Modelo inverso de un amortiguador magneto-reológico utilizando redes neuronales

Thumbnail Image
Publication date
Defense date
Journal Title
Journal ISSN
Volume Title
Google Scholar
Research Projects
Organizational Units
Journal Issue
This Final Degree Project aims to model an inverse magneto-rheological damper based on Artificial Neural Networks. The principal function of the suspension system of a vehicle is to keep the tires in contact to the road, absorbing the vibrations generated in the vehicle and giving security as well as comfort to the passengers. The main components of the suspension system are elastic components such as helical springs or the anti-roll bar and the damper. There are three types of suspension systems based on the control element: passive suspension, active suspension and semiactive suspension. Passive suspension is design to produce an optimal behavior just in some situations. Active suspension substitutes the damper and elastic elements of the conventional suspension systems by some actuators that produce a force which modifies the damper force. The main problems of this suspension system are that it is very expensive and it has a big energetic cost. To solve these problems semiactive suspension appeared. This third type of suspension system is an intermediate design between the others; it produces variable force by regulating one of the elements of the conventional suspension. Recent studies have shown that this can be possible by adapting some characteristics of the damper. The damper of a vehicle is a resistant element that reduces the vibrations generated by the engine and the movement of the vehicle to avoid damages in the chassis. Nowadays, the most used damper model is the hydraulic telescopic. This type of damper is a passive suspension system. One of the main research lines of semiactive suspension systems is the magneto- rheological damper. These dampers have a magneto- rheological fluid which has a 40 % of metallic particles. When applying a magnetic field to the damper, these metallic particles align and produce a variation in the damper force. The advantages of these dampers are the low response time and the big amount of variation possibilities that they provide. There are two models of magneto- rheological dampers: the direct model and the inverse model. The first one obtains the damper force by knowing the current that has to be applied to the damper. The inverse model estimates the current that has to be applied to the damper to get the desired force. In this project a model for an inverse magneto- rheological damper is implemented by using Artificial Neural Networks. The software used is called JavaNNS. The structure of the Neural Network used has one input layer with four neurons (each neuron corresponds to an input variable: displacement, force, frequency and velocity), many hidden layers as it is needed and one output layer with one neuron which corresponds to the current. The results obtained show that a network of three hidden layers with 20 neurons is the optimal one to this model of magneto- rheological damper.
Redes neuronales, Sistemas de suspensión, Reología, Amortiguadores
Bibliographic citation