期检查设备，若发现设备出现磨损、润滑等问题，导致生产精度、生产效率低，就可以说设备出现功能下降型故障。
在一些实施例中，管理平台可以获取设备执行每个生产任务的实际耗时，基于每个生产任务执行的实际耗时与对应的标准耗时，确定多个耗时差，基于多个耗时差，确定设备是否存在功能下降型故障。
实际耗时是指实际完成生产任务花费的时间。例如，在生产制造过程中，凸轮组件安装设备中的抓取机构抓取一个组件花费20s，这20s即为实际耗时。实际耗时可以通过人工记录，也可以通过服务器获取等其他方式记录。
标准耗时是指在设备正常运行下，完成生产任务需花费的时间。例如，凸轮组件安装设备中的抓取机构正常抓取一个组件只需10s，这10s即为标准耗时。标准耗时可以通过多种方式确定。例如，可以将一段时间内完成一个生产任务的平均时间值作为标准耗时，也可以将用户规定的完成一个生产任务的时间作为标准耗时等。
耗时差可以是实际耗时与标准耗时的绝对值。例如，抓取机构抓取一个组件的实际耗时为20s，而标准耗时为10s，那么两者间的耗时差为10s。耗时差可以通过多种方式确定。例如，利用服务器中的处理设备计算、人工统计等。
在一些实施例中，功能下降型故障可以基于多个耗时差来确定。例如，耗时差超过预设的差值阈值的次数超过预设的次数阈值时，即认为设备存在功能下降型故障。
步骤330，响应于设备存在功能下降型故障，发出预警并对设备进行相应修复。
预警可以是出现异常情况时的警示信息。预警的方式可以有多种，例如灯光预警、声音预警等。
修复可以是对设备进行修理，使其恢复正常的功能。修复的方式也可以有多种。例如，可以通过人工对故障设备进行修复。又如，可以通过设备自检、自我修复。例如，利用设备服务器中的检修指令中的自修复指令进行设备的自修复操作等。
本说明书的一些实施例，通过基于至少一个生产任务执行情况，判断设备是否存在功能下降型故障，并针对存在功能下降型故障的设备提前预警、及时处理，可以确保在最佳时刻或者提前发现和解决设备功能下降型故障问题，进而确保设备安全和稳定，保障设备的使用寿命和生产效率。
图4是根据本说明书一些实施例所示的用于设备功能下降型故障预警的工业物联网控制方法的示例性流程400的流程图。如图4所示，一种用于设备功能下降型故障预警的工业物联网的控制方法包括以下步骤：
步骤410，当制造设备执行制造时，上传单件制造参数数据至对应的传感网络平台的分平台；单件制造参数数据至少包括该制造设备在单件制造时的总耗时数据。
传感网络平台的分平台可以采用网关服务器等。关于传感网络平台的分平台的更多内容参见图2及其描述。
步骤420，传感网络平台的分平台将单件制造参数数据转换成管理平台可识别的数据文件并发送至对应的管理平台的分平台。
管理平台的分平台可以设置为第二服务器的子服务器等。关于管理平台的分平台的更多内容参见图2及其描述。
步骤430,管理平台的分平台接收数据文件并提取总耗时数据，基于总耗时数据，按照单件制造时间顺序依次计算相邻两个总耗时数据的时差，将所有时差按照单件制造时间顺序顺次排序形成时差数据集，并将数据文件、时差数据集存储后发送至管理平台的总平台。
步骤440，管理平台的总平台接收时差数据集后，并基于时差数据集进行设备功能下降型故障分 析，并基于分析结果执行步骤450或步骤460。
在一些实施例中，管理平台的总平台接收时差数据集后，并基于时差数据集进行设备功能下降型故障分析，包括：选择按照单件制造时间顺序依次计算相邻两个时差的绝对值的差值；当差值出现负值，且差值连续出现负值的次数大于管理平台的总平台设定的阈值时，判定设备功能下降型故障存在，确定分析结果为异常；否则判定设备功能下降型故障不存在，确定分析结果为正常。
在一些实施例中，管理平台的总平台接收时差数据集后，并基于时差数据集进行设备功能下降型故障分析，进一步包括：管理平台的总平台存储有阈值表，每个管理平台的分平台均对应阈值表中唯一的阈值；当设备功能下降型故障分析时，管理平台的总平台对每个管理平台的分平台的时差数据集进行分析，调取阈值表中对应阈值与差值连续出现负值的次数进行差值计算；基于差值计算结果：若差值连续出现负值的次数小于阈值，则判定设备正常；若差值连续出现负值的次数大于阈值，则判定设备存在故障，即可执行故障分析。
步骤450，响应于分析结果为正常时，管理平台的总平台删除时差数据集，并等待分析重新上传的时差数据集。
步骤460，响应于分析结果为异常时，将数据文件、异常结果数据及时差数据集合并成分析数据上传至对应的服务平台的分平台。
步骤470，服务平台的分平台接收分析数据并基于数据文件获取异常时间节点，将异常时间节点、异常结果数据及时差数据集作为打包数据发送至服务平台的总平台。
在一些实施例中，服务平台的分平台接收分析数据并基于数据文件获取异常时间节点，将异常时间节点、异常结果数据及时差数据集作为打包数据发送至服务平台的总平台，具体为：所述单件制造参数数据还包括制造设备在单件制造时的制造开始时刻；当制造设备上传单件制造参数数据时，将单件制造时的制造开始时刻与总耗时数据进行关联并上传；服务平台的分平台接收分析数据后提取数据文件，将时差数据集中出现负值的差值对应的时差提取出来，并基于该时差获取对应的多个总耗时数据；基于多个总耗时数据，获取多个总耗时数据对应的多个制造开始时刻；按照单件制造时间顺序将多个制造开始时刻顺序排序形成所述异常时间节点。
异常时间节点可以有助于判断制造设备出现异常的时刻及时间段，进而可以根据异常时间节点来进行辅助判断异常原因，当用户平台接收打包数据后，可以提取异常时间节点进而判断异常出现的时间和原因，以便于异常能够针对性处理。
步骤480，服务平台的总平台接收打包数据并存储，基于打包数据中的异常结果数据进行故障分级，并将分级后对应的级别信息发送至用户平台。
故障分级是指对设备的发生故障的严重程度划分等级。例如，机械领域中一般设备故障可由低到高分级为轻微型E级、普通型D级、一般型C级、严重型B级以及重大型A级。
在一些实施例中，针对智能制造设备的使用情况，确定了一种故障分级方法，具体为：将连续出现负值的对应差值中，绝对值最大的差值设为T1，绝对值最小的差值设为T2，连续出现负值的具体次数设为N，并设分级基准为F，则分级基准F满足：
F＝(T1-T2)/N   (1)
设对应制造设备中总耗时数据中，允许出现的绝对值最大的差值为T1’，允许连续出现负值的次数为N’，则允许分级标准为F’并且F’满足：
F’＝T1’/N’   (2)
将公式(1)和公式(2)相除获得分级基数Q：
Q＝F/F’
当0＜Q≤0.2时，所述故障分级为普通型D级；
当0.2＜Q≤0.6时，所述故障分级为一般型C级；
当0.6＜Q≤0.8时，所述故障分级为严重型B级；
当0.8＜Q时，所述故障分级为重大型A级。
每个智能制造设备根据运行时长、使用年限及对应不同的加工参数、条件等，都会存在不同的故障，并且故障出现后一般多体现在产品制造上，制造设备即可根据产品制造情况来辅助对故障进行分级。基于此，本实施例通过获取的连续出现负值的对应差值，来计算实际制造过程中的一个分级基准，然后通过不同制造设备允许的不同的连续出现负值的对应差值情况，来获取该制造设备允许的一个允许分级基准，允许分级基准认为是该制造设备最大的分级基准，即最大允许的故障级别，再通过实际标准与允许标准的一个数值大小，来确定实际标准对应属于哪一类故障级别，进而可以针对不同制造设备来进行精准分级判断。
在一些实施例中，如果故障分级为严重型B级或重大型A级时，说明制造设备故障已经较为严 重，为了避免用户平台在处理上的滞后性和延时性，本实施例中当所述故障分级为严重型B级或重大型A级时，服务平台的总平台将分级后对应的级别信息发送至用户平台的同时执行：服务平台的总平台基于故障分级发出预警指令至对应的服务平台的分平台、管理平台的总平台；管理平台的总平台接收预警指令并基于预警指令调取预警指令数据包，将预警指令、预警指令数据包一并发送至对应的管理平台的分平台、传感网络平台的分平台，所述预警指令数据包存储于管理平台的总平台；传感网络平台的分平台将预警指令、预警指令数据包转换成对象平台可识别的组态文件并发送至对应的对象平台；对象平台基于组态文件执行预警操作。
通过以上步骤，当所述故障分级为严重型B级或重大型A级时，服务平台的总平台可以提前用户平台下发对应级别的预警指令并进行相应的预警操作，例如报警、远程警示等，从而可以提前进行相应的预警处理，进而可以提前进行故障处理或减少故障处理的滞后性。
在一些实施例中，当所述用户平台获取级别信息后，基于级别信息，所述用户平台发出检修指令至服务平台的总平台，检修指令至少包括一个自修复子指令；服务平台的总平台接收检修指令并基于检修指令进行指令解析，获得至少一个自修复子指令，将至少一个自修复子指令发送至对应的服务平台的分平台；服务平台的分平台接收自修复子指令并将自修复子指令发送至管理平台的总平台；管理平台的总平台接收自修复子指令，并获取对应的指令代码数据包，将指令代码数据包与自修复子指令关联后一并发送至管理平台的分平台；所述指令代码数据包预存于管理平台的总平台；管理平台的分平台接收指令代码数据包与自修复子指令后发送至对应传感网络平台的分平台；传感网络平台的分平台将指令代码数据包与自修复子指令转换成对象平台可识别的组态文件后发送至对应的对象平台，所述对象平台基于自修复子指令调取指令代码数据包内的指令代码数据执行自修复。
在一些实施例中，当用户平台获取级别信息后，可以根据故障级别提前下发一些指令，用于消除或辅助消除异常和故障。例如，因制造设备自身控制系统错误，可能导致制造设备在抓取零部件、定位零部件及装配度出现因精度问题而需要多次执行动作来完成，导致单件制造时间增加，而这类故障一般是低级别故障，并可通过设备自检、自修复等方式自动处理，无需进行进一步的故障处理。
在一些实施例中，当检修指令对应不同的执行时刻时，服务平台的总平台在解析后将执行时刻写入对应的自修复子指令；当对象平台基于自修复子指令调取指令代码数据包内的指令代码数据后，在对应执行时刻执行自修复，以减少制造设备的任务影响产品制造。
本说明书的一些实施例，通过对设备故障分级并进行预警，可以提前进行故障处理，避免故障程度严重化。同时基于故障级别信息，提前下发自修复指令使设备主动自修复，可以减少故障处理复杂程度及时间、减少故障处理对生产效率的影响。
图5是根据本说明书一些实施例所示的基于合格度确定设备功能下降型故障流程500的示例性流程图。以下简称流程500，如图5所示，流程500可以包括以下步骤。
需要说明的是，流程500可以应用于设备功能下降型故障预警的多种场景。在一些实施例中，流程500可以应用于汽车零部件生产线，例如，汽车发动机配件生产线、汽车制动系配件生产线、电器仪表配件生产线等。管理平台可以基于汽车零部件生产设备的至少一个生产任务中，所生产的汽车零部件的合格零部件的比率，确定每个生产任务的实际合格率。管理平台进一步基于多个生产任务的实际合格率与预设的标准合格率进行对比分析，确定相应的汽车零部件生产设备的多个合格度，并基于多个合格度确定生产设备是否存在功能下降型故障。
对于本领域的技术人员来说，在了解该系统的原理后，可能在不背离这一原理的情况下，将系统移用到其他任何合适的场景下。
以下将以食品生产场景为例对流程500进行具体说明。
步骤510，基于至少一个生产任务的产出食品中合格食品的比率，确定至少一个生产任务中每个生产任务的实际合格率。
合格食品可以指生产任务中产出的满足各项预设指标的食品。例如，产出的食品满足长度、形状、体积、颜色、味道等指标的要求的食品。在一些实施例中，预设指标可以包含产品的膨化度、形状等参数。
合格率可以指设备的某个生产任务所产出的食品中合格食品的数量占总数量的比例。例如，合格率为0.95等。
管理平台可以基于某段时间内(如预设的生产周期)至少一个生产任务的合格率确定相应生产任务的实际合格率。如将各个生产任务在一个生产周期内的合格率作为该生产任务的实际合格率。
在一些实施例中，管理平台可以通过产出食品与标准食品的各项指标的对应关系确定产出食品是否为合格食品。例如，预设各类标准食品数据表，管理平台可以通过对实际产出食品的各项指标与该数据表中的各项指标进行对比分析，确定产出食品是否为合格食品。
在一些实施例中，产出食品是否合格相关于产出食品的膨化度符合度和形状符合度。管理平台可以对产出食品进行图像识别，基于产出食品与标准食品的图像，确定产出食品的膨化度符合度和形状符合度。例如，管理平台可以通过对摄像装置获取的产出食品的实时图像与预设的产出食品的标准图像进行对比分析，确定产出食品的膨化度和形状是否满足标准要求，以进一步确定产出食品是否为合格食品。
膨化度符合度可以指产出食品的体积与标准食品的体积之比符合要求的程度。当膨化度符合度越接近1，则表示产出食品的体积越符合标准食品的体积。在一些实施例中，可以预设膨化度符合度的范围，如[0.9，1.2]区间，当产出食品的膨化度符合度在该范围内，如0.95，则表示产出食品的膨化度符合度符合预设要求。
形状符合度可以指产出食品的形状与标准产出食品的形状的符合程度。例如，雪米饼的标准形状应当为圆形，当实际产出的雪米饼的形状为长方形，则可以认为其形状符合度低。
在一些实施例中，管理平台可以基于第一图像识别模型确定产出食品的膨化度符合度和形状符合度。
第一图像识别模型可以指用于确定产出食品的膨化度符合度和形状符合度进而判断产出食品是否合格的模型。第一图像识别模型可以为训练好的机器学习模型。例如，卷积神经网络、深度神经网络模型或其他自定义的模型结构等中的任意一种或组合。
在一些实施例中，第一图像识别模型可以包括特征提取层和判断层。特征提取层可以包括两个卷积神经网络模型，用于分别提取产出食品的实时图像的食品特征和标准图像中的食品特征，其中，两个卷积神经网络模型具有相同的初始参数，且参数共享。判断层可以是深度神经网络模型，用于确定产出食品的膨化度符合度和形状符合度，进而判断产出食品是否合格。
在一些实施例中，管理平台可以通过摄像装置获取生产任务中的产出食品的实时图像以及预设的产出食品的标准图像，并输入至第一图像识别模型，基于特征提取层的两个卷积神经网络模型分别对实时图像和标准图像进行处理，分别输出产出食品特征向量和标准食品特征向量。食品特征向量可以用于表示食品的膨化度(如长度、宽度、高度特征等)和形状(如轮廓、区域)特征。
进一步的，管理平台将产出食品特征向量和标准食品特征向量输入至判断层，通过判断层对上述两个食品特征向量进行处理，例如，判断层可以计算产出食品和标准食品的食品特征向量的向量距离(如欧式距离)，当向量距离小于预设阈值时，表示产出食品与标准食品相似度满足预设要求，从而确定产出食品的膨化度和形状满足标准的要求，从而确定产出食品为合格食品。又如，判断层还可以将产出食品的膨化度以及形状特征分别与标准食品的膨化度以及形状特征进行比对，以输出产出食品的膨化度符合度和形状符合度，再基于产出食品的膨化度符合度和形状符合度是否都符合预设条件，进而判断产出食品是否为合格食品。
在一些实施例中，第一图像识别模型可以通过特征提取层和判断层的联合训练获取。训练样本可以为历史生产任务的产出食品的多组图像，以及相应的食品的标准图像。训练样本的标签可以基于需要第一图像识别模型的输出进行设定，如若需要输出为产出食品的膨化度符合度和形状符合度，则训练样本的标签可以为历史生产任务的产出食品的实际膨化度符合度和形状符合度。又如，若需要模型直接输出产品是否为合格食品的判断结果，则训练样本的标签可以为历史产出图像所对应的产出食品是否合格的结果。标签可以基于人工标注。
以下以标签为历史产出图像所对应的产出食品是否合格为例，标签可以为1或0，1表示合格，0表示不合格。在训练时，管理平台可以基于训练样本的标签和判断层输出的结果建立损失函数，对模型的参数进行更新。其中，特征提取层的两个卷积神经网络模型可以同步更新。管理平台基于损失函数迭代更新第一图像识别模型的参数，直到预设条件被满足时训练完成，得到训练好的第一图像识别模型。其中，预设条件可以是损失函数小于阈值、收敛，或训练周期达到阈值。
本说明书一些实施例，通过第一图像识别模型对实时的产出食品的图像与标准食品的图像进行处理，有助于快速确认产出食品的膨化度符合度和形状符合度是否符合预设要求，从而提高对产出食品是否为合格食品的判断效率。
在一些实施例中，产出食品是否合格还相关于产出食品的单位面积的上料的均匀程度。例如，管理平台可以基于产出食品的不同区域的情况(如外观、颜色等)，确定产出食品的均匀程度，进而判断产出食品是否合格。
上料可以指在生产某类食品过程中需要加入一定数量的食品辅料的操作。例如，生产薯条，则需要在薯条表面加入预设数量的番茄粉和/或其它辅料。
均匀程度可以指在生产过程中食品的单位面积上的辅料的均匀程度。示例性的，当均匀程度低，即食品中不同区域的上料量差异较大时，则上料量大的区域可能颜色比较集中且接近于辅料的颜色，上料量小的区域颜色可能比较接近于产出产品本身的颜色。均匀程度可以是一个[0-1]区间的数值，如0.8，值越 大表示越均匀。可以预设均匀程度阈值，如0.7。当产出食品的均匀程度大于该阈值时，表示产出的食品满足均匀程度要求。
在一些实施例中，管理平台可以基于第二图像识别模型确定产出食品的均匀程度，进而基于产出食品的均匀程度确定产出食品是否合格。
第二图像识别模型可以指用于确定产出食品的均匀程度进而判断产出食品是否合格的模型。第二图像识别模型可以为训练好的机器学习模型。
在一些实施例中，第二图像识别模型可以包括卷积神经网络模型。用于提取实时图像中产出食品的不同区域的颜色分布特征，以确定产出食品的均匀程度。管理平台可以通过摄像装置获取生产任务中产出食品的实时图像，并输入至第二图像识别模型，通过第二图像识别模型对实时图像进行处理，输出产出食品的均匀程度。
在一些实施例中，第二图像识别模型可以通过训练获得。训练样本可以为历史生产任务的产出食品的多组图像。训练样本的标签可以为历史产出图像所对应的产出食品的均匀程度，如0.5等，训练标签可以进行人工标注。在训练时，管理平台可以基于训练样本的标签和第二图像识别模型的输出建立损失函数，对模型的参数进行更新，并基于损失函数迭代更新第二图像识别模型的参数，直到预设条件被满足训练完成，得到训练好的第二图像识别模型。其中，预设条件可以是损失函数小于阈值、收敛，或训练周期达到阈值。
本说明书一些实施例，通过第二图像识别模型对实时的产出食品的图像进行处理，有助于快速、实时地确认产出食品的均匀程度是否符合预设要求，从而提高对产出食品是否为合格食品的判断效率，减少人工分析带来的时间和精力的消耗。
步骤520，基于每个生产任务的实际合格率与对应的标准合格率，确定设备的多个合格度。
标准合格率可以指预设的合格率。标准合格率可以基于食品的种类、生产难度等生产经验确定，例如，可以预设0.95作为标准合格率。
合格度可以指生产任务的合格率满足合格要求的程度。例如，标准合格率是0.95，若某个生产任务的实际合格率是0.95及以上，则认为合格率满足要求，相应的该生产任务的合格度可以为1；若该生产任务的实际合格率低于0.95，则认为合格率未满足要求，相应的该生产任务的合格度也相应降低。
合格度可以通过各种计算方式确定。例如，合格度可以通过实际合格率与标准合格率的比值确定。如某生产任务的实际合格率为0.92，标准合格率为0.95，则可以认为该生产任务的合格度为0.92/0.95＝0.968。
步骤530，基于多个合格度，确定设备是否存在功能下降型故障。
在一些实施例中，管理平台可以获取设备的多个生产任务的合格度，进而得到该设备的多个合格度，基于多个合格度确定该设备是否存在功能下降型故障。例如，可以预设合格度阈值，管理平台可以计算多个合格度的平均值，当该平均值小于预设的合格度阈值时，确定该设备存在功能下降型故障。
在一些实施例中，管理平台可以通过故障概率预测模型确定设备是否存在功能下降型故障的概率。更多说明参见图6及其描述。
通过引入产出食品的合格度来判断生产设备是否存在功能下降型故障，从生产设备的产出质量情况进行考虑具有实际意义。另外，借助第二图像识别模型确定产品是否合格，有助于提高对食品质量的判断的速度，提升判断设备是否存在功能下降型故障的效率的同时，减少人工检测的成本。
图6是根据本说明书一些实施例所示的故障概率预测模型的示例性示意图。
在一些实施例中，管理平台可以通过故障概率预测模型620预测故障概率，故障概率预测模型620为机器学习模型。例如，循环神经网络模型、卷积神经网络或其他自定义的模型结构等中的任意一种或组合。
故障概率预测模型620可以指用于预测设备存在功能下降型故障的概率的模型。在一些实施例中，故障概率预测模型620包括参数扩充层630和概率预测层670。
参数扩充层630可以指用于处理设备相关的特征的处理层。例如，参数扩充层630可以用于处理设备产出食品的合格度特征、设备执行生产任务的耗时特征等。
在一些实施例中，参数扩充层630可以包括第一特征层640。第一特征层640可以包括第一时序模型641和第一嵌入层643。其中，第一时序模型641可以是长短期记忆网络模型，第一嵌入层643可以是深度神经网络模型。
在一些实施例中，管理平台可以将合格度序列611输入至第一特征层640，通过第一特征层640对合格度序列611进行处理，输出第一特征向量661。
合格度序列611可以指截止到当前时间，设备的历史多个生产任务的合格度根据时间顺序排列所构成的序列。例如，(0.9，0.9，0.6)表示截止到当前时间的3个历史时刻(如T1、T2、T3)的合格度所构成的序列，其可以是一个向量表示。合格度序列611可以基于管理平台通过设备的历史多个生产任务的执行 情况获得。
第一特征向量661可以指合格度特征的向量表示。第一特征向量661包括历史多个时刻的合格度以及预测的未来多个时刻的合格度。例如，3个历史时刻T1、T2、T3和未来2个时刻T4、T5的合格度所构成的第一特征向量661为(0.9，0.9，0.6，0.94，0.92)。
第一时序模型641可以指用于预测未来多个时刻的合格度的模型。在一些实施例中，第一时序模型可以对管理平台输入至第一特征层640的合格度序列611进行处理，输出多个未来时刻的合格度642。例如，基于3个历史时刻T1、T2、T3的合格度序列，输出未来2个时刻T4、T5的合格度0.94、0.92。需要说明的是，多个未来时刻的合格度642也可以是按照时间顺序的序列形式。
第一嵌入层643可以指用于对合格度序列611和多个未来时刻的合格度642进行处理的处理层。在一些实施例中，第一嵌入层643输入合格度序列611以及多个未来时刻的合格度642，输出第一特征向量661。
本说明书一些实施例，通过将预测的合格度作为后续概率预测层的输入以预测功能下降型故障的概率，可以减少设备实际进行测试的次数，降低测试成本，同时增加概率预测层的预测准确度。
概率预测层670可以指用于预测设备存在功能下降型故障概率的处理层。概率预测层670可以是深度神经网络模型。
在一些实施例中，管理平台将第一特征向量661及设备的设备信息663输入至概率预测层670，通过概率预测层670对第一特征向量661和设备信息663进行处理，输出设备存在故障的概率680。
在一些实施例中，管理平台可以将合格度序列611输入至故障概率预测模型620，通过参数扩充层630的第一特征层640对合格度序列611进行处理，输出第一特征向量661。第一特征向量661作为概率预测层670的输入，通过概率预测层670对第一特征向量661和设备信息663进行处理，输出设备存在故障的概率680。在一些实施例中，可以预设概率阈值，例如0.7。响应于故障概率预测模型620输出的设备存在故障的概率680大于预设的概率阈值，确定相应的设备存在功能下降型故障。
在一些实施例中，故障概率预测模型620可以通过训练获取。训练样本包括多组历史合格度序列。多组训练样本可以通过历史生产数据获得。例如，训练样本可以是设备的过去一年里若干个连续的生产任务的合格度构成的合格度序列。训练样本的标签可以是每一组样本所对应的设备故障情况。标签可以基于人工进行标注或其它可行的方式标注。如0表示设备发生了故障，1表示设备未发生故障。管理平台可以将训练样本中的多组合格度序列611输入至故障概率预测模型620，基于故障概率预测模型620的输出与标签构建损失函数，并基于损失函数同时迭代更新初始的第一时序模型641、第一嵌入层643以及概率预测层670的参数，直到预设条件被满足训练完成，得到训练好的故障概率预测模型620。其中，预设条件可以是损失函数小于阈值、收敛，或训练周期达到阈值。
在一些实施例中，故障概率预测模型620还可以包括第二特征层650。第二特征层620可以包括第二时序模型651和第二嵌入层654。其中，第二时序模型651可以是长短期记忆网络模型，第二嵌入层654可以为深度神经网络模型。
在一些实施例中，管理平台可以将耗时差序列612输入至第二特征层650，通过第二特征层650对耗时差序列612进行处理，输出为第二特征向量662。
耗时差序列612可以指截止到当前时间的历史多个生产任务的耗时差根据时间顺序所构建的序列。例如，(0s，0s，0s，1s，0s)表示截止到当前时刻的5个历史生产任务的耗时差序列，其可以是一个向量表示。耗时差序列可以基于管理平台通过设备的历史多个生产任务的耗时情况获得。关于耗时差的说明参见图3的相关内容。
第二特征向量662可以指耗时差特征的向量表示。第二特征向量包括历史多个生产任务的耗时差以及预测的未来多个生产任务的耗时差。例如，5个历史生产任务和未来3个生产任务的耗时差所构成的第二特征向量为(10s，1s，0s，0s，0s，2s，3s，0s)。其中，前5个耗时差为5个历史生产任务的耗时差。
第二时序模型651可以指用于预测未来多个时刻生产任务的耗时差的模型。在一些实施例中，第二时序模型651可以对管理平台输入至第二特征层650的耗时差序列612进行处理，输出多个未来时刻的耗时差652。例如，基于上述的5个历史时刻T1、T2、T3、T4、T5的耗时差序列(10s，1s，0s，0s，0s)，输出未来3个时刻T6、T7、T8的耗时差(2s，3s，0s)。
第二嵌入层654可以指用于对耗时差序列612和多个未来时刻的耗时差652进行处理的处理层。在一些实施例中，管理平台可以将耗时差序列612以及多个未来时刻的耗时差652输入至第二嵌入层654，通过第二嵌入层654的处理，输出第二特征向量662。
在一些实施例中，响应于故障概率预测模型620包括第二特征层650，则概率预测层670的输入还包括第二特征层650输出的第二特征向量662。概率预测层670可以对第一特征向量661、第二特征向量662以及设备相关信息663进行处理，输出设备存在故障的概率680。
本说明书一些实施例，通过引入生产任务的历史耗时差以及预测的未来的耗时差数据对设备功能下降型故障的概率进行预测，可以提高预测的准确性。
在一些实施例中，第一特征层640的输出还包括第一置信度664，第二特征层650的输出还包括第二置信度665。相应的，概率预测层670的输入则还包括第一置信度664和第二置信度665。
第一置信度可以指预测的多个未来时刻的合格度642的可信程度。第一置信度可以表示为一个(0，1)区间内的数值，如0.8，数值越大表示可信程度越高。可以理解的是，当预测未来的时刻距离当前时刻越远，其置信度越低。
在一些实施例中，第一特征层640还可以包括第一置信度计算模块644。第一置信度664可以通过第一置信度计算模块644的计算获得。计算的方式可以基于合格度序列611与预测的未来时刻的合格度的数量的关系进行确定。例如，预设计算公式：第一置信度＝合格度序列维度/(合格度序列维度+未来时刻的合格度的数量)。示例性的，合格度序列维度为5，需要预测的未来时刻的合格度的数量为3，即通过第一时序模型基于5个历史数据预测未来时刻的3个生产任务的合格度，则第一置信度＝5/(5+3)＝0.625。需要说明的是，每个生产任务的执行具有相应的生产周期，当预测未来时刻的合格度数量越多，表示预测的未来的时间的跨度越大，相应的第一置信度会越低。另一方面，当采用的历史合格度的数量越多，表示数据支持得越充分，相应的第一置信度会越高。
第二置信度可以指预测的多个未来时刻的耗时差的可信程度。第二置信度可以表示为一个(0，1)区间内的数值，如0.8，数值越大表示可信程度越高。可以理解的是，当预测未来的时刻距离当前时刻越远，其置信度越低。
在一些实施例中，第二特征层650还可以包括第二置信度计算模块653。第二置信度665可以基于第二置信度计算模块653获得。计算的方式可以基于耗时差序列的维度与预测的未来时刻的耗时差的数量的关系进行确定。例如，预设计算公式：第二置信度＝耗时差序列维度/(耗时差序列维度+未来时刻的耗时差的数量)。与第一置信的计算方式同理，此处不再赘述。
本说明书一些实施例，通过在对设备功能下降型故障的概率进行预测时，考虑了合格度数据以及耗时差数据的置信度的影响，可以进一步提高预测的准确性。
在一些实施例中，故障概率预测模型620可以通过参数扩充层630与概率预测层670的联合训练获得。训练样本包括多组历史生产任务的合格度序列611，以及与每一组合格度序列611所对应的生产任务的耗时差序列612，训练样本的标签可以基于设备对应的故障的情况进行人工标注。
管理平台可以将训练样本中的多组合格度序列611、耗时差序列612输入至参数扩充层630。其中，合格度序列611输入至第一特征层640，同时，耗时差序列612输入至第二特征层650。
在第一特征层640中的处理中，合格度序列611输入至第一时序模型641，输出多个未来时刻的合格度642；多个未来时刻的合格度642与合格度序列611作为第一嵌入层643的输入，经过第一嵌入层643的处理得到第一特征向量661，同时，第一置信度计算模块644基于多个未来时刻的合格度642与合格度序列611通过预设的计算公式得到第一置信度664。
与此同时的，在第二特征层650中的处理中，耗时差序列612输入至第二时序模型651，输出多个未来时刻的耗时差652；多个未来时刻的耗时差652与耗时差序列612作为第二嵌入层653的输入，经过第一嵌入层643的处理得到第二特征向量662，同时，第二置信度计算模块653基于多个未来时刻的耗时差652与耗时差序列612通过预设的计算公式得到第二置信度665。
进一步的，管理平台将第一特征层640输出的多组第一特征向量661、第一置信度664和第二特征层输出的多组第二特征向量662、第二置信度665作为概率预测层670的训练样本数据，同时，概率预测层670的训练样本还包括设备相关信息663。
管理平台将上述多组概率预测层670的训练样本输入至概率预测层670，并基于概率预测层670的输出与标签构建损失函数，基于损失函数同时迭代更新初始的第一特征层640、初始的第二特征层650以及概率预测层670的参数。其中，第一特征层640的参数更新包括更新第一时序模型641和第一嵌入层643的参数；第二特征层650的参数更新包括更新第二时序模型651和第二嵌入层653的参数。参数迭代更新的终止条件可以是预设条件被满足，则训练完成，得到训练好的第一特征层640、第二特征层650以及概率预测层670，并最终获取故障概率预测模型620。其中，预设条件可以是损失函数小于阈值、收敛，或训练周期达到阈值。
通过参数扩充层630与概率预测层670进行联合训练获得故障概率预测模型620，有助于降低获取训练样本复杂度，提高训练的效率。
本说明书一些实施例，通过故障概率预测模型对设备的功能下降型故障进行预测，有助于对可能存在功能下降型故障的设备进行预警和提前防范。
本说明书提供了一种用于设备功能下降型故障预警的工业物联网及其控制方法，通过基于五平台 结构搭建工业物联网，并采用前分平台式布置与独立式布置相结合的平台布置方法，可以确保数据传输时的独立，也便于对数据进行分类与处理。通过对设备功能下降型故障进行分级，并基于故障分级情况，服务平台的总平台自行下发故障处理命令，可以提前对设备进行故障预警，确保生产线能正常运转，进而达到保障生产效率的目的。
上文已对基本概念做了描述，显然，对于本领域技术人员来说，上述详细披露仅仅作为示例，而并不构成对本说明书的限定。虽然此处并没有明确说明，本领域技术人员可能会对本说明书进行各种修改、改进和修正。该类修改、改进和修正在本说明书中被建议，所以该类修改、改进、修正仍属于本说明书示范实施例的精神和范围。
同时，本说明书使用了特定词语来描述本说明书的实施例。如“一个实施例”、“一实施例”、和/或“一些实施例”意指与本说明书至少一个实施例相关的某一特征、结构或特点。因此，应强调并注意的是，本说明书中在不同位置两次或多次提及的“一实施例”或“一个实施例”或“一个替代性实施例”并不一定是指同一实施例。此外，本说明书的一个或多个实施例中的某些特征、结构或特点可以进行适当的组合。
此外，除非权利要求中明确说明，本说明书所述处理元素和序列的顺序、数字字母的使用、或其他名称的使用，并非用于限定本说明书流程和方法的顺序。尽管上述披露中通过各种示例讨论了一些目前认为有用的发明实施例，但应当理解的是，该类细节仅起到说明的目的，附加的权利要求并不仅限于披露的实施例，相反，权利要求旨在覆盖所有符合本说明书实施例实质和范围的修正和等价组合。例如，虽然以上所描述的系统组件可以通过硬件设备实现，但是也可以只通过软件的解决方案得以实现，如在现有的服务器或移动设备上安装所描述的系统。
同理，应当注意的是，为了简化本说明书披露的表述，从而帮助对一个或多个发明实施例的理解，前文对本说明书实施例的描述中，有时会将多种特征归并至一个实施例、附图或对其的描述中。但是，这种披露方法并不意味着本说明书对象所需要的特征比权利要求中提及的特征多。实际上，实施例的特征要少于上述披露的单个实施例的全部特征。
一些实施例中使用了描述成分、属性数量的数字，应当理解的是，此类用于实施例描述的数字，在一些示例中使用了修饰词“大约”、“近似”或“大体上”来修饰。除非另外说明，“大约”、“近似”或“大体上”表明所述数字允许有±20％的变化。相应地，在一些实施例中，说明书和权利要求中使用的数值参数均为近似值，该近似值根据个别实施例所需特点可以发生改变。在一些实施例中，数值参数应考虑规定的有效数位并采用一般位数保留的方法。尽管本说明书一些实施例中用于确认其范围广度的数值域和参数为近似值，在具体实施例中，此类数值的设定在可行范围内尽可能精确。
针对本说明书引用的每个专利、专利申请、专利申请公开物和其他材料，如文章、书籍、说明书、出版物、文档等，特此将其全部内容并入本说明书作为参考。与本说明书内容不一致或产生冲突的申请历史文件除外，对本说明书权利要求最广范围有限制的文件(当前或之后附加于本说明书中的)也除外。需要说明的是，如果本说明书附属材料中的描述、定义、和/或术语的使用与本说明书所述内容有不一致或冲突的地方，以本说明书的描述、定义和/或术语的使用为准。
最后，应当理解的是，本说明书中所述实施例仅用以说明本说明书实施例的原则。其他的变形也可能属于本说明书的范围。因此，作为示例而非限制，本说明书实施例的替代配置可视为与本说明书的教导一致。相应地，本说明书的实施例不仅限于本说明书明确介绍和描述的实施例。

Claims (19)

  1. 一种用于设备功能下降型故障预警的工业物联网，其特征在于，包括管理平台，所述管理平台被配置为执行以下操作：
    获取设备在预设时间段执行的至少一个生产任务的执行情况；
    基于所述至少一个生产任务的执行情况，判断所述设备是否存在功能下降型故障；
    响应于所述设备存在所述功能下降型故障，发出预警并对所述设备进行相应修复。
  2. 根据权利要求1所述的用于设备功能下降型故障预警的工业物联网，其特征在于，还包括用户平台、服务平台、传感网络平台和对象平台，所述用户平台、所述服务平台、所述管理平台、所述传感网络平台和所述对象平台从上到下依次交互；
    所述服务平台和管理平台均采用前分平台式布置，所述传感网络平台采用独立式布置；所述前分平台式布置是指对应平台设置有一个总平台和多个分平台，多个分平台分别存储和处理下层平台发送的不同类型或不同接收对象的数据，一个总平台对多个分平台的数据进行汇总后存储和处理，并传输数据至上层平台，所述独立式布置是指传感网络平台对不同对象平台的数据采用不同的分平台进行数据存储、数据处理和/或数据传输；所述对象平台被配置为智能制造的制造设备；
    当制造设备执行制造时，上传单件制造参数数据至对应的传感网络平台的分平台；单件制造参数数据至少包括该制造设备在单件制造时的总耗时数据；
    传感网络平台的分平台将单件制造参数数据转换成管理平台可识别的数据文件并发送至对应的管理平台的分平台；
    管理平台的分平台接收数据文件并提取总耗时数据，基于总耗时数据，按照单件制造时间顺序依次计算相邻两个总耗时数据的时差，将所有时差按照单件制造时间顺序顺次排序形成时差数据集，并将数据文件、时差数据集存储后发送至管理平台的总平台；
    管理平台的总平台接收时差数据集后，并基于时差数据集进行设备功能下降型故障分析，并基于分析结果执行：
    当分析结果为正常时，管理平台的总平台删除时差数据集，并等待分析重新上传的时差数据集；
    或，当分析结果为异常时，将数据文件、异常结果数据及时差数据集合并成分析数据上传至对应的服务平台的分平台；
    服务平台的分平台接收分析数据并基于数据文件获取异常时间节点，将异常时间节点、异常结果数据及时差数据集作为打包数据发送至服务平台的总平台；
    服务平台的总平台接收打包数据并存储，基于打包数据中的异常结果数据进行故障分级，并将分级后对应的级别信息发送至用户平台。
  3. 根据权利要求2所述的用于设备功能下降型故障预警的工业物联网，其特征在于，管理平台的总平台接收时差数据集后，并基于时差数据集进行设备功能下降型故障分析具体为：
    管理平台的总平台接收时差数据集后，选择按照单件制造时间顺序依次计算相邻两个时差的绝对值的差值；
    当差值出现负值，且差值连续出现负值的次数大于管理平台的总平台设定的阈值时，判定设备功能下降型故障存在，确定分析结果为异常；否则判定设备功能下降型故障不存在，确定分析结果为正常。
  4. 根据权利要求3所述的用于设备功能下降型故障预警的工业物联网，其特征在于，所述管理平台的总平台存储有阈值表，每个管理平台的分平台均对应阈值表中唯一的阈值；
    当设备功能下降型故障分析时，管理平台的总平台对每个管理平台的分平台的时差数据集进行分析，调取阈值表中对应阈值与差值连续出现负值的次数进行差值计算。
  5. 根据权利要求3所述的用于设备功能下降型故障预警的工业物联网，其特征在于，服务平台的分平台接收分析数据并基于数据文件获取异常时间节点，将异常时间节点、异常结果数据及时差数据集作为打包数据发送至服务平台的总平台，具体为：
    所述单件制造参数数据还包括制造设备在单件制造时的制造开始时刻；当制造设备上传单件制造参数数据时，将单件制造时的制造开始时刻与总耗时数据进行关联一并上传；
    服务平台的分平台接收分析数据后提取数据文件，将时差数据集中出现负值的差值对应的时差提取出来，并基于该时差获取对应的多个总耗时数据；
    基于多个总耗时数据，获取多个总耗时数据对应的多个制造开始时刻；
    按照单件制造时间顺序将多个制造开始时刻顺序排序形成所述异常时间节点。
  6. 根据权利要求2所述的用于设备功能下降型故障预警的工业物联网，其特征在于，当所述用户平台获取级别信息后，基于级别信息，所述用户平台发出检修指令至服务平台的总平台，检修指令至少包括一个自修复子指令；
    服务平台的总平台接收检修指令并基于检修指令进行指令解析，获得至少一个自修复子指令，将至少一个自修复子指令发送至对应的服务平台的分平台；
    服务平台的分平台接收自修复子指令并将自修复子指令发送至管理平台的总平台；
    管理平台的总平台接收自修复子指令，并获取对应的指令代码数据包，将指令代码数据包与自修复子指令关联后一并发送至管理平台的分平台；所述指令代码数据包预存于管理平台的总平台；
    管理平台的分平台接收指令代码数据包与自修复子指令后发送至对应传感网络平台的分平台；
    传感网络平台的分平台将指令代码数据包与自修复子指令转换成对象平台可识别的组态文件后发送至对应的对象平台，所述对象平台基于自修复子指令调取指令代码数据包内的指令代码数据执行自修复。
  7. 根据权利要求6所述的用于设备功能下降型故障预警的工业物联网，其特征在于，
    当检修指令对应不同的执行时刻时，服务平台的总平台在解析后将执行时刻写入对应的自修复子指令；
    当所述对象平台基于自修复子指令调取指令代码数据包内的指令代码数据后，在对应执行时刻执行自修复。
  8. 根据权利要求1所述的用于设备功能下降型故障预警的工业物联网，其特征在于，所述基于所述至少一个生产任务的执行情况，判断所述设备是否存在功能下降型故障包括：
    获取所述设备执行每个生产任务的实际耗时；
    基于所述每个生产任务执行的所述实际耗时与对应的标准耗时，确定多个耗时差；
    基于所述多个耗时差，确定所述设备是否存在所述功能下降型故障。
  9. 根据权利要求1所述的用于设备功能下降型故障预警的工业物联网，其特征在于，所述基于所述至少一个生产任务的执行情况，判断所述设备是否存在功能下降型故障包括：
    基于所述至少一个生产任务的所述产出产品中合格产品的比率，确定所述至少一个生产任务中每个生产任务的实际合格率；
    基于所述每个生产任务的所述实际合格率与对应的标准合格率，确定所述设备的多个合格度；
    基于所述多个合格度，确定所述设备是否存在功能下降型故障。
  10. 根据权利要求9所述的用于设备功能下降型故障预警的工业物联网，其特征在于，所述基于所述多个合格度，确定所述设备是否存在功能下降型故障包括：
    通过故障概率预测模型预测故障概率；所述故障概率预测模型为机器学习模型，所述故障概率预测模型包括：参数扩充层和概率预测层；所述参数扩充层用于基于所述多个合格度确定所述设备的第一特征向量，所述概率预测层用于基于所述第一特征向量及所述设备的设备信息，预测所述设备的所述故障概率；
    响应于所述故障概率大于概率阈值，确定所述设备存在所述功能下降型故障。
  11. 用于设备功能下降型故障预警的工业物联网的控制方法，其特征在于，
    基于用于设备功能下降型故障预警的工业物联网的管理平台实现；
    所述控制方法包括：
    获取设备在预设时间段执行的至少一个生产任务的执行情况；
    基于所述至少一个生产任务的执行情况，判断所述设备是否存在功能下降型故障；
    响应于所述设备存在所述功能下降型故障，发出预警并对所述设备进行相应修复。
  12. 根据权利要求11所述的控制方法，所述用于设备功能下降型故障预警的工业物联网包括从上到下依次交互的用户平台、服务平台、管理平台、传感网络平台和对象平台；其特征在于，
    所述服务平台和管理平台均采用前分平台式布置，所述传感网络平台采用独立式布置；所述前分平台式布置是指对应平台设置有一个总平台和多个分平台，多个分平台分别存储和处理下层平台发送的不同类型或不同接收对象的数据，一个总平台对多个分平台的数据进行汇总后存储和处理，并传输数据至上层平台，所述独立式布置是指传感网络平台对不同对象平台的数据采用不同的分平台进行数据存储、数据处理和/或数据传输；所述对象平台被配置为智能制造的制造设备；
    所述控制方法包括：
    当制造设备执行制造时，上传单件制造参数数据至对应的传感网络平台的分平台；单件制造参数数据至少包括该制造设备在单件制造时的总耗时数据；
    传感网络平台的分平台将单件制造参数数据转换成管理平台可识别的数据文件并发送至对应的管理平台的分平台；
    管理平台的分平台接收数据文件并提取总耗时数据，基于总耗时数据，按照单件制造时间顺序依次计算相邻两个总耗时数据的时差，将所有时差按照单件制造时间顺序顺次排序形成时差数据集，并将数据文件、时差数据集存储后发送至管理平台的总平台；
    管理平台的总平台接收时差数据集后，并基于时差数据集进行设备功能下降型故障分析，并基于分析结果执行：
    当分析结果为正常时，管理平台的总平台删除时差数据集，并等待分析重新上传的时差数据集；
    或，当分析结果为异常时，将数据文件、异常结果数据及时差数据集合并成分析数据上传至对应的服务平台的分平台；
    服务平台的分平台接收分析数据并基于数据文件获取异常时间节点，将异常时间节点、异常结果数据及时差数据集作为打包数据发送至服务平台的总平台；
    服务平台的总平台接收打包数据并存储，基于打包数据中的异常结果数据进行故障分级，并将分级后对应的级别信息发送至用户平台。
  13. 根据权利要求12所述的控制方法，其特征在于，管理平台的总平台接收时差数据集后，并基于时差数据集进行设备功能下降型故障分析具体为：
    管理平台的总平台接收时差数据集后，选择按照单件制造时间顺序依次计算相邻两个时差的绝对值的差值；
    当差值出现负值，且差值连续出现负值的次数大于管理平台的总平台设定的阈值时，判定设备功能下降型故障存在，确定分析结果为异常；否则判定设备功能下降型故障不存在，确定分析结果为正常。
  14. 根据权利要求13所述的控制方法，其特征在于，所述故障分级具体为：
    将连续出现负值的对应差值中，绝对值最大的差值设为T1，绝对值最小的差值设为T2，连续出现负值的具体次数设为N，并设分级基准为F，则分级基准F满足：
    F＝(T1-T2)/N  (1)
    设对应制造设备中总耗时数据中，允许出现的绝对值最大的差值为T1’，允许连续出现负值的次数为N’，则允许分级标准为F’并且F’满足：
    F’＝T1’/N’  (2)
    将公式(1)和公式(2)相除获得分级基数Q：
    Q＝F/F’
    当0＜Q≤0.2时，所述故障分级为普通型D级；
    当0.2＜Q≤0.6时，所述故障分级为一般型C级；
    当0.6＜Q≤0.8时，所述故障分级为严重型B级；
    当0.8＜Q时，所述故障分级为重大型A级。
  15. 根据权利要求14所述的控制方法，其特征在于，当所述故障分级为严重型B级或重大型A级时，服务平台的总平台将分级后对应的级别信息发送至用户平台的同时执行：
    服务平台的总平台基于故障分级发出预警指令至对应的服务平台的分平台、管理平台的总平台；
    管理平台的总平台接收预警指令并基于预警指令调取预警指令数据包，将预警指令、预警指令数据包一并发送至对应的管理平台的分平台、传感网络平台的分平台，所述预警指令数据包存储于管理平台的总平台；
    传感网络平台的分平台将预警指令、预警指令数据包转换成对象平台可识别的组态文件并发送至对应的对象平台；
    对象平台基于组态文件执行预警操作。
  16. 根据权利要求11所述的控制方法，其特征在于，所述基于所述至少一个生产任务的执行情况，判断所述设备是否存在功能下降型故障包括：
    获取所述设备执行每个生产任务的实际耗时；
    基于所述每个生产任务执行的所述实际耗时与对应的标准耗时，确定多个耗时差；
    基于所述多个耗时差，确定所述设备是否存在所述功能下降型故障。
  17. 根据权利要求11所述的控制方法，其特征在于，所述基于所述至少一个生产任务的执行情况，判断所述设备是否存在功能下降型故障包括：
    基于所述至少一个生产任务的所述产出产品中合格产品的比率，确定所述至少一个生产任务中每个生产任务的实际合格率；
    基于所述每个生产任务的所述实际合格率与对应的标准合格率，确定所述设备的多个合格度；
    基于所述多个合格度，确定所述设备是否存在功能下降型故障。
  18. 根据权利要求17所述的控制方法，其特征在于，所述基于所述多个合格度，确定所述设备是否存在功能下降型故障包括：
    通过故障概率预测模型预测故障概率；所述故障概率预测模型为机器学习模型，所述故障概率预测模型包括：参数扩充层和概率预测层；所述参数扩充层用于基于所述多个合格度确定所述设备的第一特征向量，所述概率预测层用于基于所述第一特征向量及所述设备的设备信息，预测所述设备的所述故障概率；
    响应于所述故障概率大于概率阈值，确定所述设备存在所述功能下降型故障。
  19. 一种计算机可读存储介质，其特征在于，所述存储介质存储计算机指令，当所述计算机指令被处理器执行时实现权利要求11～18中任一项所述的方法。
PCT/CN2022/106031 2022-07-15 2022-07-15 用于设备功能下降型故障预警的工业物联网、方法及介质 WO2024011601A1 (zh)

Priority Applications (3)

Application Number Priority Date Filing Date Title
PCT/CN2022/106031 WO2024011601A1 (zh) 2022-07-15 2022-07-15 用于设备功能下降型故障预警的工业物联网、方法及介质
CN202280007664.3A CN116940956A (zh) 2022-07-15 2022-07-15 用于设备功能下降型故障预警的工业物联网、方法及介质
US18/329,679 US20230315062A1 (en) 2022-07-15 2023-06-06 Industrial internet of things, mthods and mediums for early warning of descending function fault of equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2022/106031 WO2024011601A1 (zh) 2022-07-15 2022-07-15 用于设备功能下降型故障预警的工业物联网、方法及介质

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US18/329,679 Continuation US20230315062A1 (en) 2022-07-15 2023-06-06 Industrial internet of things, mthods and mediums for early warning of descending function fault of equipment

Publications (1)

Publication Number Publication Date
WO2024011601A1 true WO2024011601A1 (zh) 2024-01-18

Family

ID=88194072

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/106031 WO2024011601A1 (zh) 2022-07-15 2022-07-15 用于设备功能下降型故障预警的工业物联网、方法及介质

Country Status (3)

Country Link
US (1) US20230315062A1 (zh)
CN (1) CN116940956A (zh)
WO (1) WO2024011601A1 (zh)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117096817B (zh) * 2023-10-12 2024-03-26 南方电网数字电网研究院有限公司 继电器、继电器修复方法、装置、计算机设备
CN117270514B (zh) * 2023-11-22 2024-01-26 南京迅集科技有限公司 基于工业物联网的生产过程全流程故障检测方法
CN117835291B (zh) * 2024-03-04 2024-05-10 济南光路科技有限公司 一种基于物联网的数据管理系统及方法
CN117873007B (zh) * 2024-03-11 2024-05-24 成都秦川物联网科技股份有限公司 基于工业物联网的制造流程管理方法、系统、设备及介质
CN117891223A (zh) * 2024-03-14 2024-04-16 成都秦川物联网科技股份有限公司 一种基于工业物联网的设备管理方法、系统、设备及介质

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105677554A (zh) * 2016-02-24 2016-06-15 上海和鹰机电科技股份有限公司 基于互联网平台的远程维护方法
US20190130284A1 (en) * 2017-10-27 2019-05-02 International Business Machines Corporation Interactive Feedback and Assessment Experience
CN112217283A (zh) * 2020-10-12 2021-01-12 云南电网有限责任公司电力科学研究院 一种基于物联网的电力设备状态在线监测系统
CN114449023A (zh) * 2022-04-11 2022-05-06 成都秦川物联网科技股份有限公司 双前分平台式工业物联网及其控制方法
CN114488988A (zh) * 2022-04-14 2022-05-13 成都秦川物联网科技股份有限公司 用于生产线平衡率调控的工业物联网及控制方法
CN114742487A (zh) * 2022-06-13 2022-07-12 成都秦川物联网科技股份有限公司 基于工业物联网的生产任务管控方法及系统
CN114742254A (zh) * 2022-06-10 2022-07-12 成都秦川物联网科技股份有限公司 用于流水线设备故障处理的工业物联网及其控制方法
CN114741454A (zh) * 2022-06-10 2022-07-12 成都秦川物联网科技股份有限公司 用于巡检数据处理的工业物联网及其控制方法

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105677554A (zh) * 2016-02-24 2016-06-15 上海和鹰机电科技股份有限公司 基于互联网平台的远程维护方法
US20190130284A1 (en) * 2017-10-27 2019-05-02 International Business Machines Corporation Interactive Feedback and Assessment Experience
CN112217283A (zh) * 2020-10-12 2021-01-12 云南电网有限责任公司电力科学研究院 一种基于物联网的电力设备状态在线监测系统
CN114449023A (zh) * 2022-04-11 2022-05-06 成都秦川物联网科技股份有限公司 双前分平台式工业物联网及其控制方法
CN114488988A (zh) * 2022-04-14 2022-05-13 成都秦川物联网科技股份有限公司 用于生产线平衡率调控的工业物联网及控制方法
CN114742254A (zh) * 2022-06-10 2022-07-12 成都秦川物联网科技股份有限公司 用于流水线设备故障处理的工业物联网及其控制方法
CN114741454A (zh) * 2022-06-10 2022-07-12 成都秦川物联网科技股份有限公司 用于巡检数据处理的工业物联网及其控制方法
CN114742487A (zh) * 2022-06-13 2022-07-12 成都秦川物联网科技股份有限公司 基于工业物联网的生产任务管控方法及系统

Also Published As

Publication number Publication date
CN116940956A (zh) 2023-10-24
US20230315062A1 (en) 2023-10-05

Similar Documents

Publication Publication Date Title
WO2024011601A1 (zh) 用于设备功能下降型故障预警的工业物联网、方法及介质
US11340591B2 (en) Predictive maintenance and process supervision using a scalable industrial analytics platform
US10257310B2 (en) Industrial data analytics in a cloud platform
CN113112086B (zh) 一种基于边缘计算和标识解析的智能生产系统
US11348023B2 (en) Identifying locations and causes of network faults
CN106598791B (zh) 一种基于机器学习的工业设备故障预防性识别方法
US20140336795A1 (en) Remote assistance via a cloud platform for industrial automation
CN113614666A (zh) 用于检测和预测工业过程自动化系统中的故障的系统和方法
US20140337277A1 (en) Industrial device and system attestation in a cloud platform
CN111047082A (zh) 设备的预警方法及装置、存储介质和电子装置
US20140336791A1 (en) Predictive maintenance for industrial products using big data
US10444746B2 (en) Method for managing subsystems of a process plant using a distributed control system
US11321581B2 (en) Detecting and mitigating anomalies and degradation associated with devices and their operations
CN110749462B (zh) 一种基于边缘计算的工业设备故障检测方法及系统
CN116107282B (zh) 一种基于企业应用集成的工业机器人预测性维护系统
CN113313280B (zh) 云平台的巡检方法、电子设备及非易失性存储介质
US20220057788A1 (en) End to end smart manufacturing architecture for operational efficiency and quality control
WO2024065988A1 (zh) 设备监测方法和系统、电子设备、存储介质
EP3843338B1 (en) Monitoring and analyzing communications across multiple control layers of an operational technology environment
JP2023547849A (ja) ラベルなしセンサデータを用いた産業システム内の稀な障害の自動化されたリアルタイムの検出、予測、及び予防に関する、方法または非一時的コンピュータ可読媒体
CN117390529A (zh) 多因素溯源的数据中台信息管理方法
CN115695150A (zh) 一种基于分布式异构融合组网设备检测方法及装置
CN110794799A (zh) 应用于工业生产的具有故障诊断功能的大数据系统
Larrinaga et al. Implementation of a reference architecture for cyber physical systems to support condition based maintenance
US10896079B2 (en) Common gateway platform

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: 22950719

Country of ref document: EP

Kind code of ref document: A1