Wyniki wyszukiwania - BazTech

1

Comparison of the adaptive and neural network control for LWR 4+ manipulators: simulation study

Woliński Łukasz

Archive of Mechanical Engineering

|

2020

|

LXVII, nr 1

111--121

EN

This paper deals with two control algorithms which utilize learning of their models’ parameters. An adaptive and artificial neural network control techniques are described and compared. Both control algorithms are implemented in MATLAB and Simulink environment, and they are used in the simulation of a postion control of the LWR 4+ manipulator subjected to unknown disturbances. The results, showing the better performance of the artificial neural network controller, are shown. Advantages and disadvantages of both controllers are discussed. The usefulness of the learning algorithms for the control of LWR 4+ robots is discussed. Preliminary experiments dealing with dynamic properties of the two LWR 4+ robots are reported.

2

Identiﬁcation of water treatment plant based on feedforward neural network

Mohamed A. F., Radwa H. Z.

Control and Cybernetics

|

2017

|

Vol. 46, no. 3

247--258

EN

Coagulation process is the main process in conventional water treatment process sequence. It inﬂuences the following treatment process aspects: maintaining plant eﬃciency and increasing the quality of the produced water. This is accomplished by adding chemicals to raw water, such as alum sulphate. To secure the appropriate plant performance, a mathematical model is proposed in this paper for the coagulation unit, followed by the development of the control strategy. Classic PID and neural network based controller regulating the process are used. Tests were performed, based on the real data for water treatment, using MATLAB/SIMULINK. Simulation results showed better values for both settling time and overshoot in the case of using neural network based controller than PID.

3

Neural network control design considerations for the active damping of a smart beam

Cupiał P., Łacny Ł.

Journal of Theoretical and Applied Mechanics

|

2015

|

Vol. 53 nr 4

767--774

EN

In this study, possible options for the active damping of a smart beam with piezoelectric patches using neural network control algorithm, are presented. The algorithms used for the control are Neural Direct Inverse and Feedback Linearisation (NARMA-L2). Additionally, several possible modifications used for the purpose of improving the control, such as different values of control gain or sampling time of the training data, as well as step-wise control are tested.

4

Intelligent control algorithm for ship dynamic positioning

Meng W., Sheng L. H., Qing M., Rong B. G.

Archives of Control Sciences

|

2014

|

Vol. 24, no. 4

479--497

EN

Ship motion in the sea is a complex nonlinear kinematics. The hydrodynamic coefficients of ship model are very difficult to accurately determine. Establishing accurate mathematical model of ship motion is difficult because of changing random factors in the marine environment. Aiming at seeking a method of control to realize ship positioning, intelligent control algorithms are adopt utilizing operator's experience. Fuzzy controller and the neural network controller are respectively designed. Through simulations and experiments, intelligent control algorithm can deal with the complex nonlinear motion, and has good robustness. The ship dynamic positioning system with neural network control has high positioning accuracy and performance.

5

Predictive neural network controller for hydrostatic transmission control

Nawrocka A., Kwaśniewski J.

Mechanics / AGH University of Science and Technology

|

2008

|

Vol. 27, no. 2

62-65

EN

In this paper predictive neural network controller is applied to control rotational speed of hydraulic engine hydrostatic transmission. It describes the functional structure and design principle of predictive neural network controller. This predictive neural network control method provides control linear and nonlinear systems. In this case feed forward network with one hidden layer was used. The control objective is to minimize a control cost function. Computer simulations are provided for illustration and verification. As control plant the hydrostatic transmission was chosen. The control by using the hydraulic drive is the time variant process, consequential to inconstancy parameters of hydraulic liquid (for instance: kinematic viscosity, compressibility and thickness). The efficiency of pump control is realized by use the electrohydraulic servo - system distinguish itself by a nonlinear characteristic. Finally, a predictive neural network method to control this kind of object is successfully applied.

PL

W artykule zaprezentowano koncepcję regulatora predykcyjnego bazującego na sieciach neuronowych, wykorzystanego do sterowania prędkością obrotową silnika hydraulicznego przekładni hydrostatycznej. Regulator zastosowany w predykcyjnym układzie regulacji był strojony przy użyciu sztucznych sieci neuronowych. Do badań wykorzystano sieć jednokierunkową, jednowarstwową. Zasadniczym zadaniem układu jest minimalizacja określonej funkcji kryterialnej. Ostatnia część artykułu jest zestawieniem wyników badań zaproponowanego układu regulacji do sterowania obiektem nieliniowym. Jako obiekt do badań wykorzystano przekładnię hydrostatyczną. Zbudowana jest ona z pompy o zmiennej wydajności i silnika o stałej chłonności. Do sterowania wydajnością pompy wykorzystano elektrohydrauliczny układ sterujący. Układ ten składa się z siłownika hydraulicznego sprzężonego tłoczyskiem z wychylnym wirnikiem pompy oraz elektrohydraulicznego serwozaworu przepływowego. Obiekt ze względu na swoje właściwości jest nieliniowy i niestacjonarny.

6

Experimental four-legged walking robot : conception, design and control

Grepl R., Bezdíček M., Švehlák M., Chmelíček J.

Elektronika : konstrukcje, technologie, zastosowania

|

2004

|

Vol. 45, nr 8-9

63-65

EN

The aim of project presents in this paper was to build a low-cost experimental walking robot. Robot is four-legged, three d.o.f. each leg. Small Hitec servos are used in the design. Control unit is composed by two Atmel Atmega microcontrolers - one for the control of servodrives, the main for communication to PC, kinematic model and other tasks.

PL

Celem projektu zaprezentowanego w artykule było zbudowanie taniego doświadczalnego robota kroczącego. Robot wyposażony jest w cztery nogi, a każda z nich posiada trzy stopnie swobody. W konstrukcji robota zostały wykorzystane małe serwomechanizmy firmy Hitec. Zespół sterownika składa się z dwóch mikrokontrolerów Atmel Atmega: podrzędny steruje serwonapędami, a nadrzędny odpowiada za komunikację z komputerem osobistym, zastosowanie modelu kinematycznego oraz inne zadania.