53 results about "Non-Gaussianity" patented technology

The invention discloses a process monitoring method which is based on non-Gaussian component extraction and support vector description. The method comprises the following steps: read-in of training data and data to be diagnosed, data preprocessing, establishment of a principal component analysis model, particle swarm optimization algorithm, non-Gaussian projection calculation, support vector data description, residual analysis, principal component estimation, fault detection and the model updating. By the method, the non-Gaussian components can be automatically extracted from operating data of an industrial process, thus avoiding the disadvantage that the conventional statistical process monitoring method assumes that data is subject to normal distribution, and the non-Gaussian projection algorithm based on the particle swarm optimization algorithm ensures the maximization of the non-Gaussian properties of the extracted independent components, and avoids the problem that the independent component analysis method is easy to be involved in the locally optimal solution. Compared with the conventional statistical process monitoring method, the method can find abnormity in time, effectively reduce the rate of false alarm, and obtain better monitoring effect.

The invention discloses a just-in-time learning soft measurement modeling method based on a Bayes Gaussian mixture model, and belongs to the field of complex industrial process modeling and soft measurement. The method is used for a time varying industrial process with nonlinearity and non-Gaussianity; through a strategy of updating localities in real time online, a Bayes information criterion isadopted to determine an optimal Gaussian ingredient number; when new test data comes, a posterior probability that the new test data belongs to each Gaussian ingredient is calcuated, a Mahalanobis distance between the new test data and training data is solved, and the posterior probability and the Mahalanobis distance are blended to serve as a similarity index; and finally, one group of data withthe highest similarity is selected from the original training sample to establish a current GPR (Gaussian Process Regression) model, and model output prediction is carried out to achieve an effect onimproving product quality and lowering production cost.

Blind polarization demultiplexing algorithms based on complex independent component analysis (ICA) by negentropy maximization for quadrature amplitude modulation (QAM) coherent optical systems are disclosed. The polarization demultiplexing is achieved by maximizing the signal's non-Gaussianity measured by the information theoretic quantity of negentropy. An adaptive gradient optimization algorithm and a Quasi-Newton algorithm with accelerated convergence are employed to maximize the negentropy. Certain approximate nonlinear functions can be substitutes for the negentropy which is strictly derived from the probability density function (PDF) of the received noisy QAM signal with phase noise, and this reduces the computational complexity. The numerical simulation and experimental results of polarization division multiplexing (PDM)-quadrature phase shift keying (QPSK) and PDM-16QAM reveal that the ICA demultiplexing algorithms are feasible and effective in coherent systems and the simplified ones can also achieve equivalent performance.

The invention discloses a fault detection method based on a particle swam optimization kernel independent-component analysis model. Kernel learning skills and a particle swam optimization algorithm are combined and used in the method, a traditional independent component analysis method is expanded into a modeling method capable of directly processing non-linear process data, and a corresponding fault detection model is established on the basis. The fault detection method specifically comprise the following steps: firstly, an original training data matrix is converted into a kernel matrix through a kernel function, and centralization processing is carried out; secondly, non-linear independent components are solved by utilizing iteration of the particle swam optimization algorithm and are ranked according to non-Gaussian sizes; and, finally, the non-linear fault detection model is established, and online fault detection is implemented. Compared with a traditional method, the method of the invention prevents a whitening pretreatment process so that a condition that original data information loses or is distorted cannot occur. Furthermore, the method of the invention is not limited to establish the fault detection module, but can be applied to other fields related to non-linear data signal source separation.

The invention discloses a long-distance pipeline pressure monitoring method based on an ensemble modified ICA-KRR algorithm. The method comprises the following steps: 1) constructing a long-distance pipeline pressure monitoring data matrix; 2) calculating a variable P belonging to Rm x m; 3) extracting a component matrix T=PTX; 4) whitening the extracted component matrix T to obtain a whitened result; 5) calculating a matrix S=CTZ; 6) calculating a matrix Cn; 7) calculating a separation matrix W belonging to Rd x m and a mixing matrix A belonging to Rm x d; 8) obtaining source signals of independent components, wherein the independent relationship between the source signals of the independent components is reflected by a non-Gaussian property, the non-Gaussian property is quantized by a negative entropy function, and the negative entropy function can select three non-quadratic functions; 9) constructing three kinds of component importance evaluation standards; 10) blending to form a two-layer comprehensive learning strategy; 11) forming 9 component selection models; 12) obtaining a weight coefficient w; 13) obtaining regression fault signal data y; 14) and calculating a leakage position d. The method can realize real-time monitoring and accurate positioning of the leakage position on a long-distance pipeline.

The invention discloses a novel mall-current grounding system transient line selection method, and the method comprises the following steps: starting a fault line selection logic; collecting the fault transient current data of all branch circuits on a bus; calculating zero sequence fault transient currents of the circuits and bus; calculating the spectral kurtosis of the circuits and bus; and carrying out line selection according to the spectral kurtosis of the circuits and bus. The method is very strong in anti-noise capability, and can completely inhibit white noise theoretically. When a fault angle is smaller or a grounding resistance is big, a fault transient current is smaller, thereby causing line selection through a first half wave method, an energy method, a wavelet method and other methods based on a transient component to be difficult. The method provided by the invention takes the non-stationarity and nongaussianity of signals as characteristic values for judgment, avoids the above impact, achieves the correct line selection, and improves the accuracy of line selection.

The invention discloses a non-Gaussian submarine landform type identification method based on a multi-fractal spectrum characteristic. The method comprises the following steps of 1) calculating depthdistribution skewness and kurtosis according to submarine depth measurement data and determining whether a landform is a non-Gaussian landform; 2) calculating the multi-fractal spectrum characteristicof the non-Gaussian landform; 3) using the multi-fractal spectrum characteristic as an original variable, and applying a factor analysis method to extract a landform factor; 4) according to the landform factor, using a support vector machine to design a landform type classifier; and 5) calculating landform depth distribution skewness and kurtosis of a landform to be identified, determining the nongaussianity of the landform, calculating the multi-fractal spectrum characteristic of the non-Gaussian landform and the landform factor, and using the designed classifier to identify a landform type.The method has advantages that the method is simple, a calculated amount is small, an identification accuracy is high, and manpower is saved and so on. The method is suitable for non-Gaussian submarine landform type identification.

A parametric family of multivariate density functions formed by mixture models from univariate functions of the type exp(-|x|<beta>) for modeling acoustic feature vectores are used in automatic recognition of speech. The parameter beta is used to measure the non-Gaussian nature of the data. beta is estimated from the input data using a maximum likelihood criterion. There is a balance between beta and the number of data points that must be satisfied for efficient estimation.

The invention provides a wireless communication system frequency spectrum sensing method based on non-gaussianity measure, and relates to the technical field of frequency spectrum sensing. The method includes the specific steps that a corresponding base-band equivalent discrete-time signal is obtained by means of a wireless signal received by a cognitive user antenna and is sectioned, so that sectional sub signals are obtained; fast fourier transform is carried out on the sectional sub signals, and a power spectrum of the received signal is obtained through calculation; by means of a harr wavelet function, two-layer wavelet multi-resolution decomposition is carried out on the power spectrum of the signal received by the cognitive user and a wavelet coefficient vector of the power spectrum of the received signal is constructed and obtained; then a steepness value of the wavelet coefficient vector is calculated; a non-gaussianity measure testing statistical magnitude of the wavelet coefficient vector of the power spectrum of the received signal is obtained through calculation; the test statistical magnitude is compared with a detection threshold value, if the test statistical magnitude is larger than the detection threshold value, it is indicated that an authorized user signal exists and otherwise, it is indicated that an authorized user does not exist. According to the method, known noise statistic information is not needed, good frequency spectrum sensing performance is attained while a low signal-to-noise ratio is achieved, and a small-scale small-power wireless communication device can be effectively detected.

The invention discloses a nonlinear system state deviation evolution method based on differential algebra and Gaussian sum. The method includes forecasting the terminal state of a nonlinear system according to a differential algebra method, and representing it as a high-order Taylor expansion polynomial related to original-state deviation; determining a sub Gaussian distribution covariance matrix, and fitting a Gaussian sum model for each sub Gaussian distribution through shooting; calculating the high-order central moment of the sub Gaussian distributions; determining the mean value and the covariance matrix of each sub Gaussian distribution at the terminal moment, and providing a terminal-state deviation distribution probability density function in the form of the Gaussian sum. The method can be extended to any designated-order deviation evolution accuracy automatically, manual derivation of high-order partial derivatives of kinetic equations is not needed, the method is applicable to long-term forecasted nonlinear system deviation evolution analysis problems with high nonlinearity, the remarkable efficiency advantages of the method can be still maintained as compared to the Monte Carlo simulation method while deviation distribution and non-Gaussianity thereof are described accurately, and accordingly, the method has the advantages of convenience in use and high calculation accuracy.

The invention discloses a device fault detection method based on acoustic features, and relates to the technical field of fault detection. An interference sound separation calculation method comprisesthe following steps: S11, taking kurtosis as measure of signal nongaussianity. According to the device fault detection method based on the acoustic features, a sound of a power transformation devicein working is collected through a microphone array; independent source sound signals are extracted by an interference sound separation technology based on negative entropy independent component analysis; dimensionality reduction processing is performed on sound data based on two-dimensional principal component analysis to extract main frequency spectrum feature information; the sound signals are classified based on a multi-value decision diagram to realize fault diagnosis; and finally the power device fault location based on weighted multi-signal classification is performed. The purpose of real-time on-line fault detection is achieved; the device fault type and information can be judged accurately and timely; and while the working efficiency is improved, the use by a user is facilitated.

The invention discloses an intermittent process fault monitoring method based on fourth-order moment singular value decomposition, and is used for solving data nonlinearity and non-Gaussianity broughtby nonlinearity in an intermittent process. The method comprises two stages of "offline modeling" and "online monitoring", wherein the stage of "offline modeling" comprises the following steps that:firstly, carrying out data standardization, carrying out fourth-order moment processing, and combining fourth-order moment matrixes; and then, carrying out singular value decomposition, and simplifying the obtained matrix to make a preparation for monitoring. The stage of "online monitoring" comprises the following steps that: carrying out standardization on online data, carrying out fourth-ordermoment processing, and combining the fourth-order moment matrixes; and then, calculating a statistical amount, a residual error and a corresponding control line; and finally, using the statistical amount to monitor a generation process, and giving an alarm when faults appear. The method fully considers the nonlinearity and the non-Gaussianity of the data of the intermittent process, reduces the false alarm rate of a normal stage, reduces the false alarm rate of a fault stage, quickens response speed and has a high practical value.

A method and apparatus are provided for adaptive channel code selection according to a non-Gaussianity of a channel in a wireless communication system. A method includes, if a non-Gaussianity of the channel falls within a first range, encoding transmission data using a first channel code; and if the non-Gaussianity of the channel falls within a second range, encoding the transmission data using a second channel code. The first channel code and the second channel code have different degree distributions of repetition codes.

The invention discloses a low-rise building enclosure member wind disaster loss analysis method considering a typhoon duration effect. The method is used for evaluating wind-induced damage of a low-rise building enclosure structure. In the calculation process, determining the internal pressure for the current time step in combination with the house trepanning working condition; comparing the impact force of the throwing object with the impact bearing force of the throwing object to determine the damage to the window caused by the impact of the throwing object; comparing the load extreme value with the pressure bearing capacity to determine enclosure member damage caused by wind pressure; determining a new opening to update the internal pressure; carrying out damage analysis again until no new hole appears; entering the next time step until all the time steps in the typhoon duration are analyzed, and obtaining statistical values of the failure probability and the loss rate of each component in each time step. According to the invention, the time-holding effect of the typhoon is considered, meanwhile, factors such as non-Gaussian property, randomness and spatial correlation of the load of the enclosure component are considered, the loss condition in the typhoon is evaluated more practically, and the calculation efficiency is greatly improved.

The invention relates to a causal network inference method based on kurtosis. The method comprises the steps that exogenous variables are selected based on the kurtosis, a causal order is found layer by layer, and a cutting edge is tested by a cutting edge using the least square method, so that a complete causal network is output. The causal network inference method based on kurtosis is not sensitive to the degree of non-Gaussian of disturbance variables and has strong stability, can still be able to maintain a high recognition rate especially when the non-Gaussian is weak, and has low complexity, kurtosis calculation is needed for each variable itself, the method is a direct estimate way, and the causal network recognition accuracy is high.

The invention relates to a seismic signal Q value estimation method based on non-Gaussianity maximization. The method comprises the following steps of S1, obtaining a seismic signal, researching a phase distortion process of the seismic signal, and analyzing a relation between the phase distortion signal and the non-Gaussianity; S2, scanning a quality factor Q value of the seismic signal, and performing phase correction; and S3, calculating the non-Gaussianity of the seismic signal after the phase correction, and when the non-Gaussianity is maximized, taking the quality factor Q value at the moment as an estimation result. Compared with the prior art, the method has the advantages that it is not needed to assume a white noise reflection coefficient or extract a reference wavelet, only theQ value scanning needs to be subjected to the phase correction, and the accuracy of the Q value is measured by utilizing the non-Gaussianity maximization; and stable Q value distribution can be obtained, so that high-resolution seismic data processing and interpretation are facilitated.