Unmanned cluster system reliability digital twin model evolution method

Abstract

The application discloses an unmanned cluster system reliability digital twin model evolution method, which effectively integrates evolution elements, evolution time and evolution strategies of the unmanned cluster system reliability digital twin model, realizes evolution and update of the digital twin model, and further supports construction of the unmanned cluster system digital twin body. The steps are as follows: node intelligent agents, subsystem intelligent agents and component intelligent agents are sequentially constructed to describe component elements and functional principles of the unmanned cluster system; evolution elements of the unmanned cluster system reliability digital twin model are screened with system fault occurrence and propagation as the core; evolution time of the unmanned cluster system reliability digital twin model is captured by comparing model simulation data and real-time monitoring information; and evolution strategies of the unmanned cluster system reliability digital twin model are formulated based on an intelligent agent dynamic knowledge base, including: correcting intelligent agent attribute parameters, adjusting intelligent agent interaction relationships, and adding or deleting node intelligent agents.

Description

Technical Field

[0001] This invention provides a method for the evolution of a digital twin model of reliability for unmanned swarm systems, belonging to the field of reliability engineering technology. Background Technology

[0002] Currently, thanks to the rapid development of sensor networks, distributed communication, and big data technologies, the application characteristics of digital twins have gradually transitioned to multi-level and multi-element features. A digital twin of an unmanned swarm system is a technical means of creating a virtual simulation model of the swarm's physical entity in a digital way, and using historical monitoring information, real-time operational data, and algorithm models to simulate, verify, predict, and control the entire lifecycle of the physical entity. Its basic workflow during operation can be divided into four parts: information acquisition, state perception, operation and maintenance decision-making, and implementation. From a virtual perspective, state perception and operation and maintenance decision-making are usually based on the digital twin, and essentially, they are an aggregation of a series of models and their evolution processes. Meanwhile, swarm state perception involves multiple aspects, including functional performance and reliability. Reliability is achieved by capturing system-wide and local degradation and fault information through models, enabling the processing and display of relevant data.

[0003] However, research on digital twins of unmanned swarm systems is currently uneven, with most work focusing only on describing the system's functional performance and evolution. To achieve precise operation and maintenance of unmanned swarm systems and ensure their reliable and stable operation, it is necessary to propose corresponding theoretical methods to bridge the gap in the field of reliability digital twins. Therefore, this invention comprehensively considers the typical characteristics of unmanned swarm systems, such as distributed deployment, reconfigurability, and autonomous collaboration, and proposes an evolution method for a reliability digital twin model of unmanned swarm systems. This method can provide technical support for the development and application of digital twin technology in the field of unmanned swarm system reliability. Summary of the Invention

[0004] The purpose of this invention is to provide an evolution method for a digital twin model of unmanned swarm system reliability, so as to effectively integrate the evolution elements, evolution timing and evolution strategy of the digital twin model of unmanned swarm system reliability, realize the evolution and update of the digital twin model, and thus support the construction of the digital twin of the unmanned swarm system.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] An evolution method for a reliability digital twin model of an unmanned swarm system mainly includes the following steps:

[0007] S100: Construct node intelligent agents, subsystem intelligent agents and component intelligent agents in sequence to describe the constituent elements and functional principles of the unmanned swarm system;

[0008] S200: Focusing on the occurrence and propagation of system failures, screening the evolutionary elements of the reliability twin model of unmanned swarm systems;

[0009] S300: Compare model simulation data with real-time monitoring information to capture the evolution timing of the reliability twin model of unmanned swarm systems;

[0010] S400: Based on the agent's dynamic knowledge base, formulate an evolution strategy for the reliability twin model of unmanned swarm systems.

[0011] S401; Correct agent attribute parameters;

[0012] S402: Adjust the interaction relationship between intelligent agents;

[0013] S403: Add or remove node agents.

[0014] In step S100, based on the multi-level characteristics of the unmanned swarm system, node agents, subsystem agents, and component agents are constructed from top to bottom to describe the system's constituent elements and functional principles. Node agents simulate the operation mode of a single system; subsystem agents abstract the key subsystems of a node; and component agents characterize the functional differences of the components contained within a subsystem. Based on this, a dynamic knowledge base is configured for each agent to store the monitored node operation modes, as well as subsystem and component fault degradation information, and to generate and update knowledge to support model evolution.

[0015] In step S200, focusing on component failure occurrence and fault propagation within and outside nodes, the potential evolutionary elements of the unmanned swarm system reliability twin model are determined sequentially from the perspectives of model parameter updates, interaction relationship adjustments, and the addition or removal of agents. Based on this, the focus shifts from the inside out to parameter correction of the component failure model, adaptive adjustment of the system fault propagation path, and online addition or removal of node agents. Furthermore, evolutionary elements are selected specifically according to the specific task scenario of the unmanned swarm system.

[0016] In step S300, based on the determined evolutionary elements, real-time monitoring information and model simulation data are dynamically compared to determine the evolution timing of the unmanned swarm system reliability twin model. For parameter correction of the component failure model, a relative percentage error is used to measure the difference between the model simulation data and the actual monitoring data, and an error threshold is referenced to trigger updates to the agent attribute parameters. For adaptive adjustment of the system failure propagation path, changes in agent interaction relationships are triggered based on the differences in the actual node operating modes before and after. For online addition and removal of node agents, the addition and removal of node agents are triggered based on the activation or deactivation of the actual nodes.

[0017] In step S400, based on the agent's dynamic knowledge base and by integrating evolution timing, a reliability twin model evolution strategy is formulated for the corresponding unmanned swarm system evolution elements:

[0018] In step S401, the component fault model parameter correction strategy is triggered based on the relative percentage error measurement result. The component agent attribute parameters are corrected using measured component degradation fault information based on Bayesian update theory.

[0019] In step S402, based on the differences in the actual node operating mode before and after, an adaptive adjustment strategy for the fault propagation path is triggered. After the default node operating mode is adjusted, knowledge updates must be performed simultaneously. First, the dynamic knowledge base of the node agent is searched to see if the operating mode is already stored. If it exists, the agent interaction relationship is adjusted based on the associated relevant knowledge. If it does not exist, knowledge related to the node operating mode is generated and updated, and the agent interaction relationship is adjusted based on the updated knowledge, thereby achieving adaptive adjustment of the fault propagation path.

[0020] In step S403, based on the difference in the actual node activation state before and after, an online addition / removal strategy for node agents is triggered. When an actual node is detected to have left the unmanned cluster system, the corresponding node agent and its associated subsystem agents and component agents are adjusted to a silent state. Based on the cold / hot backup status of the removed node, it is determined whether the component fault model continues to run over time. Conversely, when a node joins the unmanned cluster system, the relevant agents are activated and the knowledge base is copied. The maximum set of node operating modes in the current multi-agent model knowledge base is marked, and the marked knowledge base is sequentially copied to the activated node agents and their associated subsystem agents and component agents. Finally, the interaction relationships between the agents are updated.

[0021] Compared with existing model evolution methods, the beneficial effects of this invention are as follows: By using the unmanned swarm system reliability digital twin model evolution method developed by this invention, model evolution elements can be selected in a targeted manner based on the actual operating characteristics of the unmanned swarm system. At the same time, the model evolution timing can be captured based on real-time system monitoring data, and corresponding reliability digital twin model evolution strategies can be formulated using a dynamic knowledge base. This achieves effective integration of the evolution elements, evolution timing, and evolution strategies of the unmanned swarm system reliability digital twin model, realizing the orderly evolution and updating of the model to support the construction of the unmanned swarm system digital twin. Attached Figure Description

[0022] Figure 1 A flowchart illustrating the evolution method of a digital twin model for the reliability of an unmanned swarm system provided by this invention;

[0023] Figure 2 This invention provides a digital twin model for the reliability of unmanned aerial vehicle (UAV) clusters. Detailed Implementation

[0024] The following will refer to the appendix. Figure 1 With appendix Figure 2 Specific embodiments of the invention are described in detail below. While specific embodiments of the invention have been discussed, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, the embodiments are provided to enable a more thorough understanding of the invention and to fully convey the information of the invention to those skilled in the art.

[0025] This invention provides a method for the evolution of a digital twin model of unmanned aerial vehicle (UAV) swarm reliability, the flowchart of which is as follows. Figure 1 As shown, it includes:

[0026] S100: Based on the multi-level characteristics of UAV swarms, node agents, subsystem agents, and component agents are constructed from top to bottom to describe the system's components and functional principles. Node agents simulate the operation mode of a single UAV; subsystem agents abstract the key subsystems of a single UAV; and component agents characterize the functional differences of the components contained within the UAV. On this basis, a dynamic knowledge base is configured for each agent to store the monitored UAV operation modes, as well as fault degradation data of subsystems and components, and to perform knowledge generation and updates to support model evolution.

[0027] Example 1: such as Figure 2 The illustrated UAV swarm consists of three UAVs. Each UAV comprises a subsystem including an airframe, attitude control module, cooperative control module, power module, and propulsion module. The propulsion module can be further broken down into an ESC (Electronic Speed ​​Controller), motor, and propellers. Based on this, a reliability twin model of the UAV swarm is constructed. Node agents simulate the formation flight mode and manual control mode of individual UAVs. Within each node agent, five subsystem agents are built to abstract the key subsystems of the UAVs. Furthermore, within the propulsion module-subsystem agent, three component agents are built to characterize the functional differences of the component ESC, motor, and propellers. On this basis, a dynamic knowledge base is configured for each agent to store the monitored UAV operating modes and fault degradation data of subsystems and components, and to perform knowledge generation and updates to support model evolution.

[0028] S200: Focusing on component failure occurrence and fault propagation within and outside nodes, this paper identifies potential evolutionary elements of the UAV swarm reliability twin model from the perspectives of model parameter updates, interaction relationship adjustments, and agent additions / removals. Based on this, it focuses from the inside out on parameter correction of the component failure model, adaptive adjustment of the system fault propagation path, and online addition / removal of node agents, and selects evolutionary elements specifically according to the particular UAV swarm mission scenario.

[0029] Following Example 1: This drone swarm is used to perform joint strike missions. During the formation flight, penetration, attack, and return-to-base processes, the drones may be shot down by enemy air defense fire. Therefore, based on the above mission characteristics, from the perspective of agent addition and subtraction, the online addition and subtraction of node agents is specifically selected as an evolutionary element of this drone swarm.

[0030] S300: Based on the determined evolutionary factors, it dynamically compares real-time monitoring information with model simulation data to determine the evolution timing of the UAV swarm reliability twin model. For parameter correction of component failure models, it uses relative percentage error to measure the difference between model simulation data and actual monitoring data, and triggers agent attribute parameter updates with reference to error thresholds. For adaptive adjustment of system fault propagation paths, it triggers changes in agent interaction relationships based on the differences in actual node operating modes. For online addition and removal of node agents, it triggers the addition and removal of node agents based on the activation or deactivation of actual nodes.

[0031] Continuing the previous example: The evolutionary element of this drone swarm is the online addition and removal of node agents. Real-time monitoring information is dynamically compared with model simulation data to determine the evolution timing of the drone swarm reliability twin model. Regarding the online addition and removal of node agents, the addition and removal of node agents are triggered by the activation or deactivation of actual drones.

[0032] S400: Based on the agent's dynamic knowledge base and incorporating evolution timing, it formulates a reliable digital twin model evolution strategy for the evolutionary elements of UAV swarms.

[0033] S401: Based on the relative percentage error metric, trigger the component fault model parameter correction strategy. Based on Bayesian update theory, use measured component degradation fault information to correct the component agent attribute parameters.

[0034] S402: Based on the differences in the actual UAV operation mode before and after, an adaptive adjustment strategy for the fault propagation path is triggered. By default, after the UAV operation mode is adjusted, knowledge updates need to be carried out simultaneously. First, it checks whether the node agent's dynamic knowledge base has stored the operation mode: if it exists, the agent's interaction relationship is adjusted based on the associated relevant knowledge; if it does not exist, the relevant knowledge of the UAV operation mode is generated and updated, and the agent's interaction relationship is adjusted based on the updated knowledge, thereby realizing the adaptive adjustment of the fault propagation path.

[0035] S403: Based on the differences in the actual UAV activation state before and after activation, trigger the online addition and removal strategy of node agents. When it is detected that an actual UAV has left the UAV cluster, the corresponding node agent and its associated subsystem agents and component agents are adjusted to a silent state. Based on the cold and hot backup status of the removed node, determine whether the fault model of the included component continues to run over the simulation time. Conversely, when a UAV joins the UAV cluster, the relevant agents are activated and the knowledge base is copied. The maximum set of node operating modes in the current multi-agent model knowledge base is marked, and the marked knowledge base is copied sequentially to the activated node agents and their associated subsystem agents and component agents. Finally, the interaction relationships of the agents are updated.

[0036] Continuing the previous example: During a joint strike mission, a drone swarm detected that one of its drones was shot down by enemy air defenses during its return flight. The drone transitioned from an active to an inactive state, triggering model evolution. This moment is marked, and the corresponding node agent, along with related subsystem agents and component agents, are placed into a silent state, with their interaction relationships updated. Furthermore, considering that the drone is now in a cold backup state due to being completely shot down, the component failure models will no longer run continuously over simulation time.

[0037] Although embodiments of the present invention have been described above in conjunction with the accompanying drawings, the present invention is not limited to the specific embodiments and application fields described above. The specific embodiments described above are merely illustrative and instructive, and not restrictive. Those skilled in the art can make many other forms based on the guidance of this specification and without departing from the scope of protection of the claims of the present invention, and all of these are within the scope of protection of the present invention.

Claims

1. A method for evolving a digital twin model of reliability for an unmanned swarm system, characterized in that, include: S100: Construct node intelligent agents, subsystem intelligent agents and component intelligent agents in sequence to describe the constituent elements and functional principles of the unmanned swarm system; S200: Focusing on the occurrence and propagation of system failures, the evolutionary elements of the reliability twin model of unmanned swarm systems are screened; starting from the perspectives of model parameter updates, interaction relationship adjustments, and the addition or removal of agents, potential evolutionary elements of the reliability twin model of unmanned swarm systems are identified; based on this, the focus is shifted from the inside out to the parameter correction of component failure models, adaptive adjustment of system failure propagation paths, and online addition or removal of node agents, and evolutionary elements are selected in a targeted manner according to the specific task scenarios of the unmanned swarm system. S300: By comparing model simulation data with real-time monitoring information, it captures the timing of the evolution of the reliability twin model of the unmanned swarm system; for the parameter correction of the component failure model, it uses relative percentage error to measure the difference between the model simulation data and the actual monitoring data, and triggers the update of the agent attribute parameters with reference to the error threshold; For adaptive adjustment of system fault propagation paths, changes in agent interaction relationships are triggered based on the differences in actual node operating modes before and after; for online addition and removal of node agents, addition and removal of node agents are triggered based on whether the actual node is activated or not. S400: Based on the agent's dynamic knowledge base, formulate an evolution strategy for the reliability twin model of unmanned swarm systems. S401: Correct agent attribute parameters; S402: Adjust the interaction relationship between intelligent agents; S403: Add or remove node agents.

2. The method for evolving a digital twin model of reliability for an unmanned swarm system according to claim 1, characterized in that: In the adjustment of the intelligent agent interaction relationship described in S402, an adaptive adjustment strategy for the fault propagation path is triggered based on the differences in the actual node operation mode before and after. After the default node operating mode is adjusted, knowledge updates need to be carried out simultaneously. First, check whether the node agent dynamic knowledge base has stored the operating mode. If it exists, adjust the agent interaction relationship based on the associated relevant knowledge. If it does not exist, then the knowledge related to the node operation mode is generated and updated, and the interaction relationship of the intelligent agent is adjusted based on the updated knowledge, thereby realizing the adaptive adjustment of the fault propagation path.

3. The method for evolving a digital twin model of reliability in an unmanned swarm system according to claim 1, characterized in that: In the addition and deletion of node agents described in S403, the online addition and deletion strategy of node agents is triggered based on the difference in the actual node activation state before and after. When it is detected that the actual node has left the unmanned cluster system, the corresponding node agent and the associated subsystem agent and component agent are adjusted to a silent state. Based on the cold and hot backup status of the removed node, it is determined whether the fault model of the included component continues to run with the simulation time. Conversely, when a node joins an unmanned cluster system, the relevant intelligent agent is activated and the knowledge base is copied. Mark the largest set of node operation modes in the current multi-agent model knowledge base, and copy the marked knowledge base to the activated node agent and the associated subsystem agent and component agent in sequence. Finally, update the interaction relationship of the agents.

Images