85 results about "Covariance function" patented technology

The invention relates to a condition monitoring data stream anomaly detection method, in particular to a condition monitoring data stream anomaly detection method based on an improved gaussian process regression model. The problem that an existing method for processing monitoring data stream anomaly detection is poor in effect is solved. The method comprises the steps that firstly, the historical data sliding window size is determined; secondly, the types of a mean value function and a covariance function are determined; thirdly, the hyper-parameter initial value is set to be the random number from 0 to 1; fourthly, q data closest to the current time t are extracted; fifthly, the gaussian process regression model is determined; sixthly, prediction is conducted by means of the nature of the gaussian process regression model; seventhly, PI of normal data at the time t+1; eighthly, monitoring data are compared with the PI; ninthly, whether the real monitoring data need to be marked to be abnormal or not is judged; tenthly, beta (xt+1) corresponding to the monitoring value at the time t+1 is calculated; eleventhly, the real value or prediction value and the t+1 are added into DT; twelfthly, new DT is created. The condition monitoring data stream anomaly detection method based on the improved gaussian process regression model is applied in the field of network communication.

The invention discloses a Gaussian process regression-based method for predicting state of health (SOH) of lithium batteries, relates to a method for predicting the SOH of the lithium batteries, belongs to the fields of electrochemistry and analytic chemistry and aims at the problem that the traditional lithium batteries are bad in health condition prediction adaptability. The method provided by the invention is realized according to the following steps of: I. drawing a relation curve of the SOH of a lithium battery and a charge-discharge period; II, selecting a covariance function according to a degenerated curve with a regeneration phenomenon and a constraint condition; III, carrying out iteration according to a conjugate gradient method, then determining the optimal value of a hyper-parameter and bringing initial value thereof into prior distribution; IV, obtaining posterior distribution according to the prior part; V, obtaining the mean value and variance of predicted output f' without Gaussian white noise; and VI, together bringing the practically predicted SOH of the battery and the predicted SOH obtained in the step V into training data y to obtain the f', then determining the prediction confidence interval and predicting the SOH of the lithium battery. The method provided by the invention is used for detecting lithium batteries.

The invention discloses a partial least squares-based Gaussian regression soft measurement modeling method. The method can be applied to industrial processes with relatively strong time-varying characteristic, coupling, nonlinearity, hysteresis and other complex characteristics. The method comprises the following steps of: firstly, carrying out dimensionality reduction on multi-element input dataon the basis of a partial least squares method, and selecting proper score vectors as input of a Gaussian process regression model; secondly, selecting and combining covariance functions, and constructing different types of Gaussian process regression soft measurement models to predict output data; and finally, evaluating prediction ability of the models by using test set data. Modeling results ofpaper-making wastewater treatment process data prove that a partial least squares-based dimensionality reduction technology for measured variables can improve the prediction ability of the Gaussian process regression model; and the Gaussian process regression models constructed by different covariance functions provide multiple options for effluent indexes, so that the method is more suitable forcomplex and changeable paper-making wastewater treatment environment.

Disclosed is a method for estimating carrier frequency of a PSK (phase shift keying) signal in Alpha-stable distribution noise. The method includes: valuing a cyclic covariance function of the received PSK signal with the Alpha-stable distribution noise; valuing a cyclic covariance spectrum by subjecting the cyclic covariance function to Fourier transform; extracting a section with cyclic frequency being epsilon=0Hz according to the obtained cyclic covariance spectrum; searching peak values of a positive semi-axis and a negative semi-axis of the section respectively to find a positive frequency value and a negative frequency value corresponding to the peak values, and valuing a mean value as the estimated value of the carrier frequency after absolute values are valued. The method is capable of estimating the carrier frequency of the PSK signal in the Alpha-stable distribution noise and is good in estimating performance in an environment of low signal to noise ratio.

The invention provides a method for evaluating the influence of speckle coherence on the ranging accuracy of a single-photon laser radar. The method comprises the steps of setting the parameters of the single-photon laser radar system, calculating the autocorrelation function of a receiving aperture of the single-photon laser radar system and the normalized covariance function on the intensity ofthe receiving aperture to calculate the speckle degree of freedom M; calculating the average signal photon number Ns according to the laser radar equation, and calculating the total noise rate Nn of the laser radar system; differentiating time through the detection probability based on the root mean square pulse width [Sigma]s of the laser pulse to obtain the detection probability density functionfs(t) of the echo signal with respect to time t, and obtaining the mean value shown in the description and the variance Var of the time when the detector detects the target point; and obtaining the influence of the speckle coherence on the ranging accuracy of the single-photon laser radar according to the drift error Ra and the random error Rp. The invention has good compatibility, can provide guidance for the system parameter design of the laser radar, improve the detection probability and reduce the ranging error as much as possible under the restraint of satisfying the false alarm probability.

The invention discloses a LightGBM fault diagnosis method based on improved Bayesian optimization. The LightGBM fault diagnosis method comprises the following steps: 1) determining hyper-parameters needing to be optimized by a LightGBM model and a hyper-parameter value range; 2) improving the Bayesian optimization algorithm to obtain an improved Bayesian optimization algorithm GP-ProbHedge; 3) selecting an optimal hyper-parameter combination of the fault diagnosis model by using the method in the step 2) in combination with a five-fold cross validation mode; and 4) constructing an improved Bayesian optimization LightGBM fault diagnosis model, and giving a model iteration process and an optimization result. By adopting the technology, compared with the prior art, according to the invention,an improved Bayesian optimization algorithm is provided to carry out optimization selection on parameters of a fault model; by improving an acquisition function of a traditional Bayesian optimizationalgorithm and a covariance function of a Gaussian process of the traditional Bayesian optimization algorithm, an improved Bayesian optimization LightGBM fault diagnosis method is provided, and equipment faults are diagnosed and predicted.

The invention discloses a frequency domain laser speckle imaging based blood flow velocity measuring method. The frequency domain laser speckle imaging based blood flow velocity measuring method comprises the following steps of illuminating laser beams on a measured object, imaging the measured object through an imaging system, collecting an original speckle image of the measured object through an image sensor, transferring dynamic speckle intensity with single pixel points being in a time domain of the collected original speckle image into a frequency domain, calculating the power spectral density, performing polynomial fitting on the power spectral density to obtain a smooth curve, transferring the smooth curve into the time domain through Fourier transform, calculating an autocovariance function of the pixel points and performing normalization, establishing a blood velocity measuring model, obtaining a relationship between the covariance function and the blood velocity and finally performing fitting to obtain a blood velocity value. The frequency domain laser speckle imaging based blood flow velocity measuring method has the advantages of not only eliminating static noise, improving the blood velocity measuring accuracy, avoiding influences from imaging environmental factors such as intensity and illumination angles and improving the measuring stability.

The invention discloses a geostatistics-based wind power station wind speed spatio-temporal data modeling method. The method comprises the following steps: 1, constructing a space structure matrix of a wind power station according to geographical location information of fans, space convariance function of wind power station wind speed and variation function of wind power station wind speed to represent the spatial correlation of input wind speed of each fan; 2, performing layered modeling on the input wind speed of each fan by a universal Kriging method and a Bayesian algorithm and estimating model parameters by adopting Gibbs sampling; and 3, predicting the forward P steps of the wind speed of each fan to acquire simulation samples of P step forward prediction distribution of the wind speed of each fan, and sampling and averaging to acquire the optimal prediction result of the wind speed of each fan, wherein P is more than or equal to 1. By the method, the time and space correlation of wind speed data of different fan positions is comprehensively analyzed based on the physical characteristics of wind, so that a more accurate prediction model is built for the whole wind power station, and the prediction result is better than that of the traditional method.

The invention provides a GPR lithium battery health state prediction method based on a neural network kernel function. The method comprises the steps that a covariance function is determined based on the neural network kernel function to construct a GPR prediction model; a mean function in the GPR prediction model and hyperparameters in the covariance function are initialized; a logarithmic maximum likelihood estimation function is used to optimize the hyperparameters; and training data and test data are input into the GPR prediction model to acquire the value of the test data. According to the lithium battery health state prediction method provided by the invention, the accuracy and the precision of battery SOH value prediction are high, and the uncertainty is reduced.

The invention provides a lithium battery SOH prediction method based on neural network and periodic kernel functions GPR. The method comprises that a covariance function is determined on the basis of the neural network kernel function and the periodic kernel function, and a GPR prediction model is constructed; a mean value function in the GPR prediction model and a hyper-parameter in the covariance function are initialized; a logarithm maximum likelihood estimation function is used to optimize a hyper-parameter; and training data and test data are input to the GPR prediction model to obtain a value of the test data. Via the lithium battery SOH prediction method, the SOH value of a battery can be predicted accurately and precisely, and the uncertainty is lower.

Signal processing technology for assessing dynamic system similarity for fault detection and other applications is based on time- and frequency-domain time series analysis techniques and compares the entire autocorrelation structure of a test and reference signal series. The test and reference signals are first subjected to similar pre-processing to help guarantee signal stationarity. Pre-processing may include formation of multivariate signal clusters, filtering and sampling. Multivariate periodograms or autocovariance functions are then calculated for each signal series. Test statistics are computed and assessed to determine the equality of the test and reference signals. When the difference between sample autocovariance functions or periodograms of such signals exceeds a preselected threshold value, fault detection signals and/or related diagnostic information are provided as output to a user.

The invention discloses a wind power combination probability prediction method considering evaluation index conflicts. The method is characterized by comprising the steps of determining a decomposition parameter K through variational mode decomposition optimized on the basis of the law of conservation of energy, decomposing an original wind power signal into a series of intrinsic mode function components, removing an intrinsic mode function with the minimum amplitude, and combining the remaining intrinsic mode functions to obtain a wind power sequence after fluctuation and randomness are reduced; constructing an input feature set containing 96-dimensional historical features by using the wind power sequence, and constructing different GPR models by using 10 covariance functions; calculating the area grey correlation closeness based on the five indexes by adopting an area grey correlation decision-making method so as to comprehensively evaluate the performance of each prediction model and solve the conflict between evaluation indexes; and calculating the weights of different GPR probability prediction models in the combined model according to the area grey correlation closeness, constructing the combined model, and carrying out wind power probability combined prediction by using the combined probability prediction model.

The invention relates to the field of nonlinear time-varying system optimization control, in particular to a control strategy for achieving multivariable PID in a PLS frame on the basis of a Gaussianprocess model. The problems that by means of existing control strategies, interaction among multiple loops cannot be eliminated, a built model is not complete, and the control performance is poor aresolved. The control strategy for achieving multivariable PID in the PLS frame on the basis of the Gaussian process model comprises the steps that 1, parameters of a PID controller are given; 2, the GP(Gaussian process) model is used for achieving the control strategy of multivariable PID, wherein firstly, an MIMO system is subjected to decoupling, secondly, the model uncertainty is improved, thirdly, a covariance function is chosen, and the parameters are optimized; 3, the PID controller is set, wherein a gradient-based optimization algorithm is used for using the GP model for adjusting the PID controller. The control strategy for achieving multivariable PID in the PLS frame on the basis of the Gaussian process model has the advantages that the GP model is used for providing a predictionvariance, and the prediction reliability of a local random area is shown; cross coupling effects brought by a multivariable control process are taken into account, through the PLS frame, the MIMO system is decoupled into single loops, and then based on the GP model, the PID controller parameters are adjusted separately.

The invention provides a steady direction of arrival estimation method based on a sparse representation and covariance fitting. The steady direction of arrival estimation method based on the sparse representation and the covariance fitting mainly comprises the steps that a sparse spatial spectrum representation model is established according to antenna array receipt signals, a parameterization representation is carried out on model errors, and an optimization problem is established according to a covariance fitting criterion. The obtained problem is non convex optimization, therefore, the transformation and solution of the problem can be performed through equivalent transformation, parameter increase and step-by-step solution. Without considering the model errors at first, an original problem can be simplified as a convex optimization problem. The convex optimization problem is quickly solved by an existing method, and an initial solution is obtained. Iteration solving is carried out on the original problem, a model error parameter is estimated and an initial estimation is renewed. The steady direction of arrival estimation method based on the sparse representation and the covariance fitting can obtain a precise DOA estimation by low complexity.

The invention discloses a beam domain Root-MUSIC method based on covariance correction. The method comprises the steps of building a uniform circular array far-field narrow-band signal model, and converting a spatial domain array output model to a beam domain in order to meet a needed Van der Monte-de structure; constructing a generalized linear combination covariance matrix, wherein the matrix isobtained by combining a traditional sample covariance matrix and a priori knowledge matrix; obtaining an initial DOA estimated value and a spatial domain steering vector by utilizing the covariance;solving a wave beam domain conversion error matrix caused by few sensors and an error matrix caused by subspace leakage under low snapshot; and searching an optimal correction factor to continuously reduce an error between the sample covariance and a true value, and finally obtaining a new direction of arrival by using the new covariance matrix. According to the method, the problem of subspace leakage under low snapshots and the beam domain conversion error caused by a small number of sensors are considered at the same time, so that the error between a sample covariance matrix and an ideal value can be remarkably reduced, and the estimation precision is improved.

The invention discloses a stepped frequency spectrum sensing method based on energy and covariance detection, relates to the frequency spectrum sensing field of cognitive radio, and aims at solving the problem that a present energy detection method is low in the detection accuracy. Energy detection is carried out at a user end; if it is detected that a main user uses an authorized frequency spectrum, a cognitive user keeps silent; if a result of energy detection is a frequency spectrum cavity, secondary detection, namely covariance detection, is needed; and a result of the covariance detection is still that the authorized frequency spectrum is not used, the cognitive user can occupy the authorized frequency spectrum for communication. Thus, advantages of energy detection and covariance detection are effectively integrated; when the channel state is sound, energy detection which is easy to implement is carried out to sense the frequency spectrum; and when the signal to noise ratio is low, secondary detection is carried out by utilizing the high statistic characteristics of covariance detection, and thus, the detection accuracy is improved.

The invention relates to a coprime array-to-uniform linear array conversion-based DOA (Direction of Arrival) estimation method and device. The method includes the following steps that: array element positions are arranged in an ascending order according to a coprime array model, differencing processing is performed on any two array element positions, so that a delay set is obtained, and a difference set table is formed; the difference set is traversed, all coordinates requiring delay are found out in a sparse covariance matrix, corresponding covariance function values are subjected to statistical averaging, autocorrelation function values requiring delay are calculated; a Toepltz matrix is constructed according to the autocorrelation function values, element values on diagonals parallelto the main diagonal of the Toepltz matrix are identical; and a spatial spectrum search function is constructed based on the MUSIC algorithm, the Toepltz matrix is solved, when a spatial spectrum search vector is consistent with a signal steering vector, a correct DOA estimation value is found out. The device includes an external RAM, a digital signal processor, and an output driving and display circuit. The method and device of the present invention do not require a multiplier and are simple in operation, can achieve the same effect as a spatial smoothing matrix, and can greatly reduce the degrees of freedom.

The invention discloses a least-square collocation model of a satellite gravity gradient tensor diagonal three-component inversion earth gravity field and a modeling method, and belongs to the field of satellite gravity measurement. The method comprises the steps that starting from an orthogonality rule of a spherical harmonics function, Txx and Tyy components and Tzz, Txx and Tyy components are combined separately, practical calculation formulas of auto-covariance and cross covariance functions among the satellite gravity gradient tensor diagonal three-components and covariance functions of the components and gravitational potential coefficients are derived, and then the least-square collocation model of the Txx and Tyy two-component and Tzz, Txx and Tyy three-component inversion earth gravity fields is obtained. Accordingly, the theoretical basis is provided for fully utilizing the satellite gravity gradient tensor diagonal three-component inversion earth gravity field, and data resources of a gravity gradient measurement satellite GOCE launched by the ESA are fully utilized.

The invention discloses a time delay estimation-based laser ranging method, which is used for ranging through the time difference between an echo signal and a transmitted signal. The method comprises the following steps of: (a) acquiring the transmitted signal and the echo signal; (b) performing time delay estimation according to the transmitted signal and the echo signal to obtain a time delay estimation value; and (c) adding the time delay estimation value to the echo signal to obtain the distance which needs to be measured and calculated. According to the laser ranging method provided by the invention, the time delay of the echo signal with low signal-to-noise ratio relative to the transmitted signal is directly obtained from continuous modulated laser signals by adopting a time delay estimation technology. According to the laser ranging method provided by the invention, the laser ranging is performed by a robustness-based time delay estimation algorithm, a noise signal is modeled by Alpha stable distribution, and the time delay between the echo signal and the transmitted signal is solved by adopting a fractional lower order covariance function and related functions.

A method for estimating Doppler spread in mobile wireless communication devices, for example in CDMA or W-CDMA cellular communication systems, with improved noise immunity. The Doppler spread estimation is based on an estimated value of an autocorrelation or autocovariance at a first lag (210) and at a second lag (220), the magnitude of which is greater than the first lag. A first ratio is determined (250) between a first difference (230) and a second difference (240). The estimated Doppler spread is generally proportional to a square root (260) of the first ratio, and is scaled (270) by a multiplicative factor that depends on whether the estimated function is an autocorrelation or autocovariance function.

A method for diagnosing life of manufacturing equipment having a rotary machine, includes: measuring reference time series data for characteristics before deterioration of the manufacturing equipment occurs; finding a reference auto covariance function based on the reference time series data; extracting a reference variation caused by variations of the process condition and power supply from the reference auto covariance function, and calculating a cycle of the reference variation; measuring diagnostic time series data for the characteristics in a sequence to be measured of the manufacturing equipment; finding a diagnostic auto covariance function based on the diagnostic time series data; and determining the life of the manufacturing equipment from the diagnostic auto covariance function using a component with a cycle shorter than a cycle of the reference variation.

A method for predict wireless network includes the following steps: the Gaussian process regression model is established by selecting several network communication index variables and corresponding target index variables, a covariance function is determined in the form of a Gaussian kernel function to solve a reasonable value of each parameter in the covariance function, a first value to be predicted at any time point and a first cross-sectional data vector of a required independent variable are set to obtain a corresponding joint distribution function to calculate an estimation value of the first value to be predicted; a second value to be predicted in any period of time and a second cross-sectional data vector of a desired independent variable are set, and an estimation value of the second value to be predicted is calculated. The invention utilizes Gaussian process regression method to carry out regular mining on the historically accumulated network live broadcast data, constructs adata model, and can effectively predict the change situation of the network index of the target in a certain period of time in the future, thus providing more effective data reference for optimizing radio network resource allocation and performance optimization.

The invention provides a multi-action sequence mechanism reliability assessment method. Through reliability assessment, the mechanism system reliability, key action sequences and key factors can be determined. The method mainly comprises the five steps of analyzing action sequences of a mechanism, and for the action sequences, analyzing motion function related fault modes in sequence; for the fault modes of the action sequences, building reliability calculation models; according to relationships between the action sequences and faults, building mechanism system reliability models; by utilizinga covariance function, analyzing correlation among the fault modes of the action sequences; and by utilizing an improved one-time multi-dimensional normal method and considering the correlation of the action sequences, performing mechanism system reliability and sensitivity assessment.

The invention discloses a non-stationary random dynamic response analysis method considering geometric nonlinearity of a structure. The method comprises the steps of (1) performing calculation according to non-stationary random dynamic loads to obtain a mean value and an auto-covariance matrix of the loads; (2) calculating an eigenvalue and an eigenvector of the auto-covariance matrix, and obtaining truncation orders of the eigenvalue and the eigenvector; (3) decomposing the non-stationary random dynamic loads based on KL expansion and Latin hypercube sampling methods to obtain a group of random samples of the non-stationary random dynamic loads; and (4) establishing a finite element model of the structure, calculating the random samples of the non-stationary random dynamic loads as response functions of the loads by adopting a transient analysis method, and performing calculation according to the response functions to obtain variance and auto-covariance functions. According to the method, the limitation of conventional non-stationary random dynamic response analysis only for a linear structure is overcome.

The invention discloses a method extracting fault anomaly from geophysical prospecting gravity anomaly and belongs to exploration geophysics in the earth sciences. An existing method for separating regional anomaly and local anomaly (a trend analysis method, a frequency domain filtering method, a moving average method, upward extension and the like) has the defects that the boundary effect is obvious and spatial structural characteristics of the regional anomaly are not considered, and accordingly the separated and extracted regional anomaly is not high in accuracy. For obtaining a high-accuracy method for separating the regional anomaly and local anomaly of the fault, a structure function method is given. Under the condition that a spatial distribution geometrical characteristic, a covariance function and the regional anomaly form of gravity observation data are known, the fault region anomaly is obtained by means of the optimal, linear and unbiased interpolation estimation method. The trend analysis method is only a particular case of the structure function method. The structure function method is suitable for separation and extraction of the fault gravity anomaly with a linear structure. An experiment proves that the structure function method is effective and feasible on the aspect of implementation.

The invention provides a lithium battery state of health prediction method based on a neural network and Maternard kernel function GPR. The method comprises the steps that a covariance function is determined based on a neural network kernel function and a Maternard kernel function so as to construct a GPR prediction model; the mean value function in the GPR prediction model and the hyper-parameter in the covariance function are initialized; the hyper-parameter is optimized by using a logarithmic maximum likelihood estimation function; and training data and test data are inputted to the GPR prediction model so as to acquire the value of the test data. According to the lithium battery state of health prediction method, the prediction accuracy and precision of the battery SOH value are enabled to be high, and the uncertainty is low.