Intelligent weak current system fault prediction method and system based on edge calculation

Abstract

The invention discloses an intelligent weak current system fault prediction method and system based on edge computing, and belongs to the technical field of edge computing, and the method comprises the following steps: S1, carrying out the fault correlation analysis of a dependency relationship and a fault between micro-services in an intelligent weak current system according to a knowledge graph; s2, completing dependency relationship analysis and fault propagation direction pre-judgment tasks among multiple micro-services of the intelligent weak current system; s3, performing intelligent dynamic micro-service unit combination on the system according to business requirements, and dividing the whole logic processing process and the data communication process into one or a plurality of state machines; and S4, based on adaptive service orchestration and fault isolation, and based on an incremental learning and model evolution management mechanism, completing management of complex service orchestration update. The prediction model combines the incidence relation between the complex intelligent weak current systems, predicts the possible directions of multiple faults at the same time, and automatically judges, isolates and repairs fault services according to the propagation directions of the multiple faults.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent weak current system fault prediction methods, and particularly to an intelligent weak current system fault prediction method and system based on edge computing. Background Art

[0002] The current microservice application scenarios require the support of big data technology. In fault diagnosis and prediction, since microservices are deployed in a large number of distributed devices and communicate with each other, the data volume is huge, the data access is frequent, and the CPU resources are severely consumed. Conventional big data technologies cannot meet the requirements of high processing business time and complex business processes. With the gradual popularization of weak current intelligent systems in industrial production, in order to facilitate the management and monitoring and operation and maintenance of weak current systems, the fault prediction model for weak current systems has become an inevitable research direction.

[0003] However, the fault transmission of intelligent weak current systems has latency, and there are dependencies between services. If the data flow of fault data cannot flow in the designed manner, it will affect the next judgment. However, since the data flow of fault data is random, for complex system structures, it is necessary to design an analysis and prediction technology for strong coupling microservice fault propagation and fault determination with strong fault tolerance to achieve the purpose of real-time processing of fault occurrence information and provide a solid foundation for service fault diagnosis and decision support.

[0004] In addition, although the existing real-time fault data acquisition devices can improve the fault judgment accuracy by collecting multi-terminal fault data in real time, they are not applicable to the problems of strong coupling microservices, and the influence of the data flow on strong coupling microservices also needs to be considered. Moreover, due to the large number of microservices, when a large number of fault points need to be uploaded in microservices, the dependency relationship between fault points will affect the processing speed of the overall fault prediction method, bringing inconvenience to subsequent fault analysis and data backup. Summary of the Invention

[0005] In order to make the prediction method more accurate, the present invention proposes an intelligent weak current system fault prediction method and system based on edge computing. Aiming at the complex service interaction and dependency relationships formed by multi-service and multi-data interaction methods of intelligent weak current systems, the automatic fault association mode based on machine learning is used to perceive the complex service association relationships and dependency relationships, and a describable complex service fault propagation direction is formed, effectively solving the problems of unknowability and unpredictability of weak current system faults.

[0006] The present invention provides a method for predicting faults in an intelligent weak current system based on edge computing, including the following steps: Step S1. Perform fault correlation analysis on the dependency relationships and fault propagation between microservices in the intelligent weak current system according to a knowledge graph; Step S2. Complete the analysis of the dependency relationships between multiple microservices in the intelligent weak current system and the task of predicting the fault propagation direction based on the service - to - service dependency relationship analysis and fault propagation prediction method of a directed acyclic graph (DAG); S3. Combine the system into intelligent dynamic microservice units according to business requirements. For each microservice unit in the actual system, it includes two parts: a state machine and a data flow. The state machine is a microservice unit logically defined based on the microservice splitting of the system. According to its logical processing process and the data communication logic between microservice units during the logical processing process, the entire logical processing process and data communication process are divided into one or several state machines, and the state machine is used as the basic unit for microservice orchestration and management; Step S4. Complete the management of complex service orchestration updates based on adaptive service orchestration and fault isolation, and an incremental learning and model evolution management mechanism.

[0007] Further, it also includes the following steps: Step S1 also includes collecting the data flows of various service interactions between microservices to implement a microservice process oriented to the data flow; generating decision information from the microservice process oriented to the data flow, extracting fault data from the data flow, and classifying the fault data; where the microservice process includes the storage of dependency relationships between microservice units and a fault propagation model; finally, converting the classified data into corresponding decision types; the service interaction is the data interaction between microservices, and the interaction methods include data transmission of distributed service execution, asynchronous communication of publish / subscribe services, and batch - processing workflows and stateful distributed transaction operations.

[0008] Further, the task of predicting the fault propagation direction is predicted through a fault prediction model, and the specific steps are as follows: Predict the fault propagation direction through the dependency relationship data between service pairs, so as to predict a series of possible fault propagation directions. Specifically: The dependency relationship analysis between associated microservices is mainly described by a directed acyclic graph (DAG). Each microservice is defined as a node, and the directed edge of the node represents the mutual dependency relationship between services. The entire system is a DAG graph composed of multiple interconnected microservices; Define \(s_0\) and \(s_n\) as the starting service node and the ending service node of the fault propagation path \(P\) respectively, and the number of object nodes covered by the fault propagation path \(P\) is \(n\); The starting service node generates an initial fault propagation path \(p_0\) and adds \(p_0\) to a queue \(r\) with a global sorting of 1; When \(r\) is not empty, dequeue a fault propagation path \(p\) from the head of the queue \(r\), and then select an uncovered object node \(n\) from \(p\) i , and add \(n\) iThe covered service node is added to the fault propagation path p; at this time, n is updated i The number of covered nodes. If n i is the end service node, then the fault propagation path p is added to the result fault propagation path set G; if n i is an intermediate service node, then another covered object is taken; when all the covered objects in the fault propagation path p are marked as covered objects, the fault propagation path p is added to the result fault propagation path set G, and then the next uncovered object is selected; if p is an empty set, then it is judged whether the queue r is empty. If it is empty, the fault prediction is exited; if it is not empty, the next object in the queue r is continued to be processed, and the above steps are repeated; the fault propagation path set G is the predicted fault propagation direction.

[0009] Furthermore, there are multiple concurrent relationships between the state machines in step S3, including sequential, parallel, aggregative, selective, and nested; among them, in the sequential state, the microservice needs to go through multiple states from the start state to the end state; in the parallel state, multiple microservices can be executed simultaneously and output results; in the aggregative state, one state needs to aggregate the states of multiple microservices; in the selective state, different selections are made according to the different output results of a microservice; the nested state is a complex state in which multiple states are nested with each other, and the output of each state can cause the switching of the output state.

[0010] Furthermore, the state machine includes states, input response processes, and state migrations.

[0011] An intelligent weak current system fault prediction system based on edge computing, including a fault propagation direction prediction module and a microservice unit management module; the fault propagation direction prediction module includes a model training module for training a prediction model for the intelligent weak current system prediction; a service distributed fault point prediction module for predicting multiple fault points where the intelligent weak current system fails through the prediction model; an incremental prediction model production module for performing incremental training and model evolution on each microservice unit through the microservice unit management; the microservice unit management module includes: a fault management module, an isolation policy monitoring module for fault isolation between microservice units; a fault analysis module for data interaction between microservice units to perform fault location and fault repair; a history management module for recording the prediction history data generated by the prediction model; a model management module for storing the trained prediction model and its related information.

[0012] Furthermore, it also includes a rule calculation module, which includes a business rule management module and a prediction rule management module.

[0013] Further, the business rule management module is used to collect data streams and convert the data streams into data of corresponding decision types; the prediction rule management module is used to perform association judgment and logical judgment between microservices.

[0014] Further, the collection and conversion of data streams by the business rule management module: after the business rule management unit collects data from the data stream, it further formats the data into the standard format of the corresponding decision type; and sends it to the prediction rule management unit; the business rule management module includes a mapping rule management unit and a standard format definition unit. Mapping rule management is the rule management in which the source data and the standard format data have a mapping relationship. When the system is extended to new types of data, the mapping between the source data and the standard format is also required; the standard format definition unit defines a set of standard data formats, and the standard format definition unit includes data entity definition, field definition, and the definition of the association relationship between data entities.

[0015] Further, the prediction rule management module includes: a dependency analysis sub-module that calculates the dependencies between microservices; a dependency storage sub-module that stores the dependencies between services obtained by the dependency analysis sub-module and forms a service dependency graph, where each microservice is a node, and the dependency relationship between the node and the service is a solid or dashed line graph between services, and the dashed line and the solid line represent the data stream and the control command stream respectively; a fault propagation direction prediction sub-module that analyzes the occurrence of faults in sequence according to the dependencies between services; a data stream analysis sub-module that analyzes the data stream between microservices; a fault relationship pre-judgment that pre-judges the fault propagation direction through the dependency relationship and data communication relationship between services.

[0016] Further, the microservice unit management module further includes a microservice orchestration module, and the specific steps are as follows: S1. Orchestrate the state machine configuration table according to the state machine orchestration model and the dependencies between state machines, and load the state machine configuration table into the microservice orchestration engine, and the microservice orchestration engine binds each state machine according to the configuration table; S2. Start the state machine execution process of the microservice orchestration engine, and automatically execute the state machine execution process according to the state machine orchestration model and the dependencies between state machines, and incorporate the data stream and operation process between microservices into the state machine control; S3. When triggering the state node migration, the state machine calculates the data access and data output involved after the state migration, and analyzes the state machine dependency relationship, and matches the dependency relationship between state machines with other state machines, thereby triggering the state machine migration of the dependent state machine.

[0017] Furthermore, it also includes a microservice update management module, which consists of a microservice process engine module, a fault isolation strategy module, and a model evolution module; the microservice process engine module is responsible for the logical processing of microservice units by each state machine during the microservice orchestration process, that is, the control and management of microservice units by the state machine, including start, stop, rollback, synchronization, and asynchrony, and the management object is the state machine; the fault isolation strategy module is responsible for isolating the fault point according to the detected abnormal conditions of each microservice, in accordance with the established fault propagation direction and fault propagation priority; the model evolution module updates and evolves the prediction model; when a new state machine is generated in the state machine management data, or a new prediction model is generated by the model evolution module, it will trigger the update of model evolution; during the update, when a certain state machine is added or deleted, it will trigger the update of the state machine configuration table; when other data in the state machine configuration table is added, deleted, or modified, it will trigger the update of the corresponding model of the state machine configuration.

[0018] Advantages and beneficial effects of the present invention: Compared with the prior art, the prediction model of the present invention combines the correlation relationships between complex intelligent weak current systems, dynamically constructs the system to simultaneously predict the possible directions of multiple faults, and automatically judges, isolates, and repairs faulty services according to multiple fault propagation directions. It fully considers the actual business characteristics and the ease of use of system construction, and can comprehensively analyze large-scale microservices. In addition, the present invention realizes data flow prediction oriented to data flow to accurately analyze the faults of each microservice unit in the system, and truly applies the complex service correlation analysis of complex intelligent weak current systems to practice. Description of the drawings

[0019] Figure 1 It is a flowchart of the intelligent weak current system fault prediction method based on edge computing of the present invention.

[0020] Figure 2 It is a flowchart of the fault propagation direction prediction algorithm of the present invention.

[0021] Figure 3 It is a structure diagram of the intelligent weak current system fault prediction system based on edge computing of the present invention.

[0022] Figure 4 It is a logic block diagram of the microservice unit management module of the present invention.

[0023] Figure 5 It is a logic block diagram of the rule calculation module of the present invention.

[0024] Figure 6 It is a logic block diagram of the microservice update management module of the present invention.

[0025] Figure 7 It is a logic block diagram of the microservice orchestration module of the present invention. Detailed Implementation Manner

[0026] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the following combines the accompanying drawings and preferred embodiments to specifically describe the implementation manner, structure, features, and their effects in detail. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0027] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs. Embodiment

[0028] Refer to Figure 1 , the intelligent weak current system fault prediction method based on edge computing provided in this embodiment includes the following steps: Step S1. Perform fault correlation analysis on the dependency relationship and fault propagation between microservices in the intelligent weak current system according to the knowledge graph; Step S2. Complete the dependency relationship analysis and fault propagation direction prediction tasks between multiple microservices in the intelligent weak current system based on the service - to - service dependency relationship analysis and fault propagation prediction method of the DAG (Directed Acyclic Graph); Step S3. Combine the system into intelligent dynamic microservice units according to business requirements. For each microservice unit in the actual system, it includes two parts: a state machine and a data stream. The state machine is a microservice unit logically defined based on the microservice splitting of the system. According to its logical processing process and the data communication logic between microservice units during the logical processing process, the entire logical processing process and data communication process are divided into one or several state machines, and the state machine is used as the basic unit for microservice orchestration and management; Step S4. Complete the management of complex service orchestration updates based on adaptive service orchestration and fault isolation, and based on incremental learning and model evolution management mechanisms.

[0029] The method proposed by the present invention makes full use of the advantages of edge computing and decomposes the complex fault prediction task into four key steps. First, in step S1, the knowledge graph technology is used to conduct fault correlation analysis on the microservices in the intelligent weak current system. For example, in the intelligent weak current system of a large commercial complex, there may be multiple subsystems such as an access control system, a video surveillance system, and a fire alarm system. Through the knowledge graph, the complex relationships between various microservices in these subsystems can be clearly described. For instance, the identity verification service of the access control system may have a close association with the visitor management system, while the fire alarm system may have a linkage relationship with the air conditioning control system. In this way, the mutual influence between various parts of the system can be more comprehensively understood, laying a foundation for subsequent fault prediction.

[0030] Next, in step S2, an analysis method based on a directed acyclic graph (DAG) is introduced. This method is particularly suitable for describing the complex dependency relationships in the intelligent weak current system. For example, in the weak current system of an intelligent building, each microservice can be regarded as a node in the DAG, and the dependency relationships between services are represented by directed edges. In this way, the upstream and downstream services in the system can be clearly shown, enabling more accurate prediction of the fault propagation direction. For example, if an abnormality is detected in a certain service in the energy management system, it can be quickly determined which downstream services, such as the lighting control or air conditioning system, may be affected, and corresponding preventive measures can be taken.

[0031] Step S3 introduces the concept of combining dynamic microservice units, which is of great significance in the intelligent weak current system. By dividing the system into a state machine and a data stream, the complex system can be managed more flexibly. For example, in an intelligent parking lot system, functions such as vehicle entry recognition, vacant space allocation, and toll processing can be divided into different microservice units. Each unit contains its specific state machine (such as the vacant space allocation service may include states such as searching for vacant spaces, reserving vacant spaces, and confirming allocations) and data stream (such as license plate information, vacant space information, etc.). This division method makes the system easier to expand and maintain.

[0032] Finally, in step S4, by introducing an adaptive service orchestration and fault isolation mechanism, as well as an incremental learning and model evolution management mechanism, the system can continuously optimize its own performance. For example, if the system detects that the performance of a certain microservice (such as the face recognition service in the access control system) starts to decline, it can automatically adjust the service orchestration, temporarily redirect some requests to the standby fingerprint recognition service, and at the same time start a fault isolation program to prevent the problem from spreading. In addition, by continuously learning new data, the system can continuously optimize its prediction model and improve the accuracy of fault prediction.

[0033] This method not only improves the accuracy of fault prediction but also greatly enhances the flexibility and maintainability of the system. By promptly predicting and isolating potential faults, the system downtime can be significantly reduced, and the reliability and stability of the entire intelligent weak current system can be improved.

[0034] The step S1 further includes collecting the data streams of various service interactions between microservices to implement a microservice process oriented to data streams; generating decision-making information from the microservice process oriented to data streams, extracting fault data from the data streams, and classifying the fault data; where the microservice process includes the storage of dependencies between microservice units and a fault propagation model; finally, converting the classified data into corresponding decision types after classification; the service interaction is the data interaction between microservices, and the interaction methods include data transmission of distributed service execution, asynchronous communication of publish / subscribe services, and workflow and stateful distributed transaction operations of batch processing.

[0035] In an intelligent weak current system, the collection and analysis of data streams are crucial. For example, in the weak current system of an intelligent office building, it may be necessary to collect service interaction data between multiple subsystems such as access control systems, lighting control systems, and air conditioning systems. This data may include employees' entry and exit records, lighting status in each area, indoor temperature changes, etc.

[0036] By implementing a microservice process oriented to data streams, the operating state of the system can be better understood. For example, by analyzing the data streams of the access control system and the lighting system, it can be found whether there are delays or errors in the process of the lights automatically turning on after an employee enters the office. Valuable decision-making information can be generated from these data streams, such as whether it is necessary to adjust the response time of certain services.

[0037] When extracting fault data, some abnormal patterns may be concerned. For example, if it is found that the air conditioning system frequently switches on and off in a short period of time, this may indicate a fault in the temperature sensor. Machine learning algorithms such as support vector machines (SVMs) or random forests can be used to classify these fault data. In practical applications, the kernel function of the SVM may be set to the radial basis function (RBF), where the parameter γ can be set to 0.1 and the penalty parameter C can be set to 1. The selection of these parameters is based on a large amount of experimental data and can achieve good classification results in most cases.

[0038] Finally, the classified data is converted into corresponding decision types. For example, for fault data classified as high risk, a decision type of emergency repair may be generated; while for low-risk fault data, a decision type of regular inspection may be generated.

[0039] This method can accurately extract valuable information from a vast amount of data streams and transform it into executable decisions. This not only improves the accuracy of fault prediction but also helps system administrators better understand the operating status of the system, thereby making more informed decisions.

[0040] See Figure 2 , the task of predicting the fault propagation direction in step S2 is predicted through a fault prediction model. The specific steps are as follows: The fault propagation direction is predicted through the dependency relationship data between services, so as to predict a series of possible fault propagation directions. Specifically: The dependency relationship analysis between related microservices is mainly described by a directed acyclic graph (DAG). Each microservice is defined as a node, and the directed edge of the node represents the mutual dependency relationship between services. The entire system is a DAG graph composed of multiple interconnected microservices; Define \(s_0\) and \(s_n\) as the starting service node and the ending service node of the fault propagation path P respectively. The number of object nodes covered by the fault propagation path P is N, and the initial defined value of N is 0; The starting service node generates an initial fault propagation path \(p_0\) and adds \(p_0\) to a queue r sorted globally as 1; When r is not empty, dequeue a fault propagation path p from the head of the queue r, and then select an uncovered object node n from p i , and add the service nodes covered by n i to the fault propagation path p; At this time, update the number of covered nodes of n i \(N = N + 1\). If n i is the ending service node, then \(N = N - 1\), and add the fault propagation path p to the result fault propagation path set G; If n i is an intermediate service node, then select another covered object; When all covered objects in the fault propagation path p are marked as covered objects, add the fault propagation path p to the result fault propagation path set G, and then continue to select the next uncovered object; If p is an empty set, then judge whether the queue r is empty. If it is empty, exit the fault prediction; If it is not empty, continue to process the next object in the queue r and repeat the above steps; The fault propagation path set G is the predicted fault propagation direction.

[0041] The present invention details the implementation method of the task of predicting the fault propagation direction based on DAG. In an intelligent weak current system, this method is particularly useful. For example, consider an intelligent weak current system in a large shopping mall, which includes multiple subsystems such as an access control system, a video surveillance system, and a fire protection system. The microservices in each subsystem can be regarded as a node in the DAG, and the dependency relationship between services is represented by a directed edge.

[0042] Specifically, assume that an abnormality is detected in the authentication service (referred to as node A) in the access control system. The goal is to predict how this fault might spread. First, set node A as the starting service node, and then start constructing possible fault propagation paths.

[0043] It may be found that the authentication service directly affects the visitor management service (node B) and the elevator control service (node C). Therefore, two initial paths: A->B and A->C are generated and added to the queue r. Then, the possible extensions of these paths will continue to be explored. For example, it may be found that the visitor management service is also related to the parking lot management service (node D), so a new path A->B->D is generated.

[0044] In this way, all possible fault propagation paths can be gradually constructed. In practical applications, a depth limit may be set, such as exploring at most 5 layers, to avoid overly long paths. In addition, a weight can be assigned to each edge to represent the probability of fault propagation. For example, it may be considered that the probability of the authentication service's fault propagating to the visitor management service is 0.8, while the probability of propagating to the elevator control service is only 0.3.

[0045] This method can not only help predict the possible fault propagation paths but also help identify key nodes and weak links in the system. For example, if it is found that a certain service node appears in most of the high-probability fault propagation paths, then this node may be a key point of the system and requires special attention and protection. In this way, system optimization and resource allocation can be carried out more targeted, thereby improving the reliability and stability of the entire intelligent low-voltage system.

[0046] There are various concurrent relationships among the state machines in step S3, including sequential, parallel, aggregative, selective, and nested types; among them, in the sequential state, the microservice needs to go through multiple states from the start state to the end state; in the parallel state, multiple microservices can be executed simultaneously and output results; in the aggregative state, one state needs to aggregate the states of multiple microservices; in the selective state, different choices are made according to the different output results of one microservice; the nested state is a complex state in which multiple states are nested with each other, and the output of each state can cause the switching of the output state.

[0047] The present invention details various concurrent relationships among the state machines, which is of great significance in the intelligent low-voltage system. Let's illustrate the application of these concurrent relationships through a comprehensive intelligent building management system.

[0048] First, sequential states are very common in intelligent low-voltage systems. For example, in the visitor management process, there may be the following sequence: authentication -> access permission confirmation -> elevator authorization -> destination floor recording. Each step is an independent microservice, and they are executed in a strict order. Only after the previous state is completed can the next state be entered. This sequential design ensures the security and traceability of visitor management.

[0049] Parallel states are very useful when multiple tasks need to be processed simultaneously. For example, when an employee swipes their card to enter the office building, the system may trigger multiple microservices simultaneously: the access control system records the entry time, the lighting system adjusts the workstation lighting according to personal preferences, and the air conditioning system starts to adjust the temperature, etc. These services can be executed in parallel, greatly improving the system's response speed and efficiency.

[0050] Aggregate states are useful when making decisions that require considering multiple factors. For example, in an intelligent fire protection system, the system may need to consider multiple factors such as smoke concentration, temperature change, and carbon dioxide concentration simultaneously. Only when the states of these microservices all meet certain conditions (such as the smoke concentration exceeding the threshold, the temperature rising sharply, and the carbon dioxide concentration being abnormal) will the system trigger a fire alarm. This aggregate design can effectively reduce false alarms and improve the reliability of the system.

[0051] Selective states allow the system to react differently according to different situations. For example, in an intelligent elevator system, according to the current passenger flow and elevator usage, the system may select different dispatching strategies. During peak hours, the system may choose a zoning operation strategy to improve efficiency; while during off-peak hours, it may choose an energy-saving operation strategy to reduce energy consumption.

[0052] Finally, nested states can be used to handle more complex scenarios. For example, in a comprehensive building management system, there may be a large state machine to manage the operating state of the entire building (such as normal operation, energy-saving mode, emergency state, etc.), and within each large state, there may be multiple small state machines nested to manage specific subsystems. For example, in an emergency state, the state machine of the fire protection system may be activated, while the state machines of other non-critical systems may be temporarily stopped.

[0053] This diverse design of concurrent relationships brings significant beneficial effects. First, it greatly improves the flexibility and adaptability of the system, enabling it to handle various complex scenarios and requirements. Second, it optimizes the system's performance, improving the response speed through parallel processing and the accuracy of decision-making through aggregation and selection mechanisms. Finally, it improves the maintainability and scalability of the system. Different concurrent relationships can be flexibly combined according to needs, facilitating system upgrades and expansions.

[0054] The state machine includes states, input response processes, and state transitions. The present invention further refines the components of the state machine. In the intelligent weak current system, this design is of great significance. Let's take the intelligent lighting control system as an example to illustrate.

[0055] First, the state is the core concept of the state machine. In the intelligent lighting system, there may be multiple states, such as off, on, dimming, energy-saving mode, etc. Each state represents a specific working mode of the lights.

[0056] The input response process is the reaction mechanism of the state machine to external stimuli. In the example, the inputs may include signals from the human body sensor, data from the light intensity sensor, time information, etc. For example, when the human body sensor detects someone entering, this input will trigger the response process of the state machine.

[0057] The state transition defines how the state machine transitions from one state to another. In the intelligent lighting system, the state transition rules are as follows: When it is detected that someone enters and the current light is insufficient, it transitions from the off state to the on state.

[0058] When it is detected that someone has left for more than 5 minutes, it transitions from the on state to the off state.

[0059] When it is detected that the ambient light changes, the brightness is adjusted in the on state and it enters the dimming state.

[0060] When the time enters the preset energy-saving period, it transitions from any other state to the energy-saving mode.

[0061] The beneficial effects of this design are significant. First, it makes the behavior of the system predictable and controllable. By clearly defining the states, input response processes, and state transition rules, the performance of the system in different situations can be accurately controlled. Second, this design improves the flexibility and scalability of the system. The functions of the system can be easily extended or modified by adding new states, defining new input response processes, or modifying the state transition rules. Finally, this design also facilitates the debugging and maintenance of the system. When the system exhibits abnormal behavior, the problem can be quickly located by checking the current state, the most recent inputs, and the state transition records.

[0062] In practical applications, a probability threshold may be set for each state transition to handle possible uncertainties. For example, a threshold of 0.9 may be set, and only when the confidence levels of the detection results of the human body sensor for three consecutive times exceed 0.9, will the state transition from off to on be triggered. This can effectively reduce incorrect operations caused by accidental interferences. Embodiment

[0063] Based on the above edge computing intelligent weak current system fault prediction method, combined with Figures 3 - 7 , an embodiment of the present invention further provides an edge computing-based intelligent weak current system fault prediction system, including a fault propagation direction prediction module 1 and a microservice unit management module 2; the fault propagation direction prediction module 1 includes a model training module 3 for training a prediction model for the intelligent weak current system; a service distributed fault point prediction module 4 for predicting multiple fault points where the intelligent weak current system fails through the prediction model; an incremental prediction model production module 5 for performing incremental training and model evolution on each microservice unit through microservice unit management; the microservice unit management module 2 includes: a fault management module 6 and an isolation policy monitoring module 7 for fault isolation between microservice units; a fault analysis module 8 for data interaction between microservice units to perform fault location and fault repair; a history management module 9 for recording the prediction history data generated by the prediction model; and a model management module 10 for storing the trained prediction model and its related information.

[0064] In an embodiment of the present invention, the weak current system of a large intelligent commercial complex can be considered as an application scenario. In this scenario, the fault propagation direction prediction module 1 plays a core role. For example, when there is an abnormality in the air conditioning system of the shopping mall, this module can predict how the fault may affect related subsystems such as the lighting system and the elevator system. The model training module 3 uses historical operation data to train the prediction model and may find that during the peak summer period, when the outdoor temperature exceeds 35°C and lasts for more than 3 hours, the probability of a central air conditioning system failure will rise to 15%. Such accurate prediction helps the shopping mall management to take preventive measures in advance, such as adjusting the air conditioning operation parameters or starting the standby system.

[0065] The service distributed fault point prediction module 4 further refines the prediction and may point out that there is an 80% probability that the main air conditioning unit in Building 3 will fail within the next 2 hours. Such specific prediction enables maintenance personnel to conduct targeted inspections and maintenance, greatly improving work efficiency. At the same time, the incremental prediction model production module 5 ensures that the system can continuously learn and adapt to new situations. For example, when a new intelligent parking lot management system is added to the shopping mall, this module will collect the operation data of the new system and perform incremental training on the existing model to enable it to accurately predict faults related to the new system.

[0066] The microservice unit management module 2 is responsible for specific fault management and system maintenance work. When the fault management module 6 and the isolation policy monitoring module 7 detect potential faults, they will take prompt actions. For example, if it is predicted that the security system in a certain area of the shopping mall may fail, the system may automatically activate the standby camera and temporarily increase the security patrol frequency in that area to ensure that safety is not affected.

[0067] The fault analysis module 8 accurately locates faults and provides repair suggestions by analyzing the data interaction between microservice units. In practical applications, if multiple elevators report anomalies simultaneously, this module may infer that there is a problem with the central control system rather than a single elevator fault. Such precise diagnosis can greatly shorten the fault repair time and reduce the impact on mall operations.

[0068] The historical management module 9 and the model management module 10 ensure the traceability and maintainability of the system. By saving the prediction history and model versions, long-term performance analysis can be carried out to continuously improve the prediction algorithm. For example, by analyzing the prediction data of the past year, it may be found that the prediction accuracy of the system for sudden faults is only 70%, which provides a clear direction for further optimizing the algorithm.

[0069] The beneficial effects brought by this system design are multi-faceted. First of all, it significantly improves the accuracy and timeliness of fault prediction, enabling preventive measures to be taken before faults occur, which is crucial for commercial complexes that need to operate 24 / 7. Secondly, through the microservice architecture and incremental learning, the system has strong flexibility and scalability, and can adapt to the changing business needs of the mall. Finally, the perfect management module ensures the reliability and maintainability of the system, which is conducive to the long-term stable operation of the mall's weak current system.

[0070] It also includes a rule calculation module 11, which contains a business rule management module 12 and a prediction rule management module 13; the business rule management module 12 is used to collect data streams and convert the data streams into data of corresponding decision types; the prediction rule management module 13 is used to perform association judgment and logical judgment between microservices: the collection and conversion of data streams by the business rule management module 12 are as follows: after the business rule management unit collects data from the data stream, it further formats the data into the standard format of the corresponding decision type; and sends it to the prediction rule management unit; the business rule management module 12 includes a mapping rule management unit and a standard format definition unit. Mapping rule management is the rule management that has a mapping relationship between source data and standard format data. When the system is extended to new types of data, the mapping between source data and standard format is also required; the standard format definition unit defines a set of standard data formats, and the standard format definition unit contains data entity definitions, as well as field definitions and association relationship definitions between data entities.

[0071] In the scenario of an intelligent commercial complex, the rule calculation module 11 plays a crucial role. The business rule management module 12 is responsible for collecting various data streams, such as passenger flow data, energy consumption data, security system data, etc., and converting these data into standardized decision-type data. For example, when collecting passenger flow data during peak hours, the business rule management module 12 may convert it into decision-type data on the degree of congestion for subsequent analysis and prediction.

[0072] The prediction rule management module 13 is responsible for analyzing the correlation and logical relationships between these standardized data. For example, it may find that when the passenger flow reaches a certain threshold (such as more than 10,000 person-times per hour), the loads on the elevator system and air conditioning system will increase significantly, and the probability of failure will rise accordingly. This correlation analysis provides an important basis for fault prediction.

[0073] The data collection and conversion process of the business rule management module 12 is very crucial. For example, when collecting raw elevator usage data, the business rule management unit will format it into a standard format, which may include fields such as usage frequency, load, running time, etc. This standardized processing enables data from different sources to be uniformly analyzed and processed.

[0074] The role of the mapping rule management unit is to ensure that data from different sources can be correctly mapped to the standard format. For example, when a new intelligent parking lot system is added to the mall, the mapping rule management unit will establish new mapping rules to map the parking lot data (such as parking space occupancy rate, vehicle entry and exit frequency, etc.) into the standard data format of the system. This flexible mapping mechanism enables the system to easily adapt to new data sources.

[0075] The standard format definition unit ensures the data consistency of the entire system. For example, it may define a data entity for device status, including fields such as "device ID", operating status, load level, etc., and define the association relationships between this entity and other entities (such as fault records, maintenance plans, etc.). This standardized data structure greatly improves the efficiency and accuracy of data processing.

[0076] The beneficial effects brought by this design are significant. First, it realizes the standardization and unified management of data, greatly improving the efficiency and accuracy of data processing. Second, through the flexible mapping mechanism, the system has strong scalability and can easily adapt to new data sources and business requirements. Finally, the standardized data format lays a solid foundation for subsequent analysis and prediction work, improving the prediction accuracy and reliability of the entire system.

[0077] The prediction rule management module 13 includes: a dependency analysis sub-module that calculates the dependencies between microservices; a dependency storage sub-module that stores the dependencies between services obtained by the dependency analysis sub-module and forms a service dependency graph, where each microservice is a node, and the dependency relationship between the node and the service is a solid or dashed line graph between services, and the dashed and solid lines represent data flow and control command flow respectively; a fault propagation direction prediction sub-module that analyzes the sequence of fault occurrences based on the dependencies between services; a data flow analysis sub-module that analyzes the data flow between microservices; and a fault relationship pre-judgment that pre-judges the fault propagation direction through the dependency relationship and data communication relationship between services.

[0078] In the weak current system of an intelligent commercial complex, the role of the prediction rule management module 13 is crucial. First, the dependency analysis sub-module calculates the dependencies between microservices by analyzing the interaction situations between them. For example, in the intelligent lighting system of a shopping mall, there may be the following dependencies: the central control service depends on the environmental perception service and the energy management service, and these two services respectively depend on the sensor services on each floor.

[0079] Next, the dependency storage sub-module stores these dependencies and forms a service dependency graph. In this graph, each microservice is a node. For example, the central control service, the environmental perception service, the energy management service, etc. are all independent nodes. The connections between nodes represent their dependencies. Among them, the solid line may represent the control command flow, such as the switch command sent by the central control service to the lighting systems on each floor; the dashed line may represent the data flow, such as the light intensity data sent by the sensors on each floor to the environmental perception service.

[0080] The fault propagation direction prediction sub-module uses this dependency graph to analyze possible fault propagation paths. For example, if the environmental perception service fails, this sub-module may predict that this fault will affect the decision-making of the central control service, and may further lead to abnormalities in the entire lighting system.

[0081] The data flow analysis sub-module focuses on analyzing the data interaction between microservices. For example, it may find that the sensor data on a certain floor is abnormally frequent or the data volume suddenly decreases, which may indicate a problem with the sensor network on that floor.

[0082] Finally, the fault relationship pre-judgment sub-module comprehensively considers the dependency relationship and data communication relationship between services to pre-judge the possible fault propagation direction. For example, if abnormal data is detected in the energy management service, combined with the previously established dependency relationship, the system may pre-judge that this fault may affect the energy-saving control function of the entire lighting system.

[0083] The beneficial effects brought about by this design are multi-faceted. First of all, it provides a comprehensive perspective to understand the complex relationships among various microservices in the system, which helps to more accurately predict the propagation path of faults. Secondly, by visualizing the dependency relationships as graphs, system administrators can intuitively understand the system structure, facilitating quick problem location. Moreover, this dependency-based analysis method enables the system to better handle complex fault scenarios, improving the accuracy of fault prediction and handling.

[0084] The microservice unit management module 2 further includes a microservice orchestration module, and the specific steps are as follows: S1. Orchestrate the state machine configuration table according to the state machine orchestration model and the dependency relationships among state machines, and load the state machine configuration table into the microservice orchestration engine. The microservice orchestration engine binds each state machine according to the configuration table; S2. Start the state machine execution process of the microservice orchestration engine. According to the state machine orchestration model and the dependency relationships among state machines, automatically execute the state machine execution process, and incorporate the data flow and operation process among microservices into state machine control; S3. When triggering a state node migration, the state machine calculates the data access and data output involved after the state migration, analyzes the state machine dependency relationships, and matches them with the dependency relationships between state machines of other state machines, thereby triggering the state machine migration of the dependent state machine.

[0085] In the weak current system of an intelligent commercial complex, the role of the microservice orchestration module cannot be ignored. First of all, in step S1, the system will orchestrate the state machine configuration table according to the predefined state machine orchestration model and the dependency relationships among state machines. For example, for the air conditioning system in a shopping mall, the following state machines may be defined: temperature control state machine, humidity control state machine, energy consumption management state machine, etc. There are complex dependency relationships among these state machines. For example, the output of the temperature control state machine may affect the decision-making of the energy consumption management state machine.

[0086] The system will orchestrate these relationships into a configuration table and then upload it to the microservice orchestration engine. Suppose the state transition threshold of the temperature control state machine is set as: when the detected temperature change exceeds 2°C, trigger a state transition. The selection of this threshold is based on the balance consideration of the comfort level and energy efficiency of the shopping mall environment.

[0087] Next, in step S2, the microservice orchestration engine will start the state machine execution process. For example, when the temperature sensor detects that the temperature has risen by 3°C, the temperature control state machine will transition from the stable state to the cooling state. During this process, the system will automatically control relevant microservices, such as increasing the power of the refrigeration equipment and adjusting the air supply speed.

[0088] In step S3, when a state node migration is triggered, the system will perform a series of complex calculations and analyses. Continuing with the above example, when the temperature control state machine transitions from the stable state to the cooling state, the system will calculate the data access involved in this state migration (such as reading environmental parameters like the current temperature and humidity) and data output (such as sending new control instructions to the refrigeration equipment). At the same time, the system will also analyze the impact of this state migration on other state machines. For example, the energy consumption management state machine may need to transition from the energy-saving state to the high-load state accordingly.

[0089] The beneficial effects of this design are significant. First, it enables intelligent collaboration between microservices, allowing complex systems to operate more flexibly and efficiently. Second, managing microservices through state machines greatly simplifies the system complexity and improves the system's maintainability. Finally, this state machine-based management approach enables the system to better handle complex scenario changes, enhancing the stability and reliability of the entire low-voltage electrical system.

[0090] It also includes a microservice update management module, which contains a microservice process engine module, a fault isolation strategy module, and a model evolution module; the microservice process engine module manages the logical processing of each state machine on the microservice unit during the microservice orchestration process, that is, the control and management of the microservice unit by the state machine, including start, stop, rollback, synchronization, and asynchrony; the management object is the state machine; the fault isolation strategy module is responsible for isolating the fault point according to the detected abnormal situations of each microservice, in accordance with the established fault propagation direction and fault propagation priority; the model evolution module updates and evolves the prediction model; when a new state machine is generated in the state machine management data, or the model evolution module generates a new prediction model, it will trigger the update of the model evolution; during the update, when a certain state machine is added or deleted, it will trigger the update of the state machine configuration table; when other data in the state machine configuration table is added, deleted, or modified, it will trigger the update of the model corresponding to the state machine configuration.

[0091] In the low-voltage electrical system of an intelligent commercial complex, the microservice update management module plays a crucial role in ensuring that the system can continuously optimize and adapt to new requirements. First, the microservice process engine module is responsible for managing the control of each state machine on the microservice unit. For example, in the security system of a shopping mall, when an abnormal situation is detected, the microservice process engine module may start the emergency response state machine, stop the daily patrol state machine at the same time, and synchronize the relevant data to the central control system.

[0092] Specifically, assume that the system detects suspicious behavior captured by a camera in a certain area of the mall. The microservice process engine module will immediately start the abnormal behavior analysis state machine, which may include the following states: initial analysis, in-depth identification, alarm triggering, etc. At the same time, it may pause or adjust the operation of other non-critical state machines to ensure that system resources are prioritized for handling potential security threats.

[0093] The fault isolation strategy module is responsible for quickly isolating the fault point when an abnormal situation is detected. For example, if it is found that the lighting system in a certain area of the mall frequently malfunctions, this module may decide to temporarily isolate the lighting control system in this area and start the backup lighting at the same time to ensure that the normal operation of the mall is not affected. When implementing the isolation strategy, the system will consider the propagation direction and priority of the fault. For example, if the fault occurs in the power system, its priority may be set to the highest because it may affect the operation of the entire mall.

[0094] The role of the model evolution module is to continuously update and optimize the prediction model. For example, when a new intelligent parking lot management system is added to the mall, the model evolution module will collect the operation data of the new system and update the existing fault prediction model. This process may involve adding new features, adjusting model parameters, etc. Suppose the new parking lot system introduces a new feature of parking space occupancy rate. The model evolution module will evaluate the impact of this feature on fault prediction. If it is found that it can significantly improve the prediction accuracy (such as from 85% to 90% originally), it will be incorporated into the updated model.

[0095] When a new state machine or prediction model appears in the system, it will trigger an update of the model evolution. For example, if the mall decides to introduce a new energy management system, the system may add a new energy optimization state machine. At this time, the model evolution module will automatically trigger the update process, adjust the existing state machine configuration table, which may include adding new state transition rules, adjusting the priority of existing state machines, etc.

[0096] Similarly, when a certain state machine is deleted, such as when the mall decides to phase out the old HVAC system, the relevant state machine may be removed. At this time, the system will automatically update the state machine configuration table to ensure that other relevant state machines (such as temperature control, energy management, etc.) can adapt to this change.

[0097] The beneficial effects brought by this design are multi-faceted. First of all, it makes the system highly adaptable and scalable, capable of easily coping with changes in business requirements. Secondly, through timely fault isolation and model updates, the reliability and stability of the system are greatly improved. Finally, this ability of dynamic evolution ensures that the system can continuously optimize its performance and always remain in the best state. For example, through continuous learning and updates, the accuracy rate of the system's fault prediction may gradually increase from the initial 80% to over 95%, greatly reducing the risk of unexpected downtime and improving the operational efficiency of the entire commercial complex.

[0098] It should be noted that the above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A fault prediction method for an intelligent weak current system based on edge computing, characterized in that, It includes the following steps: Step S1. Perform fault correlation analysis on the dependencies and fault propagation among microservices in the intelligent weak current system according to the knowledge graph; Step S2. Complete the analysis of the dependencies among multiple microservices in the intelligent weak current system and the task of predicting the fault propagation direction based on the service - to - service dependency analysis and fault propagation prediction method of the DAG (Directed Acyclic Graph); S3. Combine the system into intelligent dynamic microservice units according to business requirements. For each microservice unit in the actual system, it includes two parts: a state machine and a data flow; The state machine is a microservice unit logically defined based on the microservice splitting of the system. According to its logical processing process and the data communication logic among microservice units during the logical processing process, the entire logical processing process and data communication process are divided into one or several state machines, and the state machine is used as the basic unit for microservice orchestration and management; Step S4. Manage the update of complex service orchestration based on adaptive service orchestration and fault isolation, and based on incremental learning and model evolution management mechanisms.

2. The method for predicting faults in an intelligent weak current system based on edge computing according to claim 1, wherein , It also includes the following steps: Step S1 further includes collecting the data flows of various service interactions among microservices to implement the microservice process oriented to the data flow; generating decision information from the microservice process oriented to the data flow, extracting fault data from the data flow, and classifying the fault data; where the microservice process contains the storage of dependencies among microservice units and the fault propagation model; finally, converting the classified data into corresponding decision types; the service interaction is the data interaction among microservices, and the interaction methods include data transmission for distributed service execution, asynchronous communication for publish / subscribe services, and batch - processing workflows and stateful distributed transaction operations.

3. A fault prediction method for an intelligent weak current system based on edge computing according to claim 1, wherein The prediction task of the fault propagation direction in step S2 is predicted through a fault prediction model, and the specific steps are as follows: The fault propagation direction is predicted through the dependency relationship data between services, so as to predict a series of possible fault propagation directions. Specifically: The dependency relationship analysis between microservices with an association relationship is mainly described by a directed acyclic graph (DAG). Each microservice is defined as a node, and the directed edge of the node represents the mutual dependency relationship between services. The entire system is a DAG graph composed of multiple interconnected microservices; Define \(s_i\) and \(s_j\) as the starting service node and the ending service node of the fault propagation path P respectively, and the number of object nodes covered by the fault propagation path P is N, and the initial defined value of N is 0; The starting service node generates an initial fault propagation path \(p_0\) and adds \(p_0\) to a queue r with a global sorting of 1; When r is not empty, dequeue a fault propagation path p from the head of the queue r, and then select an uncovered object node n from p i , and add the service node covered by n i to the fault propagation path p; At this time, update the number of nodes covered by \(n_i\) to \(N = N + 1\). If n i is the ending service node, then \(N = N - 1\), and add the fault propagation path p to the result fault propagation path set G; If n i is an intermediate service node, then select another covered object; When all the covered objects in the fault propagation path p are marked as covered objects, add the fault propagation path p to the result fault propagation path set G, and then continue to select the next uncovered object; If p is an empty set, then judge whether the queue r is empty. If it is empty, exit the fault prediction; If it is not empty, continue to process the next object in the queue r and repeat the above steps; The fault propagation path set G is the predicted fault propagation direction.

4. The intelligent weak current system fault prediction method based on edge computing according to claim 2, wherein , There are multiple concurrent relationships among the state machines in Step S3, including sequential, parallel, aggregative, selective, and nested types; among them, in the sequential state, the microservice needs to go through multiple states from the start state to the end state; in the parallel state, multiple microservices can be executed simultaneously and output results; in the aggregative state, one state needs to aggregate the states of multiple microservices; in the selective state, different choices are made according to the different output results of a microservice; the nested state is a complex state in which multiple states are nested with each other, and the output of each state can cause the switching of the output state.

5. A fault prediction method for an intelligent weak current system based on edge computing according to claim 4, characterized in that The state machine includes states, input response processes, and state migrations.

6. An intelligent weak current system fault prediction system based on edge computing, characterized in that, It includes a fault propagation direction prediction module and a microservice unit management module; the fault propagation direction prediction module includes a model training module for training and predicting the prediction model of the intelligent weak current system; a service distributed fault point prediction module for predicting multiple fault points where the intelligent weak current system fails through the prediction model; An incremental prediction model production module for performing incremental training and model evolution on each microservice unit through microservice unit management; The microservice unit management module includes: a fault management module, an isolation policy monitoring module for fault isolation between microservice units; a fault analysis module for data interaction between microservice units to perform fault location and fault repair; a history management module for recording the prediction historical data generated by the prediction model; a model management module for storing the trained prediction model and its related information.

7. An intelligent weak current system fault prediction system based on edge computing according to claim 6, characterized in that, It also includes a rule calculation module, which contains a business rule management module and a prediction rule management module; the business rule management module is used to collect data streams and convert the data streams into data of corresponding decision types; the prediction rule management module is used to perform association judgment and logical judgment between microservices: The collection and conversion of data streams by the business rule management module are as follows: After the business rule management unit collects data from the data stream, it further formats the data into the standard format of the corresponding decision type; and sends it to the prediction rule management unit; the business rule management module includes a mapping rule management unit and a standard format definition unit. Mapping rule management is the rule management in which the source data and the standard format data have a mapping relationship. When the system is extended to new types of data, the mapping between the source data and the standard format is also required. The standard format definition unit defines a set of standard data formats, which includes data entity definition, field definition, and the definition of the association relationship between data entities.

8. The fault prediction system of an intelligent weak current system based on edge computing according to claim 7, characterized in that, The prediction rule management module includes: a dependency analysis sub-module for calculating the dependencies between microservices; a dependency storage sub-module for storing the dependencies between services obtained by the dependency analysis sub-module and forming a service dependency relationship graph, where each microservice is a node, and the dependency relationship between the node and the service is a solid or dashed line graph between services. The dashed line and the solid line represent the data stream and the control command stream respectively; a fault propagation direction prediction sub-module for analyzing the sequence of fault occurrences based on the dependencies between services; a data stream analysis sub-module for analyzing the data stream between microservices; a fault relationship pre-judgment for predicting the fault propagation direction through the dependency relationship and data communication relationship between services.

9. The fault prediction system of an intelligent weak current system based on edge computing according to claim 8, wherein, The microservice unit management module also includes a microservice orchestration module, and the specific steps are as follows: S1. Orchestrate the state machine configuration table according to the state machine orchestration model and the dependencies between state machines, and load the state machine configuration table into the microservice orchestration engine. The microservice orchestration engine binds each state machine according to the configuration table; S2. Start the state machine execution process of the microservice orchestration engine. According to the state machine orchestration model and the dependencies between state machines, automatically execute the state machine execution process, and incorporate the data stream and operation process between microservices into the state machine control; S3. When triggering the state node migration, the state machine calculates the data access and data output involved after the state migration, and analyzes the state machine dependency relationship, and matches it with the dependency relationship between state machines of other state machines, thereby triggering the state machine migration of the dependent state machine.

10. An intelligent weak current system fault prediction system based on edge computing according to claim 9, characterized in that, It also includes a microservice update management module, which contains a microservice process engine module, a fault isolation policy module, and a model evolution module; the microservice process engine module performs logical processing on microservice units by each state machine during microservice orchestration, that is, the control and management of microservice units by the state machine, including start, stop, rollback, synchronization, and asynchrony; The management object is the state machine; the fault isolation policy module is responsible for isolating the fault points according to the detected abnormal situations of each microservice, in accordance with the established fault propagation direction and fault propagation priority; The model evolution module updates and evolves the prediction model; When a new state machine is generated in the state machine management data, or a new prediction model is generated by the model evolution module, it will trigger the update of model evolution; during the update, when a certain state machine is added or deleted, it will trigger the update of the state machine configuration table; when other data in the state machine configuration table is added, deleted, or modified, it will trigger the update of the model corresponding to the state machine configuration.

Images