Informatica

The present research shows the implementation of a virtual sensor for fault detection with the feature of recovering data. The proposal was implemented over a bicomponent mixing machine used for the wind generator blades manufacture based on carbon fiber. The virtual sensor is necessary due to permanent problems with wrong sensor measurements. The solution proposed uses an intelligent model able to predict the sensor measurements, which are compared with the measured value. If this value belongs to a specified range, it is valid. Otherwise, the prediction replaces the read value. The process fault detection feature has been added to the proposal, based on consecutive erroneous readings, obtaining satisfactory results.

We propose an adaptive inverse control scheme, which employs a neural network for the system identification phase and updates its weights in online mode. The theoretical basis of the method is given and its performance is illustrated by means of its application to different control problems showing that our proposal is able to overcome the problems generated by dynamic nature of the process or by physical changes of the system which originate important modifications in the process. A comparative experimental study is presented in order to show the more stable behavior of the proposed method in several working ranks.

The aim of this study is to predict the energy generated by a solar thermal system. To achieve this, a hybrid intelligent system was developed based on local regression models with low complexity and high accuracy. Input data is divided into clusters by using a Self Organization Maps; a local model will then be created for each cluster. Different regression techniques were tested and the best one was chosen. The novel hybrid regression system based on local models is empirically verified with a real dataset obtained by the solar thermal system of a bioclimatic house.