CN114048952B

CN114048952B - Iron-making plant safety situation sensing method based on edge internet of things technology and neural network

Info

Publication number: CN114048952B
Application number: CN202111194017.1A
Authority: CN
Inventors: 白久君; 李富春; 李梅娟; 陈雪波; 赵莹; 白梓辰; 白梓洋
Original assignee: University of Science and Technology Liaoning USTL
Current assignee: University of Science and Technology Liaoning USTL
Priority date: 2021-10-13
Filing date: 2021-10-13
Publication date: 2023-02-17
Anticipated expiration: 2041-10-13
Also published as: CN114048952A

Abstract

The invention provides an iron mill safety situation perception method based on an edge internet of things technology and a neural network, which comprises the following steps: s1: decomposing a production system and a character ring system of an iron-making plant into subsystems, and then coupling, isomerizing, fusing and pooling the decomposed two systems; s2: collecting data; s3: constructing an edge Internet of things platform; s4: predicting an artificial neural network; s5: and realizing the safety situation awareness of the iron works. The traditional iron-making process is decomposed into a molecular system, the molecular system is coupled with a 'human', 'object' and 'ring' system newly built in an iron-making plant, heterogeneous fusion among all system equipment is carried out through a gateway, real-time data of all production processes are collected and uploaded to a cloud to form a data pool which can be used in a butt joint mode, an artificial neural network is used for calculating and analyzing data of the data pool, and therefore perception prediction of the safe operation condition of the whole system is achieved.

Description

Iron-making plant safety situation sensing method based on edge internet of things technology and neural network

Technical Field

The invention relates to the technical field of industrial safety, in particular to a method for sensing the safety situation of an iron-making plant based on an edge internet of things technology and a neural network.

Background

Reviewing the history of industrial development, from industry 1.0 to industry 3.0, through mechanization, electrification and automation, the current industry 4.0 is intelligent, and the internet of things is the core key for realizing the industry 4.0. For the industry serving as national economic life and foundation, the application of the technology of the internet of things to the traditional industrial field is a mainstream direction of the current industrial development and is a revolution of the traditional industry. The emergence of technologies represented by the internet of things, cloud computing and edge computing pushes the traditional industry to move towards informatization and intellectualization, and under the background, data interconnection, sharing and fusion gradually become the mainstream trend of industrial development.

Taking the iron and steel industry as an example, in a traditional iron-making enterprise, daily production activities are carried out around a blast furnace, and the iron and steel industry also relates to associated systems such as a feeding system, a furnace top system, an air supply system, an iron slag processing system, a gas dust removal system and the like besides a blast furnace body system, the systems are mutually independent and only responsible for overall management of subsystems under the systems without mutual communication, and meanwhile, management and operating personnel of an iron-making plant are free from the systems, so that the whole safety perception is hardly formed from the change of the operation change of the systems of the existing iron-making process.

In addition, the traditional iron-making enterprises, as one of the enterprises with the highest industrial accident rate in the industrial field, have the advantages of complex production process, long process and multiple and wide major hazard sources. And because the personnel safety guarantee of the steel enterprises is closely related to the production process, the personnel in the steel enterprises can face the dangers of high kinetic energy, high potential energy and high thermal energy, and can face other dangers of toxicity, harmful substances, flammability, explosiveness and the like.

Therefore, the traditional iron and steel enterprises are improved by using the internet of things and the edge computing technology, and a set of security situation perception system is constructed on the basis of the improvement, which is necessary for the iron and steel enterprises.

Disclosure of Invention

In order to overcome the defects in the background art, the invention provides a method for sensing the safety situation of the iron-making plant based on an edge internet of things technology and a neural network, and designs a method for constructing a set of system for sensing the safety situation of the iron-making plant by applying the technologies such as edge internet of things, an artificial neural network and situation sensing under the traditional iron-making plant operation system. The traditional iron-making process is decomposed into a molecular system, the molecular system is coupled with a 'human', 'object' and 'ring' system newly built in an iron-making plant, heterogeneous fusion among system devices is carried out through a gateway, real-time data of each production process are collected and uploaded to a cloud to form a data pool which can be used in butt joint, and an artificial neural network is used for calculating and analyzing data of the data pool, so that perception prediction of the safe operation condition of the whole system is achieved.

In order to achieve the purpose, the invention adopts the following technical scheme:

a safety situation perception method for an iron mill based on an edge Internet of things technology and a neural network comprises the following steps:

s1: decomposing a production system and a character ring system of an iron-making plant into subsystems, and then coupling, isomerizing, fusing and pooling the decomposed two systems;

s2: data acquisition: establishing a gateway connection relation of the Internet of things of the iron works production system and the character ring system, acquiring data of each subsystem in the iron works production system and the character ring system by using various sensors, and storing the acquired data in a built storage server to form a data pool for acquiring a large amount of information;

s3: constructing an edge Internet of things platform;

s4: predicting an artificial neural network;

s5: and realizing the perception of the safety situation of the iron works.

Further, in step S1, the subsystems of the iron works production system include a blast furnace body system, a feeding and feeding system, a furnace top system, an air supply system, an iron slag treatment system, and a gas dust removal system; decomposing the ironworks production systems into their respective subsystems;

the subsystem of the figure ring system comprises a person system, an object system and a ring system, wherein the person system comprises a person identification subsystem, a person monitoring subsystem and a person positioning subsystem; the system of the object comprises a vehicle, material and facility subsystems; the system of the ring includes natural, production, energy consumption subsystems.

Further, in step S1, the coupling between the two systems is specifically: sharing all data of subsystems of the two systems, so as to achieve coordinated linkage, and when a safety accident happens to one system of the iron plant production system, responding to a 'human' system in a plant area, positioning personnel in a dangerous area, and prompting the personnel to leave the dangerous area quickly; meanwhile, the system about the 'things' responds to the evacuation of important materials and closes related facilities; the system about the "ring" isolates the hazardous area; the harmfulness is reduced to the minimum by coordinating and linking all the systems.

Further, in step S1, the pooling specifically is: and (3) separating hardware of each subsystem from the original system, uniformly incorporating the hardware into a resource pool of the factory, and performing control and resource combination by the cloud server.

Further, in step S1, the heterogeneous integration is to connect devices based on different protocols in the respective subsystems of the two systems to the platform of the internet of things through a unified protocol standard for centralized management.

Further, the construction of the edge internet of things platform in the step S3 is specifically as follows: upgrading the subsystems and sub-systems of the iron and steel plant production system and the character ring system into functional edge nodes in a mode of building an edge server and an edge service platform, wherein the edge nodes can longitudinally control the subsystems of the edge nodes and can also perform transverse information interaction with other edge nodes in an edge network mode; the edge nodes are unified at the cloud end to construct an edge internet of things service management system, so that a cloud-edge-end three-layer architecture is formed.

Further, the step S4 specifically includes the following steps:

s41: constructing a safety evaluation index system, namely determining an object to be evaluated and the content of the object to be evaluated;

s42: extracting data about the safety condition of each evaluation object stored in the data pool constructed in the step S2 through a data interface, extracting information influencing system safety such as danger sources, danger types and the like from the data, and determining the condition of the system to form a safety analysis table;

s43: calculating the safety analysis table through an evaluation algorithm, and taking the calculation result (namely the risk index h of each subsystem) as a basis for reflecting the risk degree of each subsystem;

s44: and (5) taking the calculation result of the S43 as expected output of the neural network, and realizing automatic evaluation and prediction of the risk index h of the subsystem through learning training of the neural network.

Further, the evaluation objects determined in S41 are: taking the subsystem as a first-class index and the subsystem as a second-class index as an evaluation object; the content of the evaluation comprises:

(1) "number of dangerous patterns", i.e. the number of types of possible safety incidents of the system;

(2) the second index is the number of defects of equipment and facilities;

(3) the third index is 'the number of defects of the operating environment';

(4) the fourth indicator is "length of hazardous exposure".

Further, S44 specifically includes:

the input nodes of the input layer of the neural network comprise six indexes of danger mode number, equipment defect number, environment defect number, danger exposure duration, intrinsic safety level and substance risk coefficient, six types of index data of all subsystems are intercepted once by taking one month as a time node, and an input matrix of a neural network training set is formed and used for training the neural network;

the output result of the output layer of the neural network is the risk index h of all the subsystems.

Further, the step S5 specifically includes the following steps:

(1) calculating the risk index H of the whole system, wherein the calculation formula is as follows:

h _i is the subsystem risk index, n _i As the number of subsystems, E _i Length of exposure for risk points;

(2) calculating an enterprise initial security risk level S (K-1): calculated according to the following formula:

wherein G is a constant and Y is the negative injury rate of thousands of people; the larger the injury rate of thousands of people is, the lower the initial safety risk level is;

(3) and (3) judging the enterprise risk management and control capability C: on the basis of percentage control, each system of the enterprise is checked according to the items on the enterprise risk control scoring table, so that the enterprise risk control capability is scored;

(4) calculating the control ability index of the iron works: the specific calculation formula is as follows:

b (k) = α C- β H α and β are constants, α =0.068, β =0.55

The management and control index is used for representing dynamic change of the enterprise risk level, when B (K) is larger than 0, the system risk level is rising, and when B (K) is smaller than 0, the system risk level is falling;

(5) and calculating the comprehensive security risk level S (k) of the enterprise, wherein the specific calculation formula is as follows:

S(K)＝S(K-1)+B(K)

(6) carrying out security risk level grading according to the enterprise comprehensive security risk level S (k), wherein the grading formula is as follows:

s1 and S2 are interval critical values of grade division, M is the number of actual staff, and P is the expected injury rate of thousands of people.

Compared with the prior art, the invention has the beneficial effects that:

1) The method provided by the invention can respectively carry out edge intelligent upgrade on the traditional iron-making process and newly added 'human', 'object' and 'ring' systems in a factory area, and the systems are coupled with each other by analyzing the relationship among the systems to form a new system, thereby realizing cooperative and unified management.

2) The method provided by the invention can be used for collecting, analyzing and preprocessing the log information of the iron works through the terminal edge service node and the edge gateway, monitoring the safety state data of each system of the iron works in real time, and acquiring, understanding, displaying and early warning the elements which can cause the safety accidents of the iron works, thereby effectively reducing the occurrence of casualty accidents and simultaneously reducing the economic loss and unnecessary maintenance cost brought by enterprise safety.

Drawings

FIG. 1 is a general flow diagram of the method of the present invention;

FIG. 2 is an exploded view of the iron mill production system of the present invention;

FIG. 3 is an exploded view of the human, object and ring system of the present invention;

FIG. 4 is a diagram of the safety situation awareness system of the ironworks based on the edge Internet of things technology according to the present invention;

FIG. 5 is a schematic diagram of heterogeneous fusion in accordance with the present invention;

FIG. 6 is a gateway protocol conversion schematic of the present invention;

FIG. 7 is a relational diagram of a data acquisition gateway and an Internet of things gateway of the present invention;

FIG. 8 is a data flow diagram of the data acquisition gateway of the present invention;

FIG. 9 is a flow chart of the gateway computing module data processing of the present invention;

FIG. 10 is a data collection gateway software flow diagram of the present invention;

FIG. 11 is an edge node layout of the present invention;

FIG. 12 is a motor data acquisition flow diagram of an embodiment of the present invention;

FIG. 13 is a diagram illustrating overall data analysis by a classifier in a dangerous operating mode of a motor according to an embodiment of the present invention;

FIG. 14 is an evaluation chart of the intrinsic safety level of the present invention;

FIG. 15 is a diagram of a classical three-layer neural network.

Detailed Description

The following detailed description of the present invention will be made with reference to the accompanying drawings.

As shown in fig. 1, a method for sensing the safety situation of the iron works based on the edge internet of things technology and the neural network is characterized by comprising the following steps:

s1: decomposing a production system and a figure ring system of an iron works into subsystems, and then coupling, isomerizing and pooling the two decomposed systems;

s3: constructing an edge Internet of things platform;

s4: predicting an artificial neural network;

s5: and realizing the perception of the safety situation of the iron works.

1. Specific embodiment of step S1

Iron-making production in China is almost all blast furnace production flow, and although various direct reduction and smelting reduction processes are developed continuously, blast furnace processes keep monopoly due to the advantages of maturity and high efficiency. It can be seen from the figure that blast furnace ironmaking has bulky main and auxiliary systems which are interconnected and cooperate with each other to provide enormous productivity.

The whole process of blast furnace ironmaking can be summarized as follows: under the condition of lowest energy consumption, the processes of reduction, slagging, heat transfer, slag iron reaction and the like are efficiently completed through the controlled forward sinking of furnace materials and the reverse movement of coal gas flow, and liquid metal products with ideal chemical components and temperature are obtained. In which each part of the process is very complicated, involving numerous technologies, and the production process is accompanied by a large number of physical and chemical changes.

Blast furnace production needs to be carried out in a sealed blast furnace, people cannot directly observe changes in the furnace, and only can indirectly observe and know through instruments and meters. Therefore, a computer control system for the specific production flow of the blast furnace main body system and each auxiliary system is designed. Taking a hot blast stove system as an example, in order to determine an optimal combustion system, an expert acquires data such as combustion waste gas components, waste gas temperature, furnace top combustion temperature and the like through internet of things terminal equipment, and realizes the functions of automatically adjusting combustion air and coal gas quantity, determining optimal furnace change time, automatically prompting and printing reports according to the parameters through computer analysis. Compared with the traditional production process, the computer control system has obvious advantages in the aspects of saving energy, keeping air supply, maintaining air pressure and improving the efficiency of the air furnace. Although computer-controlled production systems have significant advantages, they are relatively decentralized, neither interacting with each other, nor behaviorally affecting each other, due to the nature of the traditional iron-making process, and they are only responsible for their specific tasks.

In order to solve the problems, the decomposition, the coupling, the pooling and the heterogeneous fusion of the systems of the traditional iron-making production process are needed.

1. System decomposition

The system decomposition is to divide the original production process more carefully, and aims to modularize, unify and embody the iron-making production process and facilitate centralized management. The concept of the sub-system and the sub-system is adopted for division. The blast furnace body and the auxiliary system contained in the original iron-making process are kept as sub-systems. All the production links of iron making are contained under the sub-systems, the sub-systems have clear functions, and the parts of the iron making production process are combined together to form a complete process flow. Through the analysis of the iron-making process, the number of the divided subsystems is six, and the divided subsystems are respectively a blast furnace body system, a feeding system, a charging system, a furnace top system, an air supply system, an iron slag processing system and a coal gas dust removal system. The subsystem is an operating system which is obtained by vertically dividing a subsystem in more detail and is specific to a specific link. Taking an air supply system as an example, the system not only can be used for conveying combustion-supporting hot air to a blast furnace, but also is a huge system and comprises complicated links such as furnace changing, parallel air supply, various damping-down operations, hot blast furnace burning control and the like, and the links are taken as subsystems and are unified under the subsystems, so that a molecular system overall framework is formed, and a precondition is created for subsequent heterogeneous fusion.

As shown in fig. 2, which is an exploded view of a production system of an iron works, the subsystems of the production system of the iron works comprise a blast furnace body system, a feeding and feeding system, a furnace top system, an air supply system, an iron slag treatment system and a gas dust removal system; FIG. 3 is an exploded view of the decomposition of the subsystems of an iron mill production system into their respective subsystems. For example: the blast furnace body system is decomposed into: monitoring, electrical, cooling, etc. The furnace roof system is decomposed into: material distributing, material charging, material bell and other systems.

FIG. 3 is an exploded view of a "people", "things" and "rings" system, wherein the sub-system of the people ring system includes a people system, a things system and a rings system, and the people system includes a personnel identification, personnel monitoring and personnel positioning sub-system; the system of the object comprises a vehicle, material and facility subsystems; the system of the loop includes natural, production, and energy consumption subsystems.

The traditional iron-making production process system only considers the operation conditions of all systems in the production process, peripheral factors such as people, objects and rings are less considered, and in order to achieve the perception of the overall safety situation, the people need to consider in multiple directions, wherein an internet of things monitoring system taking people, objects and rings as main factors is also part of the important consideration of the people, objects and rings. We decompose the "people", "things", "ring" system into relatively independent vertical element subsystems in the same way by analyzing the surrounding environmental factors. Taking a system of a factory area related to a person as an example, in order to ensure the safety of the person in the factory area, the identity of the person is generally required to be identified, and the person is ensured to have the authority of entering a certain production area; secondly, the dynamic trajectory needs to be tracked and positioned to grasp real-time information of people.

2. System coupling

The system coupling means two systems which are communicated with each other and have different phase differences, people adopt certain measures to guide and strengthen the two systems, promote benign and positive interaction and mutual influence of the two systems, and excite the intrinsic potentials of the two systems, thereby realizing the complementary advantages and the common promotion of the two systems. After the system decomposition is completed, we couple the two major systems. And establishing association between the original iron-making system and the newly-built human, object and ring system.

As shown in fig. 4, the two types of systems that establish the association can share data, so as to achieve coordinated linkage. When a safety accident occurs to a certain system in a factory, the factory responds to the system of people, locates the personnel in the dangerous area and prompts the personnel to leave the dangerous area quickly; meanwhile, the system about the 'things' responds to the evacuation of important materials and closes related facilities; the system about the "ring" isolates the hazardous area. The harmfulness is reduced to the minimum by coordinating and linking the systems. In fig. 4, the box numbered 1 is an edge gateway.

3. Heterogeneous fusion

The heterogeneous integration is to connect devices based on different protocols to an internet of things platform through a unified protocol standard for centralized management. Because the production process of the iron works involves a plurality of subsystems, and the types of regional equipment contained in each subsystem are different, so that the equipment connection modes are different, and the data transmission mode is complicated, a brand new internet of things system is constructed through the internet of things technology to connect the subsystems which are numerous and relatively dispersed. After connection, the new Internet of things system has hundreds of millions of connection points and hundreds of billions of information points, so that the iron works can achieve the functions of interconnection, cooperation and control of everything.

The internet of things technology is a key core for realizing the interconnection of 'everything' in a factory area, however, in the field of internet of things information perception, because the application scene of the wireless sensor network technology has particularity, the wireless sensor network technology is generally applied to a local area, communication between networks cannot be achieved, and remote data transmission is not suitable, so that each system forms an information isolated island, and real and comprehensive interconnection and intercommunication cannot be achieved. The gateway is a link for connecting the wireless sensor network and the traditional communication network, and completes the conversion of protocols among the wireless sensor network, the traditional communication network and different types of networks. The key point for realizing heterogeneous integration is realizing wireless networking through a gateway and connecting the wireless networking with a platform.

The gateway is not a uniform individual, but has a plurality of varieties, and the functions of the gateway can be correspondingly expanded according to different production requirements so as to meet the requirements of daily production. For example, there are a temperature and humidity gateway dedicated to collect temperature and humidity information, and an internet of things connection gateway dedicated to negative protocol conversion.

In the field of iron works, a proper gateway is selected according to equipment and operation characteristics of each subsystem, and the gateway is laid on each subsystem so as to form physical connection with the subsystem. After the physical connection is established, all the subsystems are connected through a networking module of the gateway by the same domain name, and a wireless sensing network is established. Then, the wireless sensing network is connected with the network of the factory through the protocol conversion function of the gateway, so that all the gateways connected with the subsystems are connected to the factory cloud platform

The specific principle of gateway connection is shown in fig. 5. The Bluetooth network is a network set formed by all devices which are connected through an embedded Bluetooth module to complete data receiving, sending and transmitting; the zigbee network is an information network formed by various sensing devices connected through a zigbee protocol; the infrared networking network is a set of a series of devices controlled by an infrared remote sensing technology. The heterogeneous devices and the networks form an independent whole based on respective internal protocols, data transmission and exchange are only carried out in the own network, and communication cannot be carried out between the networks due to the difference of the protocols. And the internet of things gateway can realize the connection between heterogeneous devices through the conversion of the internal protocol, as shown in fig. 6.

4. Equipment pooling

The device pooling is to separate hardware of each sub-system and each sub-system from a primary system, uniformly incorporate the hardware into a resource pool of a factory area, and perform control and resource combination by a cloud server. In the traditional iron-making process, a fixed production mode can be formed once each system is built due to the fixity of hardware resources, different production requirements are difficult to meet by building a new system through the transfer of the hardware resources, only slight change can be made on the original system, or the new system is built, and a large amount of hardware resources are wasted. While device pooling provides the possibility for flexible deployment of hardware. In the design, the integration of the heterogeneous equipment in the factory is realized through the gateway of the internet of things, hardware resources of all sub-systems and sub-systems can be separated from the original fixed system through the combination of the internet of things and the equipment, the hardware resources are integrated into the heterogeneous integration monitoring platform of the internet of things in the cloud, the hardware resources can form a hardware resource pool through integration, and then all the architectures are only required to be built and applied on the resource pool, so that the dynamic hardware resources are realized. According to different requirements, related hardware resources are transferred through the cloud to control and integrate, and a new system and application are built on the integrated resources to deliver the resources in a service mode. Compared with the traditional mode, the pooled hardware resources are more flexible, the characteristic of hardware solidification of the traditional iron-making system is broken through, and infinite possibilities can be created.

2. Specific embodiment of step S2

In step S2, we will perform the acquisition of the relevant data. The original intention that the thing networking concept was put forward is that the messenger has RFID radio frequency technology, infrared sensing technology and other novel sensing equipment that have the perception ability and can seamlessly insert the thing networking to realize the collection and the transmission of information, further reach the intelligent recognition of production process and the control management of personnel's equipment. The purpose of applying the technology of the internet of things to iron and steel enterprises is not beyond the scope. For the construction of a security situation awareness system, data acquisition is an important link, and a security situation awareness three-layer architecture defines data acquisition as follows: the industrial data acquisition is characterized in that different systems, equipment and products are accessed through the existing communication means, and on the premise that the acquired data meet deep and large-scale conditions, the data base required by the sensing platform is constructed according to unique edge processing and protocol conversion of heterogeneous data. Through the link of the internet of things heterogeneous fusion, access to different systems, equipment and products through the existing communication means is achieved, and conditions are created for data acquisition. For a data acquisition link, a zigbee sensor network is taken as an example, and data collection is realized through a zigbee data acquisition gateway laid on equipment. The gateways are of various types, modules can be added and deleted according to different requirements, and different from the internet of things gateway mentioned in the step S1, the integration among heterogeneous network devices is realized through conversion of internet of things gateway negative protocol, and the acquisition of zigbee data acquisition gateway negative data can be connected with hundreds of zigbee data acquisition gateways. The relationship between the two is shown in FIG. 7.

The gateway is used as data acquisition terminal equipment and can acquire basic data of each subsystem equipment which is physically connected with the gateway, wherein the basic data comprises measurement data of temperature, humidity, gas concentration and the like which need to be measured; operation and maintenance data such as equipment operation time, start-stop times and the like, and real-time data such as equipment state and the like. The edge gateway reads data periodically according to set time, processes the data, stores the data and displays the data.

As shown in fig. 8, the sensing module on the data acquisition gateway uploads the acquired related data to the local storage module of the gateway for storage through the RTDB timing service according to the set time, and uploads the data to the operation processing module for operation, and is connected to the local host computer through the web interface for display. The Web interface and the operation processing module can also call related data from the storage module, the data processed by local operation is uploaded to the gateway of the Internet of things in wireless transmission modes such as Bluetooth, wiFi, zigbee and lora, and the like, and is uploaded to the cloud end after protocol conversion.

Inside the gateway, not all collected data is uploaded to the cloud platform, and a core processor of the operation center inside the gateway has some simple data processing capability, as shown in fig. 9.

Three types of data (equipment state, measurement data and operation and maintenance state) collected by the sensor need to be correspondingly processed in the gateway, firstly, analog-to-digital conversion is carried out, and analog signals are converted into digital signals which can be processed by a machine; unpacking the data, removing a data head and a data tail, and only leaving the data which can be directly processed; the three types of data reflecting the visual state of the equipment are stored through a storage module on one hand, and the required information is obtained through calculation through an algorithm on the other hand. Taking a motor as an example, the method collects measurement data including current, voltage, temperature and the like, operation and maintenance data including motor running time, starting and stopping times and the like, and real-time data including motor states and the like. Evaluating a score according to a fuzzy algorithm or an expert learning method and the like for reflecting the safety condition of equipment, and then outputting only one datum reflecting the overall safety of the motor; and finally, packaging and uploading the obtained final data to a cloud platform. Therefore, the obtained data on the cloud platform side is not the original data of voltage, current, start-stop times and operation and maintenance states, but the unique data of the safety state of the equipment is directly displayed. Fig. 10 is a data collection gateway software flow diagram.

The safety state data of all the devices uploaded through the gateway can be stored in the cloud background storage server to form a huge data pool, and on one hand, the data in the data pool is displayed at the cloud and on the other hand, the data is provided for local calling through the resource access interface so as to be convenient for subsequent further processing.

3. Specific embodiment of step S3

In step S3, we will proceed with the building of the edge platform. The core is that the original iron works subsystems (such as a blast furnace body system, a pulverized coal injection system and the like) are upgraded into functional edge nodes in a mode of building an edge server and an edge service platform, and the edge nodes can longitudinally control the directly-belonging subsystems and can also perform transverse information interaction with other edge nodes in an edge network mode. The edge nodes are unified at the cloud end to construct an edge internet of things service management system, so that a cloud-edge-end three-layer architecture is formed.

By edge computing technology is understood computations that occur at the edge of the network. In a traditional computing mode, terminal devices such as a gateway are directly connected with a cloud server, the terminal devices directly transmit data to a data center or cloud located in a central position, the cloud server uniformly processes the data and sends instructions, and the mode can process a large amount of complex data. However, with the increase of terminal devices, the burden of the cloud server gradually increases, and due to the increase of data volume, the speed of processing data is also reduced, so that higher time delay is generated, and some requirements for time delay in an industrial production link cannot be met. In a normal industrial production process, some data which needs to be processed in time is always accompanied, and if the time delay is large, irreversible loss can be caused. Therefore, the problems can be well solved by putting some simple data analysis processing functions into the edge server which is close to the terminal.

For iron works, when an object system is established, more and more devices are connected into the network, and the cloud server faces significant pressure. In addition, according to the basic requirements of factory production safety protection, in order to protect the production process and the safety of workers, the server is required to have high calculation speed, accurate data processing and break decision making. Therefore, it is necessary to construct an edge internet of things service management system in the iron works.

The production process of the iron works has the characteristics of one center and multiple systems, is matched with an edge calculation framework, and is very suitable for the deployment of edge calculation. The iron-making process consists of a blast furnace body system and a series of auxiliary systems, wherein in the step S1, the systems are divided into nine types of subsystems, and in the deployment link of edge calculation, the nine types of subsystems are used as basic units for building an edge system on the basis. The method comprises the steps of taking an original subsystem as an edge node, adding an edge server, and constructing an edge platform through a system environment formed by designing networks, calculation, storage, application and the like on the edge node, thereby providing near-end service for the subsystem. The edge platform is constructed to disperse large services which are originally and completely processed by the cloud or the central node to the edge service node, functions of core control parts of original systems of the iron works are integrated into the edge internet of things server, and the purpose that the internal affairs of the system are processed nearby at the position closest to the terminal is achieved. As shown in fig. 11.

After the design of the edge platform is completed, all edge nodes are connected to the cloud platform according to the S1 heterogeneous fusion principle, an edge internet of things service management system is constructed at the cloud end, and overall management of edge services at the cloud end is achieved. Thereby realizing the design of a cloud-edge-end three-layer architecture.

Firstly, the construction of the edge platform realizes the longitudinal 'penetrating' deep penetration from the end to the edge node to the cloud, and the edge node is used as the intermediate link of the end and the cloud, so that the longitudinal intelligent control on the subsystem can be realized, and the data can be uploaded to the cloud platform for overall decision making. Secondly, breaking through barriers of information interaction between edge nodes, and connecting the edge nodes in pairs to form an edge information interaction network, thereby ensuring the flow of data and the closed loop of control. The whole system is designed in a shape of a cross, and through interconnection and intercommunication of longitudinal and transverse dimensions, forwarding and processing time in data transmission is shortened, end-to-end time delay is reduced, network bandwidth pressure is relieved, service response capacity is enhanced, and powerful support is provided for construction of an overall security situation perception system of a steel plant.

4. Specific embodiment of step S4

The construction of the security situation awareness system not only comprises the construction of the hardware system in the steps S1, S2 and S3, but also comprises the construction of a security situation awareness model and the prediction of a neural network.

The construction of the security situation awareness model is divided into four main parts as follows:

(1) Data acquisition: the industrial data acquisition is characterized in that different systems, equipment and products are accessed through the existing communication means, and on the premise that the acquired data meet deep and large-scale conditions, the data base required by the sensing platform is constructed according to unique edge processing and protocol conversion of heterogeneous data.

(2) Characteristic extraction: after a large amount of data is collected in the first step, useful data is extracted from the data to perform corresponding preprocessing work, and data preparation is performed for the next work. Data acquisition and feature extraction are the bottommost layer of the whole security situation perception system, and data preparation is carried out.

(3) And (4) situation evaluation, wherein the situation evaluation is mainly to perform data fusion processing on the associated events and perform association identification from multiple aspects such as time, space, protocol and the like. In short, the risk assessment and the risk level judgment are carried out on the current time by combining data information.

(4) And (3) safety early warning, namely after a large amount of safety state data are extracted through the steps, the system can evaluate and predict the current safety state and the future safety state according to a specified standard, and further provides a corresponding analysis report and safety state early warning processing.

In step S4, we will use the artificial neural network to achieve the security posture awareness of the iron works. Firstly, a security situation awareness model is constructed. The first two steps (data acquisition and feature extraction) of the construction of the security situation awareness model are realized through steps S1, S2 and S3. And then, in step S4, a third step of situation assessment link design is performed.

For the situation assessment link, the invention comprehensively utilizes the methods of a safety check list method, a forecast risk analysis method, a fault type and influence analysis and the like to carry out the overall situation assessment of the iron works. Firstly, constructing a safety evaluation index system, namely determining an object to be evaluated and specific contents of the object to be evaluated; then, extracting data about the safety condition of each evaluation object stored in the data pool constructed in the step S2 through a data interface, analyzing information affecting system safety such as a danger source, a danger type and the like from the data, and determining the condition of the system to form a safety analysis table; then, the safety analysis table is calculated based on a certain rule through an evaluation algorithm, so that an evaluation value reflecting the safety situation of the system is obtained, and the purpose of evaluating the safety situation is achieved. And finally, the defects of manual prediction are simplified through the learning training of the neural network, and the automatic evaluation prediction of the safety situation is realized.

The method used by the dynamic safety risk assessment model of the industrial enterprise is adopted, and the new system constructed by the method is used as a frame to be modified appropriately, so that the safety situation assessment problem of the iron and steel plant designed by the user is realized. For the selection of the evaluation object, the design of the system is based on the architecture of the subsystems and the subsystems, so the subsystems are used as the first-class indexes, and the subsystems are used as the second-class indexes and are used as the evaluation object of the design. The evaluation content is considered from four aspects, namely:

(1) the first indicator is "number of dangerous modes", i.e., the total number of times the system is operated in dangerous mode.

(2) The second index is "the number of defects in equipment and facilities".

(3) The third index is "number of defects in work environment".

(4) The fourth indicator is "length of hazardous exposure".

After the index system is established, the data of the indexes needs to be acquired. And (3) calling safety data of each sub-system and each sub-system in one year from a data pool designed in S2 through a data interface, analyzing the safety data according to a statistical method, counting the number of types of safety accidents of each system in one year, determining the number of defects of equipment and facilities of each system by adopting an equipment defect analysis method, counting the number of operating environment defects in one year through environment data collected by environment systems in human, object and environment systems, and forming the dangerous exposure duration by extracting equipment log information. The safety analysis table is formed after the index system and the data are obtained, and is as follows:

TABLE 1 safety analysis Table

Taking a motor device in an electrical subsystem under a blast furnace body subsystem as an example, six indexes of a neural network input matrix are explained. Three types of data of a motor are collected on a data collection gateway, namely operation and maintenance states, measurement data and environmental data, and the data collection gateway is specifically divided into the following aspects, namely motor temperature, input voltage, input current, motor rotating speed, motor running time, starting and stopping times, running stop states, environmental temperature, environmental humidity, noise, corrosive gas, dust, air pressure, safe intrinsic horizontal value and the like, the data are collected through various sensors connected to the data collection gateway, and the safe intrinsic horizontal value is collected by adopting an RFID technology. According to the design of the data acquisition part in the step S2, the data are subjected to local calculation, local storage and local display on the data acquisition gateway side, the operation mode of the motor is uploaded to the cloud platform as a result of the local calculation, and the operation mode is stored, displayed and analyzed on the cloud platform side. The operation mode is divided into safety and danger, the information of the primary motor is automatically uploaded by the gateway at every certain time, and the data of the motor obtained by the user at the cloud platform side is as follows:

as shown above, the number of dangerous operation modes occurring in one month is the number of dangerous operation modes of the motor, according to the formula:

the number of dangerous modes of the electrical subsystem = the number of dangerous modes of the motor + the number of dangerous modes of other equipment

The number of dangerous modes of the electric subsystem can be calculated, namely the first type of index data input by the neural network.

The equipment defect number and the environmental defect number index data are acquired by the following design, a primary dangerous operation mode of the motor equipment is taken as an example for explanation, and when the motor is in the dangerous operation mode, a data flow diagram 12 of the whole designed system is shown. The main index data analysis is performed by a classifier, which is a simple judgment program based on some logic, and can judge and classify the input data similar to a NAND gate. The classifier adopts a layered design, and each layer carries out logic judgment of different functions, thereby realizing data classification. The first layer classifies data indexes and judges which type of data is 13 data; the second layer needs to classify the data types into two categories of equipment data and environment data so as to further judge the segments; the third layer needs to perform interval judgment, sets a reasonable interval of each index data, and judges whether the input index is in the reasonable interval; and finally judging the defect type through the fourth layer.

As shown in fig. 13, the classifier analyzes the overall data in the dangerous operation mode of the motor, and it can be obtained whether the motor is in dangerous operation caused by equipment defects or environmental defects. And counting the dangerous defect number and the environmental defect number of all equipment in the electrical subsystem to obtain the second and third indexes input by the neural network.

And the fourth type is the dangerous exposure time, the data is simple in record and not too much in design, the time interval from the occurrence of the danger to the solution of the danger is the dangerous exposure time of the equipment, and the dangerous exposure time can be intuitively obtained through a common equipment log.

After the safety analysis table is constructed, the evaluation value is calculated through an evaluation algorithm. The method comprises the following steps:

(1) Calculating a risk evaluation index according to the data of the safety analysis table;

(1) calculating the safety cost level h _s . The safety intrinsic level is the capacity of effectively preventing accidents from occurring from the source through various means in a broad sense, and intrinsic safety level evaluation is carried out from three aspects, firstly, people are the key for realizing the intrinsic safety of enterprises; secondly, the equipment guarantees the intrinsic safety of enterprises; finally, the environment is the push to achieve intrinsic safety for the enterprise. Therefore, the safety cost level is considered from three aspects of human, equipment and environment. FIG. 14 is an evaluation diagram of intrinsic safety level.

(2) Calculating a device risk coefficient k ₁ . The danger coefficient of the equipment passes through the formula

k ₁ = 5% number of device defects. And (6) performing calculation.

(3) Calculating an environmental hazard coefficient k ₂ . Environmental risk coefficient is given by formula

k ₂ = 5% of environmental defects. And (6) performing calculation.

(4) Judging the danger coefficient k of the material ₃ The evaluation criteria are as follows:

TABLE 2 evaluation table of substance Risk

(5) According to the formula h = h _s (1+k ₁ )(1+k ₂ )(1+k ₃ ) And calculating the risk index h of each subsystem.

The risk assessment indexes of each system calculated through the step (1) are shown in the following table:

TABLE 3 risk assessment index Table for each system

(2) Prediction of risk index h by artificial neural networks

Referring to the artificial neural network structure of fig. 15, the artificial neural network prediction process is divided into two parts, one is a forward propagation process of data and the other is a backward propagation process of errors. Firstly, an input matrix and an expected output are given, the input matrix is led into a neural network, an actual value is generated according to a calculation rule in the neural network, an error between the actual value and the expected value is calculated, the error is propagated reversely, and the weight and the threshold value in the neural network are continuously improved and optimized, so that the actual output is closer to the expected output. The process of continuously modifying the weight threshold is a training process of the neural network, relatively optimal weight and threshold are obtained inside the neural network after a large amount of training, an input matrix is given to the neural network again, and the neural network generates an actual value with small error from an expected value according to a trained rule to serve as a prediction result of the neural network. Comprehensively considering, the BP neural network optimized by the particle swarm optimization is used as an artificial neural network model of the experiment to predict. The BP neural network is an algorithm which continuously adjusts the weight and the threshold value of the network through an error back-propagation principle so as to enable the next output of the network to be closer to the expected output, is a really usable artificial neural network model, belongs to a feedback type neural network, is mature in theory and performance, and has strong nonlinear mapping capability and a flexible network structure. The particle swarm algorithm is derived from the research on the predation behavior of the bird swarm, and through swarm iteration, particles follow the optimal particles in a solution space to perform global search. The BP neural network is optimized by using the global search capability of the particle swarm algorithm, and the good global optimization capability of the particle swarm algorithm can be combined with the good local optimization capability of the BP algorithm, so that the generalization capability and the learning performance of the neural network are improved, and the accuracy of prediction is improved. And taking an error calculation formula as an optimization function of the particle swarm optimization, calculating the minimum error through swarm iteration, substituting the minimum error into the BP neural network to be used as the minimum error of the BP neural network for error back propagation, and training the weight and the threshold between each layer of the neural network through the error back propagation to achieve the purpose of optimization.

The artificial neural network (PSO-BP neural network) is specifically designed as follows:

an input layer: the input layer considers six indexes of danger mode number, equipment defect number, environment defect number, danger exposure duration, intrinsic safety level and substance danger coefficient, so that the number of the input layer nodes is 6. And intercepting six types of index data of 24 subsystems once by taking a month as a time node, and totally intercepting data for 1-11 months to form a 6-row 264-column input matrix to form a neural network training set for training a neural network. The training set is as follows:

TABLE 4 neural network training set input

And (3) converting the actual value of the risk index h calculated in the step (1) into a numerical matrix with 1 row and 264 columns as expected output of the neural network training set.

TABLE 5 neural network training set expectation outputs

And intercepting six types of index data of 24 subsystems in the 12 th month as a test set for testing. As follows:

TABLE 6 neural network test set input

The training set and the test set constructed by the method have different dimensions, and a large error is formed when the training set and the test set are directly input into the neural network for prediction, so that normalization processing is required before the training set and the test set are input into the neural network to enable the dimension of each index data to be unified within a range of (0, 1). The normalization processing formula is as follows:

where x is the original data, x _min ，x _max The minimum and maximum values of the data. The training set and the test set which are subjected to the normalization processing are data sets which can be directly processed by the neural network.

Hidden layer: there is no clear regulation for the number of hidden layer neurons in the neural network model construction, which is generally greater than the number of input layer neurons, and therefore, we select the number of hidden layer neurons as 12 in order to ensure the training accuracy and shorten the training time. The activation function between neurons selects the s-function. The activation function is a function that maps net activation amount to output, and activation is required because the input data may not be of the same order of magnitude as the expected value. The output of the S function is between (0, 1), the output range is limited, the optimization is stable, and in addition, the function is a continuous function and is convenient to derive, so that the S function can be used as an activation function of the neural network in the text.

The training times are designed to be 1000 times, the learning rate is 0.01, and the target minimum error is 0.000001.

And (3) an output layer: the neural network outputs the risk index h of 24 subsystems, is a 24-column numerical matrix with 1 row and 1 column, and has dimension of 1, so that the number of neurons in an output layer is 1, the dimension of the output numerical matrix is between (0 and 1), and the output numerical matrix can be data which intuitively reflects the risk index h of the system only by carrying out inverse normalization.

Through neural network training and testing, the risk index h of each subsystem can be obtained in real time, although the obtained risk index h can reflect the overall safety condition of the system only through further calculation, a large amount of calculation processes are saved through the application of the neural network, a large amount of collected index data are processed into data reflecting the risk index of the subsystem, and convenience is provided for subsequent safety situation perception of the whole system.

The 24 subsystem risk indices calculated from the index data of month 12 in the above example are calculated algorithmically as follows:

TABLE 7 hazard index Table for 24 subsystems in test set

5. Specific embodiment of step S5

Iron works integral safety situation perception realization

(1) And calculating the risk index H of the whole system. The risk index H of each subsystem predicted by the neural network cannot be completely used as a basis for evaluating the safety level of the system, and further processing is needed, namely the risk index H is calculated firstly. Is calculated by the formula

h _i Is the subsystem risk index, n _i As the number of subsystems, E _i The risk index of the demonstrated example is calculated as the exposure time of the risk point

(2) And calculating the initial security risk level S (K-1) of the enterprise. Calculated according to the following formula:

wherein G is a constant of 10 and Y is the negative injury rate of thousands of people. Assuming that the injury rate of thousands of people in the iron works is 3/1000, the initial safety risk level of the iron works is 58.09. The greater the rate of thousands of people getting loaded, the lower the initial level of safety risk.

(3) And judging the enterprise risk management and control capability C. And on the basis of the percentage control, each system of the enterprise is checked according to the items on the enterprise risk control scoring table, so that the enterprise risk control capability is scored. The experiment assumes that the iron works risk management and control capability C =66.

(4) And calculating the control capability index of the iron works. The specific calculation formula is as follows:

b (k) = α C- β H α and β are constants, α =0.068, β =0.55

The regulatory index is used for representing the dynamic change of the enterprise risk level, when B (K) is larger than 0, the system risk level is rising, and when B (K) is smaller than 0, the system risk level is falling. Through calculation, based on the above experimental data, the iron works management and control ability index B (k) =1.40 referred to herein.

(5) And calculating the comprehensive security risk level S (k) of the enterprise. The specific calculation formula is as follows:

S(K)＝S(K-1)+B(K)

the comprehensive safety risk level S (K) =59.49 of the experiment is calculated

Through the five steps of calculation, the safety risk level calculation result of the system is shown in table 8:

TABLE 8 calculation of System Risk level

(6) And grading the security risk level according to the enterprise comprehensive security risk level S (k). The safety risk levels were divided as shown in the following table

TABLE 9 Security Risk rankings

S1 and S2 are interval critical values, and the calculation formula is as follows:

in the formula, M is the number of actual staff, P is the expected injury rate of thousands of people, and we assume that the company has 3500 people, the expected injury rate of thousands of people is 0.002, and calculated S1=67.912, S2=63.725, while the comprehensive risk level of our system is S =59.49, so the system fails.

In step S5, iron works safety situation awareness is achieved. By the comprehensive risk level of the system, the safety level of the iron works can be judged, when the safety level of the iron works is low, a reverse reasoning method is adopted to trace the source of the factors causing the low safety level of the system, the source of the dangerous accident is positioned, and therefore corresponding measures are adopted to achieve accurate handling of the accident. Firstly, judging the safety level of the iron works according to the comprehensive risk level; secondly, if the safety level is low, the safety analysis table established in the step S4 is derived, index data with low safety risk level caused by abnormality are checked, taking the number of equipment defects of the cooling subsystem of the blast furnace body as an example, if the number of the equipment defects is found to be high through checking of the safety analysis table, an instruction is sent, and all equipment data of the original cooling system forming the number of the equipment defects in the safety analysis table are called out; finally, the specific equipment can be accurately positioned through the investigation of the original data, so that the equipment is pertinently subjected to fault analysis, after a result is analyzed, a maintenance scheme is formulated according to the fault reason, then the equipment can be maintained in a local maintenance mode according to human resources and the operation risk degree, and the maintenance equipment can also be remotely controlled through the edge internet of things service management system constructed in the step S3. The cloud terminal issues an instruction, the instruction is issued to a subsystem to which the accident belongs through an edge service management platform, and the subsystem instruction is converted into an instruction which can be identified by a terminal through a protocol of the Internet of things gateway, so that industrial control equipment or relay equipment on the data acquisition gateway is controlled to make corresponding actions. By tracing the dangerous accident, the final safety problem of the iron works is solved.

Through the design of the five parts S1-S5, the architecture scheme of the iron works security situation perception system is finally realized. Through the design of the step S1, the barriers of information interaction among subsystems are opened, people, objects and rings are considered in the whole framework of the system, and the problems of factory area everything and physical connection and flexible system scheduling are solved through the pooling design; by designing gateway hardware and software in the step S2, the problem of data acquisition of heterogeneous equipment is solved, and a data pool is established at the cloud end for storing and calling mass data; by the design of the step S3, the construction problem of the factory edge platform is realized, the subsystem is used as the edge node to carry out the specific architecture of the edge platform, the problem of real-time data processing is solved, and meanwhile, the coordinated linkage of the edge nodes is realized through the design of an interactive network among the edge nodes; by the design of S4, the construction of the security situation perception model and the neural network are applied to the security situation perception system, the data processing process is optimized, and the steps of data analysis and calculation are simplified; by way of example of step S5, the basic principle of hazard source backtracking is introduced. And finally, completing the construction of the iron works safety situation perception system.

The method of the system is characterized in that:

(1) the decomposition and fusion design of the whole system and the coupling with people, objects and rings;

(2) a cross-shaped architecture design consisting of a longitudinal cloud-edge-end three-layer architecture and transverse edge network coordination linkage;

(3) designing a safety analysis table in the construction of a safety situation perception model and evaluating an intrinsic safety level h in a calculation method;

(4) the application of the PSO-BP artificial neural network in a system.

The deep combination of the traditional industry and the internet of things is the core for realizing the 4.0 industry in the early days and is the core layout of the national industrial system, taking an iron plant as an example, the combination of the internet of things technology and the edge computing technology collects and preprocesses data of the traditional iron making process flow, the overall evaluation is carried out through a safety risk evaluation model, the prediction is carried out through an artificial neural network, and the safety regulation and control are realized through the analysis of the prediction result. The method is a reasonable construction scheme of the iron works safety situation perception system.

The above embodiments are implemented on the premise of the technical solution of the present invention, and detailed embodiments and specific operation procedures are given, but the scope of the present invention is not limited to the above embodiments. The methods used in the above examples are conventional methods unless otherwise specified.

Claims

1. A method for sensing the safety situation of an iron-making plant based on an edge Internet of things technology and a neural network is characterized by comprising the following steps:

s1: decomposing a production system and a figure ring system of an iron works into subsystems, and then coupling, isomerizing, fusing and pooling the two decomposed systems;

s2: data acquisition: establishing a gateway connection relation of the Internet of things of the iron works production system and the character ring system, acquiring data of subsystems in the iron works production system and the character ring system by using various sensors, and storing the acquired data in a built storage server to form a data pool with a large amount of acquired information;

s3: constructing an edge Internet of things platform;

s4: predicting an artificial neural network;

s5: realizing the safety situation awareness of the iron works;

in the step S1, the iron works production system comprises a blast furnace body system, a feeding and feeding system, a furnace top system, an air supply system, an iron slag treatment system and a coal gas dust removal system; decomposing the ironworks production system into their respective subsystems; the method comprises the following specific steps:

in the safety situation perception method, the system decomposition is to divide the original production process more carefully, so as to modularize, unify and embody the iron-making production process and facilitate centralized management; the thinking of subsystems and subsystems is adopted for division; keeping a blast furnace body and an auxiliary system which are included in the original iron-making process as sub-systems; all the production links of iron making are contained under the subsystems, the subsystems have clear functions and are respectively responsible for one part of the iron making production process, and the combination of the subsystems is a complete process flow;

the person ring system comprises a person system, an object system and a ring system, wherein the person system comprises a person identification subsystem, a person monitoring subsystem and a person positioning subsystem; the system of the object comprises a vehicle, material and facility subsystems; the system of the ring comprises a nature subsystem, a production subsystem and an energy consumption subsystem;

in step S1, the coupling between the two systems is specifically: sharing all data of subsystems of the two systems, so as to achieve coordinated linkage, and when a safety accident happens to one system of the iron plant production system, the system of the plant personnel generates response to position the personnel in the dangerous area and prompt the personnel to leave the dangerous area quickly; meanwhile, the system of the object generates response, withdraws important materials and closes related facilities; the system of rings responds to isolate the hazardous area; the harm is reduced to the minimum by coordinating and linking the systems;

the step S4 specifically includes the following steps:

s42: extracting data about the safety condition of each evaluation object stored in the data pool constructed in the step S2 through a data interface, extracting information influencing the safety of the system from the data, and determining the condition of the system to form a safety analysis table; the information influencing the system safety comprises a danger source and a danger type;

s43: calculating the safety analysis table through an evaluation algorithm, and taking the calculation result, namely the risk index h of each subsystem as a basis for reflecting the risk degree of each subsystem;

2. The iron mill safety situation awareness method based on the edge internet of things technology and the neural network as claimed in claim 1, wherein in the step S1, the pooling specifically comprises: and (3) separating the hardware of each subsystem from the original system, uniformly bringing the hardware into a resource pool of the factory, and performing control and resource combination by the cloud server.

3. The method for sensing the safety situation of the iron works based on the edge internet of things technology and the neural network as claimed in claim 1, wherein in the step S1, the heterogeneous integration is performed by connecting devices based on different protocols in the subsystems of the two systems to the edge internet of things platform for centralized management through a unified protocol standard.

4. The iron mill safety situation awareness method based on the edge internet of things technology and the neural network according to claim 1, wherein the edge internet of things platform in the step S3 is specifically constructed by: upgrading subsystems and subsystems of a production system and a character ring system of an iron and steel plant into functional edge nodes in a mode of building an edge server and an edge service platform, wherein the edge nodes can longitudinally control the subsystems of the iron and steel plant and can also perform transverse information interaction with other edge nodes in an edge network mode; and unifying the edge nodes in a cloud side to construct an edge Internet of things platform, so as to form a cloud-edge-end three-layer architecture.

5. The method for sensing the safety situation of the iron works based on the edge internet of things technology and the neural network as claimed in claim 1, wherein the objects of the evaluation determined in S41 are: taking the subsystem as a first-class index and the subsystem as a second-class index as an evaluation object; the content of the evaluation comprises:

(1) The first indicator is "number of dangerous modes", i.e. the total number of times the system is operated in dangerous mode;

(2) The second index is the number of defects of equipment and facilities;

(3) The third index is 'the number of defects of the operating environment';

(4) The fourth indicator is "length of hazardous exposure".

6. The iron mill safety situation awareness method based on the edge internet of things technology and the neural network according to claim 1, wherein the S44 specifically includes: