18 results about "Linear methods" patented technology

The invention discloses a chemical data missing process fault detection method and system, and the method comprises the steps: converting two-dimensional data into a three-dimensional tensor through employing MDT, and carrying out the linear prediction and nonlinear prediction of an incomplete value through employing linear smooth CP and CP-SAE, and carrying out the reconstruction of normal data. And extracting a residual error and a feature space of the data in the data missing process through CP-SAE. And combining features extracted by smooth CP-decomposition and CP-SAE, and establishing three statistical magnitudes to realize fault detection. According to the method, the MDT and smooth CP decomposition method is used for solving the problems of data missing and time delay of data sample sampling, and the timeliness is high. The CP-SAE feature extraction robustness is high, and compared with a conventional linear method, the precision is high, and the response speed is high. The method is simple to operate, does not need repeated operation, and can extract the key features of the data while reconstructing the complete data set only by operating on the incomplete data set. The missing data can be effectively supplemented, the process monitoring requirement can be met, and the precision is high.

The invention discloses a storage and calculation integrated method and device for matrix exponentiation, and the method comprises the steps: inputting a to-be-exponentiated matrix A and a power k into a storage and calculation integrated circuit; an optimal operation sequence for calculating the k-th power of a matrix A to be exponentiated is determined through a dynamic programming algorithm, the optimal operation sequence is combined with a linear method and a flat method to minimize the total calculation time, and corresponding matrix multiplication operations are sequentially executed in a storage and calculation integrated circuit according to the optimal operation sequence, completing the calculation of the k-th power of the matrix A to be exponentiated to obtain a calculation result; and outputting a calculation result.

The invention discloses a deformation error compensation method for a dilator in total knee arthroplasty, and aims to solve the problem of measurement errors caused by deformation of a cantilever structure. The method comprises the following steps of: firstly, performing parameterized dynamic modeling and simulation on an external execution mechanism through finite element software, calibrating a model by combining actual measurement data of a physical prototype, and constructing a high-precision deformation data set; then, the data set is used for training a BP neural network optimized through a genetic algorithm, and the network takes the applied force and the displacement measurement value which are collected in real time as input and outputs a real displacement value after compensation. And finally deploying the trained lightweight network model in an embedded processing unit of the spreader to realize real-time error compensation. According to the method, complex nonlinear deformation errors are effectively compensated, the average absolute error is reduced to 0.056 mm, the real-time response delay is smaller than or equal to 5 ms, the method is remarkably superior to a traditional linear method, more accurate gap measurement is provided for an operation, and the method can be further used for forming closed-loop control and improving the intellectualization and precision of the operation.

This application relates to the field of simulation and calculation technology, and in particular to a nonlinear galloping simulation, risk assessment method and medium for ice-wind composite contact networks. By introducing an unsteady aerodynamic correction term related to the rate of change of angle of attack and a geometric stiffness matrix updated in real time with tension, this application successfully constructs an unsteady aerodynamic negative damping mechanism, which can simulate the negative aerodynamic damping effect that traditional linear methods cannot capture. This significantly improves the prediction accuracy of the galloping start-up wind speed and the maximum amplitude, and solves the technical problem that traditional quasi-steady linear models cannot accurately predict the large-amplitude galloping characteristics of ice-covered contact networks. It realizes accurate simulation and assessment of start-up wind speed, galloping amplitude and risk level, and is suitable for the analysis of the survivability of contact networks under extreme ice-wind conditions.

The invention discloses a nuclear power sensor data reconstruction method and system based on a Seq2Seq network model, and belongs to the technical field of intelligent operation and maintenance of nuclear power sensors. Firstly, steady-state normal operation data and fault state data are obtained through a PCtran nuclear power station simulation platform; a copula entropy is used as a correlation measurement method, and a sensor combination (except the target sensor itself) associated with the target sensor is identified under a steady-state working condition; based on a correlation structure, training the established Seq2Seq model by using associated sensor combination and target sensor data in normal data; and performing data reconstruction of the target sensor data under the fault state on the Seq2Seq model by using the associated sensor combination in the fault data and the target sensor data. According to the method, the limitation of a traditional linear method is broken through, the indicating value of the target sensor under the fault condition is reconstructed by utilizing a sensor combination associated with the target sensor, and meanwhile, the model is combined with the two-way LSTM to capture the spatial-temporal characteristics, so that end-to-end high-precision modeling is realized.

The invention provides a chemical component analysis and quality targeted improvement method for tobacco leaves in a high-quality producing area and an adjacent producing area, which comprises the following steps of: acquiring content data of 70 kinds of constant and semi-trace chemical components in the tobacco leaves, obtaining variables with high significance by using step-by-step screening, and removing high-correlation variables from the variables; a discrimination model of the high-quality production area and the adjacent production area is established through multiple linear methods, the component importance is determined according to a discrimination function, the result is more stable and accurate, and the credibility is high. The invention also provides a quality targeting improvement method, which comprises the following steps: searching high-correlation chemical components in 70 constant and semi-trace chemical component data based on modeling chemical components, and establishing a high-correlation component list; and preparing a quality improvement solution based on the high-correlation component list to regulate and control the components or the high-correlation components of the components in the tobacco leaves so as to achieve the purpose of improving the quality of the tobacco leaves.

The application relates to the air conditioning technical field, and provides an air conditioner refrigerant quantity prediction method, an air conditioner and a storage medium, which comprise the following steps: acquiring all key parameter data corresponding to different refrigerant quantities of an air conditioner; the key parameter data comprises suction superheat data and discharge superheat data; all the key parameter data is segmented to form a plurality of segments; the ratio of the key parameter data corresponding to each segment to all the key parameter data is calculated; the ratio is input into a refrigerant quantity prediction value model to calculate a predicted refrigerant quantity; and the refrigerant quantity prediction value model is obtained by training based on historical key parameter data. The patent adopts a multi-dimensional linear method based on a large amount of random data, adopts the proportion of key parameters for analysis, and has higher accuracy.

The invention relates to the technical field of computer vision and signal processing, in particular to a multidimensional color image filling method and system based on quaternion tensor. According to the method, RGB values of an image are coded into pure quaternions through quaternion tensor representation, the pure quaternions are organized into a quaternion tensor, then inherent complex nonlinear characteristics in visual data are captured through nonlinear transformation in a quaternion domain, then two novel regularization terms are introduced by constructing a unified optimization model, and finally, the visual data are obtained. According to the method, global low-rank performance and local smoothness are jointly coded, finally, the model is solved through a nonlinear alternating direction multiplier method, effective fusion of quaternion representation, nonlinear transformation, low-rank performance and smoothness prior is achieved, and the model has higher expression ability and robustness compared with a traditional linear method. Therefore, higher-quality and more stable multi-dimensional color image filling is realized.

The application discloses a non-harmonic blade dynamic stress field reconstruction method, system, medium and equipment, and relates to the technical field of blade vibration monitoring. The application firstly ensures high fidelity of a numerical model by minimizing the deviation of measured vibration displacement and finite element analysis vibration displacement, so as to obtain a multi-order modal dynamic stress field formed by multi-order modal dynamic stresses on each grid node of a non-harmonic blade model through finite element analysis, and determine a mistuning coefficient of vibration frequency of the non-harmonic blade as a non-harmonic quantity according to a reference frequency under a designed blade motion state, and then depth learning is performed on a complex non-linear mapping relationship of full-field multi-order modal dynamic stresses from a local tip position to the full field; when applied, the multi-order modal dynamic stress field of the target non-harmonic blade can be obtained by coupling the multi-order modal, the distortion problem of the linear method in processing the complex vibration mode of the non-harmonic blade is effectively overcome, and the precise prediction of the multi-modal coupling dynamic stress field of the target non-harmonic blade in actual operation is realized.

The application discloses a dynamic range compression method for improving image flicker, and proposes a double-threshold judgment method. Compared with a conventional linear method, the double-threshold judgment method increases a flicker threshold, and the continuity between frames can be utilized to make the image mapping curve in the linear mapping process not abrupt. By utilizing the double-threshold judgment method, the application can reduce the sharp change of frame brightness when the scene brightness is abruptly changed, and improve the anti-flicker capability of the picture when the environment is abruptly changed.

The invention discloses a lightning arrester electromagnetic transient modeling method and a lightning arrester electromagnetic transient modeling system based on piecewise linear fitting, and solves the technical problem that response lag of a model is easy to occur in a rapidly changing transient scene due to an existing lightning arrester electromagnetic transient modeling method. The method comprises the following steps: acquiring a multi-working-condition voltage and current sampling sequence of the lightning arrester, pre-processing to obtain a sequence after working condition sorting, and constructing a voltage and current characteristic curve; future, current and historical sampling voltages of each working condition are determined through electromagnetic transient simulation; segmenting the characteristic curve by adopting a piecewise linear method and calculating the basic slope of each interval; and finally, constructing an electromagnetic transient model of the lightning arrester by combining the voltage at the three moments, the segmented interval and the basic slope.

The application discloses a power system dynamic observability calculation method based on a data-driven Koopman operator, belongs to the technical field of power system dynamics and observability analysis, and comprises the following steps: establishing a power system dynamic observability analysis model based on system state variables and measured variables; converting the established dynamic state observability model into an augmented Koopman canonical coordinate system; constructing an observation matrix of the system in the augmented Koopman canonical coordinate system; and calculating the dynamic observability degree of the power system by using a matrix index in the Koopman space. The method of the application uses the data-driven Koopman operator, overcomes the inaccuracy of the model while considering the nonlinear characteristics of the power system dynamic model, effectively solves the low-precision problem of the traditional linear method and the problem of excessive calculation amount of the nonlinear method, and can realize efficient and rapid analysis and calculation of the dynamic observability of the power system.

The invention provides an electricity purchase side settlement business ticket optimization matching method, device and system, and relates to the technical field of electric power grids. According to the method, key original features are extracted from original settlement data, a deep potential structure is mined through a nonlinear structure, a feature set with high characterization force is formed by fusing an explicit and implicit feature interaction relation, and finally accurate matching is completed by combining dynamic similarity measurement and an adaptive threshold value. The method effectively overcomes the limitation that a traditional linear method is difficult to process a complex non-linear relation, improves the automation degree and accuracy of matching, reduces manual intervention and subjective errors, solves the technical problem that the matching of settlement business tickets is inaccurate by using the linear method, and improves the matching efficiency. And a reliable technical support is provided for rapid and accurate processing of the settlement service of the electricity purchasing side.

This application discloses a vibration analysis method, apparatus, medium, and equipment for an automotive electric drive system, relating to the field of electronic digital data processing technology. The method includes: analyzing excitation characteristics and gear meshing dynamic load parameters based on a dynamic model of the motor's electromagnetic excitation and gear system; using these as mechanical excitation to couple a model of the motor, gears, bearings, and housing to form a rigid-flexible coupled dynamic model; transforming this model into state-space equations; obtaining multivariable time-domain data through dimensionless transformation and piecewise linear methods; and then calculating the transmission efficiency of vibration energy among the motor, gears, bearings, and housing components based on the excitation characteristic parameters, gear meshing dynamic load parameters, and multivariable time-domain data. The key noise transmission paths of the automotive electric drive system are then located based on the magnitude of the transmission efficiency values. This application effectively solves NVH problems such as high-frequency howling under high-speed operating conditions by comprehensively analyzing the system's vibration characteristics.

This invention discloses a target localization method based on the trust region piecewise linear method, comprising the following steps: (1) Initializing the localization processing parameters. (2) Reading in the arrival angle and arrival frequency parameters of the pulse signal, as well as the coordinate position and speed of the ship. (3) Calculating the mean and standard deviation of the arrival angle and arrival frequency of the pulse signal, and normalizing the arrival angle and arrival frequency of the signal using the calculated mean and standard deviation. (4) Constructing an objective function using the normalized arrival angle and arrival frequency of the signal, and passing the coordinates of the initial iteration point into the objective function. (5) Iteratively solving the target coordinate estimation and the corresponding objective function using the trust region piecewise linear method. If the maximum gradient of the objective function on the current iteration estimated coordinates is less than the tolerance or reaches the upper limit of the number of iterations, the target coordinate estimation result of the current iteration is given.

The invention provides a multi-dimensional nonlinear LIME method for enterprise environment index prediction and interpretation, which comprises the following steps: step 1, based on a multi-dimensional enterprise environment training data set, training a CTGAN as a sampling model: a generator G is used for synthesizing a sample G (z, m) according to a condition vector m constructed by noise data z and discrete features, the discriminator D is used for judging the matching degree of a synthetic sample and a real sample, and a Warisstein generative adversarial network WGAN-GP with gradient penalty is used as a target function of the generative adversarial network; 2, generating a complex prediction model of disturbance samples and enterprise environment indexes based on a CTGAN model, and training an MNLIME interpretation model: for a to-be-interpreted sample X, synthesizing the sample by using a generator G, further filtering by using a discriminator D to obtain n disturbance samples Xi'close to the real distribution of data, and calculating the weight Pi (Xi ') of each sample based on a cosine distance; and obtaining a predicted value f (Xi ') of each disturbance sample, training the local interpretation model g, and obtaining a coefficient as each feature.

The invention discloses an output power automatic calibration method and system for a broadband radio station and a computer storage medium, and relates to the technical field of radio communication. The method aims to solve the problems of low power calibration efficiency, insufficient precision and lack of robustness in the prior art. The core of the method comprises the following steps of: after parameters are preset, circularly traversing each frequency point in a full frequency band; at each frequency point, firstly monitoring a standing-wave ratio (SWR) to ensure hardware security, and then obtaining actually measured power; dynamically selecting a linear method or a bisection method to intelligently adjust the ALC value by comparing the actually measured power with the target power until the power reaches the standard, and verifying the signal stability; and finally generating a test report. The invention further relates to a system for implementing the method and a storage medium. According to the invention, full automation of the calibration process is realized, the method has the advantages of high efficiency, good precision, strong safety and good robustness, and the production efficiency and maintenance convenience are significantly improved.