39 results about "Kernel partial least squares" patented technology

The use of machine learning for pattern recognition in magnetocardiography (MCG) that measures magnetic fields emitted by the electrophysiological activity of the heart is disclosed herein. Direct kernel methods are used to separate abnormal MCG heart patterns from normal ones. For unsupervised learning, Direct Kernel based Self-Organizing Maps are introduced. For supervised learning Direct Kernel Partial Least Squares and (Direct) Kernel Ridge Regression are used. These results are then compared with classical Support Vector Machines and Kernel Partial Least Squares. The hyper-parameters for these methods are tuned on a validation subset of the training data before testing. Also investigated is the most effective pre-processing, using local, vertical, horizontal and two-dimensional (global) Mahanalobis scaling, wavelet transforms, and variable selection by filtering. The results, similar for all three methods, were encouraging, exceeding the quality of classification achieved by the trained experts. Thus, a device and associated method for classifying cardiography data is disclosed, comprising applying a kernel transform to sensed data acquired from sensors sensing electromagnetic heart activity, resulting in transformed data, prior to classifying the transformed data using machine learning.

The invention relates to a decentralized process monitoring method, which comprises the steps as follows: in step one, data is acquired; in step two, acquired data is processed through blocking partition; in step three, partitioned data is mapped to feature space; in step four, a systematic process is processed through modeling; and in step five, failure detection and identification are performed by utilizing principal component statistic T<2> of observed data and residual error statistic SPE of observed data. The decentralized process monitoring method has the advantages as follows: the method is suitable for a complex large-scale systematic process; the advantages of kernel partial least square method and blocking partial least square method can be achieved; the complexity of nonlinear process analysis is reduced at the same time; and the identification capability is enhanced.

The invention discloses a soft measurement method for load parameters of a mill. The method adopts various different signal decomposition technologies to decompose the original cylinder vibration signal and vibration-sound signals into a series of sub-signals based on different perspectives for multi-scale and unsteady characteristics of the cylinder vibration and vibration-sound signals. The selected sub-signal spectrum and the original signal spectrum are taken as multisource multi-scale information to construct a soft measurement model for the load parameters of the mill. A global optimization selective ensemble kernel partial least squares (GOSENKPLS) based on an adaptive genetic algorithm (AGA) and a branch-and-bound (BB) algorithm is adopted to carry out optimization selection on structural parameters and learning parameters of candidate sub-models and a selective ensemble model (SEN model), so that the effectively selective fusion of the multisource multi-scale signals is realized. According to the method, the accuracy for soft measurement of the load parameters of the mill can be improved.

The invention relates to an industrial process fault diagnosis method based on a direction kernel partial least square. The method is characterized in that historical normal data of an input variable and an output variable of an industrial process is acquired, wherein a fault is easily generated in the industrial process; an operation based on the direction kernel partial least square is performed on the historical normal data; a control limit of Hotelling statistics of the historical normal data and a control limit of a squared prediction error of the historical normal data are calculated; sampling data of the input variable of the industrial process is collected and the operation based on the direction kernel partial least square is performed on the sampling data so as to acquire process monitoring statistics of the sampling data and a squared prediction error of the sampling data are obtained; when the process monitoring statistics control limit of the sampling data or the squared prediction error of the sampling data exceeds the control limit, the sampling data possesses one kind of fault; historical fault data of a known fault type is acquired; reconstruction based on the Hotelling statistics and reconstruction based on the squared prediction error are performed on the historical fault data of the known fault type and a fault type of the sampling data is determined.

The invention discloses a mill load parameter soft measuring method based on a virtual sample. The mill load parameter soft measuring method comprises the steps of acquiring a multi-dimension time domain sub-signal of mill cylinder vibration and vibration sound sample signal by means of ensemble empirical mode decomposition (EEMD) technology, and performing further processing for obtaining high-dimension spectral data with different time dimensions; then constructing a feasibility-based planning (FBP) model based on the high-dimension spectral data according to an improved selective integrated kernel partial least squares (IGASEN-KPLS) method, and generating a new virtual sample based on priori knowledge and an FBP model; then obtaining a mixed sampling model after mixing the new virtual sample with a true training sample, performing adaptive selection of a multiple-dimension spectral characteristic by means of a mutual information (MI) based characteristic selecting method, constructing a soft measuring model by means of the selected spectral characteristics and performing soft measurement.

The invention discloses a dynamic process monitoring method based on nonlinear autocorrelation rejection of variables, and aims to reject nonlinear autocorrelation of each measured variable of sampledata and establish a nonlinear dynamic process monitoring model on the basis. The method comprises steps as follows: firstly, a nonlinear regression model between the sample data and multiple delay measurement data of the sample data with a kernel partial least-squares-algorithm; secondly, a model error is taken as a new monitored object, the process monitoring model is established with a PCA (principal component analysis) algorithm, and fault monitoring is implemented. Compared with a traditional method, the method has the advantages that the error is taken as the monitored object, the error's capacity of reflecting the abnormal change condition of nonlinear autocorrelation characteristics is utilized, furthermore, by means of absence of time sequence autocorrelation of error data, convenience is provided for following establishment of the process monitoring model based on the PCA algorithm. Therefore, the method is more suitable for the dynamic process monitoring.

The invention discloses a K-PLS regression model based traditional Chinese medicine tongue image color correction method and belongs to the fields of computer image processing and traditional Chinese medicine inter-discipline. According to the kernel partial least square regression model based traditional Chinese medicine tongue image color correction method, tongue images in different light environments are corrected to be at unified standard so as to overcome the problem that tongue image colors are inconsistent in showing due to illumination difference of a tongue image instrument and enable a digitalized traditional Chinese medicine tongue image analysis result to be objective and accurate. By means of the K-PLS regression model based traditional Chinese medicine tongue image color correction method, the tongue image instrument is utilized to conduct shooting and acquisition on a standard color chart, then color code samples of color codes in images are selected, standard chromatic values of all colors in the color codes are used as target values to perform K-PLS regression model training so as to obtain a training model for a color code actual acquisition value and the color code standard value, the training model is used for conducting color regression model on actual tongue images shot in the shooting environments, and finally traditional Chinese medicine tongue images with corrected colors are obtained.

The invention discloses a near infrared spectrum (NIR) detection method for insulation coating layer thickness of silicon steel. The method is characterized in that an acousto-optic tunable filter (AOTF) NIR analyzer is used for acquiring the NIR of an insulation coating layer of a silicon steel surface, a sample database containing insulation coating layer NIR and coating layer thickness standard value is created, an insulation coating layer thickness analytical mode which is composed of a pre-processing module and a kernel partial least square (KPLS) module is created based on the insulation coating layer sample database, NIR data of the sample database is input into the pre-processing module, the output of the pre-processing module is input into the KPLS module, the thickness value of the insulation coating layer is output after processing, and acquired NIR data of the insulation coating layer of the silicon steel surface is input into the insulation coating layer thickness analytical mode to obtain the thickness value of the insulation coating layer. The invention expands the application field of the NIR analyzer, realizes the thickness detection of semi-organic insulation coating layer of silicon steel surface, has the advantages of high speed, accuracy and reliability, and the like.

The invention discloses a software failure time forecasting method based on a kernel partial least squares regression algorithm. Through the application of a kernel function technology, the problem of software reliability forecast is converted to the problem of recession estimation, and the kernel partial least squares regression algorithm is used for resolving the problem of the software reliability forecast. Through fully consideration of a small sample property of the software reliability forecast, the situations that the size of observational variables is bigger than that of observational samples and multicollinearity exists among the variables can be overcome by using the kernel function technology, and so that a model 'overfitting' situation arises in modeling approaches such as a neural network does not occur. By means of the software failure time forecasting method based on the kernel partial least squares regression algorithm, model parameters are automatically and continuously adjusted to fit the dynamic change in a failure process, therefore adaptive forecasting of the software reliability is achieved, and the adaptive capability of a software failure forecasting model is improved effectively.

The invention belongs to the technical field of convexity prediction, and particularly relates to a plate convexity prediction method based on kernel partial least squares (KPLS) and a support vectormachine combined. The plate convexity prediction method comprises the following steps: S1, field data are collected through a high-precision monitoring device; S2, the collected data are preprocessed;S3, a KPLS regression prediction model is established; and S4, a KPLS-SVM plate convexity prediction model is established. By taking a data-driven algorithm as a mathematical tool, abnormal values can be removed from a large quantity of collected field rolling process data, a convexity prediction model for a strip steel continuous-rolling plate based on a KPLS method and the support vector machine combined is established, the convexity of the strip steel continuous-rolling plate is predicted, the established model is optimized through a particle swarm optimization algorithm, and the prediction precision of the convexity of the strip steel continuous-rolling plate is further improved. The plate convexity prediction method is used for predicting the convexity of the strip steel continuous-rolling plate.

The invention discloses a target tracking algorithm based on an image set. The target tracking algorithm based on the image set comprises an initialization stage, a training stage used for building a target model, a test stage used for testing the central position of a specific target object according to the built target model and an updating stage. According to the target tracking algorithm based on the image set, the image set serves as a basic unit of a training sample and a test sample in the target tracking process, the target tracking problem is considered as a multi-class classification problem under the framework of the kernel partial least square method, and therefore the problem of drifting of a tracker can be effectively relieved.

The invention discloses a simplifying soft measurement method for a primary variable in a production process integrating KPLS (Kernel Partial Least Squares) and FNN (False Nearest Neighbors). The method is characterized by comprising the following steps: determining n original auxiliary variables possibly related to the primary variable, collecting value data of the n original auxiliary variables and the primary variable, and forming a sample set; respectively calculating weighted values of the n original auxiliary variables by using a method of integrating the KPLS and the FNN; forming an original auxiliary variable sequence; modeling and determining the best auxiliary variable according to a minimum mean square error (MSE); and acquiring a simplifying soft measurement model. According to the method, an auxiliary variable set containing the auxiliary variables with the least number can be found for modeling the primary variable on the basis of the best modeling effect, so that the simplifying soft measurement on the primary variable can be realized.