Multi-agent sensor fault diagnosis method and system based on adaptive learning

Abstract

The application relates to the technical field of multi-agent systems, and provides a multi-agent sensor fault diagnosis method and system based on adaptive learning, which comprises the following steps: acquiring control signals and outputs of each multi-agent subsystem; based on the control signals and the outputs, an adaptive observer is used to obtain a state vector of the adaptive observer; by comparing the state vector of the adaptive observer with a state vector of the multi-agent subsystem, fault information of a sensor corresponding to the multi-agent subsystem is obtained; wherein the adaptive observer ensures that a tracking error and a weight vector error are bounded, the tracking error is a difference value between an output estimation of the multi-agent subsystem and an output of the multi-agent subsystem, and the weight vector error is a difference value between an estimated weight of a neural network in the adaptive observer and an ideal value; and the final consistent boundedness of identification errors is ensured.

Description

A Multi-Agent Sensor Fault Diagnosis Method and System Based on Adaptive Learning Technical Field

[0001] This invention belongs to the field of multi-agent system technology, and particularly relates to a multi-agent sensor fault diagnosis method and system based on adaptive learning. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] The concept of multi-agent systems originated from human observations of biological phenomena in nature. In order to survive better, organisms in nature develop and expand their populations through reproduction, and gather together to collect food and defend against external invasions, thereby greatly increasing their competitiveness in nature.

[0004] Similar to a single organism within a population, the concept of an intelligent agent has been proposed by researchers. An intelligent agent can be a software program existing in a virtual environment or an entity residing in the real environment. It can obtain dynamic environmental information through sensors and modify the surrounding environment by performing corresponding actions to meet expected design requirements. A multi-agent system is composed of multiple simple intelligent agents that meet consistency conditions, organically combined. These agents interact with each other through network information channels and cooperate to complete a certain autonomous and coordinated movement.

[0005] Multi-agent systems still face many challenges, the most significant being system security and reliability. While multi-agent systems are more powerful and better able to acquire complex global dynamic information than single-agent systems, their more complex topologies can significantly increase the probability of failures, especially sensor failures. Since sensors are more fragile and easily damaged than actuators, and the performance of multi-agent systems heavily relies on sensor output feedback, sensor failure can cause the entire system to malfunction and gradually collapse, potentially leading to major catastrophic accidents. Summary of the Invention

[0006] To address the technical problems mentioned above, this invention provides a multi-agent sensor fault diagnosis method and system based on adaptive learning. Through adaptive learning technology, the eventual consistency and boundedness of the identification error are guaranteed.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] The first aspect of the present invention provides a multi-agent sensor fault diagnosis method based on adaptive learning, comprising:

[0009] Acquire the control signals and outputs of each multi-agent subsystem;

[0010] Based on the control signal and output, an adaptive observer is used to obtain the state vector of the adaptive observer. By comparing the state vector of the adaptive observer with the state vector of the multi-agent subsystem, the fault information of the sensor corresponding to the multi-agent subsystem is obtained.

[0011] The adaptive observer ensures that the tracking error and the weight vector error are bounded. The tracking error is the difference between the output estimate of the multi-agent subsystem obtained by the adaptive observer and the output of the multi-agent subsystem. The weight vector error is the difference between the estimated weights of the neural network in the adaptive observer and the ideal values.

[0012] Furthermore, the adaptive observer employs a radial basis function neural network to learn the portion of the control signal trajectory increase caused by sensor failure.

[0013] Furthermore, by selecting the observer gain and calculating the time derivative of the Lyapunov function, the boundedness of the tracking error and the weight vector error is verified.

[0014] Furthermore, the Lyapunov function is constructed as follows:

[0015]

[0016] in, For tracking error, L i =diag{l i ,l i 2 ,...,l i n}, and l i A positive constant; K i To construct a matrix; For the weight vector error; P i It is a positive definite matrix.

[0017] Furthermore, by comparing the residual error between the state vector of the adaptive observer and the state vector of the multi-agent subsystem with a set value, the fault information of the sensor corresponding to the multi-agent subsystem is obtained.

[0018] Furthermore, the model of the i-th multi-agent subsystem is as follows:

[0019]

[0020] Among them, A ob For the given matrix, B ob and C ob Let x be a given vector. i y is the state vector of the i-th subsystem. i It is the output of the i-th subsystem, u i Let q be the control signal for the i-th subsystem. i () is the nonlinear function of the i-th subsystem, F i (u i The fault is due to a sensor malfunction. o i Let β() represent the time of occurrence of the unknown fault in the i-th subsystem, and let β() be the step function. Represents the state vector x i The derivative with respect to time t.

[0021] Furthermore, for the control signal, the nonlinear function of the multi-agent subsystem satisfies the local Lipschitz continuity condition.

[0022] A second aspect of the present invention provides a multi-agent sensor fault diagnosis system based on adaptive learning, comprising:

[0023] The data acquisition module is configured to acquire the control signals and outputs of each multi-agent subsystem.

[0024] The fault diagnosis module is configured to: based on the control signal and output, use an adaptive observer to obtain the state vector of the adaptive observer, and obtain the fault information of the sensor corresponding to the multi-agent subsystem by comparing the state vector of the adaptive observer with the state vector of the multi-agent subsystem;

[0025] The adaptive observer ensures that the tracking error and the weight vector error are bounded. The tracking error is the difference between the output estimate of the multi-agent subsystem obtained by the adaptive observer and the output of the multi-agent subsystem. The weight vector error is the difference between the estimated weights of the neural network in the adaptive observer and the ideal values.

[0026] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the multi-agent sensor fault diagnosis method based on adaptive learning as described above.

[0027] A fourth aspect of the present invention provides a computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the multi-agent sensor fault diagnosis method based on adaptive learning as described above.

[0028] Compared with the prior art, the beneficial effects of the present invention are:

[0029] This invention provides a multi-agent sensor fault diagnosis method based on adaptive learning. It uses an adaptive method to diagnose sensor faults and ensures the eventual consistency and boundedness of identification errors through adaptive learning technology.

[0030] This invention provides a multi-agent sensor fault diagnosis method based on adaptive learning. The observer uses a radial basis function neural network to approximate sensor faults. The radial basis function neural network has a simple structure and has a better approximation ability than other feedforward networks. It can approximate any uncertain nonlinear system with arbitrary accuracy. Attached Figure Description

[0031] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0032] Figure 1 is a flowchart of the multi-agent sensor fault diagnosis method based on adaptive learning according to Embodiment 1 of the present invention. Detailed Implementation

[0033] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0034] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0035] Example 1

[0036] This embodiment provides a multi-agent sensor fault diagnosis method based on adaptive learning, as shown in Figure 1, which specifically includes the following steps:

[0037] Step 1: Obtain the control signal u for each nonlinear multi-agent subsystem with sensor faults. i and output y i .

[0038] It should be noted that, unless otherwise specified, in this embodiment, the nonlinear multi-agent subsystem or multi-agent subsystem or subsystem refers to a nonlinear multi-agent subsystem that has experienced a failure.

[0039] In this system, a nonlinear multi-agent subsystem corresponds to one or more sensors.

[0040] The model of the i-th nonlinear multi-agent subsystem is as follows:

[0041]

[0042] in, C ob =[1 0 … 0] It is the state vector of the i-th subsystem when the fault occurs. Represents the state vector x i The derivative of y with respect to time t i ∈R is the output of the i-th subsystem. Let be the input vector (control signal) of the i-th subsystem. It is a smooth nonlinear function. Due to sensor malfunction, T o i Let represent the time when the unknown fault of the i-th subsystem occurs, and t represent the t-th time.

[0043] Considering the possibility of sudden sensor failure, design β(t,T) o i Let ) be the step function, as follows:

[0044]

[0045] Considering sensor failure and measurable signals (e.g., control signal u) i Related to this, the control signal trajectory is denoted as φ(u i And the following assumptions are made:

[0046] Assumption 1: For control signal u i ∈U, the nonlinear function term q of the subsystem i (x i ,u i It satisfies the local Lipshitz continuity condition, that is:

[0047]

[0048] Where q0 is q i (x i ,u i A local Lipschitz constant, It is the state vector of the adaptive observer.

[0049] Assumption 2: Control signal trajectory φ(u) i It is a periodic or cyclical orbit.

[0050] Step 2: Based on the control signal u of each nonlinear multi-agent subsystem i and output y iAn adaptive observer is used to obtain the state vector of the adaptive observer. The adaptive observer ensures that the tracking error and the weight vector error are bounded, and the tracking error is the output estimate of the multi-agent subsystem. With the output y of the multi-agent subsystem i The difference, the weight vector error is the estimated weight of the neural network in the adaptive observer. Compared with the ideal value W i * The difference.

[0051] The adaptive observer is an adaptive neural network observer, which is constructed as follows:

[0052]

[0053] in, It is part of the observer gain; It is to estimate the weights; for Time derivative; S i (u i ) is a regression vector described in the form of an RBF (radial basis function); L i =diag{l i ,l i 2 ,...,l i n}, and l i l is a positive constant, where n is the dimension of the state vector; i Θ is a positive constant. i Let Θ be a positive definite symmetric matrix. i =Θ i T >0; l i >0; It is the state vector of the adaptive observer; Represents the state vector of the adaptive observer The derivative of y with respect to time t; i It is the output of the i-th subsystem. The output is y i The estimated value; It is a radial basis function neural network (RBF) used to approximate sensor faults (there exists a predicted cyclic trajectory; due to sensor faults, the cyclic trajectory becomes larger, and the RBF neural network can learn this fault, capturing the portion of the trajectory increase caused by the fault); q i () is a smooth nonlinear function.

[0054] parameter for:

[0055]

[0056] Among them, P i It is a positive definite matrix that satisfies:

[0057]

[0058] Construct matrix K i for:

[0059]

[0060] For the i-th nonlinear multi-agent subsystem, the tracking error is defined as follows: and weight vector error

[0061]

[0062] Among them, W i * For an ideal neural network weight vector, It is a bounded time-varying matrix and is exponentially stable.

[0063] Based on the above equation (8), the tracking error dynamics function is defined.

[0064]

[0065] Define the weight vector error dynamics function

[0066]

[0067] Where ∈ represents the approximation error of the radial basis function neural network.

[0068] Next, consider the state vector x of the i-th nonlinear multi-agent subsystem. i Perform parameter conversion:

[0069]

[0070] and

[0071]

[0072] Based on equations (11) and (12) above, the transformed tracking error dynamics function can be obtained. as follows:

[0073]

[0074] Define α i satisfy:

[0075]

[0076] but It manifests as:

[0077]

[0078] Define G i satisfy:

[0079] G i =C ob K i +S i (u i (16)

[0080] Then the error dynamics function of the transformed weight vector can be derived.

[0081]

[0082] Consider the tracking error dynamics function If assumption 1 holds, then the observer gain l can be selected. i To achieve tracking error and weight vector error The eventual uniform boundedness.

[0083] Constructing Lyapunov functions:

[0084]

[0085] The time derivative of the Lyapunov function V is:

[0086]

[0087] Among them, W i * Let I be the ideal weight vector, and let I be the identity matrix.

[0088] For equation (13) The following definition is given, and derived from the mean value theorem:

[0089]

[0090] in,

[0091] From equation (14), we can obtain the following through algebraic operations: Based on hypothesis 1, q i Satisfying the local Lipschitz continuity condition, the above equation (20) is derived as follows:

[0092]

[0093] Among them, in l i Under the condition that >1, It is a kind of l i Irrelevant constants.

[0094] Substituting the above equation (21) into the time derivative of the Lyapunov function V, we get 2α. i T P i ζ i Part of it, the derivation process is as follows

[0095]

[0096] in, and It is K i The upper limit.

[0097] Substituting equation (21) into the time derivative of the Lyapunov function V... The derivation process for this part is as follows:

[0098]

[0099] Among them, p1 i Let P be a matrix i The smallest eigenvalue, and It is G i The upper boundary.

[0100] Therefore, combining equations (19), (22), and (23), the time derivative of function V satisfies:

[0101]

[0102] Select observer gain li Make it satisfy the following inequality:

[0103]

[0104] Define parameters satisfy:

[0105]

[0106] From this, the time derivative of the Lyapunov function V can be derived. The final form:

[0107]

[0108] Observing the above formula (27), we can see that only α i or When inequality (28) is satisfied,

[0109]

[0110] Then α i and The eventual consistency and boundedness are guaranteed as follows:

[0111]

[0112] Due to α i and Both guarantee boundedness, which means that and It is also bounded, as shown below:

[0113]

[0114] Step 3: Using the state vector of the adaptive observer The residual error is obtained by comparing the state vector of the non-faulty multi-agent subsystem with the state vector of the non-faulty multi-agent subsystem. The residual error is then compared with the set value to obtain fault estimation information (i.e., information on whether the sensor corresponding to the nonlinear multi-agent subsystem is faulty), thus ensuring the eventual uniformity and boundedness of the identification error system.

[0115] The multi-agent sensor fault diagnosis method based on adaptive learning provided in this embodiment uses an adaptive method to diagnose sensor faults. Through adaptive learning technology, the eventual consistency and boundedness of the identification error are guaranteed.

[0116] The multi-agent sensor fault diagnosis method based on adaptive learning provided in this embodiment uses a radial basis function neural network (RBN) to approximate sensor faults in the observer. The RBN has a simple structure and has a better approximation ability than other feedforward networks. It can approximate any uncertain nonlinear system with arbitrary precision.

[0117] Example 2

[0118] This embodiment provides a multi-agent sensor fault diagnosis system based on adaptive learning, which specifically includes the following modules:

[0119] The data acquisition module is configured to acquire the control signals and outputs of each multi-agent subsystem.

[0120] The fault diagnosis module is configured to: based on the control signal and output, use an adaptive observer to obtain the state vector of the adaptive observer, and obtain the fault information of the sensor corresponding to the multi-agent subsystem by comparing the state vector of the adaptive observer with the state vector of the multi-agent subsystem;

[0121] The adaptive observer ensures that the tracking error and the weight vector error are bounded. The tracking error is the difference between the output estimate of the multi-agent subsystem and the output of the multi-agent subsystem, and the weight vector error is the difference between the estimated weights of the neural network in the adaptive observer and the ideal values.

[0122] It should be noted that each module in this embodiment corresponds one-to-one with each step in Embodiment 1, and their specific implementation processes are the same, so they will not be repeated here.

[0123] Example 3

[0124] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the multi-agent sensor fault diagnosis method based on adaptive learning as described in Embodiment 1 above.

[0125] Example 4

[0126] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the multi-agent sensor fault diagnosis method based on adaptive learning as described in Embodiment 1 above.

[0127] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0128] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more blocks of the flowchart illustrations and / or one or more blocks of the block diagrams.

[0129] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more block diagrams.

[0130] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions specified in one or more flowcharts and / or one or more block diagrams.

[0131] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0132] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A multi-agent sensor fault diagnosis method based on adaptive learning, characterized in that, include: Acquire the control signals and outputs of each multi-agent subsystem, and design a step function considering sudden sensor failures; Based on the control signal and output, an adaptive observer is used to obtain the state vector of the adaptive observer. By comparing the state vector of the adaptive observer with the state vector of the multi-agent subsystem, the fault information of the sensor corresponding to the multi-agent subsystem is obtained. The adaptive observer ensures that the tracking error and weight vector error are bounded. The tracking error is the difference between the output estimate of the multi-agent subsystem obtained by the adaptive observer and the output of the multi-agent subsystem. The weight vector error is the difference between the estimated weights of the neural network in the adaptive observer and the ideal values. The boundedness of the tracking error and weight vector error is verified by selecting the observer gain and calculating the time derivative of the Lyapunov function. The fault information of the sensor corresponding to the multi-agent subsystem is obtained by comparing the residual error between the state vector of the adaptive observer and the state vector of the multi-agent subsystem with a set value.

2. The multi-agent sensor fault diagnosis method based on adaptive learning as described in claim 1, characterized in that, The adaptive observer uses a radial basis function neural network to learn the portion of the control signal trajectory that increases due to sensor failure.

3. The multi-agent sensor fault diagnosis method based on adaptive learning as described in claim 1, characterized in that, The Lyapunov function is constructed as follows: in, ； ， To track errors, ,and A positive constant; ； To construct a matrix; For the weight vector error; It is a positive definite matrix.

4. The multi-agent sensor fault diagnosis method based on adaptive learning as described in claim 1, characterized in that, No. The model of the multi-agent subsystem is as follows: in, For the given matrix, and For the given vector, It is the first The state vectors of each subsystem It is the first The output of each subsystem For the first Control signals for each subsystem It is the first Nonlinear functions of individual subsystems Sensor malfunction. Indicates the first The time of occurrence of unknown failures in each subsystem It is a step function. State vector The derivative with respect to time t.

5. The multi-agent sensor fault diagnosis method based on adaptive learning as described in claim 1, characterized in that, For the control signal, the nonlinear function of the multi-agent subsystem satisfies the local Lipschitz continuity condition.

6. A multi-agent sensor fault diagnosis system based on adaptive learning, characterized in that, include: The data acquisition module is configured to acquire the control signals and outputs of each multi-agent subsystem. Considering the possibility of sudden sensor failure, design a step function; The fault diagnosis module is configured to: based on the control signal and output, use an adaptive observer to obtain the state vector of the adaptive observer; by comparing the state vector of the adaptive observer with the state vector of the multi-agent subsystem, obtain the fault information of the sensor corresponding to the multi-agent subsystem; wherein, the adaptive observer ensures that the tracking error and the weight vector error are bounded, the tracking error is the difference between the output estimate of the multi-agent subsystem obtained by the adaptive observer and the output of the multi-agent subsystem, and the weight vector error is the difference between the estimated weights of the neural network in the adaptive observer and the ideal values; verify that the tracking error and the weight vector error are bounded by selecting the observer gain and calculating the time derivative of the Lyapunov function; obtain the fault information of the sensor corresponding to the multi-agent subsystem by comparing the residual error between the state vector of the adaptive observer and the state vector of the multi-agent subsystem with a set value.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the multi-agent sensor fault diagnosis method based on adaptive learning as described in any one of claims 1-5.

8. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the multi-agent sensor fault diagnosis method based on adaptive learning as described in any one of claims 1-5.

Images