318 results about "Nonlinear system" patented technology

A model predictive controller (MPC) for predictive control of nonlinear processes utilizing an EKF (Extended Kalman Filter) and a nominal trajectory generator. The nominal trajectory generator includes another instance of EKF, a linear corrector, and a time-varying deviation model. A nominal control trajectory can be predicted and an optimal solution for the time-varying deviation model can be computed based on an approximation of a system inverse known as signal de-convolution. The EKF can be utilized to estimate a current process state by supplying a measured output and to predict a future nominal trajectory by supplying a reference output. A Kalman smoother can also be utilized for the signal de-convolution in order to obtain enhanced trajectory estimates.

An optical control system is described which is capable of maintaining and optimizing a fiber-optic transport system within it's domain of control while interacting with other optical systems which are controlled independently. This allows the optical system to be incorporated as a building block into a larger optical network in a relatively arbitrary fashion. This provides an underlying control system for a non-linear system like optics network that is flexible and extensible.

A silicon based source for radiation in the 0.5-14 Terahertz regime. This new class of devices will permit continuously tunable, milli-Watt scale, continuous-wave, room temperature operation, a substantial advance over currently available technologies. The Silicon Terahertz Generator (STG) employs a silicon waveguide for near infrared radiation, situated within a metal waveguide for Terahertz radiation. A nonlinear polymer cladding permits two near-infrared lasers to mix, and through difference frequency generation produces Terahertz output. The small dimensions of the design greatly increase the optical fields, enhancing the nonlinear effect. The design can also be used to detect Terahertz radiation.

Methods and apparatus are disclosed to automatically generate posynonomial performance parameter models for linear and non-linear systems such as electronic circuit characteristics, based on numerical simulations, measurements or observation. The resulting models can be used for automated optimisation of the systems, e.g. optimization-based sizing of electronic circuits so that their performance meets the set of specifications imposed by the designer. They can also be used for optimizing the performance of the system such as the performance of an electronic circuit or to adjust or control the operation of the circuit. The methods and apparatus can be applied to any physical (e.g., electrical, chemical, mechanical, biological) or non physical (e.g., economical, financial, banking) system with which the input is performance values of the system rather than an analytical definition of the system, e.g. the input can be performance data obtained by simulation and / or by measurement and / or observation.

The invention relates to a neural network deep learning-based cloth defect detection method. The method includes the following steps that: (1) high-speed line scanning imaging is performed; (2) an improved BP neural network cloth defect detection algorithm is adopted to accurately detect cloth defects; and (3) a convolutional neural network deep learning-based cloth defect classification algorithm is adopted to automatically select characteristic information of defect diversity, and non-linear and systematic processing and classification are performed. According to the method of the invention, algorithms such as image correction, splicing and de-noising are realized in an imaging system through a GPU, so that high-speed and high-quality image acquisition can be realized; the improved BP neural network cloth defect detection algorithm is adopted to detect and eliminate interference factors such as dust, dirt, cotton balls and folds; and the convolutional neural network deep learning-based cloth defect classification algorithm is adopted to monitor various kinds of detects in real time, and the classification algorithm can automatically select the characteristic information of defect diversity and carry out non-linear and systematic processing and classification.

The invention discloses an amplifier digital pre-distortion device based on a dynamic fuzzy neural network. The amplifier digital pre-distortion device comprises a digital pre-distortion device, a digital analog converter, a broadband quadrature modulator, a power amplifier, an attenuation coupler, a broadband quadrature demodulator, an analog digital converter and a pre-distortion device training module. Dynamic fuzzy neural network pre-distortion processing is conducted on an inputted digital baseband signal through the digital pre-distortion device, coupling, quadrature demodulation and analog digital conversion are conducted on an analog baseband signal outputted by the power amplifier through the attenuation coupler, the broadband quadrature demodulator, the analog digital converter respectively, and then an outputted digital baseband signal is generated. The inputted digital baseband signal and the outputted digital baseband signal are synchronously inputted into the pre-distortion device training module, and the pre-distortion device training module obtains a dynamic fuzzy neural network module parameter after carrying out parameter training through an online self-organized learning algorithm and sends the dynamic fuzzy neural network module parameter to the digital pre-distortion device. The invention further provides an amplifier digital pre-distortion method. By means of the amplifier digital pre-distortion device and method, the advantages of a fuzzy system and the advantages of a neural network are combined, dynamic adjustment can be achieved according to the complexity of a non-linear system, implementation is easy, and the non-linear characteristic and the memory effect of the power amplifier can be well compensated for.

An optical control system is described which is capable of maintaining and optimizing a fiber-optic transport system within it's domain of control while interacting with other optical systems which are controlled independently. This allows the optical system to be incorporated as a building block into a larger optical network in a relatively arbitrary fashion. This provides an underlying control system for a non-linear system like optics network that is flexible and extensible.

The invention discloses an SVM (support vector machine) parameter selection method based on multi-kernel function adaptive fusion. Compared with the prior art, the method achieves the specific analysis of the characteristics of a local kernel function, a global kernel function, a mixed kernel function and a multi-kernel function. The method comprises the steps: combining all fusion coefficients, kernel function parameters and regression parameters of the multi-kernel function together to serve as a parameter state vector, thereby enabling a model selection problem to be converted into a state estimation problem of a nonlinear system; carrying out the parameter estimation through fifth-order volume Kalman filtering, and achieving the adaptive fusion of weighted coefficients of the multi-kernel function and the selection of kernel parameters and regression parameters.

The invention discloses a pavement peak-value attachment coefficient estimating method on the basis of a UKF (unscented kalman filter) and a corrected Dugoff tire model. The method is characterized by comprising the following steps: collecting each sensor signal of a vehicle in real time, estimating a slip rate and a slip angle of each wheel by utilizing a longitudinal kinetic equation and a geometric coordinate relation of the model; and transmitting the estimated slip rate, vertical force, the slip angle and the like to a UKF coefficient calculation module based on the corrected Dugoff model to obtain a coefficient vector of a nonlinear system, and transmitting the vector and real-time estimated longitudinal force to a UKF pavement peak-value attachment coefficient estimation module to solve a peak-value attachment coefficient. A vehicle state observation system is used for collecting the signals in real time, so that the real-time property of the calculation is guaranteed, and the estimation accuracy on pavement situations which are not fitted is high. By utilizing the corrected Dugoff tire model and a UKF theory, the solving process is simple, the number of operation is small, the rapidness in operation is realized, and the convergence time is short. The method is good in robustness, capable of well recognizing the pavement situations of each wheel and suitable for being used for estimating the pavement peak-value attachment coefficient in real time.

The invention discloses a water quality parameter prediction method based on a multilayer RNN and a D-S evidence theory. Aimed at a high-dimension, multi-peak, discontinuous and non-convex complex nonlinear system of the water environment, historical data of water quality parameter content is used to train a deep RNN predication model, the model is constructed, and the future content of the waterquality parameter is predicted preliminarily; and the problem of uncertainty can be solved in the mathematical theory via the D-S evidence theory. Thus, conflict solution and evidence fusion can be realized by utilizing the D-S evidence theory on the basis of the RNN model prediction result, the prediction precision of the single parameter is improved, and early warning of multiple parameters is more accurate.

The invention provides a real-time on-line control method for a miscellaneous nonlinear system. The method is characterized by comprising the following steps of: a, transmitting the current observation state information of a network controlled object to a first controller, wherein the first controller is used for providing real-time on-line control for the network controlled object to make the network controlled object asymptotically stabilized; b, acquiring control output quantity information at the current moment by using the first controller according to the current observation state information; and c, feeding the control output quantity information back to the network controlled object, wherein the network controlled object is the miscellaneous nonlinear system. The aim of asymptotically stabilizing the network controlled object by controlling the miscellaneous nonlinear system through a network is fulfilled by a method for controlling a quantization controller. Particularly, observation information is quantized and then is transmitted, so that the load of the network is lightened.