Unmanned aerial vehicle formation self-adaptive anti-interference fault diagnosis method for noise environment

Abstract

The invention discloses an unmanned aerial vehicle formation adaptive anti-interference fault diagnosis method for a noise environment, and relates to the technical field of unmanned aerial vehicle fault diagnosis, and the method comprises the steps: obtaining an observation signal; based on an adaptive particle swarm optimization algorithm, an envelope entropy of an intrinsic mode function component obtained by minimizing an observation signal after variational mode decomposition is taken as an optimization target, and a VMD parameter combination comprising a penalty factor and the number of mode components is dynamically searched and determined, so that the observation signal is decomposed again, and a VMD parameter combination comprising a penalty factor and the number of mode components is obtained. Selecting the effective intrinsic mode function component for reconstruction to obtain a noise reduction signal; based on the noise reduction signal and a formation system model, extracting formation consistency key features including a position error average consistency feature and a speed error average consistency feature; the formation consistency key features are subjected to fusion reasoning by using the confidence rule base, so that the overall fault state of the unmanned aerial vehicle formation system can be evaluated in a manner of adaptively filtering interference and keeping the reasoning process transparent.

Description

Technical Field

[0001] This invention belongs to the field of UAV fault diagnosis technology, specifically relating to an adaptive anti-interference fault diagnosis method for UAV formations in noisy environments. Background Technology

[0002] Unmanned Aerial Vehicle (UAV) swarms perform complex tasks such as reconnaissance, inspection, and logistics through coordinated flight and information exchange among their members. The robustness and reliability of these systems are crucial for mission success. As system scale increases and mission environments become more complex, swarms are susceptible to various problems during flight, including communication interference, sensor malfunctions, and power system failures. Therefore, accurate and timely fault diagnosis of UAV swarms is of paramount strategic importance for ensuring flight safety, improving mission success rates, enabling predictive maintenance, and reducing operational costs.

[0003] However, in actual flight missions, the status monitoring data of UAV formations is highly susceptible to contamination by complex environmental noise during the transmission back to the ground station via the communication link. This channel noise can cause distortion of received key signals (such as position and velocity), reduce the signal-to-noise ratio, and create a data contamination problem where noise overwhelms fault characteristics, making fault mode identification extremely difficult. Traditional fault diagnosis methods, whether relying on precise model-based mechanistic analysis, purely data-driven methods requiring massive amounts of clean labeled data, or hybrid methods attempting to combine qualitative expert knowledge, all struggle to effectively balance diagnostic robustness and model interpretability when data quality is severely degraded due to channel interference.

[0004] Especially in the context of strong noise, how to construct a diagnostic model that can adaptively filter out interference while maintaining the transparency of the inference process has become a key technical bottleneck hindering the safe and reliable operation of UAV formations. Summary of the Invention

[0005] To address the aforementioned problems in the existing technology, this invention provides an adaptive anti-interference fault diagnosis method for UAV formations in noisy environments.

[0006] The technical problem to be solved by this invention is achieved through the following technical solution: In a first aspect, the present invention provides an adaptive anti-interference fault diagnosis method for UAV formations in noisy environments, the UAV formation adaptive anti-interference fault diagnosis method comprising: Acquire noisy observation signals; Based on the adaptive particle swarm optimization algorithm, the optimization objective is to minimize the envelope entropy of the intrinsic mode function components obtained after variational mode decomposition of the observed signal. The VMD (Variational Mode Decomposition) parameter combination is dynamically searched and determined. The VMD parameter combination includes a penalty factor and the number of mode components. The observed signal is decomposed again using the VMD parameter combination, and the effective intrinsic mode function components are selected for reconstruction to obtain a denoised signal; Based on the denoised signal and the formation system model, key features of formation consistency are extracted; these key features of formation consistency include average position error consistency features and average velocity error consistency features. By using a confidence rule base to perform fusion reasoning on the key features of formation consistency, the overall fault status of the UAV formation system can be assessed.

[0007] Secondly, the present invention provides an adaptive anti-interference fault diagnosis device for UAV formations in noisy environments, the UAV formation adaptive anti-interference fault diagnosis device comprising: The acquisition module is used to acquire noisy observation signals; An optimization module is used to dynamically search and determine the VMD parameter combination of variational mode decomposition based on an adaptive particle swarm optimization algorithm, with the optimization objective of minimizing the envelope entropy of the intrinsic mode function components obtained after variational mode decomposition of the observed signal; the VMD parameter combination includes a penalty factor and the number of mode components; The reconstruction module is used to decompose the observed signal again using the VMD parameter combination, select the effective intrinsic mode function components for reconstruction, and obtain the noise-reduced signal. The extraction module is used to extract key features of formation consistency based on the denoised signal and the formation system model; the key features of formation consistency include average position error consistency features and average velocity error consistency features; The evaluation module is used to perform fusion reasoning on the key features of formation consistency using a confidence rule base to evaluate the overall fault status of the UAV formation system.

[0008] Thirdly, the present invention provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; When the processor executes a computer program stored in memory, it implements the steps of any of the above-mentioned adaptive anti-interference fault diagnosis methods for UAV formations in noisy environments.

[0009] Fourthly, the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of any of the above-described adaptive anti-interference fault diagnosis methods for UAV formations in noisy environments.

[0010] This invention provides an adaptive anti-interference fault diagnosis method for UAV formations in noisy environments. Based on the adaptive particle swarm optimization algorithm, it aims to minimize the envelope entropy of the intrinsic mode function components obtained after variational mode decomposition (VMD) of the observed signal. It dynamically searches for and determines the VMD parameter combination for VMD, and then uses this combination to re-decompose the observed signal, selecting effective intrinsic mode function components for reconstruction to obtain a denoised signal. Finally, based on the denoised signal and the formation system model, it extracts key features of formation consistency, thereby achieving adaptive noise filtering and effective extraction of key fault features.

[0011] Furthermore, this invention utilizes a confidence rule base to perform fusion reasoning on key features of formation consistency, and integrates multi-source features through an evidence reasoning mechanism to output a high-confidence fault state assessment result.

[0012] Furthermore, by integrating adaptive signal decomposition and interpretable rule base optimization, this invention effectively solves the diagnostic challenges caused by data distortion under noise interference. While ensuring the transparency and rationality of the reasoning process, it achieves high-precision, robust, and reliable diagnosis of the fault status of UAV formations, providing technical support for the safe operation and intelligent maintenance of UAV clusters in complex environments.

[0013] The present invention will now be described in further detail with reference to the accompanying drawings. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating an adaptive anti-interference fault diagnosis method for UAV formations in noisy environments, provided by an embodiment of the present invention. Figure 2 This is a schematic diagram comparing the average position error signal under different conditions; Figure 3 This is a diagram illustrating the accuracy of nine simulation experiments; Figure 4 This is a diagram of a confusion matrix; Figure 5 This is a schematic diagram of the structure of an adaptive anti-interference fault diagnosis device for UAV formations in noisy environments provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0015] The present invention will be further described in detail below with reference to specific embodiments, but the implementation of the present invention is not limited thereto.

[0016] Existing adaptive anti-interference fault diagnosis methods for UAV formations struggle to effectively balance diagnostic robustness and model interpretability when data quality is severely degraded due to channel interference. To address this issue, this invention provides an adaptive anti-interference fault diagnosis method for UAV formations in noisy environments. (See [link to relevant documentation]). Figure 1 , Figure 1 This is a flowchart illustrating an adaptive anti-interference fault diagnosis method for UAV formations in noisy environments, provided by an embodiment of the present invention. The details are as follows: Step S101: Obtain the noisy observation signal.

[0017] In this embodiment of the invention, acquiring the noisy observation signal includes: Establish a state-space mathematical model of the drone formation to obtain the formation system; The observation signals transmitted back to the ground station after the formation system is contaminated by additive noise are acquired.

[0018] Specifically, for a leaderless drone formation with sensor malfunctions, this embodiment of the invention models and analyzes it based on state-space equations, with the formation system size set as follows: Given a fleet of drones, the state-space mathematical model of the drone formation, including those with sensor faults, is as follows: ; in, express The derivative; Indicates at time , No. The state vector of each drone; Indicates at time , No. Distributed control input vectors for each UAV , express A European-style space; Indicates at time , No. The output vector of each drone, , express P The real space of dimensional numbers; Represents the sensor fault vector. , Represents the set of real numbers; , and Each represents a constant matrix of the appropriate dimension; , This indicates the total number of drones.

[0019] To achieve the goal of consistent control of the drone formation, the following distributed control protocol is designed for each drone in this embodiment of the invention: ; in, Indicates at time , No. Distributed control input vectors for each UAV; Represents a constant matrix with appropriate dimensions; Indicates at time , No. The output vector of each drone; express At that moment, the The output vector of each drone; Indicates at time , No. The drone and the first Connection status between drones ;when Time means at any given moment , No. The drone can receive the first Each drone sends its output vector to update its own input vector; when At that time, it cannot be received.

[0020] In this embodiment of the invention, to simulate additive noise interference in the communication channel between UAV formations and ground stations, the following channel model is established: ; in, This represents the noisy observation signal ultimately received by the ground station; Indicates at time , No. The output vector of each drone; This represents additive Gaussian noise.

[0021] A leaderless drone formation consisting of multiple drones, such as five drones, performs horizontal maneuvers, taking response time into account. The equation of motion for a drone can be expressed as: ; in, For the first The location coordinates of the drone; For the first The ground speed of the drone; For the first The heading angle of the drone; For the first The ground speed control command corresponding to each drone; It is a time constant; For the first The heading angular velocity of the drone; for exist Projection on the axis; for exist Projection on the axis; express The derivative; express The derivative; In this embodiment of the invention, for and Differentiation yields: ; in, express The derivative; express The derivative of .

[0022] Let the actual input vector be denoted as The virtual input vector is denoted as The formula for calculating the virtual input vector can be obtained as follows: ; Among them, superscript This represents the matrix transpose operation; for exist Projection on the axis; for exist Projection on the axis; when At this time, the actual input vector and the virtual input vector can be converted using the following conversion formula: ; Then, the first The motion state of a drone can be represented in the following standard form: ; in, For at any time , No. The state vector of each drone; For at any time , No. Distributed control input vectors for each UAV; For at any time , No. The output vector of each UAV.

[0023] In the In the standard form of the motion state formula for a UAV, a state-space mathematical model of the UAV with sensor faults can be obtained by introducing sensor faults. Among them, the first... The specific parameters of the drone are as follows: ; in, Represents the sensor fault vector; This represents the lower bound of the sensor fault vector; This represents the upper bound of the sensor fault vector; No. The consensus protocol for a single UAV using the aforementioned distributed control protocol can be converted into the following protocol: ; in, Indicates at time , No. Distributed control input vectors for each UAV; Indicates the first A vector for the formation of drones; Indicates the first A vector for the formation of drones; Represents the desired state vector. ; , All represent constant matrices with appropriate dimensions; ; in, Indicates the desired speed; The desired angle can be set by technical personnel according to requirements; This represents the formation vector of the first drone formation; This represents the formation vector of the second drone formation; This represents the formation vector of the third drone formation; This represents the formation vector of the 4th drone formation; This represents the formation vector of the 5th drone formation; The initial state vector parameters for each member are as follows: ; in, This represents the initial state parameter vector of the first UAV; This represents the initial state parameter vector of the second UAV; This represents the initial state parameter vector of the third UAV; This represents the initial state parameter vector of the fourth UAV; This represents the initial state parameter vector of the 5th UAV; To simulate performance degradation in real-world scenarios, and The severity of the faults was gradually reduced from the initial value, with an experimental duration of 150 seconds and discrete time points collected at 0.2-second intervals. Five types of UAV formation fault states were constructed, representing the severity of the faults as "optimal," "good," "moderate," "critical," and "poor."

[0024] Based on the aforementioned channel model, the first The observation signals transmitted back to the ground station by a drone can be represented as follows: ; in, The power of the original signal; For the first discrete time points The number of discrete time points; The power of the observed signal; The signal-to-noise ratio (SNR) is the intensity of the signal, including the noise. For variance; For the first Noise intensity at discrete time points ; To represent a vector with a mean of 0 and a variance of , The normal distribution; For the first A standard Gaussian random variable at discrete time points; For the first At discrete points in time, the contaminated observation signals from the ground station are transmitted back. Indicates the first At each discrete time point, the first... The output vector of each UAV.

[0025] Step S102: Based on the adaptive particle swarm optimization algorithm, the optimization objective is to minimize the envelope entropy of the intrinsic mode function components obtained after variational mode decomposition of the observed signal. The VMD parameter combination of variational mode decomposition is dynamically searched and determined. The VMD parameter combination includes the penalty factor and the number of mode components.

[0026] In this embodiment of the invention, a signal decomposition model based on variational mode decomposition can be used to solve the problem that the backhaul signal of UAV formation is contaminated by noise in the channel, which causes the fault characteristics to be submerged and affects the accuracy of diagnosis.

[0027] In this embodiment of the invention, a signal decomposition model based on variational mode decomposition is constructed, the goal of which is to adaptively decompose noisy observation signals into several sparse intrinsic mode function (IMF) components.

[0028] First, the core mathematical framework of the adaptive particle swarm optimization algorithm will be explained: The Adaptive Particle Swarm Optimization (APSO) algorithm systematically improves optimization performance by introducing Evolutionary State Estimation (ESE) into the standard PSO framework to achieve adaptive parameter adjustment and incorporating an Elite Learning Strategy (ELS) to enhance global search capabilities. Its general update formula is as follows: ; in, , This indicates the number of particles, i.e., the population size. , This represents the dimension of the solution space, i.e., the maximum number of iterations. Indicates the first The particle in the first A dimensional velocity vector; Indicates the range of inertial weights; , Both represent acceleration factors; and Indicates the first Wei Zai Two uniformly distributed random numbers generated independently within the range; Indicates the first The particle in the first The location with the best fitness found by Dimension; Indicates the first The particle in the first A dimensional position vector; This indicates the optimal position within the domain, specifically in the d-th dimension. Euclidean distance is used to measure each particle. Average distance from other particles : ; in, Indicates the first The particle in the first A dimensional position vector; Let the average distance of the global optimum be denoted as . By comparison The maximum and minimum distances are denoted as follows: and Evolutionary factors are calculated by distance. : ; Will The state strategy is divided into four groups based on the fuzzy membership function: exploration state. Development status Convergence state Exiting the state The specific formula is as follows: ; in, The membership function representing the exploration state; Membership function representing the development status; The membership function representing the convergence state; The membership function representing the exit state; Subsequently, the inertial weight range was adjusted using the aforementioned strategy. and acceleration factor , Adaptive adjustments are made. The sigmoid activation function is used for mapping. It can make The value varies Change for the sake of change; among them, Represents the set of positive real numbers. When in the exploration state, it increases. And decrease This is to search for the optimal location as many times as possible; when in development mode, slightly increase... At the same time, slightly reduced Because it slightly increases Helpful to To explore and develop, while slightly reducing This helps avoid getting trapped in local optima, meaning that the globally optimal particle doesn't always find the globally optimal position; in the convergent state, slightly increasing and Slightly increased It can guide other particles to reach the globally optimal region as much as possible, if the size is reduced It can enable rapid population convergence but can easily lead to premature convergence. To avoid this problem... Also slightly increase. In the exit state, based on the convergence state, decrease. and increase When a globally optimal particle escapes a local optimum, it may become a new guiding particle, and other particles need to follow it to the new region as quickly as possible, thereby reducing... Increase The following formula helps to achieve this goal: ; ; in, Indicates the range of inertial weights With evolutionary factors The status is dynamically adjusted; Indicates the current iteration number; Indicates the current iteration number ; Indicates the next iteration number ; hour, for ; hour, for ; Indicates the acceleration rate, used to constrain... and The maximum increment or decrease between these intervals ensures that the adjustment of the acceleration coefficient is not too abrupt; use the interval constraint and If the sum of the parameters is greater than 4.0, then the following constraint formula is applied. and Re-standardization: ; Finally, by employing a learning strategy, it is applied to help the globally optimal particle escape its local optimum when the search state is determined to be convergent. The specific formula is as follows: ; ; in, Indicates the first The dimensional elite learning strategy randomly selects a dimension with the best historical position and the highest fitness; the search range is... ; It is a normal distribution; It follows the mean; It is the standard deviation; and yes The upper and lower bounds indicate the learning scale upon reaching the new area; This represents the current iteration number; This represents the maximum number of iterations.

[0029] The penalty factor and the number of modal components for VMD parameter combination are optimized based on the APSO algorithm, and the minimum envelope entropy is used as the fitness function. According to the aforementioned general update formula, the adaptive particle swarm optimization algorithm specific to this embodiment of the invention can be obtained, including: ; ; in, Indicates the current iteration number Below, the number of modal components The velocity vector; Indicates the next iteration number Below, the number of modal components The velocity vector; Indicates the current iteration number Below, the number of modal components The position vector; Indicates the current iteration number Below, punishment factor The velocity vector; Indicates the next iteration number Below, punishment factor The velocity vector; Indicates the current iteration number Below, punishment factor The position vector; Indicates the next iteration number Below, the number of modal components The position vector; Indicates the next iteration number Below, punishment factor The position vector; This represents the number of individual historical best modal components. ; Represents the individual's historical best punishment factor ; Represents the number of global historical best modal components. ; Represents the global historical best penalty factor .

[0030] Using an adaptive particle swarm optimization algorithm, the optimal VMD parameter combination of the penalty factor and the number of mode components in variational mode decomposition is determined by dynamic search with the goal of minimizing the envelope entropy of the intrinsic mode function components obtained after variational mode decomposition of the observed signal.

[0031] In this embodiment of the invention, the observed signal is decomposed into multiple intrinsic mode function components using variational mode decomposition. Each intrinsic mode function component Performing a Hilbert transform can demodulate the corresponding instantaneous amplitude, yielding the envelope signal. It directly reflects the energy distribution of the signal at different points in time: ; in, Represents the Hilbert transform operator; For discrete time point indexes, , Indicates the number of discrete time points; It represents the imaginary unit.

[0032] To calculate information entropy, the envelope signal needs to be transformed into a discrete probability distribution. Through normalization, the envelope signal is converted into a probability mass function. : ; This operation ensures Thus satisfying the probability distribution The magnitude directly characterizes the signal energy at discrete time points. The concentration ratio on.

[0033] Based on information entropy theory, the Shannon entropy value of this probability distribution is calculated and defined as the envelope entropy of the intrinsic mode function component: ; Among them, envelope entropy It quantitatively reflects the concentration of impulse components in the intrinsic mode function components. The lower the entropy value, the more concentrated the signal energy is in a few impulse events, and the better its sparsity; the higher the entropy value, the more dispersed the signal energy is, and the more noise components there are.

[0034] In this embodiment of the invention, the fitness function of the adaptive particle swarm optimization algorithm is defined as: ; in, This represents the modal bandwidth.

[0035] The minimum envelope entropy means that the signal energy is highly concentrated at a few time points, that is, it has good sparsity, prominent impact components, and background noise and interference components are effectively suppressed. The fault characteristics have obvious peak characteristics in the time domain distribution.

[0036] The fitness function ensures that the optimization process is directly guided to finding the optimal VMD parameter combination that produces the sparsest and most representative eigenmode function components. .

[0037] Step S103: The observed signal is decomposed again using VMD parameter combination, and the effective intrinsic mode function components are selected for reconstruction to obtain the denoised signal.

[0038] In this embodiment of the invention, the observed signal is re-decomposed using VMD parameter combinations, and effective intrinsic mode function components are selected for reconstruction to obtain a denoised signal, including: The observed signal is re-decomposed using VMD parameter combinations to obtain multiple intrinsic mode function components; Calculate the correlation coefficient and envelope entropy between each intrinsic mode function component and the observed signal; The intrinsic mode function components with a correlation coefficient greater than the first preset threshold and an envelope entropy less than the second preset threshold are taken as valid intrinsic mode function components. The denoised signal is reconstructed by superimposing multiple effective intrinsic mode function components.

[0039] In VMD parameter combinations Based on this, the constrained variational problem of VMD can be expressed in the following form: ; in, Represents the first continuous time interval. Each intrinsic mode function component ; Indicates the first The center frequencies of the eigenmode function components ; It is a Dirac distribution; , The number of intrinsic mode function components obtained from the decomposition, i.e., the modal bandwidth; It is about time The derivative; It is the imaginary unit, satisfying ; It is a convolution operation; yes The square of the norm; It is an observed signal, corresponding to the aforementioned .

[0040] Introducing a quadratic penalty term to reduce the impact of Gaussian noise on the reconstructed signal and introducing Lagrange multipliers Strictly implement constraints to augment the Lagrange function. The following is introduced: ; in, Indicates the first The center frequencies of the eigenmode function components; Indicates the first The intrinsic mode function components at time 1 The amplitude at that point; It is an inner product operation; Indicates at time of ; Indicates at time of ; This indicates the upper limit of the range of values ​​for the penalty factor. .

[0041] During the decomposition process, the initial decomposition mode is obtained through initialization operations. Initial center frequency and the initial Lagrange multipliers Then, the frequency domain decomposed modes are updated using the following update formula. Center frequency and Lagrange multipliers ,in This indicates that, except for the currently updating number... Other intrinsic mode function components besides the original intrinsic mode function components. The update formula is as follows: ; ; ; in, Indicates the first In the next iteration The frequency domain form; express The frequency domain form; Indicates the first In the next iteration The frequency domain form, Indicates the first The intrinsic mode function components at time 1 The amplitude at that point; Indicates the first In the next iteration The frequency domain form; Indicates the first The frequency domain form of the Lagrange multipliers in the next iteration; Represents frequency variables; Indicates the first The center frequency in the next iteration; Indicates the first The center frequency in the next iteration; This indicates the step size for updating the Lagrange multiplier.

[0042] In this embodiment of the invention, preset parameters are introduced. Used to terminate iteration, its expression is as follows: ; Among them, the formula is as follows , , "etc" is an identifier for a frequency domain signal, indicating that the variable is the frequency domain form of the corresponding time domain signal after Fourier transform.

[0043] To screen out modes containing fault characteristics, the correlation coefficient between each intrinsic mode function (IMF) component and the observed signal is first calculated, along with its envelope entropy. IMF components with high correlation coefficients and low envelope entropies are selected as effective IMF components for signal reconstruction. Specifically, IMF components with high correlation coefficients can refer to those with correlation coefficients greater than a first preset threshold, and IMF components with low envelope entropies can refer to those with envelope entropies less than a second preset threshold. The first and second preset thresholds can be set by those skilled in the art according to their needs, and are not limited here. The correlation coefficient refers to the correlation between the IMF component and the observed signal.

[0044] In this embodiment of the invention, based on a general reconstruction formula, a denoised signal is reconstructed by superimposing multiple effective intrinsic mode function components, including: ; in, This represents the speed signal after noise reduction; Indicates the noise reduction Direction and position signals; Indicates the noise reduction Direction and position signals; Represents the effective intrinsic mode function components; This indicates a general noisy signal, i.e., the signal type currently being processed; This indicates that the general noisy signal is a speed signal; This indicates that the general noisy signal is Direction and position signals; This indicates that the general noisy signal is Direction and position signals; This represents the set of effective intrinsic mode function components.

[0045] Step S104: Extract key features of formation consistency based on the denoised signal and formation system model; key features of formation consistency include average consistency features of position error and average consistency features of velocity error.

[0046] In this embodiment of the invention, key features of formation consistency are extracted based on the denoised signal and the formation system model, including: Based on the noise-reduced signal and the formation system model, the average position consistency error and average velocity consistency error between each UAV and the desired state of the formation are calculated to obtain the key features of formation consistency.

[0047] In this embodiment of the invention, based on the acquired noise-reduced signal and formation system model, the average position consistency error and average velocity consistency error between each UAV and the desired formation state are calculated. This extracts key formation consistency features that directly reflect the formation's cooperative state and fault modes, serving as core indicators characterizing the overall operational health of the system. These features are then used as input to the subsequent confidence rule base and for fault state inference. The key formation consistency features include average position error consistency features and average velocity error consistency features.

[0048] The calculation formulas for key features of formation consistency include: ; in, Indicates the first Average consistency characteristics of the position error of each UAV; It is the first The average consistency characteristics of the speed error of each UAV; This indicates the calculation of Euclidean distance. It is the first The noise-reduced formation position vector of each drone; It is the noise-reduced formation position vector of the first drone; It is the first The expected formation position vector of each drone; It is the expected formation position vector of the first UAV; It is the desired speed signal; It is the point in time that corresponds to the convergence of the drone formation; Indicates the first A drone, at a certain time point The speed signal after noise reduction.

[0049] Step S105: Use the confidence rule base to perform fusion reasoning on the key features of formation consistency to achieve the assessment of the overall fault status of the UAV formation system.

[0050] In this embodiment of the invention, after signal denoising and formation consistency feature extraction are completed, the extracted features can be fused and reasoned using a confidence rule base (BRB) to achieve an assessment of the overall fault status of the UAV formation system.

[0051] In this embodiment of the invention, a confidence rule base is used to perform fusion reasoning on key features of formation consistency to assess the overall fault state of the UAV formation system, including: Confidence rules are constructed by taking the key features of formation consistency as prerequisite attributes. By calculating the activation weights of the confidence rules and using an evidence reasoning algorithm for fusion reasoning, the confidence distribution evaluation results of the overall fault state of the UAV formation system are output.

[0052] Using key features of formation consistency as prerequisite attributes, confidence rules are constructed. In the confidence rule base, the first... Confidence rules It can be represented as: ; in, This represents the feature transformation function, used to convert position vectors and velocity vectors into average consistent position error and average consistent velocity error; Indicates the first The noise-reduced formation position vector of each drone; Indicates the first The velocity vector of the drone; Indicates the first Reference values ​​corresponding to each prerequisite attribute ; It is the first evaluation system One assessment status, , It is the total number of assessed states; It is the first Under the confidence rule, the first The confidence level of the result corresponding to each evaluation state. , It represents the total number of confidence rules in the confidence rule base. It is the first The rule weight of each confidence rule; It is the first The attribute weights corresponding to each prerequisite attribute; It is the first The complete constraints corresponding to the confidence rules; among them. ; It is the first The expression for the non-convex constraint corresponding to the confidence rule is as follows: ; The first step in this optimization process is to determine the confidence level of the current result. The peak in the distribution, i.e., the location of the maximum value. Subsequently, according to Reorganizing the distribution: If the peak is located at the beginning or end of the sequence, the reorganized distribution will be constrained to a strictly monotonic sequence; if the peak is located in the middle of the sequence, the reorganized distribution will be constrained to a single-peaked shape centered on the peak, with the left side monotonically increasing and the right side monotonically decreasing.

[0053] Will The ascending set of reference values ​​is denoted as ,in The number of reference values ​​in each premise attribute. Total number of confidence rules. satisfy Furthermore, all confidence rules satisfy the integrity constraint. Non-convex constraints .

[0054] The confidence rule base fusion process can be divided into three steps: rule activation, rule reasoning, and parameter optimization. The specific steps are as follows: a) Confidence rule activation: according to Calculate relative to The matching degree is calculated using the following formula: ; in, express The first in One reference value; express The first in One reference value; Indicates the first The first prerequisite attribute and the first The degree of matching of each confidence rule; according to Calculate each confidence rule activation weight This activation weight reflects The degree to which something is activated to participate in subsequent evidence fusion is determined by the following formula: ; in, Indicates the first Attribute weights normalized for each feature; Indicates the first The rule weight of each confidence rule; Indicates the first The first prerequisite attribute and the first The degree of matching of the confidence rules.

[0055] b) Confidence rule fusion: After calculating the activation weights of each confidence rule, they are transformed into basic probability masses that can be used for synthesis. Define the... The confidence rule is assigned to the first The basic probability quality of each evaluation state The basic probability mass calculation formula is as follows: ; in, It is the first The total unassigned probability quality of a confidence rule represents the total uncertainty that the confidence rule fails to explicitly assign to any evaluation level; The unallocated quality is caused by the incomplete conclusions of the confidence rule itself; The unallocated quality is caused by insufficient activation weights in the confidence rule.

[0056] Based on the basic probability quality calculation formula, the Evidence Reasoning (ER) algorithm is used to aggregate all activated confidence rules. The recursive form of this algorithm is as follows: ; in, For the front Performance status in confidence rules The basic probability assignment of merging; For the front Residual basic probability assignment in the confidence rule; For the front In the confidence rule, the basic probability assignment of the merged data that did not participate in the calculation; The correction coefficient is used to ensure that the sum of probability masses equals one.

[0057] The fusion confidence score of the performance states is denoted as... The result can be obtained using the following formula: ; Therefore, the output generated by the adaptive anti-disturbance confidence rule base is an evaluation of the performance status of the UAV formation system, and its expression is as follows: ; in, Indicates the performance status of the drone formation system; Indicates utility value.

[0058] In this embodiment of the invention, the performance status of the drone formation system can be divided into five levels according to a preset range: (Optimal) (good), (medium), (critical), (Poor) When lie in When the pre-defined range is within the specified range, the overall performance status of the drone formation system is judged to be good, or when... lie in When the pre-defined range is reached, the overall performance status of the UAV formation system is judged to be poor, thereby realizing the assessment of the overall fault status of the UAV formation system.

[0059] c) Optimization of confidence rule base parameters: In one implementation, to further improve the accuracy of fault diagnosis, a step of optimizing the confidence rule base parameters can be performed. The specific process is as follows: The constraint problem of a performance evaluation system based on an adaptive anti-disturbance confidence rule base can be described as follows: ; To more clearly demonstrate the optimization process, all parameters to be optimized are integrated into a high-dimensional vector. This high-dimensional vector contains the confidence scores of all the confidence rules. and confidence rule weights and attribute weight The total dimension of the APSO optimization algorithm is ; ; in, This represents the first high-dimensional vector; Indicates the first A high-dimensional vector; This indicates the confidence level of confidence rule 1; Representing confidence rules Confidence level; Indicates the weight of the confidence rule; Indicates attribute weight.

[0060] Subsequently, based on the aforementioned APSO general update formula, an APSO optimization equation is formulated for the confidence rule base parameter vector: ; ; in, Indicates the next iteration number Next, the A velocity vector of a high-dimensional vector; Indicates the current iteration number Next, the A velocity vector of a high-dimensional vector; Represents the optimal high-dimensional vector of an individual's history; Indicates the current iteration number Next, the The position vector of a high-dimensional vector; Indicates the next iteration number Next, the The position vector of a high-dimensional vector; Represents the global historical best high-dimensional vector; .

[0061] The minimum mean squared error (MSE) is used as the fitness function for the APSO optimization algorithm, and the specific formula is as follows: ; ; in, It is the first The true output value of each training sample, i.e., the true fault state; It is the first The predicted output value of each training sample, i.e., the predicted fault state; This refers to the number of training samples. Each training sample consists of two parts: the key feature of the input sample formation consistency and the actual fault state corresponding to that feature.

[0062] By using a confidence rule base optimized with improved parameters to perform fusion reasoning on key features of formation consistency, a more accurate overall fault status assessment result for the UAV formation system can be obtained.

[0063] In this embodiment of the invention, based on the adaptive particle swarm optimization algorithm, the envelope entropy of the intrinsic mode function components obtained after variational mode decomposition of the minimum observed signal is used as the optimization objective. The VMD parameter combination of variational mode decomposition is dynamically searched and determined, and the observed signal is re-decomposed using the VMD parameter combination. The effective intrinsic mode function components are selected for reconstruction to obtain a denoised signal. Then, based on the denoised signal and the formation system model, key features of formation consistency are extracted, thereby realizing adaptive noise filtering and effective extraction of key fault features. In addition, the confidence rule base is used to perform fusion reasoning on key features of formation consistency, and multi-source features are fused through evidence reasoning mechanism to output high-confidence fault state assessment results; By integrating adaptive signal decomposition and interpretable rule base optimization, the diagnostic challenges caused by data distortion under noise interference are effectively solved. Under the premise of ensuring transparency and rationality of the reasoning process, high-precision, robust and reliable diagnosis of the fault status of UAV formations is achieved, providing core technical support for the safe operation and intelligent maintenance of UAV clusters in complex environments.

[0064] A simulation experiment of an adaptive anti-interference fault diagnosis method for UAV formations in noisy environments, provided by an embodiment of the present invention, is as follows: In the specific implementation process, the drone formation position and velocity signals that tend to be consistent between 70 seconds and 150 seconds are selected as the processing objects. The variational mode decomposition method based on the adaptive particle swarm optimization algorithm is used for noise reduction. The key parameter configurations used in the adaptive optimization process are shown in Table 1. Table 1 Adaptive Variational Mode Decomposition Parameter Configuration Table

[0065] Based on the extraction method of key features of formation consistency, attribute feature transformation is performed on the denoised formation signal with a time step of 1 to obtain the average consistency features of position error and the average consistency features of velocity error. On this basis, an input dataset containing 80 samples is constructed. Each sample consists of the average consistency features of position error and velocity error, and the corresponding fault state label. The fault states are divided into five levels: (Optimal) (good), (medium), (critical), (Poor). Taking the average location error as an example, the location errors of noiseless data, noisy data, and denoised data are as follows: Figure 2 As shown, Figure 2 This is a schematic diagram comparing the average position error signal under different conditions.

[0066] Before inputting the indicator data into the BRB, it is necessary to construct indicator reference values ​​and performance status reference values ​​based on typical operating conditions of UAV formations and domain expert knowledge, as shown in Tables 2 and 3. The initial indicator weights and rule weights of the confidence rule base are set to 1, and the initial confidence level is provided by technical personnel.

[0067] Table 2 Reference Values ​​for Indicators

[0068] Table 3 Performance Status Reference Values

[0069] The optimized Adaptive Anti-interference Belief Rule Base (BRB-AD) fault diagnosis model was validated on a dataset simulating a noisy UAV formation channel environment. The experimental setup was as follows: 70% of the noise-free data was used as the training set for model training and parameter optimization; the remaining 30% of the noisy data (SNR = 30 dB) was used as the test set to evaluate the model's diagnostic performance in real-world interference environments. The maximum number of iterations for the improved adaptive particle swarm optimization algorithm was set to 300. After nine independent repeated experiments, the model achieved an average diagnostic accuracy of 96.00% on the test set, demonstrating the good robustness and stability of this invention in noisy environments. The accuracy distribution of the nine simulation experiments is shown below. Figure 3 Show, Figure 3 This is a diagram illustrating the accuracy of nine simulation experiments. The fluctuation range is within an acceptable range, further confirming the reliability of the model's performance. For a more detailed analysis of the model's diagnostic performance, please refer to [link / reference]. Figure 4 , Figure 4This is a schematic diagram of the confusion matrix, showing the model's classification results across five fault states. The confusion matrix demonstrates that the model can achieve high-precision identification of most fault states.

[0070] In summary, based on the above experimental setup and result analysis, it can be seen that the present invention can achieve high-precision, high-stability, and interpretable diagnosis of the fault status of UAV formations in a strong noise communication environment. Its comprehensive performance is significantly better than that of traditional methods, meeting the engineering requirements for intelligent maintenance of UAV swarms in complex environments.

[0071] Based on the same inventive concept, embodiments of the present invention also provide an adaptive anti-interference fault diagnosis device for UAV formations in noisy environments, see [link to relevant documentation]. Figure 5 , Figure 5 This is a schematic diagram of a UAV formation adaptive anti-interference fault diagnosis device for noisy environments provided in an embodiment of the present invention. The UAV formation adaptive anti-interference fault diagnosis device includes: Acquisition module 501 is used to acquire noisy observation signals; Optimization module 502 is used to dynamically search and determine the VMD parameter combination of variational mode decomposition based on the adaptive particle swarm optimization algorithm, with the optimization objective of minimizing the envelope entropy of the intrinsic mode function components obtained after variational mode decomposition of the observed signal; the VMD parameter combination includes the penalty factor and the number of mode components; The reconstruction module 503 is used to re-decompose the observed signal using VMD parameter combinations, select effective intrinsic mode function components for reconstruction, and obtain a denoised signal. The extraction module 504 is used to extract key features of formation consistency based on the denoised signal and the formation system model; the key features of formation consistency include the average consistency features of position error and the average consistency features of velocity error. The evaluation module 505 is used to perform fusion reasoning on key features of formation consistency using a confidence rule base to evaluate the overall fault status of the UAV formation system.

[0072] In this embodiment of the invention, based on the adaptive particle swarm optimization algorithm, the optimization objective is to minimize the envelope entropy of the intrinsic mode function components obtained after variational mode decomposition of the observed signal. The VMD parameter combination of variational mode decomposition is dynamically searched and determined, and the observed signal is re-decomposed using the VMD parameter combination. The effective intrinsic mode function components are selected for reconstruction to obtain a denoised signal. Then, based on the denoised signal and the formation system model, key features of formation consistency are extracted, thereby realizing adaptive noise filtering and effective extraction of key fault features. In addition, the confidence rule base is used to perform fusion reasoning on key features of formation consistency, and multi-source features are fused through evidence reasoning mechanism to output high-confidence fault state assessment results; By integrating adaptive signal decomposition and interpretable rule base optimization, the diagnostic challenges caused by data distortion under noise interference are effectively solved. Under the premise of ensuring transparency and rationality of the reasoning process, high-precision, robust and reliable diagnosis of the fault status of UAV formations is achieved, providing core technical support for the safe operation and intelligent maintenance of UAV formations in complex environments.

[0073] This invention also provides an electronic device, such as... Figure 6 As shown, it includes a processor 601, a communication interface 602, a memory 603, and a communication bus 604, wherein the processor 601, the communication interface 602, and the memory 603 communicate with each other through the communication bus 604. Memory 603 is used to store computer programs; When the processor 601 executes the computer program stored in the memory 603, it implements the method steps of any of the above-mentioned UAV formation adaptive anti-interference fault diagnosis methods for noisy environments.

[0074] The communication bus mentioned in the above electronic devices can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of representation, only one thick line is used in the diagram, but this does not indicate that there is only one bus or one type of bus.

[0075] The communication interface is used for communication between the aforementioned electronic devices and other devices.

[0076] The memory may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0077] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0078] The present invention also provides a computer-readable storage medium. A computer program is stored in the computer-readable storage medium, and when executed by a processor, the computer program implements the method steps of any of the above-described adaptive anti-interference fault diagnosis methods for UAV formations in noisy environments.

[0079] Optionally, the computer-readable storage medium may be non-volatile memory (NVM), such as at least one disk storage device.

[0080] Optionally, the aforementioned computer-readable storage medium may also be at least one storage device located remotely from the aforementioned processor.

[0081] In another embodiment of the present invention, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute the steps of the method described in any of the above-described adaptive anti-interference fault diagnosis methods for UAV formations in noisy environments.

[0082] It should be noted that the terms "first," "second," etc., are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the invention.

[0083] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Furthermore, those skilled in the art can combine and integrate the different embodiments or examples described in this specification.

[0084] Although the invention has been described herein in conjunction with various embodiments, those skilled in the art will understand and implement other variations of the disclosed embodiments by reviewing the accompanying drawings and the disclosure in carrying out the claimed invention. In the description of the invention, the word "comprising" does not exclude other components or steps, "a" or "an" does not exclude a plurality, and "a plurality" means two or more, unless otherwise explicitly specified. Furthermore, while different embodiments may describe certain measures, this does not mean that these measures cannot be combined to produce good results.

[0085] The method provided in this invention can be applied to electronic devices. Specifically, the electronic device can be a desktop computer, a portable computer, a smart mobile terminal, a server, etc. No limitation is made herein; any electronic device that can implement this invention falls within the protection scope of this invention.

[0086] For the embodiments of the device / electronic device / storage medium, since they are basically similar to the method embodiments, the description is relatively simple, and relevant parts can be referred to in the description of the method embodiments.

[0087] It should be noted that the device, electronic device and storage medium in the embodiments of the present invention are respectively the device, electronic device and storage medium for the above-mentioned adaptive anti-interference fault diagnosis method for UAV formation in noisy environments. Therefore, all embodiments of the above-mentioned adaptive anti-interference fault diagnosis method for UAV formation in noisy environments are applicable to the device, electronic device and storage medium, and can achieve the same or similar beneficial effects.

[0088] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A method for adaptive anti-interference fault diagnosis of UAV formations in noisy environments, characterized in that, The adaptive anti-interference fault diagnosis method for UAV formations includes: Acquire noisy observation signals; Based on the adaptive particle swarm optimization algorithm, the optimization objective is to minimize the envelope entropy of the intrinsic mode function components obtained after variational mode decomposition of the observed signal. The VMD parameter combination of variational mode decomposition is dynamically searched and determined. The VMD parameter combination includes a penalty factor and the number of mode components. The observed signal is decomposed again using the VMD parameter combination, and the effective intrinsic mode function components are selected for reconstruction to obtain a denoised signal; Based on the denoised signal and the formation system model, key features of formation consistency are extracted; these key features of formation consistency include average position error consistency features and average velocity error consistency features. By using a confidence rule base to perform fusion reasoning on the key features of formation consistency, the overall fault status of the UAV formation system can be assessed.

2. The UAV formation adaptive anti-interference fault diagnosis method according to claim 1, characterized in that, Acquiring noisy observation signals, including: Establish a state-space mathematical model of the drone formation to obtain the formation system; The observation signals transmitted back to the ground station after the formation system is contaminated by additive noise are obtained.

3. The UAV formation adaptive anti-interference fault diagnosis method according to claim 1, characterized in that, The observed signal is re-decomposed using the VMD parameter combination, and the effective intrinsic mode function components are selected for reconstruction to obtain a denoised signal, including: The observed signal is re-decomposed using the VMD parameter combination to obtain multiple intrinsic mode function components; Calculate the correlation coefficient and envelope entropy between each intrinsic mode function component and the observed signal; The intrinsic mode function components with a correlation coefficient greater than the first preset threshold and an envelope entropy less than the second preset threshold are taken as valid intrinsic mode function components. The denoised signal is reconstructed by superimposing multiple effective intrinsic mode function components.

4. The UAV formation adaptive anti-interference fault diagnosis method according to claim 1, characterized in that, Based on the denoised signal and the formation system model, key features of formation consistency are extracted, including: Based on the noise reduction signal and the formation system model, the average position consistency error and average velocity consistency error between each UAV and the desired state of the formation are calculated to obtain the key features of formation consistency.

5. The UAV formation adaptive anti-interference fault diagnosis method according to claim 1, characterized in that, The calculation method for the envelope entropy of the intrinsic mode function components includes: ； in, Indicates the first Envelope entropy of each eigenmode function component; Indicates the first The probability mass function corresponding to each intrinsic mode function component; Indicates a discrete time point index; This indicates the number of discrete time points.

6. The UAV formation adaptive anti-interference fault diagnosis method according to claim 1, characterized in that, By using a confidence rule base to perform fusion reasoning on the key features of formation consistency, an assessment of the overall fault state of the UAV formation system is achieved, including: The key features of formation consistency are used as prerequisite attributes to construct confidence rules; By calculating the activation weights of the confidence rules and using an evidence reasoning algorithm for fusion reasoning, the confidence distribution evaluation results of the overall fault state of the UAV formation system are output.

7. A UAV formation adaptive anti-interference fault diagnosis device for noisy environments, characterized in that, The UAV formation adaptive anti-interference fault diagnosis device includes: The acquisition module is used to acquire noisy observation signals; The optimization module is used to dynamically search and determine the VMD parameter combination of the variational mode decomposition based on the adaptive particle swarm optimization algorithm, with the optimization objective of minimizing the envelope entropy of the intrinsic mode function components obtained after variational mode decomposition of the observed signal; the VMD parameter combination includes a penalty factor and the number of mode components; The reconstruction module is used to decompose the observed signal again using the VMD parameter combination, select the effective intrinsic mode function components for reconstruction, and obtain the noise-reduced signal. The extraction module is used to extract key features of formation consistency based on the denoised signal and the formation system model; the key features of formation consistency include average position error consistency features and average velocity error consistency features; The evaluation module is used to perform fusion reasoning on the key features of formation consistency using a confidence rule base to evaluate the overall fault status of the UAV formation system.

8. The UAV formation adaptive anti-interference fault diagnosis device according to claim 7, characterized in that, The acquisition module is specifically used to establish a state-space mathematical model of the UAV formation to obtain the formation system; and to acquire the observation signal transmitted back to the ground station after the formation system is contaminated by additive noise.

9. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; When a processor executes a computer program stored in memory, it implements the steps of the adaptive anti-interference fault diagnosis method for UAV formations in noisy environments as described in any one of claims 1-6.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the adaptive anti-interference fault diagnosis method for UAV formations in noisy environments as described in any one of claims 1-6.

Images