An automated performance detection method and device suitable for satellite communications

Abstract

The present invention discloses an automated performance detection method and device suitable for satellite communications, the method comprising performing three-dimensional component decomposition on a received signal, obtaining periodic feature data, mutation feature data and random feature data, outputting an enhanced reconstructed signal through feature enhancement and fusion reconstruction; then performing non-uniform sampling reconstruction and multi-dimensional feature extraction on the enhanced reconstructed signal, and calculating a performance parameter set; then performing multi-modal abnormal feature extraction and performance degradation analysis based on a dynamic benchmark model, and outputting system diagnosis results; and finally realizing parameter adjustment and control instruction generation through multi-objective optimization. The present invention can accurately identify multiple types of abnormal features, accurately evaluate the system performance status, and has the characteristics of high detection accuracy and strong adaptability.

Description

Technical Field

[0001] The present invention belongs to the field of satellite communication, and in particular to an automatic performance detection method and device suitable for satellite communication. Background Art

[0002] As an important part of modern communication networks, satellite communication systems play an irreplaceable role in global communications, emergency communications, military communications and other fields. With the continuous expansion of the scale of satellite communication systems and the increasing complexity of application scenarios, the importance of performance monitoring and fault diagnosis has become increasingly prominent. Accurate and timely detection and evaluation of the performance status of satellite communication systems is not only related to the guarantee of communication quality, but also directly affects the reliability and security of the system. Therefore, the development of efficient and reliable automated performance detection methods has become an important research topic in the field of satellite communications.

[0003] The current satellite communication performance detection methods mainly include anomaly detection based on threshold judgment, performance evaluation based on statistical analysis, and fault diagnosis based on pattern recognition. Among them, the threshold detection method determines whether the system performance is abnormal by setting a fixed or adaptive threshold; the statistical analysis method uses the statistical characteristics of historical data to evaluate the current performance status; the pattern recognition method identifies possible system faults by establishing a fault feature library. These methods have achieved certain results in practical applications and provided technical support for the operation and maintenance of satellite communication systems.

[0004] However, the existing performance detection methods still have the following technical problems: First, in terms of signal feature extraction, the existing methods are difficult to effectively handle the aliasing phenomenon of periodic features, mutation features and random features, resulting in low feature recognition accuracy; second, in the calculation of performance parameters, the coupling effect between parameters is not fully considered, affecting the accuracy of the calculation results; third, in the process of anomaly detection, the recognition ability of transient anomalies, gradual anomalies, periodic anomalies and random anomalies is insufficient, which easily leads to missed detection and false detection; fourth, in the construction of performance evaluation benchmarks, the influence of historical data is not fully utilized, resulting in the evaluation benchmark being unstable; fifth, in terms of parameter optimization, the parameter adjustment of each processing link lacks coordination, affecting the overall performance improvement of the system. The existence of these technical problems seriously restricts the automation level and detection effect of satellite communication system performance detection. Summary of the invention

[0005] The purpose of the invention is to provide an automated performance detection method and device suitable for satellite communications to solve the above-mentioned problems existing in the prior art.

[0006] Technical solution, an automated performance detection method suitable for satellite communications, comprising the following steps:

[0007] S1, obtaining a received signal and performing three-dimensional component decomposition on the received signal to obtain periodic feature data, mutation feature data and random feature data; performing enhancement processing on the periodic feature data, mutation feature data and random feature data respectively to obtain enhanced periodic feature data, enhanced mutation feature data and enhanced random feature data; fusing and reconstructing the enhanced periodic feature data, enhanced mutation feature data and enhanced random feature data, and outputting an enhanced reconstructed signal and a signal quality parameter;

[0008] S2. Based on the signal quality parameters, perform non-uniform sampling reconstruction on the enhanced reconstructed signal to obtain reconstructed sequence data; perform multi-dimensional feature extraction on the reconstructed sequence data to obtain feature matrix data; perform parameter separation calculation on the feature matrix data to obtain a performance parameter set; perform reliability evaluation and optimization based on the performance parameter set, and output an optimized performance parameter set and parameter reliability data;

[0009] S3. Based on the optimized performance parameter set and parameter reliability data, a dynamic benchmark model is constructed to obtain performance benchmark data; multimodal abnormal feature extraction is performed on the performance benchmark data to obtain abnormal feature data; based on the abnormal feature data, performance degradation analysis is performed to obtain degradation trend data; based on the degradation trend data, status diagnosis is performed to output system diagnosis results, prediction trend data and abnormal feature set.

[0010] An automatic performance detection device suitable for satellite communications, comprising:

[0011] at least one processor; and,

[0012] a memory communicatively connected to at least one of the processors; wherein,

[0013] The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement the automatic performance detection method applicable to satellite communication.

[0014] Beneficial effects: The present invention improves the accuracy of feature extraction, the reliability of parameter calculation, the sensitivity of anomaly detection and the accuracy of trend prediction, realizes fully automatic and high-precision evaluation of satellite communication system performance, can timely discover and warn of system performance anomalies, and provides reliable technical support for the maintenance and optimization of satellite communication systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 Flow chart of the method of the present invention.

[0016] Figure 2 This is a flow chart of step S1 of the present invention.

[0017] Figure 3 This is a flow chart of step S2 of the present invention.

[0018] Figure 4 This is a flow chart of step S3 of the present invention.

[0019] Figure 5 This is a physical layer frame structure diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0020] like Figure 1 As shown, the present application proposes an automated performance detection method suitable for satellite communications, comprising the following steps:

[0021] S1, obtaining a received signal and performing three-dimensional component decomposition on the received signal to obtain periodic feature data, mutation feature data and random feature data; performing enhancement processing on the periodic feature data, mutation feature data and random feature data respectively to obtain enhanced periodic feature data, enhanced mutation feature data and enhanced random feature data; fusing and reconstructing the enhanced periodic feature data, enhanced mutation feature data and enhanced random feature data, and outputting an enhanced reconstructed signal and a signal quality parameter;

[0022] S2. Based on the signal quality parameters, perform non-uniform sampling reconstruction on the enhanced reconstructed signal to obtain reconstructed sequence data; perform multi-dimensional feature extraction on the reconstructed sequence data to obtain feature matrix data; perform parameter separation calculation on the feature matrix data to obtain a performance parameter set; perform reliability evaluation and optimization based on the performance parameter set, and output an optimized performance parameter set and parameter reliability data;

[0023] S3. Based on the optimized performance parameter set and parameter reliability data, a dynamic benchmark model is constructed to obtain performance benchmark data; multimodal abnormal feature extraction is performed on the performance benchmark data to obtain abnormal feature data; based on the abnormal feature data, performance degradation analysis is performed to obtain degradation trend data; based on the degradation trend data, status diagnosis is performed to output system diagnosis results, prediction trend data and abnormal feature set.

[0024] like Figure 2 As shown, according to one aspect of the present application, step S1 is further:

[0025] S11, obtaining a received signal, decomposing the received signal by nonlinear projection transformation to obtain periodic feature data; decomposing the received signal by local feature adaptive transformation to obtain mutation feature data; processing the received signal by random process decomposition to obtain random feature data;

[0026] S12, based on the periodic feature data, construct a cyclic correlation matrix, extract the main feature vector and the eigenvalue, and obtain enhanced periodic feature data through feature enhancement processing; based on the mutation feature data, extract the signal singular point sequence, construct the local morphological feature, and obtain enhanced mutation feature data through morphological enhancement processing; based on the random feature data, construct a statistical feature model, and obtain enhanced random feature data through adaptive threshold processing;

[0027] S13. Based on the significance levels of the enhanced periodic feature data, the enhanced mutation feature data and the enhanced random feature data, the corresponding weight coefficients are calculated to obtain feature weight data; the feature weight data are weightedly combined with the enhanced periodic feature data, the enhanced mutation feature data and the enhanced random feature data to obtain an enhanced reconstructed signal; the enhanced reconstructed signal is evaluated for signal integrity, feature enhancement degree and noise suppression effect to obtain signal quality parameters.

[0028] According to one aspect of the present application, it also includes S14, obtaining enhanced reconstructed signals and signal quality parameters, constructing multi-objective evaluation indicators, calculating the optimization direction of each processing parameter, and obtaining parameter optimization data; optimizing and adjusting the processing parameters in steps S11 to S13 according to the parameter optimization data to complete the signal preprocessing process.

[0029] In one embodiment of the present application, the decomposition process is specifically as follows: F(t) = Ф(t)·P(t) +ξ(t)·M(t) +Ω(t)·R(t); wherein F(t) is the time domain function of the received signal; P(t) = ∑[λk·cos(ωkt + φk)] is the periodic characteristic component, λk is the characteristic value, ωk is the angular frequency, and φk is the phase; M(t) = ∑[αj·H(t-tj)] is the mutation characteristic component, αj is the amplitude, H(t) is the unit step function, and tj is the mutation time; R(t) = ∑[βi·N(μi,σi 2 )] is the random feature component, βi is the weight coefficient, N(μi,σi 2 ) is a Gaussian random process, μi is the mean, σi 2 is the variance; Ф(t), ξ(t), and Ω(t) are adaptive mapping functions, satisfying Ф(t) + ξ(t) + Ω(t) = 1.

[0030] Feature enhancement optimization algorithm E(t) = Ep(t) + Em(t) + Er(t); where E(t) is the feature enhancement function; Ep(t) =∑[Cp(λk)·P(t)] is the periodic feature enhancement component, P(t) is the periodic feature data, Cp(λk) is the circulant correlation matrix feature enhancement factor, and λk is the eigenvalue; Em(t) =∑[Cm(αj)·M(t)] is the mutation feature enhancement component, M(t) is the mutation feature data, Cm(αj) is the morphological edge enhancement factor, and αj is the mutation amplitude; Er(t) =∑[Cr(βi)·R(t)] is the random feature enhancement component, R(t) is the random feature data, Cr(βi) is the adaptive threshold enhancement factor, and βi is the weight coefficient; Cp(λk), Cm(αj), Cr(βi) satisfy 0 < Cp(λk), Cm(αj), Cr(βi) ≤ K_max, K_max is the maximum enhancement coefficient.

[0031] The specific process of calculating the weight coefficient is: W(t) = ∑[γi·Wi(t)]; where W(t) is the feature weight function; Wi(t) is the weight component of each feature, i∈{p, m, r} corresponds to periodic, mutation, and random features respectively; γi is the weight normalization coefficient, satisfying ∑γi = 1; Wi(t) = Si(t) / ∑Si(t), Si(t) is the feature significance function, Si(t) =ηi·Vi(t)·Ci(t), ηi is the feature importance coefficient, Vi(t) is the feature variance function, and Ci(t) is the feature correlation function.

[0032] This embodiment realizes the accurate separation of periodic features, mutation features and random features in satellite communication receiving signals through three complementary signal decomposition methods: nonlinear projection transformation, local feature adaptive transformation and random process decomposition. The three types of feature data are enhanced by cyclic correlation matrix feature enhancement, morphological edge optimization and adaptive threshold noise suppression, respectively, to improve the recognizability of various features. Through feature significance evaluation and multi-scale fusion reconstruction, the optimal combination of signal features is achieved, which not only maintains the integrity of the original signal, but also improves the signal-to-noise ratio and feature clarity of the signal. This embodiment enables the system to accurately identify and enhance the key features in satellite communication signals, provides a high-quality signal foundation for subsequent performance evaluation, and effectively solves the problems of difficulty in extracting satellite communication signal features and low signal-to-noise ratio.

[0033] According to one aspect of the present application, step S11 is further:

[0034] S111, obtaining a received signal, performing nonlinear projection transformation processing on the received signal to obtain projection transformation data; constructing a feature space using a projection basis function to obtain feature space data; mapping and decomposing the projection transformation data in the feature space data to obtain space mapping data; extracting periodic components based on the space mapping data, and outputting periodic feature data;

[0035] S112, constructing a local feature adaptive transformation matrix based on the received signal to obtain adaptive matrix data; extracting local features based on the adaptive matrix data to obtain local feature data; locating mutation points on the local feature data to obtain mutation sequence data; reconstructing features based on the mutation sequence data to output mutation feature data;

[0036] S113. Based on the received signal, a random process statistical model is constructed to obtain statistical model data; based on the statistical model data, the signal component is decomposed to obtain component sequence data; random characteristic analysis is performed on the component sequence data to obtain random characteristic data; based on the random characteristic data, feature extraction is performed to output random feature data.

[0037] In one embodiment of the present application, the process of the nonlinear projection transformation method is specifically as follows: P(x)=∫K(x, s)·f(s)ds; wherein P(x) is the projection transformation function; f(s) is the input signal; K(x, s) is the nonlinear kernel function, K(x, s) = exp(-||xs|| 2 / σ 2 ), σ is the kernel width parameter, x is the feature space coordinate, s is the signal space coordinate, and the integration interval is the signal definition domain.

[0038] The projection basis function construction model is: B(x) =∑[αk·φk(x)]; where B(x) is the projection basis function; φk(x) is an orthogonal basis function family that satisfies ∫φi(x)·φj(x)dx =Δij; αk is the basis function coefficient; k is the basis function number; Δij is the Kronecker function; the basis function selection criterion J(k) = ||f(x) - B(x)|| 2 Minimum.

[0039] The spatial mapping decomposition algorithm is specifically as follows: M(x) = T[P(x)]·B(x); where M(x) is the mapping decomposition function; T[·] is the feature transformation operator; P(x) is the projection transformation result; B(x) is the projection basis function; and the mapping constraint condition is: ∫M(x)dx =1, M(x)≥0.

[0040] The specific method for constructing the local feature adaptive transformation matrix is: L(t) = W(t)·H(t)·W_inv(t); where L(t) is the adaptive transformation matrix; W(t) is the weight matrix, W(t) = diag[w1(t), w2(t), ..., wn(t)]; H(t) is the Hessian matrix; wi(t) is the adaptive weight function, wi(t) = exp(-λi·||▽f(t)||); λi is the adjustment coefficient.

[0041] The specific algorithm for locating mutation points is: D(t) = Θ[Ψ 2 f(t) / Ψt 2 - η(t)]; where D(t) is the mutation point detection function; Θ[·] is the threshold decision operator; Ψ is the partial derivative, f(t) is the signal function; η(t) is the adaptive threshold, η(t) =μ(t) + k·σ(t); μ(t) is the local mean; σ(t) is the local standard deviation; k is the adjustment coefficient.

[0042] The local feature extraction model is specifically: F(t) = G(t)○[L(t)·S(t)]; where F(t) is the local feature function; G(t) is the Gaussian smoothing kernel; L(t) is the adaptive transformation matrix; S(t) is the signal function; ○ is the convolution operator; feature constraint: ||F(t)|| 2 ≤K, K is the upper limit constant.

[0043] The specific statistical model of random process is: R(t) =∑[pi·Xi(t)]; where R(t) is the random process function; Xi(t) is the orthogonal random process basis; pi is the probability weight, ∑pi = 1; the statistical characteristics E[R(t)] =μ(t), Var[R(t)] =σ 2 (t); μ(t) is the mean function; σ 2 (t) is the variance function.

[0044] The component decomposition algorithm is specifically as follows: C(t) = H[R(t)]; where C(t) is the component sequence; H[·] is the Hilbert-Huang transform operator; R(t) is a random process; the component orthogonality condition is: ∫Ci(t)·Cj(t)dt = 0, i≠j; and the reconstruction condition is: R(t) = ∑Ci(t).

[0045] The specific random characteristic analysis method is: A(t) = {M1(t), M2(t), ..., Mn(t)}; A(t) is the characteristic analysis set; Mi(t) is the i-th order moment function; M1(t) = E[R(t)] is the mean; M2(t) = E[(R(t)-M1(t)) 2] is the variance; higher-order moment Mi(t) = E[(R(t)-M1(t)) i ]; Characteristic determination: γ(t) = M3(t) / M2(t) 3 / 2 is the skewness, κ(t) = M4(t) / M2(t) 2 is the kurtosis.

[0046] This embodiment maps the received signal to the feature space through nonlinear projection transformation, combines the projection basis function construction and space mapping decomposition, and realizes the accurate extraction of periodic features. The local feature adaptive transformation matrix is ​​used to extract local features and locate mutation points to accurately capture the mutation features in the signal. The random feature components in the signal are effectively separated by using the random process statistical model and component decomposition technology. This embodiment fully considers the diversity characteristics of satellite communication signals, and realizes the accurate separation of signal features through different mathematical tools and signal processing methods. Compared with the traditional single decomposition method, this embodiment improves the accuracy and reliability of feature separation, provides high-quality feature data for subsequent feature enhancement processing, and effectively solves the technical problems of multiple feature aliasing and mutual interference in satellite communication signals.

[0047] According to one aspect of the present application, step S12 is further:

[0048] S121, based on the periodic feature data, construct a cyclic correlation matrix to obtain correlation matrix data; perform feature decomposition on the correlation matrix data to obtain feature decomposition data; based on the feature decomposition data, calculate the main feature vector and the eigenvalue to obtain the main feature data; based on the main feature data, construct a frequency domain enhancement function to obtain frequency domain function data; based on the frequency domain function data, perform feature enhancement and output enhanced periodic feature data;

[0049] S122, based on the mutation feature data, extract the signal singular point sequence to obtain singular sequence data; perform morphological gradient calculation on the singular sequence data to obtain gradient feature data; based on the gradient feature data, construct a local morphological operator to obtain morphological operator data; based on the morphological operator data, perform edge optimization to obtain edge optimization data; perform morphological enhancement processing on the edge optimization data to output enhanced mutation feature data;

[0050] S123. Based on the random feature data, a multi-order moment statistical model is constructed to obtain statistical model data; probability distribution estimation is performed on the statistical model data to obtain distribution feature data; based on the distribution feature data, an adaptive threshold function is constructed to obtain threshold function data; based on the threshold function data, noise suppression is performed to obtain suppressed noise data; random feature enhancement is performed on the suppressed noise data to output enhanced random feature data.

[0051] In one embodiment of the present application, the cyclic correlation matrix construction algorithm is specifically: R(τ) = E[S(t)·S(t+τ)]; wherein R(τ) is the cyclic correlation matrix; S(t) is the periodic characteristic signal; τ is the time delay; E[·] is the desired operator; matrix element Rij = R(|ti-tj|); eigendecomposition R = U·Λ·U_inv, U is the eigenvector matrix, Λ is the eigenvalue diagonal matrix, and U_inv represents the inverse matrix of matrix U.

[0052] Construct the frequency domain enhancement function: E(ω) = H(ω)·F(ω); where E(ω) is the frequency domain enhancement function; F(ω) is the signal spectrum; H(ω) is the enhancement filter, H(ω) = (1+α·|F(ω)| 2 ) / (1+β·|F(ω)| 2 ), α and β are enhancement coefficients; enhancement constraint: |H(ω)| ≤ K_max, K_max is the maximum gain.

[0053] The main feature extraction model is: P(t) = V(t)·∑[λi·ui(t)]; where P(t) is the main feature function; V(t) is the eigenvalue function; λi is the eigenvalue; ui(t) is the eigenvector; feature importance ρi = λi / ∑λi; feature selection criterion: ∑ρi≥η, η is the importance threshold.

[0054] The specific method for calculating the morphological gradient is: G(x) =Δ(f)-ε(f); where G(x) is the morphological gradient; f is the input signal; Δ(f) is the dilation operator, Δ(f) = max{f(xs)+b(s)}; ε(f) is the erosion operator, ε(f) = min{f(x+s)-b(s)}; b(s) is the structural element; s is the support set of the structural element.

[0055] The construction method of the local morphological operator is specifically as follows: M(x) =γ[φ(x)]; where M(x) is the morphological operator; γ is the morphological transformation operator; φ(x) is the local feature function; γ[φ(x)]=Δ[ε(φ(x))] is an opening operation, or γ[φ(x)]=ε[Δ(φ(x))] is a closing operation; the operator selection depends on the local feature type.

[0056] The edge optimization model is: E(x) = w(x)·▽ 2 f(x); E(x) is the edge optimization function; w(x) is the adaptive weight function, w(x) = 1 / (1+k·||▽f(x)|| 2 );▽ 2 f(x) is the Laplace operator; k is the adjustment coefficient; optimization constraint: ||E(x)||≤M, M is the upper limit value.

[0057] The multi-order moment statistical model is Mk(t) = E[(X(t)-μ) k ]; where Mk(t) is the k-order central moment; X(t) is a random signal; μ is the mean; k is the moment order; characteristic quantity: skewness γ1 = M3 / M2 3 / 2 , kurtosis γ2 = M4 / M2 2 -3; the judgment criteria are: |γ1|≤ε1, |γ2|≤ε2.

[0058] The adaptive threshold function is constructed as follows: T(x) = μ(x) + k·σ(x); where T(x) is the adaptive threshold function; μ(x) is the local mean function, μ(x)=∫w(xs)·f(s)ds; σ(x) is the local standard deviation function, σ(x)=[∫w(xs)·(f(s)-μ(x)) 2 ds] 1 / 2 ; w(x) is the weighted window function; k is the threshold coefficient.

[0059] The specific noise suppression algorithm is: S(x) = f(x)·H[|f(x)|-T(x)]; where S(x) is the suppressed signal; f(x) is the original signal; H[·] is the Heaviside step function; T(x) is the adaptive threshold function; suppression condition: signal-to-noise ratio SNR(x) = 10·log[S 2 (x) / N 2 (x)]≥SNR_min.

[0060] This embodiment realizes the precise enhancement of periodic features by constructing a circulant correlation matrix and feature decomposition, combined with the extraction of main eigenvectors and eigenvalues. The edge optimization and morphological enhancement of mutation features are performed by morphological gradient calculation and local morphological operator construction, thereby improving the clarity of mutation features. The multi-order moment statistical model and adaptive threshold function are used to achieve noise suppression and feature enhancement of random features. This embodiment fully considers the mathematical characteristics and physical meanings of different types of features, and improves the recognition of various types of features through specific enhancement algorithms. Compared with the traditional unified enhancement method, the accuracy and effectiveness of feature enhancement are improved, and high-quality enhanced feature data is provided for subsequent feature fusion, which effectively solves the technical problem of different difficulties in enhancing different types of features in satellite communication signals.

[0061] According to one aspect of the present application, step S13 is further:

[0062] S131. Based on the enhanced periodic feature data, the enhanced mutation feature data and the enhanced random feature data, a feature significance evaluation model is constructed to obtain evaluation model data; the evaluation model data is hierarchically decomposed to obtain hierarchical feature data; the significance index of each layer is calculated based on the hierarchical feature data to obtain significance index data; the significance index data is normalized and outputted as feature weight data.

[0063] S132. Obtain three types of enhanced feature data and feature weight data, establish a multi-scale fusion model, and obtain fusion model data; perform consistency check on the fusion model data to obtain consistency data; construct a feature mapping function based on the consistency data to obtain feature mapping data; use the feature mapping data to perform multi-feature fusion reconstruction, and output an enhanced reconstructed signal.

[0064] S133, obtain the enhanced reconstructed signal, perform signal integrity calculation, and obtain integrity data; perform signal fidelity analysis to obtain fidelity data; perform noise suppression effect evaluation to obtain suppression effect data; perform comprehensive calculation on the three types of evaluation data and output signal quality parameters.

[0065] In one embodiment of the present application, the feature significance evaluation model is specifically: V(f) = W(f)·D(f); wherein V(f) is the significance function; W(f) is the feature weight function, W(f) = exp(-λ·||▽f|| 2 ); D(f) is the feature discriminant function, D(f) = |f-μb| / σb; μb is the background mean; σb is the background standard deviation; λ is the scale parameter.

[0066] The hierarchical decomposition algorithm is as follows: L(x)=∑[αk·Bk(x)]; where L(x) is the hierarchical decomposition function; Bk(x) is the k-layer basis function; αk is the hierarchical weight coefficient, ∑αk = 1; inter-layer constraint: ∫Bi(x)·Bj(x)dx = 0, i≠j; reconstruction error ε= ||f(x)-L(x)|| 2 , f(x) is the characteristic function or signal of the input.

[0067] The multi-scale fusion model is: F(x, s) = ∑[wk(s)·fk(x)]; where F(x, s) is the fusion function; s is the scale parameter; fk(x) is the k-scale feature; wk(s) is the scale weight function, wk(s) = exp(-|s-sk| 2 / σ 2 ), sk is the characteristic scale, and σ is the scale bandwidth.

[0068] The specific method for constructing the feature mapping function is: M(x) = Φ[G(x)]; where M(x) is the mapping function; Φ[·] is the nonlinear mapping operator; G(x) is the feature gradient function, G(x) = ∑[pk·Ψf(x) / Ψxk]; pk is the direction weight; mapping constraint: ||M(x)|| ≤ 1.

[0069] The signal integrity calculation method is as follows: C(t) =ξ*[S(t)]; where C(t) is the integrity function; S(t) is the signal function; ξ* [·] is the integrity evaluation operator, ξ*[S(t)] = ∫|S(t)| 2 dt / ∫|S0(t)| 2 dt; S0(t) is the reference signal; evaluation threshold: C(t)≥C_min, where C_min is the minimum threshold for signal integrity evaluation.

[0070] The fidelity analysis model is F(ω) = γ·H(ω)·S(ω); where F(ω) is the fidelity function; H(ω) is the system transfer function; S(ω) is the signal spectrum; γ is the gain coefficient; and the distortion D=∫|F(ω)-S(ω)| 2 dω / ∫|S(ω)| 2 dω; Fidelity requirement: D≤D_max, D_max is the maximum threshold of distortion, and dω is the small increment of the integral variable ω.

[0071] This embodiment realizes the adaptive calculation of weights of three types of enhanced features through feature significance evaluation model and hierarchical decomposition technology. The accuracy and reliability of the feature fusion process are ensured by using multi-scale fusion model and consistency verification method. Combined with signal integrity calculation and fidelity analysis, a comprehensive evaluation of the quality of the reconstructed signal is achieved. This embodiment fully considers the importance of different features under different working conditions, and realizes the optimal combination of features through dynamic weight allocation. Compared with the traditional fixed weight fusion method, it improves the quality and reliability of signal reconstruction, provides high-quality reconstructed signals for subsequent performance evaluation, and effectively solves the technical problem of unbalanced fusion of satellite communication signal features.

[0072] like Figure 3 As shown, according to one aspect of the present application, step S2 is further:

[0073] S21. Based on the signal quality parameter, generate a non-uniform sampling sequence to obtain a sampling time sequence; based on the sampling time sequence, resample the enhanced reconstructed signal to obtain reconstructed sequence data;

[0074] S22, based on the reconstructed sequence data, construct a frequency feature mapping matrix to obtain frequency feature data; based on the reconstructed sequence data, construct a phase feature mapping matrix to obtain phase feature data; based on the reconstructed sequence data, construct an amplitude feature mapping matrix to obtain amplitude feature data; based on the reconstructed sequence data, construct a time feature mapping matrix to obtain time feature data; combine the frequency feature data, phase feature data, amplitude feature data and time feature data into feature matrix data;

[0075] S23. Based on the characteristic matrix data, a parameter coupling model is constructed, and orthogonal transformation decomposition is performed to obtain decoupling characteristic data; power parameter calculation is performed on the decoupling characteristic data to obtain power parameters; frequency deviation parameter calculation is performed on the decoupling characteristic data to obtain frequency deviation parameters; phase noise calculation is performed on the decoupling characteristic data to obtain phase noise parameters; group delay calculation is performed on the decoupling characteristic data to obtain group delay parameters; modulation characteristic calculation is performed on the decoupling characteristic data to obtain modulation characteristic parameters; power parameters, frequency deviation parameters, phase noise parameters, group delay parameters and modulation characteristic parameters are combined into a performance parameter set;

[0076] S24. Based on the performance parameter set, reliability calculation is performed on each parameter to obtain a parameter reliability index; based on the parameter reliability index, a parameter correction amount is calculated to obtain parameter correction data; based on the parameter correction data, the performance parameter set is optimized to obtain an optimized performance parameter set; the comprehensive reliability of the optimized performance parameter set is calculated to obtain parameter reliability data.

[0077] In one embodiment of the present application, the non-uniform sampling reconstruction algorithm is specifically: Rs(t) =∑[L(tk)·S(t-tk)]; wherein Rs(t) is the reconstruction sequence function; S(t) is the enhanced reconstructed signal; tk is the adaptive sampling time, tk = tk_1+Δk, Δk is the variable sampling interval, and tk_1 represents the previous sampling time; L(tk) is the local interpolation operator, L(tk) = Ik(t)·Bk(t), Ik(t) is the complexity evaluation function, and Bk(t) is the basis function; the sampling interval Δk = f(Ik(t)), and f(·) is the adaptive mapping function.

[0078] The specific multidimensional feature extraction algorithm is: F_mat(t) = [Ff(t), Fp(t), Fa(t), Ft(t)]; where F_mat(t) is the feature matrix; Ff(t) =∑[ω*f, k·Wk(t)] is the frequency feature vector, Wk(t) is the wavelet coefficient; Fp(t) =∑[ω*p, k·exp(jθk(t))] is the phase feature vector, θk(t) is the instantaneous phase; Fa(t) = ∑[ω*a, k·Ak(t)] is the amplitude feature vector, Ak(t) is the envelope function; Ft(t) = ∑[ω*t, k·Tk(t)] is the time feature vector, Tk(t) is the trend function; ω* is the feature weight coefficient.

[0079] The parameter coupling decoupling algorithm is specifically as follows: P(t) = T·D(t)·T_inv; where P(t) is the performance parameter matrix; D(t) is the decoupling feature matrix; T is the orthogonal transformation matrix, satisfying T·T_inv = I, T_inv is the inverse matrix of matrix T, and I is the unit matrix; D(t) = [Dp(t), Df(t), Dφ(t), Dτ(t), Dm(t)], which respectively represent the decoupling results of power, frequency offset, phase noise, delay, and modulation characteristics; each decoupling feature satisfies Di(t)·Dj(t) = 0, i≠j.

[0080] This embodiment realizes the non-uniform sampling reconstruction of the enhanced reconstructed signal through the signal local complexity evaluation and adaptive sampling interval generation technology, overcoming the limitations of traditional equal-interval sampling in processing complex satellite communication signals. Multi-dimensional feature extraction methods such as multi-scale wavelet decomposition, instantaneous phase analysis, envelope detection and time series decomposition are adopted to obtain comprehensive frequency, phase, amplitude and time characteristics. Through the parameter coupling model and orthogonal transformation decomposition, the accurate separation and calculation of various performance parameters are realized, including key parameters such as power, frequency deviation, phase noise, group delay and modulation characteristics. This embodiment not only improves the accuracy of signal processing, but also realizes the accurate evaluation of multi-dimensional performance parameters of satellite communication systems, solving the problem of insufficient accuracy of traditional methods in complex signal feature extraction and performance parameter calculation.

[0081] According to one aspect of the present application, step S21 is further:

[0082] S211, based on the enhanced reconstructed signal and the signal quality parameter, construct a signal local complexity evaluation model to obtain complexity data; perform segmented threshold calculation on the complexity data to obtain threshold sequence data; based on the threshold sequence data, generate an adaptive sampling interval to obtain sampling interval data; based on the sampling interval data, construct a complete sampling sequence and output a sampling time series;

[0083] S212, based on the enhanced reconstructed signal and the sampling time series, construct an interpolation basis function library to obtain basis function data; based on the basis function data, perform local analysis on the original sampling points to obtain local feature data; based on the local feature data, select the optimal interpolation basis function to obtain interpolation function data; based on the interpolation function data, perform signal reconstruction and output reconstructed sequence data;

[0084] S213, based on the reconstructed sequence data, calculating the reconstruction error index to obtain error index data; based on the error index data, performing spectrum consistency analysis to obtain consistency data; based on the consistency data, evaluating the degree of information retention to obtain information retention data; based on the error index data, consistency data and information retention data, performing reconstruction quality verification, and outputting reconstruction quality data;

[0085] S214. Based on the reconstructed sequence data and the reconstructed quality data, an adaptive optimization model is constructed to obtain optimized model data; based on the optimized model data, reconstruction parameters are adjusted to obtain first parameter adjustment data; based on the first parameter adjustment data, the reconstructed sequence data is corrected, and updated reconstructed sequence data is output.

[0086] This embodiment realizes non-uniform sampling of enhanced reconstructed signals through a signal local complexity evaluation model and an adaptive sampling interval generation technology. The accuracy and continuity of the reconstructed sequence are ensured by using an interpolation basis function library and local feature analysis. Combined with reconstruction error index calculation and spectrum consistency analysis, real-time evaluation and optimization of reconstruction quality are achieved. This embodiment fully considers the characteristic differences of satellite communication signals in different time periods, and realizes accurate reconstruction of signals by dynamically adjusting the sampling interval. Compared with the traditional equidistant sampling method, the quality and efficiency of the reconstructed signal are improved, a high-quality reconstructed sequence is provided for subsequent feature extraction, and the accuracy and efficiency problems in the sampling and reconstruction of complex communication signals are effectively solved.

[0087] According to one aspect of the present application, step S22 is further:

[0088] S221. Acquire reconstructed sequence data, construct a multi-scale wavelet basis function group, and obtain wavelet basis data; perform frequency decomposition using the wavelet basis data to obtain frequency component data; perform energy clustering analysis on the frequency component data to obtain energy distribution data; establish a frequency mapping function based on the energy distribution data to obtain frequency mapping data; perform feature extraction on the frequency mapping data and output frequency feature data.

[0089] S222, obtain reconstructed sequence data, calculate the instantaneous phase of the signal, and obtain instantaneous phase data; perform phase expansion on the instantaneous phase data to obtain phase sequence data; use the phase sequence data to construct a phase continuity index to obtain continuity index data; perform phase feature extraction based on the continuity index data, and output phase feature data.

[0090] S223, obtaining reconstructed sequence data, performing envelope detection analysis, and obtaining envelope data; normalizing the envelope data to obtain normalized amplitude data; constructing an amplitude variation feature descriptor to obtain amplitude description data; using the amplitude description data to establish an amplitude feature map, and outputting the amplitude feature data.

[0091] S224, obtaining reconstructed sequence data, establishing a time series decomposition model, and obtaining sequence decomposition data; performing trend analysis on the sequence decomposition data to obtain trend feature data; calculating local time feature parameters to obtain time parameter data; establishing a time feature mapping based on the time parameter data and the trend feature data, and outputting the time feature data.

[0092] S225, acquiring frequency feature data, phase feature data, amplitude feature data and time feature data, constructing a feature correlation matrix, and obtaining correlation data; performing feature weight calculation based on the correlation data to obtain weight coefficient data; performing weighted combination of the four types of features using the weight coefficient data, and outputting feature matrix data.

[0093] This embodiment achieves accurate extraction of signal frequency characteristics through multi-scale wavelet basis function groups and energy clustering analysis. The phase characteristics of the signal are accurately obtained by using instantaneous phase calculation and phase unwrapping technology. Combined with envelope detection and amplitude change feature description, a comprehensive characterization of the signal amplitude characteristics is achieved. Through time series decomposition and trend analysis, the time domain characteristics of the signal are obtained. This embodiment fully considers the complexity and variability of satellite communication signals, and achieves a comprehensive characterization of signal characteristics through different feature extraction algorithms. Compared with the traditional single feature extraction method, it improves the integrity and accuracy of feature characterization, provides comprehensive feature data for subsequent parameter calculation, and effectively solves the technical problem of incomplete feature extraction of satellite communication signals.

[0094] According to one aspect of the present application, step S23 is further:

[0095] S231, acquiring characteristic matrix data, constructing a multi-dimensional parameter coupling model, and obtaining coupling model data; performing orthogonal transformation decomposition using the coupling model data to obtain orthogonal decomposition data; performing feature separation processing on the orthogonal decomposition data, and outputting decoupled characteristic data.

[0096] S232, acquiring decoupling characteristic data, extracting signal power characteristics, and obtaining power characteristic data; performing noise correction on the power characteristic data to obtain corrected power data; establishing a power estimation model based on the corrected power data, and outputting power parameters.

[0097] S233, acquiring decoupling characteristic data, establishing a frequency tracking model, and obtaining frequency tracking data; performing multi-point frequency deviation calculation on the frequency tracking data to obtain frequency deviation calculation data; performing deviation compensation using the frequency deviation calculation data, and outputting frequency deviation parameters.

[0098] S234, obtaining decoupling characteristic data, constructing a phase noise analysis model, and obtaining phase noise analysis data; performing in-band and out-band noise separation on the phase noise analysis data to obtain noise separation data; calculating phase jitter using the noise separation data, and outputting phase noise parameters.

[0099] S235, obtaining decoupling characteristic data, establishing a group delay measurement model, and obtaining delay measurement data; performing dispersion compensation on the delay measurement data to obtain compensated delay data; performing delay estimation using the compensated delay data, and outputting group delay parameters.

[0100] S236, obtaining decoupling characteristic data, constructing a modulation characteristic analysis model, and obtaining modulation analysis data; performing nonlinear distortion evaluation on the modulation analysis data to obtain distortion evaluation data; calculating modulation characteristics based on the distortion evaluation data, and outputting modulation characteristic parameters.

[0101] S237, obtaining power parameters, frequency deviation parameters, phase noise parameters, group delay parameters and modulation characteristic parameters, performing parameter correlation analysis, and obtaining correlation data; performing consistency optimization on the correlation data to obtain optimized parameter data; using the optimized parameter data to perform parameter combination, and outputting a performance parameter set.

[0102] This embodiment realizes the decoupling processing of the characteristic matrix through a multi-dimensional parameter coupling model and orthogonal transformation decomposition. The power parameters of the signal are accurately calculated by using power feature extraction and noise correction technology. Combined with frequency tracking and frequency offset compensation, accurate frequency offset parameters are obtained. Accurate evaluation of phase noise is achieved through phase noise analysis and separation of in-band and out-band noise. Based on group delay measurement and dispersion compensation, the group delay characteristics of the signal are calculated. This embodiment fully considers the mutual influence between various performance parameters of the satellite communication system, and realizes the accurate acquisition of parameters through decoupling calculation. Compared with the traditional direct calculation method, the accuracy and reliability of parameter calculation are improved, reliable parameter data is provided for performance evaluation, and the technical problem of mutual interference between performance parameters is effectively solved.

[0103] like Figure 4 As shown, according to one aspect of the present application, step S3 is further:

[0104] S31. Based on the optimized performance parameter set and parameter reliability data, calculate the current state mapping value to obtain current state data; read the historical state record, perform state impact calculation, and obtain historical impact data; combine the current state data and historical impact data to build a dynamic benchmark model to obtain performance benchmark data;

[0105] S32, based on the performance benchmark data, perform feature space transformation to obtain transformed feature data; perform transient anomaly extraction on the transformed feature data to obtain transient anomaly data; perform gradual anomaly extraction on the transformed feature data to obtain gradual anomaly data; perform periodic anomaly extraction on the transformed feature data to obtain periodic anomaly data; perform random anomaly extraction on the transformed feature data to obtain random anomaly data; combine the transient anomaly data, the gradual anomaly data, the periodic anomaly data and the random anomaly data into anomaly feature data;

[0106] S33, based on the abnormal feature data, calculating the weight coefficients of various abnormalities to obtain abnormal weight data; performing weighted combination on the abnormal feature data and the abnormal weight data to obtain comprehensive abnormal data; performing time series analysis on the comprehensive abnormal data to obtain degradation trend data; performing prediction calculation based on the degradation trend data to obtain prediction trend data;

[0107] S34. Based on the degradation trend data, prediction trend data and performance benchmark data, a decision function is constructed to obtain decision indicator data; based on the decision indicator data, state classification is performed to obtain state classification data; threshold optimization is performed on the state classification data to obtain threshold parameter data; based on the state classification data and threshold parameter data, a diagnostic conclusion is generated and a system diagnostic result is output; the abnormal feature data is sorted and classified and an abnormal feature set is output.

[0108] In one embodiment of the present application, the dynamic benchmark construction algorithm is specifically: B(t)=α·Bc(t) + (1-α)·Bh(t); wherein B(t) is the performance benchmark function; Bc(t) is the current state function, Bc(t) = ∑[ωk·Sk(t)], Sk(t) is the state characteristic; Bh(t) is the historical impact function, Bh(t) = ∑[μi·Hi(t)], Hi(t) is the historical state record; α is the dynamic weight coefficient, α= f(Rc(t)), Rc(t) is the state reliability function.

[0109] The multimodal anomaly detection algorithm is specifically as follows: A(t) = [Ai(t), Ag(t), Ap(t), Ar(t)]; where A(t) is the anomaly feature set; Ai(t) is the transient anomaly function, Ai(t) = H(ΨS(t) / Ψt - Δi); Ag(t) is the gradual anomaly function, Ag(t) = ∑[βk·Tk(t)], Tk(t) is the trend term; Ap(t) is the periodic anomaly function, Ap(t) = F_inv[H(|F(S(t))| - Δp)]; Ar(t) is the random anomaly function, Ar(t) = Φ_inv[P(S(t)) - Pn]; Δi and Δp are thresholds, Pn is the normal distribution function, H( ) is the transformation function, S(t) is the signal function, β is the weight coefficient, k is the index of the trend term, F_inv is the inverse Fourier transform, F(S(t)) is the Fourier transform, Φ_inv is the inverse transformation function, and P(S(t)) is the probability distribution function of the signal S(t).

[0110] The degradation trend prediction algorithm is specifically as follows: D(t+τ) = ∑[αk·Mk(t)·exp(λk·τ)]; where D(t+τ) is the prediction trend function; τ is the prediction time; Mk(t) is the trend modal function; λk is the characteristic index; αk is the modal weight coefficient; prediction reliability R(τ) = exp(-β·τ), β is the reliability attenuation coefficient; prediction confidence interval [D(t+τ)±σ(τ)], σ(τ) is the standard deviation function.

[0111] This embodiment establishes a highly adaptable performance evaluation benchmark by building a dynamic benchmark model, combining current state data and historical impact data. Multimodal anomaly feature extraction technology is used to identify transient, gradual, periodic and random anomalies, respectively, to achieve accurate characterization of various anomaly patterns. Based on weighted fusion and time series analysis of anomaly features, the system performance degradation trend is accurately evaluated, and reliable trend prediction is achieved through a multi-scale prediction model. This embodiment not only improves the accuracy of anomaly detection, but also achieves early warning of system performance degradation, solving the problems of unstable performance evaluation benchmarks and low anomaly detection accuracy of satellite communication systems.

[0112] According to one aspect of the present application, step S31 is further:

[0113] S311. Obtain the optimized performance parameter set and parameter reliability data, establish a state feature extraction model, and obtain state feature model data; use the state feature model data to perform parameter mapping transformation to obtain mapping transformation data; reconstruct the mapping transformation data in state space to obtain state space data; calculate the current system state based on the state space data, and output the current state data.

[0114] S312. Read historical state records from a preconfigured system database to obtain historical record data; perform time series decomposition on the historical record data to obtain time series component data; establish a state evolution model using the time series component data to obtain evolution model data; calculate historical state impact based on the evolution model data, and output historical impact data.

[0115] S313, obtaining current state data and historical impact data, constructing a dynamic weight distribution model, and obtaining weight distribution data; adaptively optimizing the weight distribution data to obtain optimized weight data; and using the optimized weight data to perform state fusion to obtain state fusion data.

[0116] S314, acquiring state fusion data, establishing a dynamic benchmark generation model, and obtaining benchmark model data; performing parameter calibration on the benchmark model data to obtain calibration parameter data; using the calibration parameter data to construct a performance benchmark to obtain benchmark construction data; performing stability optimization on the benchmark construction data, and outputting performance benchmark data.

[0117] S315, obtaining performance benchmark data, establishing a benchmark evaluation model, and obtaining benchmark evaluation data; performing reliability analysis on the benchmark evaluation data to obtain reliability evaluation data; performing benchmark correction based on the reliability evaluation data, and outputting updated performance benchmark data.

[0118] This embodiment achieves accurate evaluation of the current system state through state feature extraction model and parameter mapping transformation. The impact of historical state on current performance is analyzed by using historical state recording and time series decomposition technology. Combining dynamic weight allocation and state fusion methods, a highly adaptable performance benchmark model is constructed. This embodiment fully considers the time-varying characteristics of satellite communication system performance and realizes dynamic update of the benchmark through historical data analysis and state fusion. Compared with the traditional fixed benchmark method, it improves the accuracy and adaptability of performance evaluation, provides a reliable evaluation benchmark for anomaly detection, and effectively solves the technical problem of unstable performance evaluation benchmark.

[0119] According to one aspect of the present application, step S32 is further:

[0120] S321. Based on the performance benchmark data, construct a multidimensional feature space model to obtain space model data; perform orthogonal basis decomposition on the space model data to obtain orthogonal basis data; perform feature projection transformation based on the orthogonal basis data to obtain projection mapping data; perform dimensionality reduction optimization on the projection mapping data and output transformation feature data;

[0121] S322, based on the transformed feature data, calculating the signal mutation metric to obtain mutation metric data; performing threshold segmentation on the mutation metric data to obtain segmentation boundary data; based on the segmentation boundary data, extracting the transient abnormal segment to obtain transient segment data; performing feature reconstruction on the transient segment data to output transient abnormal data;

[0122] S323, based on the transformed feature data, construct a trend analysis model to obtain trend model data; perform deviation calculation on the trend model data to obtain deviation sequence data; based on the deviation sequence data, identify the gradual abnormal interval to obtain gradual interval data; perform feature extraction on the gradual interval data to output the gradual abnormal data;

[0123] S324, based on the transformed characteristic data, performing periodic pattern decomposition to obtain periodic component data; performing spectrum analysis on the periodic component data to obtain spectrum characteristic data; based on the spectrum characteristic data, identifying abnormal periodic patterns to obtain periodic abnormal pattern data; performing feature extraction on the periodic abnormal pattern data to output periodic abnormal data;

[0124] S325, constructing a random process statistical model based on the transformed feature data to obtain statistical model data; performing a probability distribution test on the statistical model data to obtain distribution test data; identifying abnormal random patterns based on the distribution test data to obtain random abnormal pattern data; performing feature extraction on the random abnormal pattern data to output random abnormal data;

[0125] S326. Based on the transient abnormal data, the gradual abnormal data, the periodic abnormal data and the random abnormal data, the abnormal feature correlation is calculated to obtain correlation data; based on the correlation data, an abnormal feature fusion model is constructed to obtain fusion model data; based on the fusion model data, feature combination is performed to output abnormal feature data.

[0126] This embodiment realizes the feature transformation of performance data through multi-dimensional feature space model and orthogonal basis decomposition. Transient anomalies are accurately identified by using mutation metric calculation and threshold segmentation. Gradual anomaly characteristics are captured by combining trend analysis and deviation calculation. Periodic anomalies are discovered through periodic pattern decomposition and spectrum analysis. Random anomaly patterns are identified based on random process statistics and probability distribution tests. This embodiment fully considers the diversity characteristics of satellite communication system anomalies and realizes comprehensive identification of anomalies through different detection algorithms. Compared with the traditional single detection method, the comprehensiveness and accuracy of anomaly detection are improved, reliable anomaly feature data is provided for fault diagnosis, and the technical problem of complex and diverse anomaly patterns is effectively solved.

[0127] According to one aspect of the present application, step S33 is further:

[0128] S331. Obtain abnormal feature data, build a feature importance assessment model, and obtain importance model data; use the importance model data to classify abnormal types and obtain abnormal classification data; calculate the impact degree of the abnormal classification data to obtain impact degree data; calculate the weight coefficient based on the impact degree data, and output abnormal weight data.

[0129] S332. Acquire abnormal feature data and abnormal weight data, establish a multi-dimensional feature fusion model, and obtain fusion model data; perform feature interaction analysis on the fusion model data to obtain interactive feature data; perform weighted combination calculation on the interactive feature data to obtain combined feature data; perform consistency optimization on the combined feature data, and output comprehensive abnormal data.

[0130] S333. Obtain comprehensive abnormal data, establish a time series decomposition model, and obtain decomposition model data; extract trend items from the decomposition model data to obtain trend item data; identify periodic items to obtain periodic item data; perform combined analysis on the trend item data and periodic item data, and output degradation trend data.

[0131] S334. Obtain degradation trend data, build a multi-scale prediction model, and obtain prediction model data; optimize parameters of the prediction model data to obtain optimized parameter data; use the optimized parameter data to perform trend extrapolation to obtain extrapolated trend data; perform reliability assessment on the extrapolated trend data, and output prediction trend data.

[0132] S335. Obtain prediction trend data and degradation trend data, establish a prediction error assessment model, and obtain error assessment data; perform volatility analysis on the error assessment data to obtain fluctuation characteristic data; construct a prediction confidence interval based on the fluctuation characteristic data to obtain confidence interval data; combine the prediction trend data and the confidence interval data, and output the revised prediction trend data.

[0133] This embodiment realizes the weight calculation of abnormal features through feature importance evaluation model and abnormal type classification. Comprehensive abnormal features are obtained by multi-dimensional feature fusion and feature interaction analysis. Combined with time series decomposition and trend item extraction, the performance degradation trend is accurately evaluated. Reliable trend prediction is achieved through multi-scale prediction model and parameter optimization. This embodiment fully considers the impact of different types of abnormalities on system performance, and realizes accurate prediction of performance trends through dynamic weight fusion. Compared with the traditional single prediction method, it improves the accuracy and reliability of trend prediction, provides an early warning basis for system maintenance, and effectively solves the technical problem of inaccurate performance trend prediction.

[0134] According to one aspect of the present application, it also includes step S4, constructing an optimization target model based on system diagnosis results, predicted trend data, abnormal feature set and signal quality parameters to obtain optimization target data; calculating a parameter adjustment strategy based on the optimization target data to obtain second parameter adjustment data; generating control instructions based on the second parameter adjustment data to obtain a control instruction set; performing feedback processing on the control instruction set, and outputting a system parameter set and control update parameters.

[0135] According to one aspect of the present application, step S4 is further:

[0136] S41. Obtain system diagnostic results, predicted trend data and abnormal feature sets, as well as signal quality parameters, perform diagnostic feature mapping to obtain diagnostic mapping data; perform trend feature mapping to obtain trend mapping data; perform abnormal feature mapping to obtain abnormal mapping data; perform quality feature mapping to obtain quality mapping data; perform weighted combination of the four types of mapping data to obtain optimized target data.

[0137] S42. Acquire optimization target data, establish a strategy transformation model, and obtain strategy transformation data; perform nonlinear compensation on the strategy transformation data to obtain compensation strategy data; combine the strategy transformation data and the compensation strategy data to generate preprocessing parameter adjustment amounts to obtain preprocessing adjustment data; generate feature extraction parameter adjustment amounts to obtain feature adjustment data; generate diagnostic parameter adjustment amounts to obtain diagnostic adjustment data; and combine the three types of adjustment data into parameter adjustment data.

[0138] S43. Obtain parameter adjustment data, system diagnosis results and prediction trend data, build a control decision model, and obtain control decision data; classify the control decision data, generate a test continuation instruction, and obtain continuation instruction data; generate a parameter adjustment instruction, and obtain adjustment instruction data; generate a test pause instruction, and obtain pause instruction data; generate a test termination instruction, and obtain termination instruction data; combine the four types of instruction data into a control instruction set.

[0139] S44, obtain the control instruction set, parameter adjustment data and optimization target data, build a feedback processing model, and obtain feedback processing data; update the preprocessing parameters of step S1 according to the feedback processing data to obtain preprocessing update parameters; update the feature extraction parameters of step S2 to obtain feature update parameters; update the diagnosis parameters of step S3 to obtain diagnosis update parameters; combine the three types of update parameters into a system parameter set; record the time series change data of the system parameter set as control update parameters.

[0140] This embodiment realizes adaptive optimization of system parameters by constructing a multi-objective optimization model and designing a strategy transformation function. The stability and effectiveness of parameter adjustment are ensured by adopting nonlinear compensation and constraint optimization methods. The overall performance of the system is improved by coordinated optimization of preprocessing parameters, feature extraction parameters and diagnostic parameters. This embodiment not only improves the adaptive ability of the system, but also realizes closed-loop control of the performance evaluation process, solving the coordination and stability problems of traditional methods in parameter optimization and control processes.

[0141] According to one aspect of the present application, step S42 is further:

[0142] S421. Obtain optimization target data, build a multi-objective optimization model, and obtain optimization model data; perform parameter sensitivity analysis on the optimization model data to obtain sensitivity data; establish a strategy transformation function based on the sensitivity data to obtain transformation function data; use the transformation function data to perform strategy conversion and output strategy transformation data.

[0143] S422, obtain strategy transformation data, establish a system response model, and obtain response model data; perform nonlinear characteristic analysis on the response model data to obtain nonlinear characteristic data; construct a compensation function family to obtain compensation function data; use the compensation function data to modify the strategy and output compensation strategy data.

[0144] S423, obtain strategy transformation data and compensation strategy data, build a preprocessing parameter mapping model, and obtain preprocessing mapping data; perform constraint optimization on the preprocessing mapping data to obtain optimized constraint data; calculate parameter adjustment amount based on the optimized constraint data to obtain adjustment amount data; perform stability check on the adjustment amount data and output the preprocessing adjustment data.

[0145] S424, obtain strategy transformation data and compensation strategy data, establish a feature extraction parameter optimization model, and obtain feature optimization data; perform performance evaluation on the feature optimization data to obtain performance evaluation data; generate a parameter adjustment strategy based on the performance evaluation data to obtain adjustment strategy data; use the adjustment strategy data to calculate adjustment parameters and output feature adjustment data.

[0146] S425. Acquire strategy transformation data and compensation strategy data, construct a diagnostic parameter adaptive model, and obtain adaptive model data; perform accuracy analysis on the adaptive model data to obtain accuracy data; optimize parameter adjustment amounts based on the accuracy data to obtain adjustment optimization data; generate diagnostic parameters using the adjustment optimization data, and output diagnostic adjustment data.

[0147] S426. Obtain preprocessing adjustment data, feature adjustment data and diagnosis adjustment data, establish a parameter correlation model, and obtain correlation data; perform conflict detection on the correlation data to obtain conflict detection data; perform parameter coordination based on the conflict detection data to obtain coordination parameter data; use the coordination parameter data for final combination and output parameter adjustment data.

[0148] In one embodiment of the present application, the process of establishing a feature extraction parameter optimization model is specifically: O(t) = min{J[P(t)]}; wherein O(t) is the optimization objective function; P(t) is a parameter vector; J[·] is a multi-objective evaluation function, J[P(t)] = ∑[wk·Jk(P(t))], wk is a weight coefficient; optimization constraint g[P(t)]≤0, h[P(t)] = 0; parameter update rule P(t+1) = P(t) - η·grad_J[P(t)], η is the learning rate, grad_J[P(t)] is the gradient of the multi-objective evaluation function; convergence condition ||P(t+1) - P(t)|| ≤ ε, ε is a threshold.

[0149] This embodiment realizes the strategy conversion of the optimization target through multi-objective optimization model and parameter sensitivity analysis. Strategy compensation optimization is performed by using system response model and nonlinear characteristic analysis. Preprocessing parameter adjustment strategy is generated by combining preprocessing parameter mapping and constraint optimization. Feature parameter adjustment scheme is determined through feature extraction parameter optimization and performance evaluation. Diagnostic parameter settings are optimized based on diagnostic parameter adaptive model and accuracy analysis. This embodiment fully considers the mutual influence between parameters of various links of the system, and realizes the overall improvement of system performance through coordinated optimization. Compared with the traditional single optimization method, it improves the effectiveness and stability of parameter optimization, provides a reliable parameter adjustment basis for system control, and effectively solves the technical problem of uncoordinated system parameter optimization.

[0150] In one embodiment of the present application, the satellite terminal and satellite master station of the satellite communication system hardware platform are used as test objects. The satellite terminal / master station RF signal is connected to the corresponding up converter (BUC) via a cable, and the BUC outputs to the satellite simulation transponder. The model transponder sends the signal to the low noise amplifier (LNB) at the opposite end, and finally inputs it to the RF input of the opposite end through the RF cable. The physical layer frame structure is as follows Figure 5As shown, it includes a first single tone, a unique word (UW) signal, a dual tone, a binary phase shift keying (BPSK) modulated data and a second single tone. The functions of various signals are as follows: the single tone signal is used to detect the flatness of the current channel and the accuracy of the transmission power, and the frequency interval between the first single tone and the second single tone is 2MHz. The UW signal is used for signal synchronization and related parameter estimation. The dual tone is used for third-order intermodulation detection. The BPSK modulated signal is used for control information transmission and parameter estimation, wherein the control information includes the frame number, the lock signal, the transmission power of the current frame and the offset of the next frame.

[0151] In order to test the flatness of the channel, the transmitted data will artificially introduce frequency deviation, and the position of the single tone in the passband will change; in order to test the noise introduced by the system, the transmitted data contains noise. At the receiving end, the data obtained from the AD passes through two channels according to the different processing contents. One channel is down-sampled for signal capture, which mainly uses the UW signal in the transmission signal. After finding the start of a frame, time domain processing is performed, and the frame alignment signal is also sent to the other channel; at the same time, the data obtained from the AD is interpolated to ensure that the frequency domain detection bandwidth is as large as possible. After obtaining the frame alignment signal, the single tone and dual tone signals in the physical frame are intercepted for data frequency domain analysis.

[0152] After data capture is achieved, the positions of single-tone, UW signal, BPSK modulated signal and dual-tone signal in the current received data can be known, and then data can be intercepted. The first single-tone and second single-tone data are input into the power calculation module, and the UW signal and BPSK modulated signal are sent to the bit synchronization module to confirm the best sampling position and transmit the information. After bit synchronization, the BPSK data is restored and extracted to obtain the corresponding data, and then the detection values ​​such as the received power stability and the receiving and transmitting power interpolation are calculated, and finally output to the host computer.

[0153] Before frequency domain processing, the up-sampling is performed, and the synchronization information obtained by time domain processing is combined to extract data and send it to the FFT module. Different parameters are calculated for different signals. For the dual-tone signal, the maximum power, the second maximum power, the third maximum power and the corresponding frequency point position are calculated. After the BPSK signal is transformed in the frequency domain, the total power of the frequency band and the 99% power point are recorded. For the first and second single tones, the maximum value of the frequency domain and the frequency point position corresponding to the maximum value are recorded to calculate the system frequency deviation.

[0154] According to one aspect of the present application, an automatic performance detection device applicable to satellite communications includes:

[0155] at least one processor; and,

[0156] a memory communicatively connected to at least one of the processors; wherein,

[0157] The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement the automatic performance detection method applicable to satellite communication described in any one of the above embodiments.

[0158] The present invention realizes a fully automatic and high-precision evaluation of the performance of a satellite communication system through a closed-loop design of "signal feature decomposition and enhancement - performance parameter extraction and calculation - anomaly detection trend prediction - parameter optimization feedback control". The key technical problems in satellite communication signal processing and performance evaluation are solved by adopting technologies such as triple feature decomposition and enhancement, adaptive sampling reconstruction, dynamic benchmark modeling and multi-objective parameter optimization. The present invention has the advantages of accurate feature extraction, reliable parameter calculation, sensitive anomaly detection and accurate trend prediction. It can timely detect and warn system performance anomalies, and provide reliable technical support for the maintenance and optimization of satellite communication systems.

[0159] The preferred embodiments of the present invention are described in detail above; however, the present invention is not limited to the specific details in the above embodiments. Within the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all belong to the protection scope of the present invention.

Claims

1. An automated performance detection method suitable for satellite communications, characterized in that: The following steps are involved: S1, obtaining a received signal and performing three-dimensional component decomposition on the received signal to obtain periodic feature data, mutation feature data and random feature data; performing enhancement processing on the periodic feature data, mutation feature data and random feature data respectively to obtain enhanced periodic feature data, enhanced mutation feature data and enhanced random feature data; fusing and reconstructing the enhanced periodic feature data, enhanced mutation feature data and enhanced random feature data, and outputting an enhanced reconstructed signal and a signal quality parameter; S2. Based on the signal quality parameter, the enhanced reconstructed signal is non-uniformly sampled and reconstructed to obtain reconstructed sequence data; Perform multi-dimensional feature extraction on the reconstructed sequence data to obtain feature matrix data; Perform parameter separation calculation on the characteristic matrix data to obtain a performance parameter set; Based on the performance parameter set, reliability evaluation and optimization are performed, and the optimized performance parameter set and parameter reliability data are output; S3. Based on the optimized performance parameter set and parameter reliability data, a dynamic benchmark model is constructed to obtain performance benchmark data; Perform multi-modal abnormal feature extraction on the performance benchmark data to obtain abnormal feature data; Based on the abnormal feature data, the performance degradation analysis is performed to obtain the degradation trend data; based on the degradation trend data, the status diagnosis is performed to output the system diagnosis results, the prediction trend data and the abnormal feature set; Step S1 is further as follows: S11, obtaining a received signal, decomposing the received signal by nonlinear projection transformation to obtain periodic feature data; decomposing the received signal by local feature adaptive transformation to obtain mutation feature data; processing the received signal by random process decomposition to obtain random feature data; S12, based on the periodic feature data, construct a cyclic correlation matrix, extract the main feature vector and the eigenvalue, and obtain enhanced periodic feature data through feature enhancement processing; based on the mutation feature data, extract the signal singular point sequence, construct the local morphological feature, and obtain enhanced mutation feature data through morphological enhancement processing; Based on random feature data, a statistical feature model is constructed, and enhanced random feature data is obtained through adaptive threshold processing; S13, based on the significance of the enhanced periodic feature data, the enhanced mutation feature data and the enhanced random feature data, calculating the corresponding weight coefficient to obtain feature weight data; The feature weight data is weightedly combined with the enhanced period feature data, the enhanced mutation feature data and the enhanced random feature data to obtain an enhanced reconstructed signal; the enhanced reconstructed signal is evaluated for signal integrity, feature enhancement degree and noise suppression effect to obtain signal quality parameters; Step S3 is further as follows: S31. Based on the optimized performance parameter set and parameter reliability data, calculate the current state mapping value to obtain current state data; read the historical state record, perform state impact calculation, and obtain historical impact data; combine the current state data and historical impact data to build a dynamic benchmark model to obtain performance benchmark data; S32, performing feature space transformation based on the performance benchmark data to obtain transformed feature data; Performing transient anomaly extraction on the transformed feature data to obtain transient anomaly data; performing gradual anomaly extraction on the transformed feature data to obtain gradual anomaly data; performing periodic anomaly extraction on the transformed feature data to obtain periodic anomaly data; performing random anomaly extraction on the transformed feature data to obtain random anomaly data; combining the transient anomaly data, the gradual anomaly data, the periodic anomaly data and the random anomaly data into anomaly feature data; S33, based on the abnormal feature data, calculating the weight coefficients of various abnormalities to obtain abnormal weight data; performing weighted combination on the abnormal feature data and the abnormal weight data to obtain comprehensive abnormal data; performing time series analysis on the comprehensive abnormal data to obtain degradation trend data; performing prediction calculation based on the degradation trend data to obtain prediction trend data; S34, constructing a decision function based on the degradation trend data, the prediction trend data and the performance benchmark data to obtain decision indicator data; Based on the decision indicator data, the state classification is performed to obtain the state classification data; Perform threshold optimization on state classification data to obtain threshold parameter data; Based on the state classification data and threshold parameter data, a diagnostic conclusion is generated and the system diagnostic result is output; the abnormal feature data is sorted and classified, and an abnormal feature set is output.

2. The automated performance detection method for satellite communications according to claim 1, characterized in that: Step S2 is further as follows: S21. Based on the signal quality parameter, generate a non-uniform sampling sequence to obtain a sampling time sequence; based on the sampling time sequence, resample the enhanced reconstructed signal to obtain reconstructed sequence data; S22, based on the reconstructed sequence data, construct a frequency feature mapping matrix to obtain frequency feature data; based on the reconstructed sequence data, construct a phase feature mapping matrix to obtain phase feature data; based on the reconstructed sequence data, construct an amplitude feature mapping matrix to obtain amplitude feature data; based on the reconstructed sequence data, construct a time feature mapping matrix to obtain time feature data; combine the frequency feature data, phase feature data, amplitude feature data and time feature data into feature matrix data; S23. Based on the characteristic matrix data, a parameter coupling model is constructed, and orthogonal transformation decomposition is performed to obtain decoupling characteristic data; power parameter calculation is performed on the decoupling characteristic data to obtain power parameters; frequency deviation parameter calculation is performed on the decoupling characteristic data to obtain frequency deviation parameters; phase noise calculation is performed on the decoupling characteristic data to obtain phase noise parameters; group delay calculation is performed on the decoupling characteristic data to obtain group delay parameters; modulation characteristic calculation is performed on the decoupling characteristic data to obtain modulation characteristic parameters; power parameters, frequency deviation parameters, phase noise parameters, group delay parameters and modulation characteristic parameters are combined into a performance parameter set; S24. Based on the performance parameter set, reliability calculation is performed on each parameter to obtain a parameter reliability index; Based on the parameter reliability index, the parameter correction amount is calculated to obtain the parameter correction data; Based on the parameter correction data, the performance parameter set is optimized to obtain an optimized performance parameter set; Calculate the comprehensive reliability of the optimized performance parameter set and obtain parameter reliability data.

3. The automated performance detection method for satellite communications according to claim 2, characterized in that: Step S11 is further as follows: S111, acquiring a received signal, performing nonlinear projection transformation processing on the received signal to obtain projection transformation data; constructing a feature space to obtain feature space data; mapping and decomposing the projection transformation data in the feature space data to obtain space mapping data; Based on the spatial mapping data, the periodic component is extracted and the periodic characteristic data is output; S112, constructing a local feature adaptive transformation matrix based on the received signal to obtain adaptive matrix data; extracting local features based on the adaptive matrix data to obtain local feature data; locating mutation points on the local feature data to obtain mutation sequence data; reconstructing features based on the mutation sequence data to output mutation feature data; S113, constructing a random process statistical model based on the received signal to obtain statistical model data; performing signal component decomposition based on the statistical model data to obtain component sequence data; Perform random characteristic analysis on the component sequence data to obtain random characteristic data; perform feature extraction based on the random characteristic data to output random characteristic data.

4. The automated performance detection method for satellite communications according to claim 2, characterized in that: Step S12 is further as follows: S121. Based on the periodic characteristic data, construct a cyclic correlation matrix to obtain correlation matrix data; perform characteristic decomposition on the correlation matrix data to obtain characteristic decomposition data; Based on the eigendecomposition data, the main eigenvector and eigenvalue are calculated to obtain the main eigendata; Based on the main feature data, a frequency domain enhancement function is constructed to obtain frequency domain function data; based on the frequency domain function data, feature enhancement is performed to output enhanced periodic feature data; S122, based on the mutation feature data, extract the signal singular point sequence to obtain singular sequence data; perform morphological gradient calculation on the singular sequence data to obtain gradient feature data; based on the gradient feature data, construct a local morphological operator to obtain morphological operator data; based on the morphological operator data, perform edge optimization to obtain edge optimization data; perform morphological enhancement processing on the edge optimization data to output enhanced mutation feature data; S123, constructing a multi-order moment statistical model based on the random feature data to obtain statistical model data; performing probability distribution estimation on the statistical model data to obtain distribution feature data; constructing an adaptive threshold function based on the distribution feature data to obtain threshold function data; Based on the threshold function data, noise suppression is performed to obtain suppressed noise data; random feature enhancement is performed on the suppressed noise data to output enhanced random feature data.

5. The automated performance detection method for satellite communications according to claim 2, characterized in that: Step S21 is further as follows: S211, based on the enhanced reconstructed signal and the signal quality parameter, construct a signal local complexity evaluation model to obtain complexity data; perform segmented threshold calculation on the complexity data to obtain threshold sequence data; based on the threshold sequence data, generate an adaptive sampling interval to obtain sampling interval data; based on the sampling interval data, construct a complete sampling sequence and output a sampling time series; S212, based on the enhanced reconstructed signal and the sampling time series, construct an interpolation basis function library to obtain basis function data; based on the basis function data, perform local analysis on the original sampling points to obtain local feature data; based on the local feature data, select the optimal interpolation basis function to obtain interpolation function data; based on the interpolation function data, perform signal reconstruction and output reconstructed sequence data; S213, based on the reconstructed sequence data, calculating the reconstruction error index to obtain error index data; based on the error index data, performing spectrum consistency analysis to obtain consistency data; based on the consistency data, evaluating the degree of information retention to obtain information retention data; based on the error index data, consistency data and information retention data, performing reconstruction quality verification, and outputting reconstruction quality data; S214. Based on the reconstructed sequence data and the reconstructed quality data, an adaptive optimization model is constructed to obtain optimized model data; based on the optimized model data, reconstruction parameters are adjusted to obtain first parameter adjustment data; based on the first parameter adjustment data, the reconstructed sequence data is corrected, and updated reconstructed sequence data is output.

6. The automated performance detection method for satellite communications according to claim 2, characterized in that: Step S32 is further as follows: S321. Based on the performance benchmark data, construct a multidimensional feature space model to obtain space model data; perform orthogonal basis decomposition on the space model data to obtain orthogonal basis data; perform feature projection transformation based on the orthogonal basis data to obtain projection mapping data; perform dimensionality reduction optimization on the projection mapping data and output transformation feature data; S322, based on the transformed feature data, calculating the signal mutation metric to obtain mutation metric data; performing threshold segmentation on the mutation metric data to obtain segmentation boundary data; based on the segmentation boundary data, extracting the transient abnormal segment to obtain transient segment data; performing feature reconstruction on the transient segment data to output transient abnormal data; S323, based on the transformed feature data, construct a trend analysis model to obtain trend model data; perform deviation calculation on the trend model data to obtain deviation sequence data; based on the deviation sequence data, identify the gradual abnormal interval to obtain gradual interval data; perform feature extraction on the gradual interval data to output the gradual abnormal data; S324, performing periodic pattern decomposition based on the transformed characteristic data to obtain periodic component data; Performing spectrum analysis on the periodic component data to obtain spectrum feature data; identifying abnormal periodic patterns based on the spectrum feature data to obtain periodic abnormal pattern data; performing feature extraction on the periodic abnormal pattern data to output periodic abnormal data; S325, constructing a random process statistical model based on the transformed feature data to obtain statistical model data; performing a probability distribution test on the statistical model data to obtain distribution test data; identifying abnormal random patterns based on the distribution test data to obtain random abnormal pattern data; Extract features from random abnormal pattern data and output random abnormal data; S326, based on the transient abnormal data, the gradual abnormal data, the periodic abnormal data and the random abnormal data, calculating the abnormal feature correlation to obtain correlation data; based on the correlation data, constructing an abnormal feature fusion model to obtain fusion model data; Based on the fusion model data, feature combination is performed and abnormal feature data is output.

7. The automated performance detection method for satellite communications according to claim 6, characterized in that: Also includes: S4. Based on the system diagnosis results, prediction trend data, abnormal feature set and signal quality parameters, an optimization target model is constructed to obtain optimization target data; Based on the optimization target data, the parameter adjustment strategy is calculated to obtain the second parameter adjustment data; based on the second parameter adjustment data, a control instruction is generated to obtain a control instruction set; feedback processing is performed on the control instruction set to output a system parameter set and a control update parameter.

8. An automated performance testing device suitable for satellite communications, characterized in that: include: at least one processor; as well as, a memory communicatively connected to at least one of the processors; wherein, The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement the automated performance detection method applicable to satellite communications as described in any one of claims 1 to 7.

Images