Industrial sensor data analysis method and system

Abstract

The invention belongs to the technical field of industrial data analysis, and discloses an industrial sensor data analysis method and system. The method comprises the following steps that: a cloud data center sets an OPC UA instance, and constructs a data analysis model and a fault maintenance guidance model; the distributed edge cluster is used for carrying out calculation task distribution; the distributed edge cluster preprocesses a plurality of pieces of real-time industrial sensor data according to a real-time calculation task allocation scheme; the cloud data center is used for writing a plurality of preprocessed real-time industrial sensor data into an OPC UA instance, extracting corresponding real-time OPC UA dynamic data, and performing data analysis by using a data analysis model; and the cloud data center is used for carrying out fault maintenance guidance by using a fault maintenance guidance model according to the real-time data analysis result. The problems that in the prior art, the data processing and analysis capacity is insufficient, the data format is not uniform, the calculation pressure is large, and a response mechanism is lacked are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of industrial data analysis, and in particular relates to an analysis method and system for industrial sensor data. Background Art

[0002] With the popularization and application of industrial sensors in industrial production, the volume of industrial sensor data generated has increased exponentially. Industrial sensor data refers to information collected by various sensors in industrial environments. These sensors are usually used to monitor and control production processes, equipment status, environmental conditions, etc. The effective use of industrial sensor data is crucial to improving the level of industrial automation, reducing operating costs, and improving production efficiency and safety.

[0003] The existing analysis of industrial sensor data has the following defects: 1) The existing industrial sensor data analysis technology has insufficient data processing capabilities and cannot process large-scale industrial sensor data, resulting in low data analysis efficiency. In addition, it has insufficient feature analysis capabilities for complex data and is difficult to extract valuable information from massive data, resulting in inaccurate and in-time data analysis results. 2) The existing industrial sensor data analysis technology does not have a unified data format between different physical entities and industrial sensors, which results in a large amount of conversion work required during data integration and exchange, reducing processing efficiency and responding to the real-time nature of industrial sensor data analysis; 3) Existing industrial sensor data analysis technology uses centralized data processing and analysis methods, which has high hardware requirements for centralized data analysis servers, high computing pressure, and is prone to hardware crashes leading to data loss. The reliability and security of data analysis are low; 4) Existing industrial sensor data analysis technology lacks a response mechanism and is unable to generate fault repair strategies for guidance based on data analysis results, and has a low level of intelligence. Summary of the invention

[0004] In order to solve the problems of insufficient data processing and analysis capabilities, inconsistent data formats, high computing pressure and lack of response mechanism in the prior art, the present invention aims to provide an analysis method and system for industrial sensor data.

[0005] The technical solution adopted by the present invention is: A method for analyzing industrial sensor data comprises the following steps: The cloud data center configures a number of corresponding industrial sensors according to the different physical entities that need to be monitored, sets up OPC UA instances, and builds data analysis models and fault maintenance guidance models; The distributed edge cluster allocates computing tasks based on a number of real-time industrial sensor data collected by a number of industrial sensors to obtain a real-time computing task allocation plan; The distributed edge cluster pre-processes a number of real-time industrial sensor data according to the real-time computing task allocation plan, obtains a number of pre-processed industrial sensor data, and encrypts and transmits them to the cloud data center; The cloud data center writes some pre-processed real-time industrial sensor data into the OPC UA instance, extracts the corresponding real-time OPC UA dynamic data, and uses the data analysis model to perform data analysis to obtain real-time data analysis results; The cloud data center uses the fault repair guidance model to provide fault repair guidance based on the real-time data analysis results, and sends the obtained real-time fault repair plan to the distributed edge cluster.

[0006] Furthermore, the cloud data center configures a number of corresponding industrial sensors according to different physical entities that need to be monitored, sets up OPC UA instances, and builds a data analysis model and a fault maintenance guidance model, including the following steps: The cloud data center confirms the monitoring indicators of several physical entities that need to be monitored, and configures several corresponding industrial sensors according to the monitoring indicators of each physical entity; Design the OPC UA information model based on the configuration parameters of all industrial sensors of all physical entities and build the corresponding OPC UA instance in the address space of the OPC UA server; Based on the OPC UA instance, some historical OPC UA dynamic data are generated and standardized to obtain some standardized historical OPC UA dynamic data; Based on some standardized historical OPC UA dynamic data, a deep learning algorithm is used to build a data analysis model and generate some historical data analysis results; Based on the analysis results of several historical data, an enhanced reinforcement learning algorithm is used to build a fault maintenance guidance model.

[0007] Furthermore, the data analysis model is constructed based on the GCN-Attention-MLP algorithm, and the data analysis model includes a graph feature extraction module constructed based on the GCN algorithm, an attention weight module constructed based on the Attention mechanism, and a data analysis module constructed based on the MLP algorithm, which are sequentially connected; The fault maintenance guidance model is constructed based on the MPO-DQN algorithm, and the fault maintenance guidance model includes a meta-strategy optimization module constructed based on the MPO algorithm and a reinforcement learning module constructed based on the DQN algorithm. The reinforcement learning module is provided with an intelligent agent, a deep Q network and an experience replay pool. The intelligent agent is respectively connected to the deep Q network, the experience replay pool and the meta-strategy optimization module, and the meta-strategy optimization module is connected to the experience replay pool.

[0008] Furthermore, according to the configuration parameters of all industrial sensors of all physical entities, an OPC UA information model is designed, and a corresponding OPC UA instance is constructed in the address space of the OPC UA server, including the following steps: According to the object type in the configuration parameters of the industrial sensor of the physical entity, a plurality of object nodes including all the industrial sensors are set; the object nodes include a first object node of the image sensor and a second object node of the non-image sensor; According to the monitoring type in the configuration parameters, a plurality of first monitoring nodes are set for the first object node, and according to the monitoring attribute in the configuration parameters, a corresponding second monitoring attribute is set for each first monitoring node; According to the monitoring type in the configuration parameters, set a number of second monitoring nodes for the second object node, and according to the monitoring attribute in the configuration parameters, set a corresponding second monitoring attribute for each second monitoring node; Setting corresponding first node relationships for a number of first object nodes and a number of first monitoring nodes, and setting second node relationships for a number of second object nodes and a number of second monitoring nodes; Traversing all industrial sensors of all physical entities, constructing an OPC UA information model according to all first object nodes, all first monitoring nodes, all second node relationships, all second object nodes, all second monitoring nodes, and all second node relationships; According to the model metadata of the OPC UA information model, a corresponding OPC UA instance is constructed in the address space of the OPC UA server.

[0009] Furthermore, the distributed edge cluster performs computing task allocation according to a plurality of real-time industrial sensor data collected by a plurality of industrial sensors to obtain a real-time computing task allocation scheme, including the following steps: Industrial sensors collect real-time industrial sensor data of corresponding physical entities and transmit the real-time industrial sensor data to distributed edge clusters within the communication range; Distributed edge cluster, confirms the distributed edge computing gateway set participating in computing task allocation, defines the computing task set, and takes minimizing the computing task execution time and computing resource cost as the optimization goal; According to the optimization target and a number of real-time industrial sensor data collected by a number of industrial sensors, a fitness function of the swarm intelligence optimization algorithm is defined; According to the computing task set and edge computing gateway set, define the individual encoding vector format of the ICPO algorithm, and set the ICPO population parameters and maximum number of iterations of the ICPO algorithm; Based on the individual encoding vector format, ICPO population parameters and fitness function, the ICPO algorithm is used to iteratively search for the best solution and retain the best individual. If the number of iterations of iterative optimization reaches the maximum number of iterations or the fitness value of the optimal individual meets the requirements, the individual encoding vector of the optimal individual is decoded to obtain a real-time computing task allocation plan.

[0010] Furthermore, the distributed edge cluster pre-processes a number of real-time industrial sensor data according to the real-time computing task allocation scheme, obtains a number of pre-processed industrial sensor data, and encrypts and transmits the pre-processed industrial sensor data to the cloud data center, including the following steps: The distributed edge cluster distributes a number of real-time industrial sensor data to a number of corresponding target edge computing gateways according to the real-time computing task allocation plan; The target edge computing gateway cleans the allocated real-time industrial sensor data to obtain the cleaned real-time industrial sensor data. The target edge computing gateway performs graph structure conversion on a number of cleaned real-time industrial sensor data to obtain a number of real-time industrial sensor data after graph structure conversion; The target edge computing gateway aggregates the real-time industrial sensor data after graph structure conversion generated by all target edge computing gateways to obtain a number of pre-processed real-time industrial sensor data; An asymmetric encryption algorithm is used to encrypt a number of pre-processed real-time industrial sensor data to obtain encrypted a number of pre-processed industrial sensor data and transmit them to a cloud data center.

[0011] Furthermore, the target edge computing gateway performs graph structure conversion on the plurality of cleaned real-time industrial sensor data to obtain the plurality of graph structure converted real-time industrial sensor data, including the following steps: The target edge computing gateway performs pixel block division on the cleaned real-time image sensor data in the cleaned real-time industrial sensor data to obtain a plurality of real-time image pixel blocks; The real-time image information of the real-time image sensor data after cleaning and the real-time pixel block information of the real-time image pixel block are used as the first object node of the OPC UA information model; The real-time pixel information of the real-time image pixel block is used as the first monitoring node of the OPC UA information model, and the real-time pixel value of the real-time image pixel block is used as the first monitoring attribute of the OPC UA information model; The real-time spatial relationship between real-time image pixel blocks is used as the first node relationship of the OPC UA information model to obtain the corresponding real-time image sensor data after graph structure conversion; Performing graph structure conversion on the cleaned real-time non-image sensor data in the cleaned real-time industrial sensor data to obtain corresponding graph structure converted real-time non-image sensor data; A number of real-time image sensor data after graph structure conversion and a number of real-time non-image sensor data after graph structure conversion are integrated to obtain a number of real-time industrial sensor data after graph structure conversion.

[0012] Furthermore, the cloud data center writes some pre-processed real-time industrial sensor data into the OPC UA instance, extracts the corresponding real-time OPC UA dynamic data, and uses the data analysis model to perform data analysis to obtain real-time data analysis results, including the following steps: The cloud data center uses an asymmetric decryption algorithm to decrypt the encrypted preprocessed real-time industrial sensor data to obtain the decrypted preprocessed industrial sensor data; Write the decrypted pre-processed industrial sensor data into the OPC UA instance, and extract the real-time OPC UA dynamic data of the OPC UA instance when the data analysis cycle arrives; Performing standardization processing on the real-time OPC UA dynamic data to obtain standardized real-time OPC UA dynamic data, and inputting the standardized real-time OPC UA dynamic data into a data analysis model; Use the data analysis model to extract the real-time graph features of the standardized real-time OPC UA dynamic data, and weight the real-time graph features according to the preset attention weight values ​​to obtain real-time weighted features; According to the real-time weighted features, data analysis is performed to obtain corresponding real-time data analysis results.

[0013] Furthermore, the cloud data center uses the fault repair guidance model to provide fault repair guidance based on the real-time data analysis results, and sends the obtained real-time fault repair solution to the distributed edge cluster, including the following steps: The cloud data center inputs the real-time data analysis results output by the data analysis model into the fault maintenance guidance model, and extracts corresponding real-time fault maintenance guidance experience based on the real-time data analysis results; According to the real-time fault maintenance guidance experience, the parameters of the fault maintenance guidance model are adjusted to obtain an adjusted fault maintenance guidance model; According to the real-time data analysis results and the real-time fault maintenance guidance experience, the adjusted fault maintenance guidance model is used to provide fault maintenance guidance and obtain a real-time fault maintenance plan; The real-time fault repair plan is encrypted using an asymmetric encryption algorithm to obtain the encrypted real-time fault repair plan, which is then returned to the distributed edge cluster.

[0014] An industrial sensor data analysis system is used to implement an industrial sensor data analysis method. The system includes a cloud data center, a plurality of distributed edge clusters, and a plurality of industrial sensors. The cloud data center is respectively connected to the plurality of distributed edge clusters for communication. Each distributed edge cluster is respectively connected to the plurality of industrial sensors within the communication range for communication. The distributed edge cluster includes a plurality of edge computing gateways connected in a distributed manner. The plurality of industrial sensors are arranged at corresponding physical entities. The cloud data center includes an initialization unit, a data analysis unit, and a fault repair guidance unit which are connected in sequence.

[0015] The beneficial effects of the present invention are: The present invention discloses an analysis method and system for industrial sensor data. The constructed data analysis model can realize the automatic analysis of massive industrial sensor data, improve the data processing capability, and the data analysis model can mine the potential connection between a number of industrial sensor data and between the industrial sensor data and the abnormalities and failures of physical entities, extract valuable information from massive data, and improve the accuracy and timeliness of data analysis results; by configuring corresponding industrial sensors and constructing OPC UA instances, it is possible to realize effective monitoring of different physical entities, improve the pertinence and efficiency of data collection, and build a unified OPC The UA information model realizes standardized conversion between different data sources, simplifies the data processing process, and improves processing efficiency. Through the standard data interaction processing process, it can quickly respond to the writing and query of real-time data to meet real-time requirements; it adopts edge-cloud collaborative technology to delegate computing tasks to the edge computing gateway for processing, which reduces the hardware requirements and computing pressure of the cloud data center, avoids accidents such as hardware crashes that lead to data loss, and allocates computing tasks to distributed edge clusters, reducing the data processing pressure and computing resource overload of a single edge computing gateway, and improving the reliability and security of data analysis; by constructing a fault maintenance guidance model, a response mechanism is realized, which can generate fault maintenance strategies for guidance based on data analysis results, thereby improving the level of intelligence.

[0016] Other beneficial effects of the present invention will be further described in the specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1It is a flowchart of the method for analyzing industrial sensor data in the present invention.

[0018] Figure 2 It is a structural block diagram of the industrial sensor data analysis system in the present invention. DETAILED DESCRIPTION

[0019] The present invention will be further explained below in conjunction with the accompanying drawings and specific embodiments.

[0020] Embodiment 1: like Figure 1 As shown, this embodiment provides an analysis method for industrial sensor data, comprising the following steps: S1: The cloud data center configures a number of corresponding industrial sensors according to the different physical entities that need to be monitored, sets up an Open Platform Communications Unified Architecture (OPCUA) instance, and builds a data analysis model and a fault maintenance guidance model, including the following steps: S1-1: The cloud data center confirms the monitoring indicators of several physical entities that need to be monitored, and configures several corresponding industrial sensors according to the monitoring indicators of each physical entity; Physical entities include industrial production equipment, communication equipment, control equipment and other physical equipment. Industrial sensors include image sensors of industrial cameras, and non-image sensors such as temperature sensors, pressure sensors, flow sensors, liquid level sensors, position sensors, and optical sensors. S1-2: Design an OPC UA information model based on the configuration parameters of all industrial sensors of all physical entities, and build a corresponding OPC UA instance in the address space of the OPC UA server, including the following steps: S1-2-1: according to the object type in the configuration parameters of the industrial sensor of the physical entity, set a number of object nodes including all industrial sensors; the object nodes include a first object node of an image sensor and a second object node of a non-image sensor; S1-2-2: according to the monitoring type in the configuration parameters, set a plurality of first monitoring nodes for the first object node, and according to the monitoring attribute in the configuration parameters, set a corresponding second monitoring attribute for each first monitoring node; S1-2-3: according to the monitoring type in the configuration parameters, set a plurality of second monitoring nodes for the second object node, and according to the monitoring attribute in the configuration parameters, set a corresponding second monitoring attribute for each second monitoring node; For example, the monitoring type of the temperature sensor includes model, temperature and timestamp, then the second monitoring node includes a node representing the model "Model-number", a node representing the temperature "Temperature" and a node representing the timestamp "Timestamp", and the second monitoring attribute of the temperature node is "Value-float"; the monitoring type of the image sensor includes the pixel information of each image pixel block, then the first monitoring node includes several nodes representing the pixel information "PixelBlock", and the first monitoring attribute of the pixel information node is "AverageColor-int, TextureFeature"; S1-2-4: setting corresponding first node relationships for a number of first object nodes and a number of first monitoring nodes, and setting second node relationships for a number of second object nodes and a number of second monitoring nodes; S1-2-5: traverse all industrial sensors of all physical entities, and build an OPC UA information model according to all first object nodes, all first monitoring nodes, all second node relationships, all second object nodes, all second monitoring nodes, and all second node relationships; S1-2-6: construct a corresponding OPC UA instance in the address space of the OPC UA server according to the model metadata of the OPC UA information model; S1-3: Based on the OPC UA instance, generate some historical OPC UA dynamic data, and perform standardization processing to obtain some standardized historical OPC UA dynamic data; S1-4: Based on some standardized historical OPC UA dynamic data, use deep learning algorithms to build a data analysis model and generate some historical data analysis results; The data analysis model is built based on the Graph Convolutional Network (GCN)-Attention-MLP algorithm, and the data analysis model includes a graph feature extraction module built based on the GCN algorithm, an attention weight module built based on the Attention mechanism, and a data analysis module built based on the Multilayer Perceptron (MLP) algorithm, which are connected in sequence; The GCN network propagates features of OPC UA dynamic data through operations similar to convolution, extracts node features of each node data in OPC UA dynamic data, and edge features between nodes, to form graph features of OPC UA dynamic data; the Attention mechanism weights several graph features, enhances the model's attention to important graph features, and improves the accuracy and efficiency of model prediction; the MLP network, as a fully connected network, can perform accurate and efficient label prediction based on weighted features; The data analysis results include the anomaly detection results of the sensor data values, the potential fault analysis results hidden between different sensor data, and the future sensor data trend analysis results; S1-5: Based on the analysis results of several historical data, an enhanced reinforcement learning algorithm is used to build a fault maintenance guidance model; The fault maintenance guidance model is constructed based on the Meta-Policy Optimization (MPO)-Deep Q Network (DQN) algorithm, and the fault maintenance guidance model includes a meta-policy optimization module constructed based on the MPO algorithm and a reinforcement learning module constructed based on the DQN algorithm. The reinforcement learning module is provided with an intelligent agent, a deep Q network and an experience replay pool. The intelligent agent is connected to the deep Q network, the experience replay pool and the meta-policy optimization module respectively, and the meta-policy optimization module is connected to the experience replay pool. The meta-strategy optimization module is used to optimize the deep Q network parameters in the reinforcement learning module so that these parameters can quickly adapt to new and unseen tasks, improve the generalization ability of the model, and update the deep Q network based on previous learning experience even under unseen data; through meta-strategy optimization, the DQN algorithm can converge faster, reduce the amount of data and time required for training, and significantly improve learning efficiency; the reinforcement learning module generates fault repair plans based on the characteristics of the data analysis results; Based on the analysis results of several historical data, an enhanced reinforcement learning algorithm is used to build a fault maintenance guidance model, which includes the following steps: S1-5-1: The fault repair plan generation problem is used as the simulation environment of the DQN algorithm, and the network parameter generation problem of the DQN algorithm is divided into several subtasks of the MPO algorithm according to different data analysis scenarios; S1-5-2: Use the MPO algorithm to build the initial meta-strategy optimization module, input some historical data analysis results, train under some subtasks, obtain the final meta-strategy optimization module, and generate some historical meta-strategy optimization experiences; S1-5-3: Use the final meta-strategy optimization module to output the initial third initial network parameters of the deep Q network in the DQN algorithm, and use the DQN algorithm to construct an initial reinforcement learning module including the initial deep Q network based on the initial third initial network parameters; S1-5-4: Analyze several historical physical entity states included in different historical data analysis results, define a state space based on several historical physical entity states, define an action space based on preset fault repair actions, and set a reward function based on the impact of the preset fault repair actions on the physical entity state; S1-5-5: Based on the state space, action space and reward function, input some historical data analysis results, train the initial deep Q network of the initial reinforcement learning module, obtain the final reinforcement learning module, and generate some historical reinforcement learning experiences; S1-5-6: Integrate the historical meta-strategy optimization experience and the corresponding historical reinforcement learning experience into historical fault repair guidance experience, and store several historical fault repair guidance experiences into the experience replay pool of the final reinforcement learning module; S1-5-7: Integrate the final meta-strategy optimization module and the final reinforcement learning module to obtain the fault maintenance guidance model; S2: The distributed edge cluster allocates computing tasks based on a number of real-time industrial sensor data collected by a number of industrial sensors to obtain a real-time computing task allocation plan, including the following steps: S2-1: Industrial sensors collect real-time industrial sensor data of the corresponding physical entities and transmit the real-time industrial sensor data to the distributed edge cluster within the communication range; S2-2: Distributed edge cluster, confirm the distributed edge computing gateway set participating in the computing task allocation, define the computing task set, and minimize the computing task execution time and computing resource cost as the optimization goal; S2-3: Define the fitness function of the swarm intelligence optimization algorithm according to the optimization target and a number of real-time industrial sensor data collected by a number of industrial sensors; The formula is:

[0021] In the formula, is the fitness function; To optimize the objective function; To calculate the task execution time; To calculate resource costs; Refers to the parameters for the solution; Refers to parameters for a data set consisting of several real-time industrial sensor data; All are fitness weight values; S2-4: Define the individual encoding vector format of the ICPO algorithm according to the computing task set and the edge computing gateway set, and set the ICPO population parameters and maximum number of iterations of the ICPO algorithm; S2-5: Based on the individual encoding vector format, ICPO population parameters and fitness function, use the ICPO algorithm to iteratively search for the best solution and retain the best individual, including the following steps: S2-5-1: Initialize the ICPO population using the Circle chaotic mapping sequence according to the individual coding vector format and ICPO population parameters to obtain an initial ICPO population containing several initial ICPO individuals; The formula is:

[0022] In the formula, is the initial ICPO individual of the Circle chaos map, corresponding to an initial solution; is the randomly generated initial ICPO individual; It is the ICPO individual indicator; S2-5-2: Introduce a cyclic population reduction mechanism to limit the number of individuals in the ICPO population parameters and obtain updated ICPO population parameters for the next iteration; The formula is:

[0023] In the formula, For the The number of individuals in the ICPO population parameter of the iteration; For the The number of individuals in the ICPO population parameter of the iteration; is the minimum number of individuals in the ICPO population parameter; Evaluate arguments for functions; Evaluate loop parameters for a function; Evaluate loop parameters for the maximum function; t is the indicator of the number of iterations; S2-5-3: Calculate the initial fitness value of the initial ICPO individuals in the initial ICPO population according to the fitness function; S2-5-4: According to the initial fitness value and the updated ICPO population parameters, the first defense strategy, the second defense strategy, the third defense strategy and the fourth defense strategy are used to update the initial ICPO population to obtain an updated ICPO population; The formula for the first defense strategy is:

[0024] In the formula, For the updated ICPO individuals within the first defense range; The initial ICPO individual within the first defense range; is a random number based on normal distribution; is a random value in the interval [0,1]; It is the optimal solution within the first defense range; is the vector generated between the true optimal solution within the first defense range and the optimal solution randomly selected from the ICPO population; It is the ICPO individual indicator; is the iteration indicator; The formula for the second defense strategy is:

[0025] In the formula, For the updated ICPO individuals within the second defense range; The initial ICPO individual within the second defense range; is the search upper limit vector of the second defense range; is a random value in the interval [0,1]; Respectively Initial ICPO individuals; All are [1, ]; is the vector generated between the true optimal solution within the second defense range and the optimal solution randomly selected from the ICPO population; The formula for the third defense strategy is:

[0026] In the formula, For the updated ICPO individuals within the third defense range; The initial ICPO individual within the third defense range; is the search upper limit vector of the third defense range; Respectively Initial ICPO individuals; is [1, ]; The odor diffusion factor defined for the fitness function; It is a defense factor; Control parameters for search direction; The formula for the fourth defense strategy is:

[0027] In the formula, For the updated ICPO individuals within the fourth defense range; The initial ICPO individual within the fourth defense range; It is the optimal solution within the fourth defense range; All are random values ​​in the interval [0,1]; It is a defense factor; Control parameters for search direction; is the average force affecting the search direction; is the convergence speed factor; S2-5-5: Use the dynamic reverse learning algorithm to perform dynamic reverse learning on the updated ICPO population to generate a dynamic reverse ICPO population; The formula is:

[0028] In the formula, It is a dynamically reversed ICPO individual; is the decreasing inertia coefficient; are the maximum and minimum values ​​of the vector space respectively; For the updated ICPO individual; S2-5-6: According to the fitness function, calculate the fitness values ​​of all ICPO individuals in the updated ICPO population and the dynamically reversed ICPO population, take the ICPO individual with the minimum fitness value as the optimal individual, and retain the optimal individual; S2-6: If the number of iterations of iterative optimization reaches the maximum number of iterations or the fitness value of the optimal individual meets the requirements, the individual encoding vector of the optimal individual is decoded to obtain a real-time computing task allocation plan; The real-time computing task allocation scheme includes a computing task assigned to each edge computing gateway in the distributed edge cluster, and the computing task is that the edge computing gateway pre-processes a number of real-time industrial sensor data received; S3: Distributed edge cluster, according to the real-time computing task allocation scheme, pre-processes some real-time industrial sensor data, obtains some pre-processed industrial sensor data, and encrypts and transmits it to the cloud data center, including the following steps: S3-1: Distributed edge cluster, according to the real-time computing task allocation scheme, distributes a number of real-time industrial sensor data to a number of corresponding target edge computing gateways; S3-2: The target edge computing gateway cleans the allocated real-time industrial sensor data to obtain the cleaned real-time industrial sensor data; S3-3: The target edge computing gateway performs graph structure conversion on a number of cleaned real-time industrial sensor data to obtain a number of real-time industrial sensor data after graph structure conversion, including the following steps: S3-3-1: The target edge computing gateway divides the cleaned real-time image sensor data in the cleaned real-time industrial sensor data into pixel blocks to obtain a number of real-time image pixel blocks; S3-3-2: taking the real-time image information of the real-time image sensor data after cleaning and the real-time pixel block information of the real-time image pixel block as the first object node of the OPC UA information model; S3-3-3: Use the real-time pixel information of the real-time image pixel block as the first monitoring node of the OPC UA information model, and use the real-time pixel value of the real-time image pixel block as the first monitoring attribute of the OPC UA information model; S3-3-4: taking the real-time spatial relationship between the real-time image pixel blocks as the first node relationship of the OPC UA information model, and obtaining the corresponding real-time image sensor data after graph structure conversion; S3-3-5: performing graph structure conversion on the cleaned real-time non-image sensor data in the cleaned real-time industrial sensor data to obtain corresponding graph structure converted real-time non-image sensor data; S3-3-6: Integrate a number of real-time image sensor data after graph structure conversion and a number of real-time non-image sensor data after graph structure conversion to obtain a number of real-time industrial sensor data after graph structure conversion; S3-4: The target edge computing gateway aggregates the real-time industrial sensor data after graph structure conversion generated by all target edge computing gateways to obtain a number of pre-processed real-time industrial sensor data; S3-5: using an asymmetric encryption algorithm, encrypting a number of pre-processed real-time industrial sensor data, obtaining a number of encrypted pre-processed industrial sensor data, and transmitting the encrypted data to a cloud data center; S4: The cloud data center writes some pre-processed real-time industrial sensor data into the OPC UA instance, extracts the corresponding real-time OPC UA dynamic data, and uses the data analysis model to perform data analysis to obtain real-time data analysis results, including the following steps: S4-1: The cloud data center uses an asymmetric decryption algorithm to decrypt the encrypted preprocessed real-time industrial sensor data to obtain the decrypted preprocessed industrial sensor data; S4-2: write the decrypted pre-processed industrial sensor data into the OPC UA instance, and extract the real-time OPC UA dynamic data of the OPC UA instance when the data analysis cycle arrives; S4-3: performing standardization processing on the real-time OPC UA dynamic data to obtain standardized real-time OPC UA dynamic data, and inputting the standardized real-time OPC UA dynamic data into the data analysis model; S4-4: Use the graph feature extraction module of the data analysis model to extract the real-time graph features of the standardized real-time OPC UA dynamic data, and use the attention weight module to weight the real-time graph features according to the preset attention weight value to obtain the real-time weighted features; S4-5: Perform data analysis using a data analysis module based on the real-time weighted features to obtain corresponding real-time data analysis results; S5: The cloud data center uses the fault repair guidance model to provide fault repair guidance based on the real-time data analysis results, and sends the obtained real-time fault repair solution to the distributed edge cluster, including the following steps: S5-1: The cloud data center inputs the real-time data analysis results output by the data analysis model into the fault maintenance guidance model, and extracts corresponding real-time fault maintenance guidance experience based on the real-time data analysis results; S5-2: According to the real-time fault maintenance guidance experience, the parameters of the fault maintenance guidance model are adjusted to obtain an adjusted fault maintenance guidance model, including the following steps: S5-2-1: Based on the real-time meta-strategy optimization experience in the real-time fault repair guidance experience, use the meta-strategy optimization module to generate real-time initialization network parameters; S5-2-2: adjusting the reinforcement learning module according to the initialized network parameters to obtain an adjusted reinforcement learning module, i.e., an adjusted fault maintenance guidance model; S5-3: Based on the real-time data analysis results and the real-time fault maintenance guidance experience, the adjusted fault maintenance guidance model is used to provide fault maintenance guidance and obtain a real-time fault maintenance plan, including the following steps: S5-3-1: Based on the real-time reinforcement learning experience in the real-time fault maintenance guidance experience, the action space of the reinforcement learning module is updated and adjusted to obtain an updated action space ,in, For the updated action, is the action indicator, is the total number of action space dimensions; S5-3-2: Update and adjust the state space of the reinforcement learning module according to the real-time data analysis results to obtain an updated state space ,in, For the updated state, is the state indicator, is the total number of state space dimensions; S5-3-3: Input the updated state space into the adjusted reinforcement learning module, and generate the Q value of each possible action in the updated action space using the adjusted deep Q network based on the updated state space; The formula is:

[0029] In the formula, For updated status and update actions The corresponding updated Q value; Status and actions The corresponding predicted Q value; is the learning rate; is the highest predicted Q value; It is a comprehensive indicator; To update the parameters; Action-based The status Transition to the updated state The reward value of S5-3-4: Repeat the above Q value update steps until the number of iterations reaches the iteration number threshold, use the greedy strategy to select the possible action with the highest updated Q value as the execution action, and obtain the real-time fault repair plan based on the execution action; Real-time fault repair solutions include preventive maintenance actions for physical entities that have failed (periodic inspections: perform regular inspections according to the manufacturer's recommendations, including A, B, C, D inspections, etc.; component replacement: replace consumable parts based on usage time and periodic requirements; fluid replenishment: replenish fuel, hydraulic oil, lubricating oil, coolant, etc.; system testing: perform functional tests on various systems of physical entities, etc.), fault repair actions (fault diagnosis: determine the cause of the fault based on fault codes and warning information; component replacement: replace faulty or degraded components, such as engine components, sensors, electronic components, etc.; troubleshooting: repair the fault, which may include software updates, hardware adjustments, or reconfiguration of the system;), modification and upgrade actions (hardware upgrades: update the hardware components of the physical entity to improve performance or comply with new regulatory requirements; software updates: install the latest onboard software updates for the physical entity; structural modifications: modify the structure of the physical entity to extend its service life;), etc.; S5-4: Use an asymmetric encryption algorithm to encrypt the real-time fault repair plan, obtain the encrypted real-time fault repair plan, and return it to the distributed edge cluster.

[0030] Embodiment 2: like Figure 2 As shown, this embodiment provides an analysis system for industrial sensor data, which is used to implement an analysis method for industrial sensor data. The system includes a cloud data center, a plurality of distributed edge clusters, and a plurality of industrial sensors. The cloud data center is respectively connected to the plurality of distributed edge clusters for communication, each distributed edge cluster is respectively connected to the plurality of industrial sensors within the communication range for communication, and the distributed edge cluster includes a plurality of edge computing gateways connected in a distributed manner, and the plurality of industrial sensors are arranged at corresponding physical entities; Industrial sensors are used to collect real-time industrial sensor data of physical entities and transmit them to distributed edge clusters within the communication range; The distributed edge cluster allocates computing tasks according to a number of real-time industrial sensor data collected by a number of industrial sensors to obtain a real-time computing task allocation plan; pre-processes a number of real-time industrial sensor data according to the real-time computing task allocation plan to obtain a number of pre-processed industrial sensor data, and encrypts and transmits the pre-processed industrial sensor data to the cloud data center; The cloud data center includes an initialization unit, a data analysis unit, and a fault repair guidance unit connected in sequence; The initialization unit is used to configure a number of corresponding industrial sensors according to different physical entities that need to be monitored, set up OPC UA instances, and build data analysis models and fault maintenance guidance models; A data analysis unit is used to write a number of pre-processed real-time industrial sensor data into the OPC UA instance, extract the corresponding real-time OPC UA dynamic data, and use the data analysis model to perform data analysis to obtain real-time data analysis results; The fault repair guidance unit is used to provide fault repair guidance based on the real-time data analysis results and use the fault repair guidance model, and send the obtained real-time fault repair plan to the distributed edge cluster.

[0031] The present invention discloses an analysis method and system for industrial sensor data. The constructed data analysis model can realize the automatic analysis of massive industrial sensor data, improve the data processing capability, and the data analysis model can mine the potential connection between a number of industrial sensor data and between the industrial sensor data and the abnormalities and failures of physical entities, extract valuable information from massive data, and improve the accuracy and timeliness of data analysis results; by configuring corresponding industrial sensors and constructing OPC UA instances, it is possible to realize effective monitoring of different physical entities, improve the pertinence and efficiency of data collection, and build a unified OPC The UA information model realizes standardized conversion between different data sources, simplifies the data processing process, and improves processing efficiency. Through the standard data interaction processing process, it can quickly respond to the writing and query of real-time data to meet real-time requirements; it adopts edge-cloud collaborative technology to delegate computing tasks to the edge computing gateway for processing, which reduces the hardware requirements and computing pressure of the cloud data center, avoids accidents such as hardware crashes that lead to data loss, and allocates computing tasks to distributed edge clusters, reducing the data processing pressure and computing resource overload of a single edge computing gateway, and improving the reliability and security of data analysis; by constructing a fault maintenance guidance model, a response mechanism is realized, which can generate fault maintenance strategies for guidance based on data analysis results, thereby improving the level of intelligence.

[0032] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for analyzing industrial sensor data, characterized in that: The steps include: The cloud data center configures a number of corresponding industrial sensors according to the different physical entities that need to be monitored, sets up OPC UA instances, and builds data analysis models and fault maintenance guidance models; The distributed edge cluster allocates computing tasks based on a number of real-time industrial sensor data collected by a number of industrial sensors to obtain a real-time computing task allocation plan; The distributed edge cluster pre-processes a number of real-time industrial sensor data according to the real-time computing task allocation scheme, obtains a number of pre-processed industrial sensor data, and encrypts and transmits them to the cloud data center; The cloud data center writes some pre-processed real-time industrial sensor data into the OPC UA instance, extracts the corresponding real-time OPC UA dynamic data, and uses the data analysis model to perform data analysis to obtain real-time data analysis results; The cloud data center uses the fault repair guidance model to provide fault repair guidance based on the real-time data analysis results, and sends the obtained real-time fault repair plan to the distributed edge cluster.

2. The method for analyzing industrial sensor data according to claim 1, characterized in that: The cloud data center configures several corresponding industrial sensors according to the different physical entities that need to be monitored, sets up OPC UA instances, and builds data analysis models and fault maintenance guidance models, including the following steps: The cloud data center confirms the monitoring indicators of several physical entities that need to be monitored, and configures several corresponding industrial sensors according to the monitoring indicators of each physical entity; Design the OPC UA information model based on the configuration parameters of all industrial sensors of all physical entities and build the corresponding OPC UA instance in the address space of the OPC UA server; Based on the OPC UA instance, some historical OPC UA dynamic data are generated and standardized to obtain some standardized historical OPC UA dynamic data; Based on some standardized historical OPC UA dynamic data, a deep learning algorithm is used to build a data analysis model and generate some historical data analysis results; Based on the analysis results of several historical data, an enhanced reinforcement learning algorithm is used to build a fault maintenance guidance model.

3. The method for analyzing industrial sensor data according to claim 2, characterized in that: The data analysis model is constructed based on the GCN-Attention-MLP algorithm, and the data analysis model includes a graph feature extraction module constructed based on the GCN algorithm, an attention weight module constructed based on the Attention mechanism, and a data analysis module constructed based on the MLP algorithm, which are connected in sequence; The fault maintenance guidance model is constructed based on the MPO-DQN algorithm, and the fault maintenance guidance model includes a meta-strategy optimization module constructed based on the MPO algorithm and a reinforcement learning module constructed based on the DQN algorithm. The reinforcement learning module is provided with an intelligent agent, a deep Q network and an experience replay pool. The intelligent agent is respectively connected to the deep Q network, the experience replay pool and the meta-strategy optimization module, and the meta-strategy optimization module is connected to the experience replay pool.

4. The method for analyzing industrial sensor data according to claim 3, characterized in that: According to the configuration parameters of all industrial sensors of all physical entities, the OPC UA information model is designed, and the corresponding OPC UA instance is constructed in the address space of the OPC UA server, including the following steps: According to the object type in the configuration parameters of the industrial sensor of the physical entity, a plurality of object nodes including all the industrial sensors are set; the object nodes include a first object node of the image sensor and a second object node of the non-image sensor; According to the monitoring type in the configuration parameters, a plurality of first monitoring nodes are set for the first object node, and according to the monitoring attribute in the configuration parameters, a corresponding second monitoring attribute is set for each first monitoring node; According to the monitoring type in the configuration parameters, set a plurality of second monitoring nodes for the second object node, and according to the monitoring attribute in the configuration parameters, set a corresponding second monitoring attribute for each second monitoring node; Setting corresponding first node relationships for a number of first object nodes and a number of first monitoring nodes, and setting second node relationships for a number of second object nodes and a number of second monitoring nodes; Traversing all industrial sensors of all physical entities, constructing an OPC UA information model according to all first object nodes, all first monitoring nodes, all second node relationships, all second object nodes, all second monitoring nodes, and all second node relationships; According to the model metadata of the OPC UA information model, a corresponding OPC UA instance is constructed in the address space of the OPC UA server.

5. The method for analyzing industrial sensor data according to claim 4, characterized in that: The distributed edge cluster allocates computing tasks according to a number of real-time industrial sensor data collected by a number of industrial sensors, and obtains a real-time computing task allocation plan, including the following steps: Industrial sensors collect real-time industrial sensor data of corresponding physical entities and transmit the real-time industrial sensor data to distributed edge clusters within the communication range; Distributed edge cluster, confirms the distributed edge computing gateway set participating in computing task allocation, defines the computing task set, and takes minimizing the computing task execution time and computing resource cost as the optimization goal; According to the optimization target and a number of real-time industrial sensor data collected by a number of industrial sensors, a fitness function of the swarm intelligence optimization algorithm is defined; According to the computing task set and edge computing gateway set, define the individual encoding vector format of the ICPO algorithm, and set the ICPO population parameters and maximum number of iterations of the ICPO algorithm; Based on the individual encoding vector format, ICPO population parameters and fitness function, the ICPO algorithm is used to iteratively search for the best result and retain the best individual. If the number of iterations of iterative optimization reaches the maximum number of iterations or the fitness value of the optimal individual meets the requirements, the individual encoding vector of the optimal individual is decoded to obtain a real-time computing task allocation plan.

6. The method for analyzing industrial sensor data according to claim 4, characterized in that: The distributed edge cluster pre-processes a number of real-time industrial sensor data according to the real-time computing task allocation scheme, obtains a number of pre-processed industrial sensor data, and encrypts and transmits them to the cloud data center, including the following steps: The distributed edge cluster distributes a number of real-time industrial sensor data to a number of corresponding target edge computing gateways according to the real-time computing task allocation plan; The target edge computing gateway cleans the allocated real-time industrial sensor data to obtain the cleaned real-time industrial sensor data. The target edge computing gateway performs graph structure conversion on a number of cleaned real-time industrial sensor data to obtain a number of real-time industrial sensor data after graph structure conversion; The target edge computing gateway aggregates the real-time industrial sensor data after graph structure conversion generated by all target edge computing gateways to obtain a number of pre-processed real-time industrial sensor data; An asymmetric encryption algorithm is used to encrypt a number of pre-processed real-time industrial sensor data to obtain encrypted a number of pre-processed industrial sensor data and transmit them to a cloud data center.

7. The method for analyzing industrial sensor data according to claim 6, characterized in that: The target edge computing gateway performs graph structure conversion on a number of cleaned real-time industrial sensor data to obtain a number of real-time industrial sensor data after graph structure conversion, including the following steps: The target edge computing gateway performs pixel block division on the cleaned real-time image sensor data in the cleaned real-time industrial sensor data to obtain a plurality of real-time image pixel blocks; The real-time image information of the real-time image sensor data after cleaning and the real-time pixel block information of the real-time image pixel block are used as the first object node of the OPC UA information model; The real-time pixel information of the real-time image pixel block is used as the first monitoring node of the OPC UA information model, and the real-time pixel value of the real-time image pixel block is used as the first monitoring attribute of the OPC UA information model; The real-time spatial relationship between real-time image pixel blocks is used as the first node relationship of the OPC UA information model to obtain the corresponding real-time image sensor data after graph structure conversion; Performing graph structure conversion on the cleaned real-time non-image sensor data in the cleaned real-time industrial sensor data to obtain corresponding graph structure converted real-time non-image sensor data; A number of real-time image sensor data after graph structure conversion and a number of real-time non-image sensor data after graph structure conversion are integrated to obtain a number of real-time industrial sensor data after graph structure conversion.

8. The method for analyzing industrial sensor data according to claim 6, characterized in that: The cloud data center writes some pre-processed real-time industrial sensor data into the OPC UA instance, extracts the corresponding real-time OPC UA dynamic data, and uses the data analysis model to perform data analysis to obtain real-time data analysis results, including the following steps: The cloud data center uses an asymmetric decryption algorithm to decrypt the encrypted preprocessed real-time industrial sensor data to obtain the decrypted preprocessed industrial sensor data; Write the decrypted pre-processed industrial sensor data into the OPC UA instance, and extract the real-time OPC UA dynamic data of the OPC UA instance when the data analysis cycle arrives; Performing standardization processing on the real-time OPC UA dynamic data to obtain standardized real-time OPC UA dynamic data, and inputting the standardized real-time OPC UA dynamic data into a data analysis model; Use the data analysis model to extract the real-time graph features of the standardized real-time OPC UA dynamic data, and weight the real-time graph features according to the preset attention weight values ​​to obtain real-time weighted features; According to the real-time weighted features, data analysis is performed to obtain corresponding real-time data analysis results.

9. The method for analyzing industrial sensor data according to claim 3, characterized in that: The cloud data center uses the fault repair guidance model to provide fault repair guidance based on the real-time data analysis results, and sends the obtained real-time fault repair plan to the distributed edge cluster, including the following steps: The cloud data center inputs the real-time data analysis results output by the data analysis model into the fault maintenance guidance model, and extracts corresponding real-time fault maintenance guidance experience based on the real-time data analysis results; According to the real-time fault maintenance guidance experience, the parameters of the fault maintenance guidance model are adjusted to obtain an adjusted fault maintenance guidance model; According to the real-time data analysis results and the real-time fault maintenance guidance experience, the adjusted fault maintenance guidance model is used to provide fault maintenance guidance and obtain a real-time fault maintenance plan; The real-time fault repair plan is encrypted using an asymmetric encryption algorithm to obtain the encrypted real-time fault repair plan, which is then returned to the distributed edge cluster.

10. An industrial sensor data analysis system, used to implement the industrial sensor data analysis method according to any one of claims 1 to 9, characterized in that: The system includes a cloud data center, a plurality of distributed edge clusters and a plurality of industrial sensors, wherein the cloud data center is respectively connected to the plurality of distributed edge clusters for communication, each of the distributed edge clusters is respectively connected to the plurality of industrial sensors within the communication range for communication, and the distributed edge cluster includes a plurality of edge computing gateways connected in a distributed manner, and the plurality of industrial sensors are arranged at corresponding physical entities; The cloud data center includes an initialization unit, a data analysis unit, and a fault repair guidance unit which are connected in sequence.

Images