System, method, device and storage medium for monitoring the operating state of industrial equipment

Abstract

The embodiment of the application relates to the technical field of industrial internet equipment management, and discloses an industrial equipment operation state monitoring system, method, device and storage medium, the system comprises an industrial gateway, a Kafka subsystem, a storage unit and a big data analysis modeling platform, wherein the industrial gateway obtains original operation data of industrial equipment, carries out data cleaning, obtains first operation data, and transmits the first operation data to a first message node of the Kafka subsystem; the first message node transmits the first operation data to the storage unit; the big data analysis modeling platform obtains the first operation data from the first message node to execute a data monitoring task, if the first operation data is abnormal, marks the first operation data to obtain second operation data and transmits the second operation data to a second message node of the Kafka subsystem; and the second message node transmits the second operation data to the storage unit. Through the above mode, the embodiment of the application realizes real-time monitoring of the operation state of the industrial equipment.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of industrial internet equipment management, and in particular to a system and method for monitoring the running state of industrial equipment, equipment and a storage medium. BACKGROUND

[0002] With the rapid development of industrialization, the use of various industrial internet equipment (referred to as industrial equipment) is also gradually increasing. For example, due to the effective reduction of labor costs by automated production, personnel allocation is more reasonable, so the application of automated production is becoming more and more widespread, and various industrial equipment will be applied in automated production. For automated production, timely monitoring of the running state of industrial equipment can ensure the continuity and stability of production.

[0003] Currently, the running state of industrial equipment is usually monitored by manually inspecting the industrial equipment, but the manual inspection method cannot monitor the running state of the industrial equipment at all times, i.e., it cannot timely detect abnormalities in the industrial equipment. Moreover, the efficiency of manual inspection is low, and if the production scale is large and a large number of industrial equipment is applied, the labor cost will also increase. Therefore, if the running state of the industrial equipment can be efficiently monitored at all times and the running state of the industrial equipment can be obtained at any time as needed, the abnormality of the industrial equipment can be timely handled when the industrial equipment is running abnormally, thereby reducing the loss caused by the abnormal running of the industrial equipment. SUMMARY

[0004] In view of the above problems, embodiments of the present application provide a system and method for monitoring the running state of industrial equipment, equipment and a storage medium, to solve the problem that the running state of industrial equipment cannot be efficiently monitored in real time in the prior art.

[0005] According to an aspect of the embodiments of the present application, a system for monitoring running states of industrial equipment is provided. The system comprises an industrial gateway, a Kafka subsystem, a storage unit and a big data analysis modeling platform. The industrial gateway is configured to continuously acquire raw running data of the industrial equipment, perform data cleaning on the raw running data to obtain first running data, and transmit the first running data to the Kafka subsystem. If the system monitors at least two industrial equipment, the industrial gateway is further configured to aggregate all the raw running data of all the industrial equipment. The Kafka subsystem comprises a first message node and a second message node. The first message node is configured to receive the first running data and transmit the first running data to the storage unit. The big data analysis modeling platform is configured to acquire the first running data from the first message node of the Kafka subsystem to perform a data monitoring task. If the result of performing the data monitoring task is that the first running data is abnormal, the first running data is marked to obtain second running data, and the second running data is transmitted to the second message node of the Kafka subsystem. The data monitoring task is a task generated by publishing a trained model. The second message node of the Kafka subsystem is configured to receive the second running data and transmit the second running data to the storage unit. The storage unit is configured to store the first running data and the second running data. The first running data stored in the storage unit is historical data. The big data analysis modeling platform is further configured to acquire the historical data from the storage unit, and train an original model using the historical data to obtain the trained model.

[0006] In an optional manner, the system further comprises an alarm unit. The second message node of the Kafka subsystem is further configured to transmit the second running data to the alarm unit. The alarm unit is configured to generate alarm information when the second running data meets a preset alarm rule.

[0007] In an optional manner, the system further comprises a failure prediction and health management subsystem. The storage unit is further configured to transmit the first running data and the second running data to the failure prediction and health management subsystem. The alarm unit is further configured to transmit the alarm information to the failure prediction and health management subsystem. The failure prediction and health management subsystem is configured to receive the first running data, the second running data and the alarm information, generate and display a health record of the industrial equipment according to the first running data and the second running data, and display the alarm information.

[0008] In an optional mode, the system monitors the industrial equipment including the newly added industrial equipment, wherein the fault prediction and health management subsystem is further configured to determine the alarm rule of the newly added industrial equipment according to the first operation data and expert experience data of the newly added industrial equipment in response to the access of the newly added industrial equipment.

[0009] In an optional mode, the industrial equipment includes a pointer type instrument panel, and the system further includes an image acquisition unit, wherein the image acquisition unit is configured to acquire an original photo of the pointer type instrument panel; and the industrial gateway is further configured to input the original photo into the trained picture model to extract a first feature map after convolution and down-sampling processing of the original photo, obtain a super high resolution image after sampling and convolution processing of the first feature map, and acquire pointer scale data of the pointer type instrument panel from the super high resolution image as the original operation data.

[0010] In an optional mode, the loss function of the trained picture model is determined by a first loss function and a second loss function, wherein the first loss function L sr is: wherein N is the number of pixels, y i is a super high resolution image, is a reference image.

[0011] The second loss function L det is:

[0012]

[0013] wherein λ coord is a loss weight, S 2 is the number of grids, B is the number of candidate boxes generated by each grid, is whether the i-th grid j-th candidate box is responsible for detecting the target, responsible for taking 1, otherwise taking 0, if the pre-selected frame is the maximum intersection over union value with the real box of a certain target in all pre-selected frames, it is responsible, otherwise it is not responsible, the value is opposite to the value, x, y, w, h represent the center coordinates, width and height of the predicted bounding box respectively, represent the real label values corresponding to x, y, w, h respectively, C represents the confidence, which is the probability of whether there is a target in a certain grid, represents the real value corresponding to C, P represents the possibility that the target in the candidate box belongs to the category C, represents the real value corresponding to P; the loss function L sum of the trained picture model is: L sum = Lsr + βL det wherein, a and β are weights of the first loss function and the second loss function, respectively.

[0014] According to another aspect of the embodiments of the present application, a method for monitoring an operation state of an industrial device is provided. The method comprises: continuously acquiring original operation data of the industrial device, and obtaining first operation data by performing data cleaning processing on the original operation data; temporarily storing the first operation data to a first temporary storage unit, so that the first temporary storage unit transmits and stores the first operation data to a permanent storage unit, wherein the first operation data stored in the permanent storage unit is historical data; training an original model using the historical data to obtain a trained model, wherein the trained model is published as a data monitoring task to monitor whether the operation state of the industrial device is abnormal; acquiring the first operation data from the first temporary storage unit to execute the data monitoring task; if a result of executing the data monitoring task is that the first operation data is abnormal, marking the first operation data and obtaining second operation data; temporarily storing the second operation data to a second temporary storage unit, so that the second temporary storage unit transmits and stores the second operation data to the permanent storage unit.

[0015] In an optional manner, the method further comprises: acquiring the second operation data from the second temporary storage unit; generating an alarm information when the second operation data meets a preset alarm rule; and displaying the first operation data, the second operation data and / or the alarm information.

[0016] According to another aspect of the embodiments of the present application, a monitoring device for monitoring an operation state of an industrial device is provided. The monitoring device comprises: a processor, a memory, a communication interface and a communication bus, the processor, the memory and the communication interface complete communication with each other through the communication bus; the memory is used to store at least one executable instruction, and the executable instruction causes the processor to perform operations corresponding to the above-mentioned method for monitoring an operation state of an industrial device.

[0017] According to still another aspect of the embodiments of the present application, a computer readable storage medium is provided. The storage medium stores executable instructions. The executable instructions cause a monitoring device for monitoring an operation state of an industrial device to perform operations corresponding to the above-mentioned method for monitoring an operation state of an industrial device.

[0018] In the embodiments of the present application, the original operation data of the industrial equipment is continuously acquired through the industrial gateway, and the operation state of the industrial equipment can be monitored in real time by monitoring the operation data in real time. The first operation data is obtained by performing data cleaning processing on the original operation data. By performing data cleaning processing on the original operation data, redundant data can be removed, so that the first operation data is transmitted to the first message node of the Kafka subsystem. Since the redundant original operation data is removed, the amount of data transmitted is reduced, thereby improving the transmission speed and the efficiency of monitoring the operation state of the industrial equipment. The first message node transmits the first operation data to the storage unit for storage. The data stored in the storage unit is referred to as historical data. The big data analysis modeling platform obtains the historical data from the storage unit to train the original model, thereby obtaining a trained model. The trained model can publish a data monitoring task. The big data analysis modeling platform obtains the first operation data from the first message node according to the data monitoring task to execute the data monitoring task, thereby monitoring the operation state of the industrial equipment. If the first operation data is abnormal, the first operation data is marked and second operation data is obtained, and the second operation data is transmitted to the second message node. By using different message nodes to transmit the first operation data and the second operation data, the first operation data and the second operation data can be distinguished, that is, the first operation data can be directly obtained from the first message node for monitoring to monitor the operation state of the industrial equipment, thereby improving the efficiency of monitoring the operation state of the industrial equipment, and the marked second operation data stored separately can also be directly obtained for use. Further, the monitoring system can also be used to monitor multiple different types of operation data at the same time, that is, at least two industrial equipment can be monitored at the same time, and / or multiple different types of operation data of one industrial equipment can be monitored at the same time. At this time, the asynchronous processing feature of the Kafka subsystem is used, that is, different types of first operation data are transmitted to different topics in the first message node, and the trained model corresponding to each type of operation data is a consumer. Different consumers can asynchronously obtain the first operation data in different topics, that is, the purpose of simultaneously monitoring the industrial equipment is achieved. Compared with synchronously obtaining different types of first operation data for monitoring, the efficiency of monitoring is greatly improved. After storing the first operation data and the second operation data in the storage unit, the user can view the operation data as needed to obtain the operation state, and then abnormal processing can be performed in a timely manner.

[0019] The above description is only a summary of the technical solutions of the embodiments of the present application. In order to more clearly understand the technical means of the embodiments of the present application, the embodiments of the present application can be implemented in accordance with the content of the description, and in order to make the above and other purposes, features and advantages of the embodiments of the present application more obvious and easy to understand, the specific embodiments of the present application are described below. BRIEF DESCRIPTION OF DRAWINGS

[0020] The accompanying drawings are included to provide a further understanding of the application, and are incorporated in and constitute a part of this specification. The drawings illustrate embodiments of the application and, together with the description, serve to explain the principles of the application. In the drawings:

[0021] Figure 1 A structural diagram of a monitoring system of an industrial equipment running state provided by an embodiment of the application is shown;

[0022] Figure 2 A structural diagram of a monitoring system of an industrial equipment running state provided by another embodiment of the application is shown;

[0023] Figure 3 A structural diagram of a monitoring system of an industrial equipment running state provided by still another embodiment of the application is shown;

[0024] Figure 4 A structural diagram of a monitoring system of an industrial equipment running state provided by yet another embodiment of the application is shown;

[0025] Figure 5 A flowchart of a monitoring method of an industrial equipment running state provided by an embodiment of the application is shown;

[0026] Figure 6 A structural diagram of a monitoring device of an industrial equipment running state provided by an embodiment of the application is shown. DETAILED DESCRIPTION

[0027] Exemplary embodiments of the present application will be described herein below with reference to the accompanying drawings. While exemplary embodiments of the present application are illustrated, it is to be understood that the application is not limited to the embodiments described herein, which can be implemented in various forms.

[0028] At present, industrial equipment is widely used in automatic production. If the abnormal running state of the industrial equipment cannot be acquired in time, the abnormal processing cannot be performed in time, thereby causing certain loss. For example, in the field of surface mounted technology (SMT), the SMT intelligent storage equipment can accurately realize automatic material storage and arrival at the production line according to the real-time demand of the production line, and the tail material of the production line can be automatically counted back to the warehouse without personnel counting. Therefore, the SMT intelligent storage equipment becomes a key equipment in the production process. Therefore, when the SMT intelligent storage equipment runs abnormally, it will have a serious impact on production and manufacturing. However, if the running state of the SMT intelligent storage equipment can be monitored, the abnormal running of the equipment can be found in time, and the abnormal processing can be performed in time, thereby reducing the loss caused by the abnormal running of the equipment.

[0029] The inventors of this application have noticed that in the prior art, the operating status of industrial equipment is usually monitored by manual inspection. However, this monitoring method cannot monitor industrial equipment in real time and is inefficient. If the operating status of industrial equipment could be monitored in real time, abnormal operating conditions of industrial equipment could be detected in a timely manner, and abnormalities could be handled promptly, reducing losses caused by abnormal operation of industrial equipment.

[0030] Based on the above considerations, in order to efficiently monitor the operating status of industrial equipment in real time, the inventors of this application, after in-depth research, proposed an industrial equipment operating status monitoring system. This system includes an industrial gateway, a Kafka subsystem, a storage unit, and a Data Analytics (DA) platform. The industrial gateway continuously acquires raw operating data from the industrial equipment. After data cleaning and processing, first operating data is obtained and transmitted to the Kafka subsystem. The Kafka subsystem then stores this first operating data in the storage unit, where the first operating data stored is historical data. The Data Analytics platform acquires and utilizes this historical data to train an initial model, which is then deployed as a data monitoring task. The Data Analytics platform also retrieves the first operating data from the Kafka subsystem to execute the data monitoring task, thereby monitoring the operating status of the industrial equipment. If the monitoring detects an anomaly in the first operating data, it marks the first operating data and obtains second operating data, which is then stored in the storage unit for user querying. This allows for timely detection of abnormal operating statuses of the industrial equipment, enabling timely anomaly handling and minimizing losses. This system continuously collects raw operating data from industrial equipment, trains a raw model using historical data, issues data monitoring tasks using the trained model, and then acquires and monitors the operating data of the industrial equipment based on the data monitoring tasks, thereby enabling real-time monitoring of the operating status of the industrial equipment.

[0031] Figure 1 A structural diagram of an industrial equipment operating status monitoring system provided in an embodiment of this application is shown. This system is used to monitor the operating status of at least one piece of industrial equipment. Figure 1 As shown, the industrial equipment operation status monitoring system 100 includes: an industrial gateway 101, a Kafka subsystem 102, a storage unit 103, and a big data analysis and modeling platform 104.

[0032] The industrial gateway 101 is used to continuously acquire the raw operating data of industrial equipment, and after cleaning the raw operating data, obtain the first operating data, and then transmit the first operating data to the Kafka subsystem 102. If the system 100 monitors at least two industrial devices, the industrial gateway 101 is also used to summarize all the raw operating data of all industrial devices.

[0033] The gateway is used to connect two or more heterogeneous networks to enable them to communicate with each other. The industrial gateway integrates heterogeneous sensors into the industrial internet, converts protocols between various bus protocols in the industrial field and required communication protocols, and realizes data interaction between different protocols.

[0034] The industrial equipment mainly refers to industrial production equipment, including various machine tools directly used for production, such as lathes, milling machines, grinding machines, planers and other machines, and other equipment serving production, such as warehouse equipment, workshop transportation equipment, etc. The original running data refers to data capable of monitoring the running state of the industrial equipment, for example, for the SMT intelligent warehouse equipment, the air valve is a key component of the equipment, and the air valve instrument panel displays the running state of the equipment in real time, so the original running data of the SMT intelligent warehouse equipment includes the air valve instrument panel data.

[0035] Data cleaning refers to the process of re-examining and verifying data, aiming to delete duplicate information and process redundant data, thereby reducing network transmission load and improving data transmission efficiency.

[0036] The Kafka system is a high-throughput distributed publish-subscribe message system, which can process all action flow data of consumers in a website. The Kafka system includes: a producer, i.e. a message or data generator, which can publish messages to a topic of the Kafka system; a consumer, i.e. a message or data consumer, which can subscribe to a topic and process its published consumption; a consumer group, which includes one or more consumers; a topic, which is a kind of message, a directory for storing messages, and a Kafka cluster can simultaneously manage the distribution of multiple topics; a partition, which is a physical grouping of a topic, a topic can be divided into multiple partitions, and each partition is an ordered queue; a node, i.e. a Kafka node or server, is a broker, and multiple brokers can form a Kafka cluster; an offset, each partition is a sequential and immutable message queue, and can be continuously added, the messages in the partition are assigned a sequence number, called offset, and the offset in each partition is unique. In the Kafka system, each partition can only be consumed by one consumer in the same consumer group, but multiple consumer groups of different consumer groups can simultaneously consume the partition. It should be understood that in the embodiments of the present application, the consumer and the producer can be modules in an application.

[0037] Specifically, the industrial gateway can be connected to sensors, programmable logic controllers (PLCs), etc. to collect raw operation data of each industrial equipment being monitored. The collection frequency of the raw operation data of the industrial equipment can be set as needed, as long as the running state of the industrial equipment can be monitored for abnormalities. The specific collection frequency is not limited here. A data stream processing model can be deployed in the industrial gateway to clean the raw operation data, and the first operation data can be obtained after the raw operation data is cleaned. The data cleaning method is a prior art and is not described here. Then, the industrial gateway transmits the first operation data to the Kafka subsystem according to a transmission protocol, for example, according to the message queuing telemetry transport (MQTT) protocol. If the monitoring system is used to monitor the running state of at least two industrial equipments, the industrial gateway is further used to aggregate all raw operation data of all industrial equipments.

[0038] The Kafka subsystem 102 includes a first message node and a second message node, wherein the first message node is used to receive the first operation data and transmit the first operation data to the storage unit 103.

[0039] One message node is one broker, that is, the first message node and the second message node are two different brokers. The storage unit 103 is a consumer in the first consumer group of the Kafka subsystem, which can consume the first operation data. If there are multiple types of operation data in an industrial equipment that need to be monitored or multiple industrial equipments that need to be monitored, different topics in the first message node of the Kafka subsystem are used to receive the first operation data of each type or different industrial equipment. For example, three types of operation data A, B, and C need to be monitored, and the first message node of the Kafka subsystem includes three different topics, namely topic1, topic2, and topic3. Topic1 is used to receive the first operation data of type A, topic2 is used to receive the first operation data of type B, and topic3 is used to receive the first operation data of type C. The storage unit 103 can be a time series database (TSDB).

[0040] The storage unit 103 is used to store the first operation data, wherein the first operation data stored in the storage unit 103 is historical data.

[0041] The storage unit 103 can consume the first running data, i.e., acquire and store the first running data, as a consumer in a first consumer group of the Kafka subsystem.

[0042] The big data analysis modeling platform 104 is configured to acquire the historical data from the storage unit 103 and train the original model to obtain a trained model, and publish the trained model as a data monitoring task to monitor whether the running state of the industrial equipment is abnormal. The big data analysis modeling platform 104 is also configured to acquire the first running data from the first message node of the Kafka subsystem 102 to execute the data monitoring task, and if the result of executing the data monitoring task is that the first running data is abnormal, mark the first running data to obtain second running data, and transmit the second running data to the second message node of the Kafka subsystem 102.

[0043] The big data analysis modeling platform can obtain the trained model by training the original model using the data, so that the trained model can be published as a task. The big data analysis modeling platform 104, as a second consumer group of the Kafka subsystem, can consume the first running data in the first message node at the same time as the first consumer group. The trained model can publish a data monitoring task, for example, a task of monitoring the first running data in real time, or a task of monitoring the first running data at a certain time period or at a certain time. The big data analysis modeling platform 104 consumes the first running data according to the task published by the trained model, i.e., acquires the corresponding first running data at the corresponding time for monitoring, and analyzes and judges whether the running state of the industrial equipment is abnormal. If the first running data is abnormal, the first running data is marked, and second running data is obtained, so as to distinguish the first running data corresponding to the normal running state from the first running data corresponding to the abnormal running state. Then the second running data is transmitted to the second message node of the Kafka.

[0044] Specifically, the storage unit 103 and the big data analysis modeling platform 104 are respectively different Consumers in two different Consumer Groups of the Kafka subsystem, and thus can simultaneously consume the first running data in the first message node. If multiple different types of running data need to be monitored, the original model is trained by using the first running data respectively, so as to obtain different trained models, and then each trained model can correspondingly publish a task. It should be noted that the different trained models are different Consumers in a second Consumer Group of the Kafka subsystem. For example, if three different types of running data A, B and C need to be monitored, the original model is trained by using the three types of historical data respectively, so as to obtain trained model A, trained model B and trained model C, and then trained model A can publish a task of monitoring data A, trained model B can publish a task of monitoring data B, and trained model C can publish a task of monitoring data C. Trained model A is Consumer 1 in the second Consumer Group, trained model B is Consumer 2, and trained model C is Consumer 3. Thus, according to the tasks published by the trained models, Consumers 1, 2 and 3 can asynchronously acquire the first running data of the corresponding type for monitoring, and determine whether the running state of the industrial equipment is abnormal. If the result of performing the data monitoring task is that the first running data is abnormal, the first running data is marked, and second running data is obtained, and then the second running data is transmitted to the second message node of the Kafka subsystem, wherein the method of marking the data is a prior art, which is not described here.

[0045] Since the storage unit 103 and the big data analysis modeling platform 104 are respectively different Consumers in two different Consumer Groups of the Kafka subsystem, the storage unit 103 and the big data analysis modeling platform 104 can simultaneously consume the first running data in the first message node, that is, the storage unit 103 and the big data analysis modeling platform 104 can asynchronously acquire the first running data, which greatly improves the efficiency of acquiring the first running data, and thus improves the efficiency of monitoring whether the running state of the industrial equipment is abnormal. Further, different Consumers in the second Consumer Group can asynchronously consume data of different topics when simultaneously monitoring different types of first running data, which improves the efficiency of acquiring the first running data compared with synchronously acquiring the first running data, and thus improves the efficiency of monitoring the running state of the industrial equipment.

[0046] The second message node of the Kafka subsystem 102 is configured to transmit the second running data into the storage unit 103.

[0047] The second running data is obtained by marking the first running data after determining that the first running data is abnormal, and thus the second running data is transmitted into the storage unit 103 through the second message node of the Kafka subsystem, that is, the first running data and the second running data are transmitted separately through different message nodes, so that when the first running data is acquired for judgment in the big data analysis modeling platform 104, the first running data can be quickly acquired from the first message node, and compared with the case that the first running data and the second running data are all transmitted through the same message node, the first running data can be quickly and accurately acquired, thereby improving the efficiency of monitoring the running state of the industrial equipment.

[0048] In the embodiment of the present application, the original operation data of the industrial equipment is continuously acquired through the industrial gateway, and the operation state of the industrial equipment can be monitored in real time by monitoring the operation data in real time. The first operation data is obtained by performing data cleaning processing on the original operation data. By performing data cleaning processing on the original operation data, redundant data can be removed, and then the first operation data is transmitted to the first message node of the Kafka subsystem. Since the redundant original operation data is removed, the amount of data transmitted is reduced, thereby improving the transmission speed and the efficiency of monitoring the operation state of the industrial equipment. The first message node transmits the first operation data to the storage unit for storage. The data stored in the storage unit is referred to as historical data. The big data analysis modeling platform obtains the historical data from the storage unit to train the original model, thereby obtaining a trained model. The trained model can then publish a data monitoring task. The big data analysis modeling platform obtains the first operation data from the first message node according to the data monitoring task to execute the data monitoring task, thereby monitoring the operation state of the industrial equipment. If the first operation data is abnormal, the first operation data is marked and second operation data is obtained, and then the second operation data is transmitted to the second message node. By using different message nodes to transmit the first operation data and the second operation data, the first operation data and the second operation data can be distinguished, that is, the first operation data can be directly obtained from the first message node for monitoring to monitor the operation state of the industrial equipment, thereby improving the efficiency of monitoring the operation state of the industrial equipment, and the separately stored second operation data can also be directly obtained for use. Further, the monitoring system can also be used to monitor multiple different types of operation data at the same time, that is, at least two industrial equipment can be monitored at the same time, and / or multiple different types of operation data of one industrial equipment can be monitored at the same time. At this time, the asynchronous processing feature of the Kafka subsystem is used, that is, different types of first operation data are transmitted to different topics in the first message node, and the trained model corresponding to each type of operation data is a consumer. Different consumers can asynchronously obtain the first operation data in different topics, thereby achieving the purpose of monitoring the industrial equipment at the same time. Compared with monitoring different types of first operation data synchronously, the efficiency of monitoring is greatly improved. After storing the first operation data and the second operation data in the storage unit, the user can view the operation data as needed to obtain the operation state, and then abnormal processing can be performed in a timely manner.

[0049] When the operation state of the industrial equipment is abnormal, Figure 2 The structure diagram of the monitoring system for the operation state of the industrial equipment provided by another embodiment of the present application is shown. The embodiment of the present application adds an alarm unit on the basis of the above-mentioned embodiment. As Figure 2As shown, the monitoring system 100 further comprises an alarm unit 105.

[0050] The second message node of the Kafka subsystem 102 is further configured to transmit the second running data to the alarm unit 105.

[0051] Specifically, the alarm unit 105 is a consumer in a third consumer group of the Kafka subsystem, i.e., the alarm unit and the storage unit 103 belong to different consumers in different consumer groups, and thus the alarm unit and the storage unit 103 can asynchronously acquire the second running data.

[0052] The alarm unit 105 is configured to generate alarm information when the second running data meets a preset alarm rule.

[0053] The alarm rule can be set as needed. For example, since the industrial equipment can have unstable network and the like, the alarm information can be generated when three second running data are continuously acquired, i.e., the alarm information is generated when three first running data are continuously monitored to be abnormal, so as to reduce misjudgment.

[0054] In the embodiment of the application, the alarm unit consumes the second running data, and when the second running data meets the preset alarm rule, i.e., when the abnormal operation of the industrial equipment is monitored and the alarm rule is met, the alarm information is generated in time to remind the user to handle the abnormality in time and reduce the loss.

[0055] In order to acquire the running information of the industrial equipment, so as to formulate a maintenance scheme and the like, Figure 3 A structure diagram of a monitoring system of an industrial equipment running state provided by another embodiment of the application is shown, and the embodiment of the application is based on the embodiment shown in Figure 2 A Prognostics and Health Management (PHM) subsystem is added to the embodiment shown. As shown in Figure 3 As shown, the monitoring system 100 further comprises a fault prediction and health management subsystem 106.

[0056] The storage unit 103 is further configured to transmit the first running data and the second running data to the fault prediction and health management subsystem 106.

[0057] The PHM subsystem is a system engineering that realizes detection, prediction and management of the health state of a complex industrial system by using various data generated in an industrial system, through signal processing and data analysis and the like operation means. The PHM technology changes the health management of the equipment from the traditional fault management to the degradation management, and realizes zero downtime and continuous reliable operation of the equipment through predictive maintenance.

[0058] Specifically, the storage unit transmits all the stored first operation data and second operation data to the PHM subsystem, so that the PHM subsystem analyzes the health status of the industrial equipment.

[0059] The alarm unit 105 is further configured to transmit the alarm information to the PHM subsystem 106.

[0060] The alarm information is information reflecting the abnormal operation of the industrial equipment, and can include the second operation data, i.e., the abnormal operation data, the generation time of the alarm information, etc. Since the PHM subsystem can display information, the alarm information is transmitted to the PHM subsystem through the alarm unit, so that the PHM subsystem displays the alarm information, and the user can query the alarm information.

[0061] The PHM subsystem 106 is configured to receive the first operation data, the second operation data and the alarm information, and to generate and display the health record of the industrial equipment according to the first operation data and the second operation data, and to display the alarm information.

[0062] The health record is a normative and scientific record of the health management process of the industrial equipment. For example, the health record can include one or more pieces of health data of the industrial equipment, and the health data of the industrial equipment can include the first operation data, the second operation data, the change trend of the first operation data, the change trend of the second operation data, etc. Each health record belongs to only one industrial equipment, and one industrial equipment can have multiple health records.

[0063] Specifically, the PHM subsystem receives the first operation data and the second operation data sent by the storage unit, generates the health record of the corresponding industrial equipment using the first operation data and the second operation data, and displays the health record. The PHM subsystem also receives the alarm information sent by the alarm unit, and displays the alarm information.

[0064] In the embodiments of the present application, the PHM subsystem generates the health record according to the operation data of the industrial equipment, so that the user can formulate a maintenance plan for the industrial equipment according to the health record. When the industrial equipment is abnormal, the user can quickly obtain the operation parameters of the industrial equipment from the health record, so as to quickly formulate a relevant maintenance plan, greatly saving the maintenance time and reducing the loss. When the industrial equipment is abnormal, the alarm information is generated and displayed, so that the user can obtain the abnormal information in time, and then handle the abnormality in time, thereby reducing the loss.

[0065] When new industrial equipment is added and the new industrial equipment needs to be monitored, in order to quickly realize monitoring of the new industrial equipment on the basis of monitoring of the existing industrial equipment, in the embodiment of the application, the PHM subsystem 106 is further configured to, in response to access of the new industrial equipment, determine an alarm rule of the new industrial equipment according to first running data of the new industrial equipment and expert experience data.

[0066] The expert experience data refers to data formed by analysis of the industrial equipment by relying on experience and knowledge of experts.

[0067] Specifically, when the monitoring system 100 accesses the new industrial equipment, the industrial gateway obtains original running data of the new industrial equipment and transmits the original running data to the storage unit via the Kafka subsystem, and then the PHM subsystem obtains the first running data from the storage unit and determines the alarm rule in combination with the first running data and the expert experience data. For example, after obtaining the first running parameter and determining the data type of the running parameter, it is found by searching the expert experience data that the normal range of the running parameter is 70-80, and then the alarm rule is determined as alarm when the running parameter is less than 70 or greater than 80, thereby realizing monitoring of the running state of the industrial equipment.

[0068] In the embodiment of the application, for the new industrial equipment, the PHM subsystem determines the alarm rule according to the expert experience data, thereby realizing monitoring of the running state of the new industrial equipment. Further, for the monitoring system 100, since the monitoring system includes the PHM subsystem 106, the new industrial equipment can be further monitored on the basis of monitoring of the existing industrial equipment, thereby expanding the practicality of the monitoring system, and compared with other monitoring systems that can only monitor a single industrial equipment, the monitoring cost is reduced.

[0069] Since some industrial equipment includes a pointer type instrument panel and the running parameter of the instrument panel needs to be monitored, in order to accurately read the parameter of the pointer type instrument panel, Figure 4 Fig. 2 shows a structure diagram of a monitoring system of a running state of industrial equipment according to another embodiment of the application. The embodiment of the application is based on the embodiment shown in Fig. 1 and adds an image acquisition unit 107. Figure 1 As shown in Fig. 2, the monitoring system 100 further includes the image acquisition unit 107. Figure 4 As shown in Fig. 2, the monitoring system 100 further includes the image acquisition unit 107.

[0070] The image acquisition unit 107 is configured to obtain an original photo of the pointer type instrument panel.

[0071] The image acquisition unit 107 can be an industrial camera. For the pointer instrument panel data, if the pointer changes rapidly and cannot be identified by the naked eye, the pointer data of the instrument panel can be read by taking a photo of the instrument panel. For example, the air source pressure instrument panel of the SMT intelligent warehousing equipment changes during the working process of the material receiving feeder. At the moment of starting the material receiving feeder, the pointer of the instrument panel changes greatly, that is, the air valve pointer changes by 270 degrees within milliseconds, which is difficult to capture by the naked eye, so that the on-site operator cannot perceive the equipment anomaly according to the change of the air valve pointer. Therefore, the industrial camera can be used to take a photo to obtain an original photo, and then the pointer scale data of the instrument panel can be read according to the original photo.

[0072] Specifically, the pointer scale position of the pointer instrument panel can be photographed by the industrial camera, so as to obtain an original photo of the pointer instrument panel, and then the pointer scale data can be read according to the original photo.

[0073] The industrial gateway 101 is also configured to input the original photo into the trained picture model to perform convolution and down-sampling processing on the original photo to extract a first feature map, perform sampling and convolution processing on the first feature map to obtain a super high-resolution image, and obtain the pointer scale data of the pointer instrument panel as the original operation data according to the super high-resolution image.

[0074] The image resolution is a group of performance parameters for evaluating the richness of the details contained in the image, including time resolution, spatial resolution, and color scale resolution, etc., which reflects the ability of the imaging system to reflect the details of the object. Compared with low-resolution images, super high-resolution images usually contain greater pixel density, richer texture details, and higher reliability. Usually, a specific algorithm and processing flow are used to reconstruct a super high-resolution image from a low-resolution image obtained, so as to overcome or compensate for the problems of blurred imaging, low quality, and non-significant regions of interest caused by the limitations of the image acquisition system or the acquisition environment itself.

[0075] Specifically, the original photo is input into the trained picture model to obtain a super high-resolution image, and then the pointer scale data is read according to the super high-resolution image, which is the original operation data of the instrument panel. The original photo is processed by a series of convolution and down-sampling of the backbone network in the trained picture model to extract a first feature map, and then a series of sampling and convolution processing is performed on the first feature map to obtain a super high-resolution image. The backbone network can refer to the design of the residual network, that is, the DarkNet53 structure is adopted, there are 53 convolution layers, the convolution kernel size is 3*3 and 1*1, there are 5 times of compression, and finally a feature map with a size of 1 / 32 of the input picture size is obtained. Compared with the traditional residual network, the detection speed is improved by 2 times under the condition of ensuring the detection accuracy.

[0076] The picture network model follows a YOLO (you only look once) v3 (version number, representing the third edition) detection structure, detects at three scales, uses a non-maximum suppression algorithm to obtain the final detection result, and alternately performs model sampling and feature fusion, and up-sampling is performed through transposed convolution. YOLO is a deep neural network-based object recognition and positioning algorithm, and its biggest feature is fast running speed, which can be used in real-time systems. YOLOv3 is the third edition of the YOLO series target detection algorithm, and compared with previous algorithms, the precision has been significantly improved, especially for small targets.

[0077] In the embodiments of the application, the original photo of the instrument panel is acquired by using the image acquisition unit, and the picture model for reading the pointer data in the original photo is deployed on the industrial gateway, so that the industrial gateway can accurately read the pointer scale data of the instrument panel according to the original photo, that is, the monitoring system is suitable for monitoring industrial equipment including a pointer instrument panel, thereby expanding the application range.

[0078] In order to improve the accuracy of reading the pointer scale data of the pointer instrument panel, in the embodiments of the application, the loss function of the trained picture model is determined by a first loss function and a second loss function. The first loss function L sr is:

[0079]

[0080] Wherein, N is the number of pixels, y i is a reconstructed high-resolution image, is a reference image;

[0081] The second loss function is:

[0082]

[0083] Wherein, λ coord is a loss weight, S 2 is the number of grids, B is the number of candidate boxes generated by each grid, is whether the i-th grid j-th candidate box is responsible for detecting the target, responsible for taking 1, otherwise taking 0, if the pre-selected box is the maximum intersection over union value with the real box of a certain target among all pre-selected boxes, it is responsible, otherwise it is not responsible, The value of is opposite to the value of x, y, w, h respectively represents the center coordinates, width and height of the predicted bounding box, The value of is opposite to the value of x, y, w, h respectively represents the center coordinates, width and height of the predicted bounding box, represents the real label value corresponding to x, y, w, h respectively, represents the confidence, which is the probability of whether there is a target in a certain grid, represents the real value corresponding to C, P represents the possibility that the target in the candidate box belongs to the category C, represents a true value corresponding to P;

[0084] The loss function of the trained picture model is:

[0085] L sum = aL sr + bL det ;

[0086] Wherein, a and b are the weights of the first loss function and the second loss function respectively, which can be set as needed.

[0087] In the embodiments of the application, the loss function of the picture model is determined according to the first loss function and the second loss function after determining the first loss function and the second loss function respectively, so as to improve the accuracy of the model, and further improve the accuracy of reading the pointer scale data.

[0088] Due to the inconsistency of the detection scale and the cross repeatability of the candidate frame, the algorithm may generate multiple prediction frames for the same target. Therefore, in some embodiments, in order to remove redundant boundary frames, a Non Maximum Suppression (NMS) algorithm is further used to process the redundant boundary frames to ensure the accuracy of the final detection result.

[0089] Figure 5 The flowchart of the method for monitoring the running state of the industrial equipment provided by the embodiments of the application is shown. As shown in Figure 5 The method comprises the following steps:

[0090] Step 110: continuously acquiring the original running data of the industrial equipment, and obtaining the first running data after data cleaning processing of the original running data.

[0091] Step 120: temporarily storing the first running data to the first temporary storage unit, so that the first temporary storage unit transmits and stores the first running data to the permanent storage unit, wherein the first running data stored in the permanent storage unit is historical data.

[0092] Step 130: training the original model using the historical data to obtain a trained model, wherein the trained model is published as a data monitoring task to monitor whether the running state of the industrial equipment is abnormal.

[0093] Step 140: acquiring the first running data from the first temporary storage unit to execute the data monitoring task.

[0094] Step 150: judging whether the result of executing the data monitoring task is the first running data abnormal.

[0095] If yes, go to step 160; if no, go to step 110.

[0096] Step 160: Mark the first running data and obtain the second running data.

[0097] Step 170: Temporarily store the second running data to the second temporary storage unit, so that the second temporary storage unit transmits and stores the second running data to the permanent storage unit.

[0098] The monitoring method of the industrial equipment running state corresponds to the monitoring system of the industrial equipment running state, so the specific implementation mode of steps 110-170 can refer to the implementation mode of the above-mentioned system embodiment, which will not be described here.

[0099] In the embodiment of the application, the original running data of the industrial equipment is continuously obtained through the industrial gateway, so that the running state of the industrial equipment can be monitored in real time by monitoring the running data in real time. The first running data is obtained by performing data cleaning processing on the original running data. By performing data cleaning processing on the original running data, redundant data can be removed, so that the first running data is transmitted and stored in the first temporary storage unit. Since the repeated original running data is removed, the amount of data transmitted is reduced, thereby improving the transmission speed and further improving the efficiency of monitoring the running state of the industrial equipment. The first temporary storage unit transmits the first running data to the permanent storage unit for storage. The data stored in the permanent storage unit is called historical data. The historical data is obtained from the permanent storage unit to train the original model, thereby obtaining a trained model. The trained model is then published as a data monitoring task to monitor whether the running state of the industrial equipment is abnormal. According to the published data monitoring task, the first running data is obtained from the first temporary storage unit to execute the data monitoring task, thereby monitoring the running state of the industrial equipment. If the first running data is abnormal, the first running data is marked and the second running data is obtained, and the second running data is transmitted and stored in the second temporary storage unit. By using different temporary storage units to transmit and temporarily store the first running data and the second running data, the first running data and the second running data can be distinguished, that is, the first running data can be directly obtained from the first temporary storage unit for monitoring to monitor the running state of the industrial equipment, thereby improving the efficiency of obtaining the first running data and improving the efficiency of monitoring the running state of the industrial equipment. After storing the first running data and the second running data in the permanent storage unit, the user can view the running data as needed to obtain the running state, and further can timely perform abnormal processing.

[0100] When the abnormality of the industrial equipment is monitored, in order to enable the user to obtain the abnormality information in time so as to perform the abnormality processing, the embodiment of the present application provides a way of generating the alarm information when the abnormality of the industrial equipment is monitored. The embodiment of the present application is based on the embodiment shown in Figure 5 The embodiment of the present application adds the following steps:

[0101] Step a1: obtaining the second running data from the second temporary storage unit.

[0102] Step a2: generating the alarm information when the second running data meets the preset alarm rule.

[0103] Step a3: displaying the first running data, the second running data and / or the alarm information.

[0104] The monitoring method of the industrial equipment running state corresponds to the monitoring system of the industrial equipment running state, therefore, the specific implementation mode of steps a1-a3 can refer to the implementation mode of the system embodiment, which will not be repeated here.

[0105] In the embodiment of the present application, when the second running data meets the preset alarm rule, that is, when the abnormality of the industrial equipment is monitored and the alarm rule is met, the alarm information is generated in time to remind the user to perform the abnormality processing in time, thereby reducing the loss. Further, by displaying the first running data, the second running data and / or the alarm information, the user can query, so as to formulate the maintenance or maintenance plan according to the running data of the industrial equipment.

[0106] Figure 6 The structure schematic diagram of the monitoring device of the industrial equipment running state provided by the embodiment of the present application is shown, and the specific implementation of the monitoring device of the industrial equipment running state is not limited in the embodiment of the present application.

[0107] As shown in Figure 6 The monitoring device of the industrial equipment running state can include a processor 202, a communications interface 204, a memory 206, and a communications bus 208.

[0108] The processor 202, the communications interface 204, and the memory 206 can communicate with each other through the communications bus 208. The communications interface 204 is used for communicating with network elements such as clients or other servers. The processor 202 is used for executing the program 210, and specifically can execute the related steps in the above monitoring method of the industrial equipment running state.

[0109] Specifically, the program 210 can include program code, and the program code includes computer executable instructions.

[0110] The processor 202 can be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to perform one or more of the embodiments of the application. The one or more processors of the industrial equipment running state monitoring device can be of the same type, such as one or more CPUs, or of different types, such as one or more CPUs and one or more ASICs.

[0111] The memory 206 is configured to store a program 210. The memory 206 can include a high-speed RAM memory, and can also include a non-volatile memory, such as at least one disk memory.

[0112] The embodiments of the application provide a computer readable storage medium, the storage medium stores executable instructions, when the executable instructions are executed on the industrial equipment running state monitoring device, the industrial equipment running state monitoring device executes the industrial equipment running state monitoring method in any of the method embodiments.

Claims

1. A system for monitoring the operating state of an industrial plant, characterized in that The system comprises an industrial gateway, a Kafka subsystem, a storage unit and a big data analysis modeling platform, wherein, the industrial gateway is configured to continuously acquire raw operation data of industrial equipment, perform data cleaning on the raw operation data to obtain first operation data, and transmit the first operation data to the Kafka subsystem, wherein, if the system monitors at least two industrial equipment, the industrial gateway is further configured to aggregate all raw operation data of all industrial equipment; the Kafka subsystem comprises a first message node and a second message node, wherein, the first message node is configured to receive the first operation data and transmit the first operation data to the storage unit; the big data analysis modeling platform is configured to acquire the first operation data from the first message node of the Kafka subsystem to perform a data monitoring task, if the result of performing the data monitoring task is that the first operation data is abnormal, mark the first operation data to obtain second operation data, and transmit the second operation data to the second message node of the Kafka subsystem, wherein, the data monitoring task is a task published by a trained model; the second message node of the Kafka subsystem is configured to receive the second operation data and transmit the second operation data to the storage unit; the storage unit is configured to store the first operation data and the second operation data, wherein, the first operation data stored in the storage unit is historical data; the big data analysis modeling platform is further configured to acquire the historical data from the storage unit, and train an original model using the historical data to obtain the trained model; wherein, the industrial equipment comprises a pointer instrument panel, and the system further comprises an image acquisition unit, wherein, the image acquisition unit is configured to acquire a raw photo of the pointer instrument panel; the industrial gateway is further configured to input the raw photo into a trained picture model to extract a first feature map by performing convolution and down-sampling on the raw photo, obtain a super-high resolution image by performing sampling and convolution on the first feature map, and acquire pointer scale data of the pointer instrument panel from the super-high resolution image as the raw operation data; a loss function of the trained picture model is determined by a first loss function and a second loss function, wherein, the first loss function is: wherein, N is a number of pixels, is a super high resolution image, is a reference image; The second loss function is: ， in, To lose weight, For the number of grid cells, B The number of candidate boxes generated for each grid. For the first i The first grid j Each candidate bounding box is assigned a value of 1 to indicate whether it is responsible for detecting the target, and 0 otherwise. A candidate bounding box is considered responsible if it has the highest intersection-union ratio (IU) with the ground truth bounding box of a target among all candidate bounding boxes; otherwise, it is considered irresponsible. Values ​​and The values ​​are opposite. x , y , w , h These represent the center coordinates, width, and height of the predicted bounding box, respectively. , , , Indicates to respectively and x , y , w , h The corresponding real label value, C This represents the confidence level, which is the probability that a target exists in a given grid cell. Indicates and C The corresponding actual value, P This indicates that the target in the candidate box belongs to the category. C The possibility, Indicates and P The corresponding actual value; the loss function of the trained picture model is: wherein, , are weights of the first loss function and the second loss function, respectively.

2. The system of claim 1, wherein, the system further comprises an alarm unit, wherein, the second message node of the Kafka subsystem is further configured to transmit the second operation data to the alarm unit; the alarm unit is configured to generate alarm information when the second operation data meets a preset alarm rule.

3. The system of claim 2, wherein, The system further comprises a fault prediction and health management subsystem, wherein, the storage unit is further configured to transmit the first operation data and the second operation data to the fault prediction and health management subsystem; the alarm unit is further configured to transmit the alarm information to the fault prediction and health management subsystem; The failure prediction and health management subsystem is configured to receive the first operation data, the second operation data and the alarm information, generate and display a health profile of the industrial equipment according to the first operation data and the second operation data, and display the alarm information.

4. The system of claim 3, wherein, The industrial equipment monitored by the system includes the newly added industrial equipment, wherein, The failure prediction and health management subsystem is further configured to determine an alarm rule of the newly added industrial equipment according to the first operation data and expert experience data of the newly added industrial equipment in response to the access of the newly added industrial equipment.

5. A method of monitoring the operating state of an industrial plant, characterized in that The method comprises: acquiring an original photo of a pointer instrument panel of an industrial equipment; inputting the original photo into a trained picture model to extract a first feature map after convolution and down-sampling processing of the original photo, obtaining a super-high resolution image after sampling and convolution processing of the first feature map, and acquiring pointer scale data of the pointer instrument panel as original operation data according to the super-high resolution image; obtaining first operation data after data cleaning processing of the original operation data; temporarily storing the first operation data to a first temporary storage unit, so that the first temporary storage unit transmits and stores the first operation data to a permanent storage unit, wherein the first operation data stored in the permanent storage unit is historical data; training an original model using the historical data to obtain a trained model, wherein the trained model is published as a data monitoring task to monitor whether the operation state of the industrial equipment is abnormal; acquiring the first operation data from the first temporary storage unit to execute the data monitoring task; if the result of executing the data monitoring task is that the first operation data is abnormal, marking the first operation data and obtaining second operation data; temporarily storing the second operation data to a second temporary storage unit, so that the second temporary storage unit transmits and stores the second operation data to the permanent storage unit; wherein The loss function of the trained picture model is determined by a first loss function and a second loss function, wherein the first loss function is: wherein, N is the number of pixels, is a super high-resolution image, is a reference image; The second loss function is: ， in, To lose weight, For the number of grid cells, B The number of candidate boxes generated for each grid. For the first i The first grid j Each candidate bounding box is assigned a value of 1 to indicate whether it is responsible for detecting the target, and 0 otherwise. A candidate bounding box is considered responsible if it has the highest intersection-union ratio (IU) with the ground truth bounding box of a target among all candidate bounding boxes; otherwise, it is considered irresponsible. Values ​​and The values ​​are opposite. x , y , w , h These represent the center coordinates, width, and height of the predicted bounding box, respectively. , , , Indicates to respectively and x , y , w , h The corresponding real label value, C This represents the confidence level, which is the probability that a target exists in a given grid cell. Indicates and C The corresponding actual value, P This indicates that the target in the candidate box belongs to the category. C The possibility, Indicates and P The corresponding actual value; the loss function of the trained picture model is: wherein, , are weights of the first loss function and the second loss function, respectively.

6. The method of claim 5, wherein, the method further comprises: acquiring the second operation data from the second temporary storage unit; generating alarm information when the second operation data meets a preset alarm rule; displaying the first operation data, the second operation data and / or the alarm information.

7. A monitoring device of an operating state of an industrial plant, characterized by comprise: a processor, a memory, a communication interface and a communication bus, the processor, the memory and the communication interface complete communication with each other through the communication bus; the memory is configured to store at least one executable instruction, and the executable instruction causes the processor to execute the industrial equipment operation state monitoring method of claim 5 or 6.

8. A computer-readable storage medium, characterized in that, The storage medium stores at least one executable instruction, and the executable instruction causes the processor to execute the industrial equipment operation state monitoring method of claim 5 or 6.

Images