17 results about "Covariance function" patented technology

The invention discloses a power prediction method for a wind power generation system based on a stacked sparse auto-encoder network Hammerstein model, and aims to solve the problems that only nonlinear mapping is modeled and dynamic characteristics of the system are neglected and parameter coupling of the Hammerstein model leads to complex identification in an existing method, the Hammerstein model is constructed, an ARMAX model is utilized to describe a dynamic linear module, and the dynamic linear module is used to predict the power of the wind power generation system based on the stacked sparse auto-encoder network Hammerstein model. A static nonlinear module is described by stacking the sparse auto-encoder network; designing a zero-mean Gaussian signal input proxy model, realizing series module decoupling based on covariance function characteristics, and eliminating parameter coupling; a self-adaptive multi-strategy grey wolf optimization algorithm is adopted to determine the number of neurons in a network hidden layer, training is performed in combination with a sparse criterion function, and the feature extraction capability is improved. According to the method, synchronous capture of static nonlinear and dynamic linear characteristics is realized, the calculation complexity is reduced, the model identification precision, prediction precision and robustness are improved, and the method is suitable for accurate prediction of the power of a wind power system.

This invention provides an underdetermined mode identification method based on block term decomposition and adaptive kurtosis harmonic separation, belonging to the field of structural health monitoring and operating mode identification technology. The method acquires multi-channel vibration responses, calculates the covariance matrices under different time delays, and stacks them into a third-order tensor. Block term tensor decomposition is used to extract the time-domain autocovariance function sequence of each potential component. The natural frequencies and damping ratios of each component are initially estimated using local peak finding and logarithmic decay methods. The minimum truncation length is adaptively calculated based on the coupling relationship between period and damping ratio. Kurtosis values ​​are calculated for the truncated autocovariance sequence, and harmonic interference is eliminated based on a kurtosis threshold while retaining the true structural modes. Finally, mode shapes are extracted from the spatial block matrix of the retained components, outputting high-precision natural frequencies, damping ratios, and mode shapes. This invention can achieve accurate identification of dense modes and high-damped modes under complex operating conditions with severe underdeterminacy and harmonic interference, improving the robustness of operating mode parameter identification.

The invention relates to the field of machine learning, in particular to a residual service life prediction method and system based on a deep Gaussian process and meta-learning, a medium and equipment, and the method comprises the steps: obtaining original data to construct a multi-task data set; a multi-layer depth Gaussian process model is constructed based on a multi-task data set, a radial basis function kernel is used as a covariance function, and a multi-task Gaussian likelihood function is used for modeling uncertainty of RUL prediction; meta-training is carried out on a PHM data set, the model is optimized through internal circulation and external circulation, and an Adam optimizer is adopted to learn cross-task generalization weights; transferring the last layer of parameters and likelihood function parameters obtained by meta-training to a fan gearbox data set and an NASA aero-engine data set based on the same meta-learning model framework, and performing fine tuning on a support set of the test data set; and performing RUL prediction on the test sets of the fan gear data set and the NASA aero-engine gear data set by using the weights obtained after fine tuning, and outputting a prediction mean value, a variance, a confidence band and a prediction band.

The invention relates to the field of agricultural meteorology early warning methods, in particular to an agricultural meteorology early warning method based on regional microclimate data interpolation and correction. An adversarial network model is constructed by introducing a topographic physical constraint loss function and an uncertainty quantization mechanism to generate a high-resolution, physically credible and static meteorological background field, so that the generated meteorological background field is ensured to be rich in space details and strictly follow basic physical rules such as a temperature vertical declining rate and the like; and meanwhile, the uncertainty of each lattice point can be self-evaluated and is used as prior information of a Bayesian layering dynamic correction model, a spatial covariance function fusing a geographic distance and a terrain distance is constructed in the Bayesian layering model, and the purpose that the spatial covariance function of the geographic distance and the terrain distance is fused in a complex terrain area with sparse meteorological stations is achieved. And meanwhile, a meteorological analysis field with high spatial resolution and high timeliness is obtained, so that the early warning accuracy and timeliness of sudden agricultural meteorological disasters in local areas such as frost, dry and hot air and the like are remarkably improved.

A method for obtaining a strain field of a key area of a shaft wall comprises the steps that sensing optical fibers are buried in concrete of a key well section of the inner wall of the shaft in the circumferential direction of the shaft; carrying out data acquisition on the sensing optical fiber of the key well section; interference of temperature on strain measurement is eliminated through regional temperature compensation; using a covariance function to represent the correlation of the spatial data; performing interpolation through known data points; and after the predicted values of all interpolation points are calculated, all interpolation results are drawn into a strain thermodynamic diagram of the key horizon. According to the invention, the high-density sensing optical fibers are arranged in the concrete of the inner wall of the 360-degree whole area of the key well section of the shaft, so that the strain field change data can be economically and feasibly obtained in real time, and the high-density space coverage of the shaft strain monitoring sensor at the key part under the condition of low cost is met; therefore, the precision of the well wall strain field of the local well section is improved, and the precise monitoring of the strain evolution of the potential fracture risk area is realized.

The application discloses an RCS full-probability size extrapolation method and system based on an electromagnetic scattering mechanism, and the method is as follows: 1, a backscattering electric field formula under different sizes is derived according to the electromagnetic scattering mechanism, and a target RCS calculation expression of an incident wave vector is obtained according to the definition of a radar cross section (RCS); the target RCS calculation expression is converted into a variant formula used for constructing a covariance function through a coefficient method and an integral mean value theorem; 2, the variant formula is inversely transformed and then is disassembled; 3, a polynomial covariance function is used to represent a polynomial part in the inversely transformed formula, and a spectral mixture covariance function is used to represent a cosine part in the inversely transformed formula; the two kinds of covariance functions are combined into an SPFPE covariance function; 4, the SPFPE covariance function is initialized by using a random number method; 5, hyperparameters of the SPFPE covariance function are optimized with a maximum log-likelihood as an optimization target, and an SPFPE-GPR method is obtained; 6, the target RCS is extrapolated by using the SPFPE-GPR method.

The invention relates to a laser peening inherent strain prediction method based on Gaussian process regression, which comprises the following steps: carrying out a laser peening orthogonal test, and establishing model training data; carrying out normalization processing on the training data by adopting a maximum and minimum normalization method; selecting a mean value function, a covariance function and a hyper-parameter optimization algorithm, and establishing a Gaussian process regression model; training the Gaussian process regression model by using the training data, and optimizing hyper-parameters of the Gaussian process regression model; and utilizing the trained Gaussian process regression model to predict inherent strain distribution under any process parameter combination so as to realize continuous correspondence between the process parameters and the inherent strain. According to the laser peening intrinsic strain prediction method based on Gaussian process regression, intrinsic strain distribution prediction under any technological parameter combination is achieved by establishing the Gaussian process regression model and optimizing hyper-parameters through training data, and the method has the advantages that the number of tests is reduced, and prediction efficiency and accuracy are improved.

The present application relates to geophysics and geodetic data processing technical field, disclose the GNSS time series analysis and modeling method considering variable amplitude, the method comprises the following steps: S1, obtain the GNSS coordinate time series;S2, time series is modeled as the Gaussian process defined by mean function and covariance function;S3, construct the mean function describing long-term linear trend;S4, construct the product kernel by the periodic kernel and the non-periodic smooth kernel multiplication as the covariance function, to unify the periodicity and time-varying amplitude of the signal modeling;S5, the maximum likelihood estimation method is used to solve the hyperparameter in model;S6, the time series is decomposed by using the optimized model, and the time-varying amplitude periodic signal is obtained.The present application can integrally model the periodic signal and its time-varying amplitude by constructing the Gaussian process product kernel, so as to realize the accurate separation of signal component.

This invention relates to the field of steel rolling, specifically disclosing a method, system, computer equipment, and medium for predicting rail rolling force. The method includes: collecting on-site data of the target rail production line; selecting multiple reduction parameters from the third pass as model input factors; selecting corresponding rolling force parameters as model output factors; and dividing the data into training and testing sets; constructing a Gaussian process regression model, defined as f(x) ~ GP[m(x), k(x, x')], where m(x) is the mean function, k(x, x') is the covariance function, and the observation equation of the model is y = f(x) + ε, where y is the model output factor, x and x' are the model input factors, ε is Gaussian white noise, and ε ~ N(0, σ n ²), σ n ² represents the noise variance; the Gaussian process regression model is trained and optimized using the training dataset to obtain a trained prediction model; the prediction accuracy of the trained prediction model is verified using the test set; if the prediction accuracy reaches a threshold, the prediction model is used to predict the rail rolling force.

This invention provides a configuration method for RAG systems based on multi-objective Bayesian optimization, belonging to the technical field of RAG systems. This invention obtains an initial observation dataset by randomly sampling initial configurations from a structured search space using a multi-objective evaluation function. For three performance indicators, independent Gaussian process models are established to fit the mean and covariance functions of the objective function. Based on the current Pareto front and the posterior prediction distribution of the Gaussian process model, the expected hypervolume improvement acquisition function is calculated, and the configuration with the largest improvement value is selected for iterative evaluation and updating. Finally, non-dominated solutions are identified to form the Pareto optimal configuration set. This invention solves the technical problem of complex configuration parameter space and conflicts between multiple performance indicators in retrieval-enhanced generation systems, which makes it difficult to automatically obtain the Pareto optimal configuration set.

A computer-implemented method for enabling control or monitoring of a computer-controlled entity operating in an environment by predicting a future state of the computer-controlled entity and / or its environment using sensor data which is indicative of a current state of the computer-controlled entity and / or its environment. The method includes using a first neural network for approximating a drift component of a stochastic differential equation and a second neural network for approximating a diffusion component of the stochastic differential equation, and discretizing the stochastic differential equation into time steps, and obtaining time-evolving mean and covariance functions based on the discretization and determining a probability distribution of a second state of the computer-controlled entity and / or its environment therefrom. The control of the computer-controlled entity may thus be enhanced and made more efficient and reliable using the uncertainty information available from the determined probability distribution.

The application relates to a target RCS prediction method based on a Gaussian process regression and a two-dimensional agent model, which comprises the following steps: a three-dimensional rectangular block is established, and parameters are set; a data set is acquired; a composite covariance function is constructed, and a Newton gradient method is used to optimize and solve hyperparameters; a two-dimensional agent model is constructed according to the optimized hyperparameters, a covariance vector and a covariance matrix of the agent model are calculated, and a prediction result is obtained; and the prediction result is evaluated according to evaluation indexes. The two-dimensional agent model has good performance; through training of training data, a corresponding relationship between input and output is found, and a test set is used to realize a prediction function; overall, time consumed is obviously less than that of a traditional electromagnetic scattering algorithm, and calculation cost is saved; the method is not only suitable for a radar signal analysis field, but also suitable for remote sensing research, geophysical science and the like, and analysis cost is low.

The application provides a foundation pit settlement space inference method based on an improved space-time Kriging model. The method specifically comprises the following steps: S1. Collecting data to establish a multi-dimensional and staged database structure; S2. Segmenting the construction period by using the collected data, and establishing a corresponding space-time covariance function for each stage; S3. Segmenting and splicing the space-time covariance functions of multiple stages to establish a space-time joint covariance structure with geological disturbance perception capability; S4. Using the defined space-time covariance function joint covariance matrix and covariance vector; S5. Defining the settlement space inference to predict the settlement value. The application can effectively deal with the settlement mutation caused by the construction stage and geological section, has higher prediction accuracy and stability, realizes the unified modeling of different stages in the whole construction process, and significantly improves the adaptability of the model to time sequence evolution and physical field response.

This invention discloses a multi-objective optimization method and related equipment for small-sample selected area laser melting process parameters, relating to the field of metal additive manufacturing technology. The method includes: acquiring a sample dataset obtained from selected area laser melting forming experiments under a preset combination of laser power and scanning speed parameters; constructing a relative density prediction model and a crack density prediction model using a Gaussian process regression method, and training the relative density prediction model and crack density prediction model respectively using the sample dataset; wherein the covariance function used in the Gaussian process regression method is the Matrn kernel function, and a white noise kernel is introduced into the covariance function; based on the trained relative density prediction model and crack density prediction model, searching for the optimal process parameters in the process parameter space using a Bayesian optimization algorithm. This invention can efficiently and accurately achieve multi-objective optimization of selected area laser melting process parameters.

A method (200) for determining target magnetic field data d(Rβ,ω) is disclosed. The method (200) comprises selecting empirical magnetic field data d(Rω) that is associated with an influence region 105ω; fitting synthetic magnetic dipole source moments m(rω,s) to the empirical magnetic field data d(Rω): and determining target magnetic field data d(Rβ,ω) using the synthetic magnetic dipole source moments m(rω,s). A method (300) for determining target magnetic field data d(Rβ,ω) is also disclosed. The method (300) comprises selecting empirical magnetic field data d(Rω) that is associated with an influence region 105ω; determining empirical covariance data Yω based on the empirical magnetic field data d(Rω); determining an optimised value of one or more parameter of a covariance function C by fitting the covariance function C to the empirical covariance data Yω; and determining target magnetic field data d(Rβ,ω) using the covariance function C.

The invention discloses a high-dimensional signal encryption transmission method based on nonlinear adaptive modulation, which comprises the following steps of: sensing a channel state: acquiring channel noise and nonlinear distortion data, and calculating a covariance function cov (n (t)) and a transmission matrix H (t); chaotic parameter generation: generating a dynamic chaotic parameter set based on the covariance function cov (n (t)), and constructing a modulation sub-matrix A d (t); constructing a modulation matrix: generating a high-dimensional dynamic modulation matrix M (t) through tensor product operation; signal encryption mapping: mapping an original signal S raw into an encrypted signal S enc (t), and adding a noise masking item; anti-interference transmission: optimizing a target function through an APSO algorithm, and dynamically adjusting M (t) to counteract channel distortion; and decryption and restoration: the receiving end decrypts the ciphertext through the M (t) 1 inverse matrix to restore the original signal S raw. Through combination of the dynamic chaos parameter and the high-dimensional modulation matrix, channel adaptive encryption transmission is realized, the anti-interference capability and security are improved, and the method is suitable for high-dimensional signal secret communication under a complex channel.

The invention relates to an intermittent energy output modeling prediction method and system based on a Gaussian process, and the method comprises the steps: carrying out the modeling of intermittent energy output based on a Bernstein polynomial, calculating a mean value and a covariance function of the intermittent energy output, and constructing a Gaussian process prior form; obtaining a historical observation set of intermittent energy output, optimizing a hyper-parameter in a Gaussian process prior form through maximum likelihood estimation according to a noise observation point of each track sample in the historical observation set on discrete time, and solving a covariance matrix of a Bernstein coefficient; and calculating a mean value and a covariance matrix of the posterior distribution of the Bernstein coefficient according to the observed intermittent energy output at multiple moments, thereby obtaining the joint distribution of the intermittent energy output in the future test time. Compared with the prior art, the method has the advantages that the latest observed data can be effectively utilized to predict the generating capacity of the next time period, and energy scheduling and optimization with quick response are realized.