61 results about "Outlier elimination" patented technology

The invention provides an aeroengine load spectrum filtering method based on a rain-flow counting method. The aeroengine load spectrum filtering method comprises steps that (1), rain-flow filtering for original spectrum of aeroengine load spectrum is carried out to acquire rain-flow filtering spectrum; and (2), based on the rain-flow filtering spectrum, points of multiple original load spectrum are added, and the filtering spectrum based on the rain-flow method is lastly acquired. Through the method, the load loading/unloading path information can be reflected, namely, on the basis of deleting small circulation, circulation peak/valley values of the filtering spectrum and the original spectrum are guaranteed to be identical, and the path of the original spectrum is reserved to a maximum degree; singular points can be more effectively eliminated, threshold filtering for signals is carried out to eliminate noise of the signals, and signal processing efficiency is improved. The method is advantaged in that research on flight parameter pre-processing and outlier elimination in an aeroengine load spectrum compilation process is realized.

The invention belongs to the field of array signal processing, and particularly relates to an array position error correction method for the situation that information source azimuth information is biased, and the array position error is accurately corrected. According to the array position error correction method, single information sources are corrected, a receiving array can rotate precisely, the effect that the multiple information sources are corrected independently in a time division mode, and therefore a large number of samples are obtained; the phase position of an information source steering vector is estimated through a least square fitting method, array element data with large phase fitting errors are eliminated through an outlier elimination method, the phase position is estimated through the least square fitting method, and the azimuth information of the information sources is obtained; the errors of the array element data are corrected through a least square method. By means of the method for correcting the errors of the array element position, the errors of the array element position can be accurately corrected, and the method is simple and suitable for being applied to practical engineering.

The invention provides a satellite telemetry data outlier elimination pre-processing method, comprising: firstly, analyzing the structures of the telemetry frame of the satellite, determining the frame length, identifying fields such as frame synchronization heads, frame counts, check bits and check modes in the telemetry frame; secondly, creating a reading pointer, successively searching the frame synchronization heads backwards, reading frame data to cache according to the frame length, taking out the frame count and the check bits, finally, calculating check bits according to the check modes of the satellite telemetry frame, comparing, judging whether the frame count is legal according to a satellite telemetry download rule, comparing with a former frame, if the frame counts of the former and the next frames are continuous and the check bits are right, illustrating that the frame data are right data; if the frame counts of the frame data and the former frame data are discontinuous and the check bits are wrong, illustrating that the frame data are outliers and are eliminated. According to the invention, double conditions of the check bits and frame counts are compared, the outliers are judged accurately, the method is easily realized through programming, and the elimination efficiency is high.

The invention discloses a 3D point cloud denoising method based on statistical outlier and adaptive bilateral hybrid filtering, and the method comprises the steps: firstly calculating the average distance from each point in a 3D point cloud to the nearest K neighborhood point, the expected value of the average distance, a standard deviation and a distance threshold value; secondly, if judging thatthe average distance is larger than a distance threshold value, filtering is carried out, otherwise, performing retaining and 3D point cloud data obtained after primary denoising is obtained; and finally, calculating a bilateral smooth filtering factor, and processing the 3D point cloud data subjected to primary denoising by using the smooth filter to obtain final denoised 3D point cloud data. Astatistical outlier elimination filter and a self-adaptive bilateral filter are effectively combined to carry out denoising processing on 3D point cloud data. Results show that the method not only removes outlier noise points, but also removes fluctuating noise points. Meanwhile, the boundary of the denoised 3D point cloud data is smoother, and a good foundation is laid for subsequent segmentationand feature extraction of the 3D point cloud data.

The invention discloses a method for eliminating outliers based on polynomial fitting, which comprises the steps of: performing n-order polynomial fitting on the original measurement data to obtain a coefficient matrix and a fitting polynomial, and draw a rough For the scatter diagram, select the appropriate number n to perform the least squares polynomial fitting, and construct a function p(x) for the given measurement data (xi, yi) as an approximate expression of the given data (xi, yi), so that The sum of squares of error ri=p(xi)-yi is the smallest, that is, i is an integer from 0 to m. Based on polynomial fitting, this method invented a method for the computer to automatically remove the outliers in the measured data, and identified and eliminated the outliers in the observed data sequence by fitting the residual sequence of the estimated value and the observed value, which is very useful for practical engineering applications. important application value.

An embodiment of the invention provides an outlier elimination method and device. The method includes: acquiring target residual errors forming a residual error mean value of zero from a constructed Kalman filter residual error model in preset time; when the target residual errors accord with Gaussian distribution, acquiring a covariance matrix of the Kalman filter residual error model by the Kalman filter residual error model; taking a range formed by a first target value and a second target value as a target residual error range; taking the residual errors outside the target residual error range as outliers, and eliminating the outliers from the target residual error range. By adoption of the outlier elimination method, the problem of low reliability in fusion positioning can be solved.

The invention discloses a method for dynamically assessing a confidence degree of satellite telemetry data trend prediction. The method comprises the steps of 1) obtaining historical telemetry data information of a satellite telemetry parameter TX123 from a telemetry information historical database, performing outlier elimination processing on the historical telemetry data information of the satellite telemetry parameter TX123, and performing prediction to obtain a telemetry data information set; and 2) performing calculation to obtain confidence information, which comprises a confidence value, a confidence interval and a confidence degree. According to the method, a concept of the confidence information of the satellite telemetry data trend prediction is proposed and comprises the confidence value, the confidence interval and the confidence degree; through the confidence interval and the confidence degree, satellite telemetry data information trend prediction method and result can be dynamically assessed in real time; and compared with the prior art, the method has the advantage that the accuracy of the telemetry data trend prediction method can be quantitatively assessed by calculating the confidence degree.

The invention discloses a power plant circulating water pump predictive state evaluation method and system, and the method comprises the steps: carrying out the outlier elimination and denoising of original operation data of a circulating water pump unit, carrying out the classification of the denoised original operation data, and carrying out the training optimization of a circulating water pumpoperation state prediction model based on the classified original operation data; performing predicting according to the optimized circulating pump operation state prediction model to obtain predictedoperation data of the circulating water pump unit in a future time period, and calculating expected data of the circulating water pump unit in the future time period according to the predicted operation data by adopting an online state monitoring method; and knowing the equipment state according to the residual error of the expected data and the predicted operation data, and achieving predictivestate evaluation. A data-driven empirical model is used for solving the problems that a traditional physical model is difficult to establish and insufficient in precision, state prediction of the circulating water pump is visually and effectively achieved, conversion from an after-event maintenance system of the circulating water pump of the nuclear power plant to a predictive maintenance system is achieved, and therefore the safety and economical efficiency of nuclear power equipment operation are improved.

The invention discloses a three-dimensional point cloud outlier elimination method based on image segmentation, which belongs to the field of computer graphics, and comprises the following steps: setting a sampling interval, and performing uniformly sampling in a unit sphere parameter space to generate a projection direction; solving a transformation matrix according to a 313 rotation relationshipbetween the generated projection direction and a Z axis of a world coordinate system where the three-dimensional point cloud is located, and performing attitude transformation on the point cloud by utilizing the transformation matrix; calculating the image resolution of the three-dimensional point cloud after attitude transformation projected to the perspective projection virtual view; obtainingperspective projection virtual views of the three-dimensional point cloud in all projection directions; segmenting a main body part of the image in the obtained perspective projection virtual view byutilizing a main body extraction algorithm based on image segmentation; according to a visible shell technology, forming a convex hull of a three-dimensional point cloud by utilizing a side shadow contour line of a main body part in a perspective projection virtual view, and taking three-dimensional points except the convex hull of the three-dimensional point cloud as outliers to be eliminated.

The invention discloses a public energy consumption prediction method based on machine learning, and belongs to the field of energy consumption prediction. The method comprises the following steps: S101, collecting data; S102, data preprocessing: A, performing outlier elimination by adopting an MAD algorithm; B, replacing the missing value; and C, performing variable reduction by adopting a PCA algorithm; and S103, carrying out prediction modeling; a DNN deep neural network is adopted for calculation, and the DNN and more hidden layers are used together in an R software tool in a Keras libraryby utilizing the collected data. The provided method overcomes the current situation that in the prior art, public energy consumption prediction still lacks effective means, and mainly solves the problem of how to integrate a big data platform and machine learning into an intelligent system for managing energy. Efficiency of a public department is an important component of a smart city concept, and the method is implemented in an existing MERIDA intelligent system and provides accurate energy consumption prediction for users.

The invention relates to a self-adaptive PCA (Principal Component Analysis) model establishment method. A self-adaptive outlier elimination method is added into a traditional PCA model, the distance square of a projection vector-residual vector of training data in a residual space and a monitoring limit threshold are used as outlier detection bases, namely when the principal component analysis Q statistical magnitude of certain sampling data is higher than the monitoring limit threshold, the data is considered to be abnormal and determined to be an outlier and should be eliminated from a training matrix, abnormal data in all cycles is eliminated by adopting a self-adaptive cycle method, and the training matrix is repeatedly compressed until the monitoring index Q statistical magnitudes of all the samples of the training matrix are lower than the threshold. Meanwhile, the invention further relates to an ultrasonic machining partial discharge detection method based on self-adaptive PCA. The threshold of the monitoring index Q statistical magnitude is greatly reduced through outlier elimination, gross error data caused by partial discharge in the ultrasonic machining process can be well detected, the detection sensitivity and the detection efficiency are improved, and the performance of ultrasonic equipment is reflected in time.

A multi-satellite communication time difference positioning data fusion processing method disclosed by the invention is high in time difference positioning speed and high in precision. According to the technical scheme, after multi-satellite communication radiation source time difference positioning data are obtained, all positioning points are classified according to target numbers through a mirror image point elimination module, mirror image points in a positioning point classification result are eliminated through a mirror image point elimination algorithm, and non-mirror image points are output to a positioning point aggregation module; a normal point aggregation module performs trace point aggregation according to a trace point aggregation algorithm to generate aggregation points, and outputs the aggregation points to an outlier point elimination module; the outlier point elimination module performs outlier point elimination according to an outlier elimination algorithm, and outputs non-outlier points to an aggregation point filtering module; and the aggregation point filtering module filters the aggregation point traces based on an aggregation point filtering algorithm, performs fusion processing by using a covariance intersection algorithm and a k-nearest neighbor sliding window least square algorithm, and generates a final fusion result.

The invention provides a topological map node generation method based on laser point cloud distribution characteristics, and belongs to the technical field of robots. According to the invention, the method comprises the steps: acquiring three-dimensional point cloud data of a scene by using a three-dimensional laser sensor, dividing the three-dimensional point cloud data into two parts of ground point cloud and non-ground point cloud, and performing outlier elimination and clustering processing on the non-ground point cloud data so as to acquire the number of point cloud categories; performingfeasible region extraction on the ground point cloud, and performing straight line fitting on the extracted feasible region boundary so as to judge whether the current position is a corner or an intersection. A cost function is constructed according to the size of the environment range described by the acquired three-dimensional point cloud data, the number of non-ground point cloud categories, whether the current position is a corner or an intersection or not and the distance between the current position and the previous node, and meanwhile, whether the current position of the robot can generate a topological node or not is judged according to the cost value.

The invention discloses an adaptive image optical flow and RTK fusion attitude determination method. Image optical flow attitude measurement data and RTK attitude measurement data are fused by using adaptive Kalman filtering, so that error accumulation of image optical flow attitude measurement is effectively inhibited; and the image optical flow attitude measurement system is used for assisting RTK attitude measurement, so that the stability of the system and the attitude measurement precision are enhanced. By using a contrast-limited histogram equalization processing method, a cost function-based outlier elimination method and an improved optical flow motion equation, the precision of system attitude measurement is effectively improved.

The invention discloses a soft measurement method for the final-stage exhaust enthalpy of a steam turbine, and the method comprises the steps: carrying out the collection and storage of historical operation data of key characteristic variables which affect the final-stage exhaust enthalpy value of a steam turbine unit, carrying out the outlier elimination, screening of steady-state working conditions, and determining the final-stage exhaust enthalpy value of the steam turbine through the combination with an XGBoost algorithm which is subjected to particle swarm optimization (PSO). The soft measurement model for the last-stage exhaust enthalpy of the steam turbine can realize the online accurate measurement of the last-stage exhaust enthalpy value of a unit, and has important significance on the online real-time monitoring of the economic performance of the steam turbine.

The invention provides an outlier elimination method and system based on a fuzzy prediction system, and a computer related product, and solves the problem that the outlier identification accuracy and real-time performance are still not ideal because the prior art cannot adapt to the diversity of flight parameter changes. The outlier elimination method based on the fuzzy prediction system comprises the following steps: 1) constructing a time sequence fuzzy prediction system at a moment k on line; 2) obtaining a predicted value at a k + 1 moment based on the time sequence fuzzy prediction system, and calculating a residual error between the predicted value and an observed value; 3) determining whether the residual error at the k + 1 moment is an abnormal value or not according to the Dick's criterion by using the residual error sequence before the k + 1 moment; if the residual error at the k + 1 moment is an abnormal value, determining that the current observation value is an outlier, and removing the outlier; otherwise, reserving the observed value and the residual error. The method has no special requirements on signals, is suitable for outlier elimination of various signals, and is wide in application range.

The invention particularly relates to a signal-to-noise ratio estimation method applicable to a medium-voltage carrier system, which comprises the following steps of: firstly, performing FFT on a timedomain sequence of a receiving end of the medium-voltage carrier system; then calculating first-order, second-order and fourth-order statistics through first Nt pieces of data; secondly, calculatinga threshold value with outlier removed based on the Pauta criterion and a statistics result, wherein the first Nt outliers do not need to be removed; combining each frame of data, performing outlier judgment on subsequent data, removing data which do not meet conditions; finally iteratively calculating non-average second-order and fourth-order statistic information, and introducing an intermediatevariable to simplify the calculation of a signal-to-noise ratio. According to the method, system adjustment is not needed, equalization is not needed, and the signal-to-noise ratio of each subcarriercan be accurately estimated through outlier elimination.

The invention discloses a method for eliminating outliers in track measurement data. The method comprises the following steps: acquiring the track measurement data; adopting a kalman filtering innovation chi-square test method to identify outliers in the track measurement data; wherein the outliers comprise spot type outliers; and for more than a set number of spot-type outliers, adopting a cubic spline interpolation Spline algorithm and combining a kalman filtering algorithm to remove the spot-type outliers. On the basis of adopting an innovation chi-square test, outlier kalman identification effect analysis is given, and different outlier elimination schemes are given according to different classifications of isolated outliers and spot outliers; isolated outlier elimination needs to be realized by correcting a gain matrix K (k); and the spot type outliers are eliminated by adopting a Spline point supplement mode after kalman filtering identification. Error characteristics of outliers of isolated points and outliers of spots are obtained from divergence reasons of Kalman filtering, and a method for identifying the outliers is given, so that a relatively accurate quantitative definition is provided, and a scientific elimination method is established.

The invention discloses a load single-machine equipment state discrimination method driven by telemetering original data. The method comprises the following steps: receiving telemetering original datadownloaded by a satellite; according to a telemetering parameter configuration file, extracting subpackage telemetering data corresponding to the load single-machine equipment from the telemetering original data; inputting the subpackage telemetering data into a pre-trained load single-machine equipment discrimination model to obtain a load single-machine equipment state result; the load stand-alone equipment discrimination model comprising a telemetry parameter formatting processing module, an outlier elimination module and a load stand-alone equipment discrimination module. The telemeteringparameter formatting processing module is used for formatting subpackage telemetering data according to a processing strategy to obtain telemetering parameter formatted data; the outlier eliminationmodule is used for eliminating outliers in the telemetry parameter formatted data; and the load stand-alone equipment discrimination module is used for outputting a load stand-alone equipment state result according to the remote measurement parameter formatted data.

The invention provides a satellite-borne equipment health prediction method based on data self-adaption, which aims at solving the problems of realizing trend prediction of parameters and calculation of performance degradation conditions by utilizing long-term monitoring data of satellite-borne equipment, and comprises the following steps of: selecting historical parameter data of the satellite-borne equipment in a specified time period; carrying out outlier elimination, null value filling and data sampling on the historical parameter data; performing data point-to-point comparison on the sampled data to obtain a comparison value sequence corresponding to the sampling sequence; analyzing the period of the comparison value sequence, and after the comparison value sequence of the data is obtained, decomposing the comparison value sequence into a periodic component, a trend component and a random component according to a time sequence decomposition method; and performing long-term trend prediction on the periodic component, the trend component and the random component by using corresponding prediction methods, adjusting parameter values in various prediction methods according to the period of the data, and after the prediction results of the three types of data are added, comparing the prediction data with the original data to realize the attenuation performance of the equipment parameters.