Output Of Nonlinear Model Research Articles

Attitude Control and Parameter Optimization: A Study on Hubble Space Telescope

Abstract In this work, we build a satellite attitude Proportional-Integral-Derivative (PID) controlled system by using the Hubble Space Telescope (HST) parameters as a reference and tune its controller parameters using various tuning methods. First, we give the equations for the motion of a satellite. We elaborate the control structure as controller, actuator, dynamics, and kinematics subsystems and construct an external disturbance model. We use a reaction wheel assembly used in the HST with the same configuration as the actuator. We evaluate the performance of the linearization by comparing it with the nonlinear model output. By working on the linearized model, we tune the PID controller parameters using two different methods: “Model-Based Root Locus Tuning” and “Genetic Algorithm Based Tuning”. First, we obtain the controller parameters by manipulating the poles on the root locus plot of the linearized system. In addition, we use genetic algorithms to find the optimized controller values of the system. Finally, we compare the performances of the two methods based on their cost function values and find that the Genetic Algorithm-based tuned parameters are more fruitful in terms of the cost function value than the parameters obtained by the Root Locus-based tuning. However, it is found that the Root Locus-based tuning performs better in disturbance rejection.

The parameter identification method of steam turbine nonlinear servo system based on artificial neural network

The servo system in steam turbine digital electric-hydraulic control system (DEH) is affected by nonlinear factors when it is working. To accurately simulate dynamic characteristics of the DEH, a new nonlinear servo system is proposed, which has limit, dead zone and correction coefficient caused by unknown factors. The model parameters are divided into linear parameters and nonlinear parameters to be identified, respectively. Neural networks are used to identify linear parameters. The nonlinear parameters should be identified according to flow characteristic curve. To verify the validity of the proposed model and parameter identification method, the actual data of primary frequency control from a 1000 MW Ultra Supercritical Unit is adopted. Meanwhile, the linear model with no nonlinear factors is used for comparison. Where the fitting degree of valve opening is 98.445% and power is 96.986%, the output of nonlinear model coincides with actual output well. Where the relative error of stable result is 5% of valve opening and 1.58% of power, the error of linear model is larger. The simulation results of the proposed method show that the nonlinear factors of high-power units cannot be ignored and the nonlinear model of servo system is more accurate.

Finite Element Simulation of Rough Rice Kernel (Oryza sativa L.) cv. Fajer Drying

This study was conducted to present a mathematical model with numerical solutions to predict the mass distributions inside a cross sectional area of an individual Fajer variety of rough rice (Oryza Sativa L.) kernel as a function of drying time considering the effects of coupled heat and mass transfer processes. The modified Luikov's equation was used for the simulation of drying kinetics of a single rough rice grain. The applied drying models with simultaneous heat and mass diffusion in biological materials were developed by finite element method. A finite element formulation and solution of a set of nonlinear coupled conductive heat and diffusive moisture transfer equation to improve grain drying simulation of axisymmetric bodies is presented. Axisymmetric linear triangular elements with two degrees of freedom per node were used to discretize the rice grain. Moisture distribution inside the individual rough rice kernel was produced by the model. Good agreement was observed when the output of nonlinear model was compared to experimental data.

An Adaptive Nonlinear MOS Scheme for Precipitation Forecasts Using Neural Networks

A novel neural network (NN)–based scheme performs nonlinear model output statistics (MOS) for generating precipitation forecasts from numerical weather prediction (NWP) model output. Data records from the past few weeks are sufficient for establishing an initial MOS connection, which then adapts itself to the ongoing changes and modifications in the NWP model. The technical feasibility of the algorithm is demonstrated in three numerical experiments using the NCEP reanalysis data in the Alaskan panhandle and the coastal region of British Columbia. Its performance is compared with that of a conventional NN-based nonadaptive scheme. When the new adaptive method is employed, the degradation in the precipitation forecast skills due to changes in the NWP model is small and is much less than the degradation in the performance of the conventional nonadaptive scheme.

Construction of confidence intervals for neural networks based on least squares estimation

We present the theoretical results about the construction of confidence intervals for a nonlinear regression based on least squares estimation and using the linear Taylor expansion of the nonlinear model output. We stress the assumptions on which these results are based, in order to derive an appropriate methodology for neural black-box modeling; the latter is then analyzed and illustrated on simulated and real processes. We show that the linear Taylor expansion of a nonlinear model output also gives a tool to detect the possible ill-conditioning of neural network candidates, and to estimate their performance. Finally, we show that the least squares and linear Taylor expansion based approach compares favorably with other analytic approaches, and that it is an efficient and economic alternative to the nonanalytic and computationally intensive bootstrap methods.