45results about How to "Small root mean square error" patented technology

The invention provides a universal circuit breaker mechanical fault diagnosis method based on feature fusion of vibration and sound signals. The method includes steps of 1, collecting machine vibration signals and machine sound signals during an engaging and disengaging process of a universal circuit breaker; 2, adopting an improved wavelet packet threshold value denoising algorithm for denoising; 3, adopting a complementary total average empirical mode decomposition algorithm for extracting a plurality of solid mode function components reflecting state information of engagement and disengagement actions of the circuit breaker from the denoising signals; 4, determining the number Z of the solid mode function components; 5, calculating the energy ratio, the sample ratio and the power spectrum entropy as three types of features; 6, adopting a combination core principal component analysis method for performing dimension reduction on a feature sample with unified three types of features of the vibration and the sound signals and obtaining M principle components; 7, establishing a related vector machine based sequence binary tree multiple classifier model.

The invention discloses a broadband signal DOA estimation method based on a co-prime array, and the method comprises the steps: S1, designing a co-prime array structure through an antenna; S2, carrying out the sampling and discrete Fourier transform of a broadband signal received by an antenna in the co-prime array, and obtaining a frequency domain signal output model; S3, calculating an autocorrelation matrix of the frequency domain signal output model, carrying out the vectorization of the frequency domain signal output model, and obtaining a new signal model; S4, carrying out the processing of the new signal model, and obtaining a spatial smooth covariance matrix of the broadband signal; Sa5, dividing a space domain grid, constructing a dictionary, carrying out the sparse representation of the spatial smooth covariance matrix through employing the dictionaries of a plurality of frequency points of the broadband signal, and forming a multi-measurement-vector sparse representation model of a plurality of dictionaries of the broadband signal; S6, achieving the arrival direction estimation of the broadband signal in a mode of solving a sparse inverse problem through the joint sparse constraint of the sparse representation coefficients of the plurality of dictionaries. The method can improve the estimation precision of the direction angle of the broadband signal under the condition of low signal to noise ratio, and reduces the direction finding error.

The invention discloses a nonlinear system state estimation method based on Kalman filtering positioning, and proposes a strongly adaptive Kalman filtering mechanism which combines a nonlinear filtering algorithm and Kalman filtering. The method comprises the steps: carrying out the simultaneous estimation of node positions and channel parameters through employing an RSSI state estimation algorithm based on square root volume Kalman filtering, and obtaining the estimation value of a state vector; employing the Kalman filtering for further processing according to the linear change of a state equation, obtaining optimal estimation, and building a strongly adaptive square root volume Kalman filtering algorithm; giving the design steps of the strongly adaptive square root volume Kalman filtering algorithm; and calculating a theoretical square root error lower bound under a state space model based on the RSSI state estimation. The method enables an estimation result to be improved, and improves the precision. The method does not need to excessively depend on improper initial conditions, can be well adapted to a highly nonlinear system, and is not liable to enable the algorithm to be divergent and ineffective.

An SVR antifriction bearing performance degradation prediction method based on a krill-herd algorithm belongs to the field of functional approximation rotating machinery prediction methods. The method comprises the following steps: firstly analyzing time domain, frequency domain and time-frequency domain feature indexes, and proposing a feature extraction method based on combination of CEEMD and wavelet packet half-soft threshold noise reduction to perform fault diagnosis of an antifriction bearing; performing comprehensive evaluation of the fault degradation feature of the antifriction bearing for multiple feature parameters, and proposing a method of combining the LLE nonlinear feature dimension reduction method with the fuzzy C mean value; and finally, introducing the basic theory of the support vector regression machine, and proposing the prediction model of multivariable support vector regression machine based on the krill herd algorithm, optimizing parameters of the SVR, and selecting the optimal C, [sigma] parameters. The method is advantaged by high prediction precision, short calculation time, and good feature value prediction effect after clustering. The degradation process of the antifriction bearing can be precisely predicted through the abovementioned three steps.

The invention relates to a system and method for removing an ocular artifact from an electroencephalogram signal. The provided system includes a signal interception module, an ocular artifact recognition module and an ocular artifact removal module. A segment of an original electroencephalogram signal of a subject is processed by the signal interception module and sent to the ocular artifact recognition module and the ocular artifact removal module at the same time. If the segment of the electroencephalogram signal contains the ocular artifact, the ocular artifact recognized by the ocular artifact recognition module is sent as a reference signal to the ocular artifact removal module; otherwise, the reference signal based on a recursive least square adaptive filter is set to zero. Finally,the adaptive filtering technology of the ocular artifact removal module is used for achieving the on-line removal of the ocular artifact. According to the system and the method, the ocular artifact can be removed online, a large amount of useful information in the original electroencephalogram signal is retained, there is no requirement for the number of channels of the input electroencephalogram,the root mean square error of the signal is reduced further, less time is consumed, and the system is more suitable for real-time BCI scenes.

The invention discloses a towed linear array sonar subarray error mismatching estimation method, aims to realize reduction of the positioning error and can obtain the accurate orientation estimate andthe higher angle resolution. The method is characterized in that in an array manifold matrix model, a real full array manifold matrix is represented as linear combination of the displacement error amount of each subarray and the error amount of each subarray contributed to the full array manifold matrix; the position error between subarrays is introduced to a direction finding model, for a full array model with subarray displacement mismatching, the Bayesian rule is utilized to establish a data fusion model, a subarray position error vector and the target true azimuth are simultaneously solved, according to data collected by fused sensor nodes, the Bayesian algorithm is utilized to estimate the subarray displacement error and the direction of arrival to obtain the likelihood function of amulti-shot observation value; the azimuth estimate and the displacement error estimate of the root mean square error changing with the SNR are obtained through utilizing the posterior function.

The invention relates to a failure prediction method of a roller bearing based on a partial least squares extreme learning machine. The method herein includes: analyzing feature indexes, such as timedomain, frequency domain and time-frequency domain, providing a feature extraction method based on the combination of half-normal distribution and empirical wavelet denoising to perform failure diagnosis on a roller bearing so as to obtain better denoising effect owing to proximity to original signals; for multi-feature parameters, comprehensively evaluating failure attenuation features of the roller bearing, and providing a method with the combination of residual-modified ISOMAP (isometric feature mapping) nonlinear feature dimension reduction and fuzzy C-means, so that change tendency and sorting precision are improved for the roller bearing in different attenuation stages; based on the extreme learning machine theory, providing a data prediction model based on a partial least squares extreme learning machine, optimizing parameters in the ELM (extreme learning machine), selecting node quantity of an optimal hidden layer and weight value of a connection layer, and selecting a Softmaxactivation function. Therefore, prediction precision is high, calculating time is short, and post-clustering feature value detection is effective. The failure stage of the roller bearing can be precisely predicted via the above steps.

The invention discloses a vacuum pump vibration signal noise reduction method based on EEMD (ensemble empirical mode decomposition) and a wavelet threshold; the method comprises the following steps: firstly, an original signal is subjected to EEMD to obtain a plurality of IMF (Intrinsic Mode Function) components and a remainder; secondly, all the IMF components are subjected to calculation of a normalization self-correlation function, and the IMF components are divided into a signal-dominant IMF component and a noise-dominant IMF component according to the characteristic of zero attenuation ofthe self-correlation function; then, the noise-dominant IMF component is subjected to wavelet soft threshold noise reduction processing; finally, the noise-dominant IMF component subjected to waveletsoft threshold processing and the signal-dominant IMF component are subjected to signal reconstruction with the remainder so as to obtain a noise-reduced vacuum pump vibration signal. According to the method in the invention, EEMD is adopted, so that the problems of mode aliasing, end point effect and the like caused by EEMD can be overcome, noise signals in the vacuum pump vibration signals areeffectively removed, more useful signals are well reserved, and the signal-to-noise ratio of the signals is improved.

The invention discloses a relative radiation correction method based on pseudo-invariant feature point classification and layering, and belongs to the field of remote sensing image radiation correction. An existing radiation correction method is low in correction precision of relative radiation on remote sensing images including coastlines, islands and the like occupying dominant ground object areas. The method comprises the steps of 1, obtaining ground object sub-images based on remote sensing image classification; 2, determining an initial relative radiation correction model and an initial PIFs of the ground object sub-image based on nonlinear regression analysis of the spectrum; 3, based on gradient-based fine nonlinear regression analysis, determining a fine nonlinear relative radiation correction model and a fine PIFs of the ground object sub-image; 4, performing relative radiation correction on the to-be-corrected surface feature sub-image by using the refined PIFs and the refined nonlinear relative radiation correction model; and 5, synthesizing the corrected image into a complete image. The method is applied to the field of remote sensing image radiation correction.

The invention discloses a spacecraft weighted orbit determination method based on data driving, and the method comprises the steps: enabling an orbit determination sample set Z consisting of a measurement data set X and a corresponding target spacecraft orbit set Y to be equal to {X, Y}, carrying out the weighting of the measurement data, and constructing an orbit determination weighted sample set; calculating a Gram matrix of the constructed orbit determination weighted sample set, and taking an elastic network as a loss function; and calculating an optimal estimation value of a spacecraft orbit determination result y (t) based on the Gram matrix. According to the method, a complex dynamic model does not need to be constructed, the idea of machine learning is introduced, and an unknown spacecraft orbit can be estimated by learning a large number of nominal orbits with existing tags; besides, the training data and the test data have the same noise characteristics, and the sample data is taken as the training data, so that the sensitivity of the orbit determination result to the measurement noise can be reduced, and the application prospect is wide.

The invention discloses a three-factor method cuffless continuous blood pressure detection system based on an artificial neural network. The system predicts the blood pressure after three types of information of propagation time, a PPG waveform and personal characteristics are fused, and a hardware circuit includes an electrocardiogram (ECG) signal acquisition circuit based on shaped signals and oversampling, and a photoplethysmography (PPG) signal acquisition circuit based on a transimpedance amplifier circuit. The system uses oversampling and fast digital phase lock demodulation technology to simplify the circuit; filtering processing is performed on synchronously acquired PPG and ECG signals, the propagation time information (PTT) and the PPG waveform (PWPs) information are extracted, the personal characteristic parameters of the testees are recorded, and the neural network is used to establish a relationship model between PTT, PWPs and PCPs and the blood pressure; and the blood pressure value is predicted through the relationship model. Compared with a traditional cuffless blood pressure detection system, the blood pressure detection system proposed by the invention has highercorrelation and a lower root mean square error (RMSE) of blood pressure prediction results.

The invention provides a time-frequency overlapped Gaussian amplitude modulation communication signal separation method. The method comprises the following step: I, building a signal model, converting a received mixed signal into a plurality of Gaussian amplitude modulation source signals, and transforming a solving process of the mixed signal into a process of solving a multi-dimensional variable parameter; II, calculating an initial value of the multi-dimensional variable parameter with a genetic algorithm; III, calculating an optimal value of the multi-dimensional variable parameter with a minimum value search method; and IV, calculating each source signal according to the optimal value of the multi-dimensional variable parameter. Through adoption of the method, separation of single-channel time-frequency overlapped Gaussian amplitude modulation communication signals can be realized effectively. According to the method, the Gaussian amplitude modulation communication signal model is built firstly; initial values of parameters needing to be estimated are solved; the optimal values of the parameters needing to be estimated are searched with the method; and the Gaussian amplitude modulation source signals are recovered according to the obtained optimal values.

The invention discloses a compressive sensing high-resolution array processing method based on subspaces, and a convex optimization calculation function is constructed by employing the sparsity of measured signals in a sparse matrix. According to the method, the signals can be recognized in a coherent condition of the signals and noise; and in different signal-to-noise ratio conditions, the root-mean-square error of the method is smaller than that of other contrast algorithms, and the calculating time of the method and the calculating time of the conventional methods are almost the same.

The invention relates to a double-layer tubular column pulse eddy current data denoising method based on a noise model, which comprises the following steps of: establishing four noise models by analyzing the characteristics of electromagnetic noise, jitter noise, temperature noise and oil tube eccentric noise in real noise; respectively designing a network model based on a linear noise parameter,a network model based on a sine jitter noise parameter and a network model based on a Gaussian white noise parameter, adding depth weight coefficients into the three noise components, and constructingan overall deep learning model based on noise model parameters suitable for solving by a deep learning method; constructing a simulation model based on material attributes, spatial dimensions, physical field interfaces and the like of the double-layer tubular column and performing simulating to obtain a pure signal of the double-layer tubular column as a training set; training the eddy current data of different well sections according to the overall deep learning model, obtaining noise model parameters, obtaining a noise model in the pulse eddy current signal of the whole double-layer tubularcolumn, and carrying out the adaptive denoising of the double-layer tubular column according to the noise model.

The invention provides a universal circuit breaker mechanical fault diagnosis method based on feature fusion of vibration and sound signals. The method includes steps of 1, collecting machine vibration signals and machine sound signals during an engaging and disengaging process of a universal circuit breaker; 2, adopting an improved wavelet packet threshold value denoising algorithm for denoising; 3, adopting a complementary total average empirical mode decomposition algorithm for extracting a plurality of solid mode function components reflecting state information of engagement and disengagement actions of the circuit breaker from the denoising signals; 4, determining the number Z of the solid mode function components; 5, calculating the energy ratio, the sample ratio and the power spectrum entropy as three types of features; 6, adopting a combination core principal component analysis method for performing dimension reduction on a feature sample with unified three types of features of the vibration and the sound signals and obtaining M principle components; 7, establishing a related vector machine based sequence binary tree multiple classifier model.

The invention discloses a medical image enhancing method based on shear wave deformation and fuzzy contrast combination. The method includes the steps: S1 decomposing a medical image into a low frequency component and a plurality of high frequency components by shear wave deformation; S2 linearly transforming the decomposed low frequency component and denoising and enhancing the decomposed high frequency components; S3 performing shear wave inverse transformation on the linearly transformed low frequency component and the high frequency components processed by a threshold value method in the step S2 to obtain a reconstructed image; S4 enhancing fuzzy contrast of the reconstructed image obtained in the step S3 to improve global contrast of the image. The method has the advantage that the definition and the global contrast of the medical image are improved.

The invention provides a spectrum reconstruction method based on micro-nano structure optical filter array modulation and sparse matrix transformation. The method comprises the implementation steps: a micro-nano structure optical filter array is constructed; the transmission spectrum of each micro-nano structure optical filter is measured; discrete point sampling is carried out on the transmission spectrum of each micro-nano structure optical filter; the intensity of transmission light passing through each micro-nano structure optical filter is measured; a matrix equation is constructed; discrete cosine transform is carried out on the matrix equation; sparse transformation is carried out on the discrete cosine transformation matrix equation; and a spectrum reconstruction result is obtained. According to the method, the problem that in the prior art, the noise error is not processed, the obtained reconstructed spectrum is still influenced by the noise error is solved, meanwhile, the root-mean-square error of the reconstructed spectrum is effectively reduced, and simulation results show that the method effectively improves the precision of spectrum reconstruction.

The invention discloses a wind power cluster power prediction and parameter optimization method, which comprises the steps of dividing historical NWP data and historical power data into two independent data sets, and optimizing parameters in three stages; carrying out principal component analysis of the original wind speed vector, taking a principal component analysis result as input of a wind power cluster power prediction model, and respectively dividing two independent data sets into a to-be-predicted data set and a historical data set; calculating an Euclidean characteristic distance between the input data matrix of the prediction point and the historical data set; comparing the Euclidean characteristic distance with a threshold value delta to obtain a data set with the highest matching degree and a prediction data set, judging whether optimization is finished or not, and otherwise, setting a parameter value by using a variable-scale network search method to continue to optimize toobtain four parameters with the minimum overall prediction error; and controlling the three parameter values to be unchanged according to the obtained initial optimization values of the four parameters, and changing the remaining parameter value until an optimal four-parameter combination is obtained. The method is high in prediction precision and has popularization value.

The invention relates to a seawater total nitrogen and total phosphorus analysis method and system based on the absorption spectrum characteristic peak area. The problems of low detection precision and low instrument detection lower limit when a traditional single-wavelength and multi-wavelength method is used for measuring seawater with low total nitrogen and total phosphorus concentration are solved. The method mainly comprises seven parts of data acquisition, absorbance conversion, smooth filtering, characteristic position selection, characteristic peak area calculation, a regression modeland concentration data output, the available information amount of a weak absorbance signal close to the lower detection limit of the instrument can be amplified by utilizing the characteristic peak area, and the detection capability of the instrument when being close to the lower detection limit of the instrument can be improved.

The invention provides a passive positioning method based on regularization constraint weighted least square. The method is divided into two steps, in the first step, a positioning model based on the RCTLS thought is established for the TDOA / FDOA positioning problem, regularization parameters are solved based on the criterion of the minimum mean square error, and then a closed analytical solution of the model is given through mathematical derivation; and in the second step, an equation about the estimation error of the first step is established by using constraint conditions, then solving is carried out, and finally the estimation result of the first step is corrected by using the obtained solution. According to the method, the positioning precision of the positioning method based on the CTLS model can be improved, and the performance is more stable under the condition that the coefficient matrix is ill-conditioned.

The invention provides a time-frequency overlapped Gaussian amplitude modulation communication signal separation method. The method comprises the following step: I, building a signal model, converting a received mixed signal into a plurality of Gaussian amplitude modulation source signals, and transforming a solving process of the mixed signal into a process of solving a multi-dimensional variable parameter; II, calculating an initial value of the multi-dimensional variable parameter with a genetic algorithm; III, calculating an optimal value of the multi-dimensional variable parameter with a minimum value search method; and IV, calculating each source signal according to the optimal value of the multi-dimensional variable parameter. Through adoption of the method, separation of single-channel time-frequency overlapped Gaussian amplitude modulation communication signals can be realized effectively. According to the method, the Gaussian amplitude modulation communication signal model is built firstly; initial values of parameters needing to be estimated are solved; the optimal values of the parameters needing to be estimated are searched with the method; and the Gaussian amplitude modulation source signals are recovered according to the obtained optimal values.

The invention discloses a wideband signal DOA estimation method based on a coprime array, comprising: S1, designing a coprime array structure by using an antenna; S2, sampling and discrete Fourier transforming the wideband signal received by the antenna in the coprime array Transform to obtain the frequency domain signal output model; S3, calculate the autocorrelation matrix of the frequency domain signal output model and vectorize it to obtain a new signal model; S4, process the new signal model to obtain the spatial smoothing coherence of the wideband signal Variance matrix; S5, dividing the airspace network, constructing a dictionary, and using the dictionary at multiple frequency points of the broadband signal to perform sparse representation of the spatial smooth covariance matrix, forming a multi-measurement vector sparse representation model of multiple dictionaries of the broadband signal ; S6, realize the DOA estimation of broadband signals in the form of solving the sparse inverse problem through the joint sparse constraints on the multi-dictionary sparse representation coefficients. The invention can improve the estimation accuracy of the direction angle of the broadband signal under the low signal-to-noise ratio, and reduce the direction-finding error.

The invention discloses a data-driven weighted orbit determination method for spacecraft, which includes: an orbit determination sample set Z={X,Y} composed of a measurement data set X and a corresponding target spacecraft orbit set Y, and the measurement Data weighting, constructing a weighted sample set for orbit determination; calculating the Gram matrix of the weighted sample set for orbit determination, using the elastic network as a loss function, and calculating the optimal estimated value of the spacecraft orbit determination result y(t) based on the Gram matrix. The present invention does not need to construct The complex dynamic model introduces the idea of ​​machine learning, and by learning a large number of labeled nominal orbits, the orbit of unknown spacecraft can be estimated; in addition, the training data and test data have the same noise characteristics , taking this sample data as training data can reduce the sensitivity of orbit determination results to measurement noise, and has broad application prospects.

The invention discloses a compressive sensing high-resolution array processing method based on subspaces, and a convex optimization calculation function is constructed by employing the sparsity of measured signals in a sparse matrix. According to the method, the signals can be recognized in a coherent condition of the signals and noise; and in different signal-to-noise ratio conditions, the root-mean-square error of the method is smaller than that of other contrast algorithms, and the calculating time of the method and the calculating time of the conventional methods are almost the same.

The invention discloses a positioning precision improving method based on coil magnetic field regulation and control, and belongs to the technical field of magnetic positioning. An included angle is changed by regulating a magnetic moment direction of a magnetic target so as to avoid a positioning blind area, and the positioning error of the tensor magnetic positioning method is greatly reduced. In the method for avoiding the positioning blind area, the magnetic moment direction of a magnetic target is regulated and controlled by regulating and controlling the magnetic field direction of the coil, the problem that the more the coil axes are, the larger the energy consumption is is considered, the method for avoiding the positioning blind area of the two-axis coil and the three-axis coil isprovided, and a calculation basis is provided for selection of the coil axes in actual use.

The invention discloses a turbofan engine mutation fault diagnosis method based on the improved Gaussian particle filter, relates to the field of aeroengine fault diagnosis, can realize rapid diagnosis of mutation faults, has low noise level of diagnosis results, and high accuracy of estimation results. The invention includes: injecting the engine nonlinear mathematical model into the sudden failure of gas circuit components; designing an improved Gaussian particle filter algorithm based on pseudo-covariance; loading the engine nonlinear mathematical model into the improved Gaussian particle filter algorithm, and improving the gas circuit Diagnose component mutation faults and obtain diagnosis results. The invention adopts pseudo-covariance instead of covariance and Gaussian sampling instead of re-sampling, which reduces the diagnosis time, improves the diagnosis accuracy, and can realize the rapid diagnosis of gas path mutation faults within the life cycle of the engine.

The invention relates to a seawater total nitrogen and total phosphorus analysis method and system based on the characteristic peak area of ​​absorption spectrum, which solves the problem of low detection accuracy and instrument detection in traditional single-wavelength and multi-wavelength methods when measuring seawater with low concentration of total nitrogen and total phosphorus The lower limit problem mainly includes seven parts: data acquisition, absorbance conversion, smoothing filter, characteristic position selection, characteristic peak area calculation, regression model and concentration data output. Using the characteristic peak area, the weak absorbance near the detection limit of the instrument can be reduced The signal can be amplified by the amount of information, which can improve the detection ability of the instrument when it is close to the detection limit of the instrument.

The invention relates to a Raman spectrum peak identification method for optimizing a wavelet algorithm, in particular to an unknown article Raman spectrum peak identification detection and Raman spectrum peak wavelet algorithm analysis and identification method, and belongs to the field of analysis instruments.