31 results about "Nonlinear excitation" patented technology

The invention relates to a method for calculating DC (direct current) magnetic bias dynamic leakage inductance of a transformer. The method is characterized in that: the method comprises a dynamic parameter calculation step based on a transient field-circuit coupling model, and a dynamic leakage inductance calculation step based on transient field-circuit coupling; the method can establish a three-dimensional finite element magnetic field model of the transformer and solve the model to fully reflect the accuracy of magnetic field calculation under nonlinear excitation of the transformer; and the method can calculate dynamic inductance parameters according to the energy perturbation principle, calculate a transient current by combining with four-order Range-Kuttle method, and then calculate the dynamic leakage inductance based on a T-shaped equivalent circuit of the transformer. In comparison with the existing harmonic calculation method, the method provided by the invention can calculate the dynamic leakage inductance by considering the transient changes of equivalent parameters of magnetic field and circuit of the transformer and the relationship between the parameters under different DC perturbations on the basis of field-circuit coupling. Accordingly, the method provided by the invention has higher adaptability, and higher actual application value.

The invention discloses a nonlinear dynamics calculation method for a gear pair. The method comprises the steps that the feedback effect of vibration displacement to the dynamic contact condition of agear surface of the gear pair in the actual running process is considered, and the nonlinear excitation of the dynamic meshing rigidity and dynamic comprehensive meshing errors is calculated. Dynamiccontact analysis of the gear pair and a system dynamics solution process are combined, and an excitation-response-feedback loop-locked calculation process of the system is achieved. According to themethod, the nonlinear dynamics phenomenon of the gear pair due to parameters such as gear surface errors and shape correction can be acquired, the dynamics behavior of the system in the dynamic contract process can be more truly simulated, and the calculation accuracy of dynamics is improved.

The invention discloses an identity recognition method based on a dual-channel convolutional neural network, and the method comprises the steps: carrying out the training of a neural network and the identity recognition, carrying out the comprehensive training and recognition through employing two time synchronization images: a face image and a whole body posture image, avoiding the cheating of asingle factor, and achieving the high anti-interference capability and high recognition accuracy. According to the method, feature data of two channels are connected in a weighted mode through a fullconnection layer, image feature data are obtained through a plurality of convolution layers and a pooling layer, finally the classification probability is obtained through a classifier, and the maximum probability is extracted to be compared with a set threshold value to determine the recognition result. Through multiple times of convolution feature map extraction, nonlinear excitation and poolingdimension reduction processing, the dual-channel convolutional neural network control data is more flexible, and the abstraction capability and the learning capability are stronger, thereby having abetter identification effect.

The invention discloses a main and auxiliary integrated fault locating method and system based on a gradient descent algorithm. In the fault locating method, the fault locating of a main and auxiliaryintegrated network model is realized based on the gradient descent algorithm, and meanwhile, based on the above method, a fault locating system composed of a data acquisition module, an analysis module and an accuracy evaluation module is constructed. The method is verified by an IEEE standard 33-node topology example. The example is to simulate the grid fault, calculate and preprocess multi-dimensional data measured by each node through the data acquisition module, establish a multi-layer neural network by using a neural network gradient descent algorithm idea, select an appropriate excitation function for nonlinear excitation, optimize the accuracy of model locating based on the gradient descent algorithm, and provide a fault branch locating result in a fault locating calculation module. The example simulation shows that the branch where the fault occurs can be located with high accuracy. The fault locating method effectively locates the branch of a grid where the fault occurs, provides an effective decision basis for the operation and repair of the grid, and reduces the economic losses.

The invention discloses a three-phase transformer electromagnetic transient low-and-medium frequency modeling method. The method comprises the steps of selecting iron core joints as magnetic circuit division points of a transformer model, and partitioning a transformer; building a magnetic circuit equivalent model according to magnetic conductive features of materials inside the transformer of allpartitioned areas and the spatial distribution of all parts of the transformer; conducting classifying according to physical properties represented by all magnetic resistance elements, establishing acircuit model according to relevant parameters and electromagnetic duality principles of the elements and conducting simplifying according to a topological connection relation; adopting a dynamic loss model according to loss characteristics of magnetic conductive materials of the transformer to simulate the loss of a transformer iron core; combining the circuit model and the dynamic loss model, wherein the combined model is used for simulating nonlinear excitation and loss features of the transformer iron core; calculating the parameters of the circuit part of the transformer, and adopting alinear transformer model for simulation; connecting the combined model for representing the transformer iron core transient characteristics to the low-voltage side of the linear transformer model; calculating the parameters of the transformer.

The invention discloses a microbubble synergistic dynamic experiment system based on a double frequency superimposed ultrasonic pulse in a viscoelastic medium and a microbubble synergistic dynamic experiment method based on the double frequency superimposed ultrasonic pulse in the viscoelastic medium. The experiment method comprises the following steps 1) according to an actual waveform of a focal region sound field measured by a hydrophone, setting KZK equation parameters and establishing a microbubble nonlinear excitation waveform; 2) constructing a HIFU synergistic model according to viscoelasticity of a biological tissue and vibration characteristics of the microbubble in compressible fluid; 3) establishing a microbubble synergistic dynamic method based on double frequency superimposition; 4) in combination with simulation results of steps above, determining a HIFU waveform phase, sound pressure and frequency parameters to perform a thermal ablation or tissue damage experiment. According to the system and the method provided by the invention, the dynamic characteristics of the microbubble in the biological tissue are used fully to design the sound wave waveform and the phase parameter, maximum vibration of the microbubble is triggered at a focal region, a cavitation effect is enhanced, and safety and efficiency of HIFU thermal ablation and tissue damage are improved.

The invention is a DSP-based machine nonlinear excitation control device, and its character: it uses machine control special DSP chip as CPU and PLD as auxiliary processor, and consists of AC sampling circuit, CPU, pulse amplifying circuit, controlled silicon power circuit and switching quantity control circuit. It uses high-performance DSP chip to implement multiple generator excitation control laws of multiple generators relying mainly on nonlinear control and can largely enhance the speed and precision of excitation control, reduce control delay, making the whole system of excitation control have very good adjusting performance and reliability, enhancing network stability, assuring safety working of generator, reducing large-area power failure rate, and it is a new-generation generator excitation auto-control device with a huge market prospect.

An embodiment of the invention provides an optical microimaging device and method based on line scanning temporal-spatial focusing. The device comprises a laser transmitting module, a line scanning module, an optical microscopic module, a synchronous filtering module and a reconstruction module, the laser transmitting module generates nonlinear excitation pulse light, the line scanning module acquires angular dispersion and deflection angle of the excitation pulse light and enables the excitation pulse light to scan a sample, the optical microscopic module forms a focusing line focusing on twodimensions like space and time at the same time in the sample to excite the sample on the basis of nonlinear optical effect to generate a transmitting light signal; the synchronous filtering module detects the transmitting light signal of the sample and filters the transmitting light signal, and the reconstruction module performs two-dimension reconstruction according to the transmitting light signal detected by the synchronous filtering module to acquire a reconstruction image of the sample. The optical microimaging device can effectively inhibit signal crosstalk caused by stray light, has the advantages of wide field of view, high spatial resolution, high temporal resolution and high penetration depth and can provide rich information for biological dynamic process study and disease diagnosis basis.

The invention provides a method and a system for testing the nonlinearity degree of a closed-loop feedback circuit of a fiber-optic gyroscope. The method comprises the steps that a digital signal processing chip generates a modulation voltage signal, and the modulation voltage signal is output to a Y waveguide phase modulator through a digital-to-analog converter and a modulation wave driving circuit, wherein the modulation voltage signal is a superposed signal of a nonlinear excitation modulation wave signal and a fixed step wave signal; the Y waveguide phase modulator is used for convertingthe modulation voltage signal into a modulation phase signal to obtain a modulation phase difference; a photoelectric detector detects the modulation phase difference, and the modulation phase difference is output to the digital signal processing chip through a preamplifier and an analog-to-digital converter; the digital signal processing chip calculates and outputs a first demodulation value according to the modulation phase difference; and a determination module determines the nonlinear degree of a modulation coefficient of the closed-loop feedback circuit according to the first demodulationvalue. The closed-loop test of the fiber-optic gyroscope is realized, so that the accuracy of the nonlinearity test of the feedback circuit is improved.

The invention discloses a low-complexity convolutional neural network based on symbol random calculation. The system comprises a symbol random calculation module and an efficient convolution layer, apooling layer and a nonlinear excitation layer which are constructed based on the symbol random calculation module, the symbol random calculation module comprises a forward conversion module and a backward conversion module, the efficient convolution layer comprises a random processing unit and a random adder matrix, and the pooling layer is connected to the output end of the efficient convolutionlayer. According to the invention, a special sign bit is introduced; symbol random calculation is formed, the problem that the traditional random calculation can only represent positive numbers is solved; the calculation error caused by scaling coding is avoided, a novel parallel fast convolutional neural network architecture is provided based on the random calculation mode, the throughput rateof random convolution is increased while the hardware consumption is reduced, and the delay of random convolution calculation is reduced to a certain extent.

The invention discloses an extreme learning machine face dimension reduction method based on discriminant shared neighborhood preservation. The method comprises the following steps: S1, image preprocessing: performing scale normalization processing on a face image acquired by a face image acquisition module; s2, an input layer: inputting a face feature X subjected to scale normalization processing; s3, a hidden layer: selecting a nonlinear excitation function, randomly and nonlinearly mapping the single-sample face feature X to an N-dimensional feature space, and transforming to obtain a feature H (X); s4, outputting a layer: fixing a feature H (X); and S5, a sample data storing the face sample data t subjected to dimension reduction processing. The invention provides an SGRD-ELM algorithm. The SGRD-ELM not only considers the category information of the sample, but also pays attention to mining the shared neighbor information of any two sample points, and under the condition that the dimension of the sample data is reduced to 64 dimensions, the human face recognition rate is averagely improved by 10%-14% compared with that of an LLE method, and the human face recognition rate is averagely improved by 6%-8% compared with that of a US-ELM method.

The invention discloses an underwater target sonar detection method based on a CNN. The method specifically comprises the following steps of: S1, carrying out 360-degree sonar data collection on an underwater target through an active sonar technology to obtain original sonar data; S2, preprocessing the original sonar data to obtain a sonar image; S3, expanding the sonar image to obtain a sonar image data set, and dividing the sonar image data set into a training set and a test set; S4, training a CNN network model by using the training set; and S5, inputting the test set into the trained CNN model for testing to obtain a detection and recognition result. According to the method, a convolutional neural network with a single convolutional layer is adopted, a nonlinear excitation function Sigmoid is used during model training and is replaced with Softplus during actual testing, so that the signal intensity of the detection target is better displayed on the basis of simplified training.

The invention discloses an active damping device for a solid rocket engine, which comprises a test engine main body and an engine supporting structure, and the engine supporting structure supports the test engine main body; the main body of the test engine is a projectile body, the shell is of a porous sandwich layer structure and sequentially comprises an outer shell layer, a porous structure layer and an inner shell layer from outside to inside, and a heat insulation layer and a propellant are arranged in the inner shell layer; according to the engine supporting structure, one end of a sliding rail is connected with triangular leaning iron and installed on a base; one end of the thrust frame is connected with the triangular leaning iron, and the other end is connected with the front sealing head; the limiting stopper and the rear hanging bracket are connected with the sliding rail through the sliding block, one end of the spraying pipe of the shell is hung on the rear hanging bracket, and the limiting stopper is located between the thrust frame and the rear hanging bracket. Combustion chamber pressure and structural vibration caused by nonlinear excitation of an engine in a free trajectory can be weakened to a great extent, and the stability of the engine is enhanced.

The embodiment of the invention provides an information generation method, device and apparatus based on a fuzzy neural network, and the method comprises the steps: carrying out principal component analysis of an obtained dimension parameter, and determining a feature parameter; smoothing the characteristic parameters according to a nonlinear excitation function, and determining service parameterscorresponding to the characteristic parameters; processing the service parameters by using a fuzzy mapping principle to generate to-be-recommended information; and establishing a mapping relationshipbetween the to-be-recommended information and the feature parameters so as to generate recommendation information. The characteristic parameters are determined by using the principal component analysis method, and the fuzzy neural network is constructed by using the nonlinear excitation function, so that the service information suitable for the user is calculated and output according to the characteristic parameters, the neural network is constructed by using the fuzzy mapping principle, and the problem of inaccurate linear matching of a single user label for massive user data at present is solved. The purpose of reducing the labor cost can be achieved while the accuracy and efficiency of information recommendation are achieved.

The invention belongs to the field of semiconductor lasers. At present, the research on the influence of laser defects on the laser performance mainly focuses on the spectral characteristic analysis after the laser damage occurs; the change of the free electron density in the laser cavity surface material is not microscopically analyzed from the perspective of physics; the invention provides a method for calculating a damage threshold of a laser cavity surface. The incident laser irradiates the lower cavity film to generate photoionization, collision ionization and free electron attenuation; the damage threshold value of the laser cavity surface is analyzed and calculated by calculating the critical free electron density of the laser cavity surface, the influence rule of cavity surface material nonlinear excitation on free electrons is theoretically analyzed, and a theoretical model for calculating the cavity surface laser-induced damage threshold value is established.