Search for: dynamic-neural-networks
0.01 seconds

    An evolutionary optimizing approach to neural network architecture for improving identification and modeling of aircraft nonlinear dynamics

    , Article Proceedings of the Institution of Mechanical Engineers, Part G: Journal of Aerospace Engineering ; Vol. 228, issue. 12 , 2014 , p. 2178-2191 Roudbari, A ; Saghafi, F ; Sharif University of Technology
    In this paper, modified genetic algorithm has been used as a simultaneous optimizer of recurrent neural network to improve identification and modeling of aircraft nonlinear dynamics. Weighted connections, network architecture, and learning rules are features that play important roles in the quality of neural networks training and their generalizability in order to model nonlinear systems. Therefore, the main focus of this paper is to apply appropriate evolutionary methods in order to simultaneously optimize the parameters of neural networks for the improvement identification and modeling of aircraft nonlinear dynamics. To validate this study, the results have been compared with the recorded... 

    Prediction of transition point on an oscillating airfoil using Neural network

    , Article 2007 5th Joint ASME/JSME Fluids Engineering Summer Conference, FEDSM 2007, San Diego, CA, 30 July 2007 through 2 August 2007 ; Volume 2 FORA, Issue PART A , July , 2007 , Pages 33-39 ; 0791842886 (ISBN); 9780791842881 (ISBN) Soltani, M. R ; Seddighi, M ; Masdari, M ; Sharif University of Technology
    Dynamic Neural network was used to minimize the amount of data required to predict the location of transition point on a 2-D oscillatory wing. For this purpose, various experimental tests were carried out on a section of a 660kw wind turbine blade. A multi layer non linear perceptrons network was trained using the output signals of four hot films attached on the upper surface of the model. Results show that using only 50% of the test data, the trained network was able to the transition point with an acceptable accuracy. Moreover, the method can predict the transition points at any position of the wing surface for different Reynolds numbers, amplitudes and initial angles of oscillation, and... 

    Prediction and Optimization of Cure Cycle of Thick Fiber Reinforced Composite Parts Using Artificial Neural Networks

    , M.Sc. Thesis Sharif University of Technology Eghbal Jahromi, Parisa (Author) ; Shojaei, Akbar (Supervisor) ; Pishvaie, Mahmoud Reza (Supervisor)
    The Curing of thick thermoset parts of composites Experience substantial temperature overshoot, especially at the center of those thick parts and large temperature gradient exists through the whole part due to large amount of released heat, thickness and low conductivity of the composite.this leads to non-uniformity of cure, residual stress and consequently composite cracks and possibly degradation of the polymer. In this thesis, first a thick cubic geometery of glass/epoxy composite is modeled with the help ofa commercial CFd packageand sensitivity to different effective parameters of this process is investigated with the help of this model.This model is then substituted by a trained... 

    Multi-Object Tracking in Video using Graph Neural Networks

    , M.Sc. Thesis Sharif University of Technology Hosseinzadeh, Mehran (Author) ; Rabiee, Hamid Reza (Supervisor)
    Multiple object tracking refers to the detection and following of target object classes in video sequences. In this task, all objects belonging to the target classes in the video are detected simultaneously in each frame, and a unique ID is assigned to each of them throughout the video. In recent years, the use of graph neural networks for solving this problem has received significant attention because these models are suitable tools for discovering and improving the relationships between objects in the scene, which can greatly assist in better object pairing. However, there are various challenges to using graph neural networks, the most important of which is the limitation of input graph... 

    Actuator failure-tolerant control of an all-thruster satellite in coupled translational and rotational motion using neural networks

    , Article International Journal of Adaptive Control and Signal Processing ; 2018 ; 08906327 (ISSN) Tavakoli, M. M ; Assadian, N ; Sharif University of Technology
    John Wiley and Sons Ltd  2018
    The nonlinear model predictive control (MPC) approach is used to control the coupled translational-rotational motion of an all-thruster spacecraft when one of the actuators fails. In order to model the dynamical response of the spacecraft in MPC, instead of direct integration, a neural network (NN) model is utilized. This model is built of a static NN, followed by a dynamic NN. The static NN is used to find the changes of the mapping of “the demanded forces to the thrusters” and “the real torques/forces produced by the remaining thrusters” after the failure occurrence through online training. In this manner, the effect of failed thruster on the dynamics can be found and the need for... 

    A Dynamic Network Approach to Inertial Motion Capture

    , Ph.D. Dissertation Sharif University of Technology Razavi, Hamid Reza (Author) ; Alasty, Aria (Supervisor) ; Salarieh, Hassan (Co-Supervisor)
    The current study introduces algorithms for inertial motion capture which use data from 9-DOF inertial-magnetic sensor modules to estimate the position and attitude of body links. First, an algorithm is proposed which is capable of IMU calibration without the use of external equipment with less than 0.5% error. Next, extended and unscented Kalman filter-based (EKF and UKF) inertial motion capturing algorithms are introduced that utilize biomechanical constraints in addition to kinematics. In addition to real-time sensor calibration, the algorithms are capable of real-time link geometry estimation, which allows for the imposition of biomechanical constraints without a priori knowledge... 

    Neural network-based synchronization of uncertain chaotic systems with unknown states

    , Article Neural Computing and Applications ; Volume 27, Issue 4 , 2016 , Pages 945-952 ; 09410643 (ISSN) Bagheri, P ; Shahrokhi, M ; Sharif University of Technology
    Springer-Verlag London Ltd  2016
    In this paper, synchronization of chaotic systems with unknown parameters and unmeasured states is investigated. Two nonidentical chaotic systems in the framework of a master and a slave are considered for synchronization. It is assumed that both systems have uncertain dynamics, and states of the slave system are not measured. To tackle this challenging synchronization problem, a novel neural network-based adaptive observer and an adaptive controller have been designed. Moreover, a neural network is utilized to approximate the unknown dynamics of the slave system. The proposed method imposes neither restrictive assumption nor constraint on the dynamics of the systems. Furthermore, the... 

    Modelling and Prediction Air Polutants Level in Tehran Using Dynamic Neural Networks

    , M.Sc. Thesis Sharif University of Technology Khosravi, Neda (Author) ; Erhami, Mohammad (Supervisor)
    In parallel to the growing of population in Tehran metropolitan, air pollution in this city has become to a major problem. From which high concentration of pollutants have adverse effects on public health, accurate estimating and forecasting of concentrations for several days ahead, can provide the possibility to implement the management measures to reduce hazard and risks. Among the air pollution models, application of statistic models based on neural network in comparison to the traditional deterministic models are easier and less costly. In most studies, static models use a classical single MLP to predict one step ahead. For this purpose ANN models are required to estimate next value of... 

    Neuroplasticity in dynamic neural networks comprised of neurons attached to adaptive base plate

    , Article Neural Networks ; Volume 75 , 2016 , Pages 77-83 ; 08936080 (ISSN) Joghataie, A ; Shafiei Dizaji, M ; Sharif University of Technology
    Elsevier Ltd  2016
    In this paper, a learning algorithm is developed for Dynamic Plastic Continuous Neural Networks (DPCNNs) to improve their learning of highly nonlinear time dependent problems. A DPCNN is comprised of a base medium, which is nonlinear and plastic, and a number of neurons that are attached to the base by wire-like connections similar to perceptrons. The information is distributed within DPCNNs gradually and through wave propagation mechanism. While a DPCNN is adaptive due to its connection weights, the material properties of its base medium can also be adjusted to improve its learning. The material of the medium is plastic and can contribute to memorizing the history of input-response similar...