Research into fault diagnosis in machines with a wide range of variable loads and speeds, such as wind turbines, is of great industrial interest. Analysis of the power signals emitted by wind turbines for the diagnosis of mechanical faults in their mechanical transmission chain is insufficient. A successful diagnosis requires the inclusion of accelerometers to evaluate vibrations. This work presents a multi-sensory system for fault diagnosis in wind turbines, combined with a data-mining solution for the classification of the operational state of the turbine. The selected sensors are accelerometers, in which vibration signals are processed using angular resampling techniques and electrical, torque and speed measurements. Support vector machines (SVMs) are selected for the classification task, including two traditional and two promising new kernels. This multi-sensory system has been validated on a test-bed that simulates the real conditions of wind turbines with two fault typologies: misalignment and imbalance. Comparison of SVM performance with the results of artificial neural networks (ANNs) shows that linear kernel SVM outperforms other kernels and ANNs in terms of accuracy, training and tuning times. The suitability and superior performance of linear SVM is also experimentally analyzed, to conclude that this data acquisition technique generates linearly separable datasets.

The understanding of a psychological phenomena such as emotion is of paramount importance for psychologists, since it allows to recognize a pathology and to prescribe a due treatment for a patient. While approaching this problem, mathematicians and computational science engineers have proposed different unimodal techniques for emotion recognition from voice, electroencephalography, facial expression, and physiological data. It is also well known that identifying emotions is a multimodal process. The main goal in this work is to train a computer to do so. In this paper we will present our first approach to a multimodal emotion recognition via data fusion of Electroencephalography and facial expressions. The selected strategy was a feature-level fusion of both Electroencephalography and facial microexpressions, and the classification schemes used were a neural network model and a random forest classifier. Experimental set up was out with the balanced multimodal database MAHNOB-HCI. Results are promising compared to results from other authors with a 97% of accuracy. The feature-level fusion approach used in this work improves our unimodal techniques up to 12% per emotion. Therefore, we may conclude that our simple but effective approach improves the overall results of accuracy.

Information retrieval problem occurs when the target information is not available 'literally' into the set of documents. In problems in which the goal is to find 'hidden' information, it is important to develop hybrid methodologies or improve and design a new one. In this work the authors are dealing with identifying the most informative piece of data on a collection of documents, in order to obtain the best result on a posterior fuzzy clustering stage. The aim is to find similarities between the documents and a reference target, to establish relationships related to a non-literal feature. We propose to apply the well-known entropy term weighting scheme and then show a posterior different procedures to the right election of the interest data. This procedure brings the biggest amount of information within the smallest amount of data. Applying a specific selection procedure for a group of words, gives more information to differentiate and separate the documents after using the entropy weighting. This returns considerable results on the processing time and the right fuzzy clustering of the documents collection.