The detection of thermal insulation failures in buildings in operation responds to the challenge of improving building energy efficiency. This multidisciplinary study presents a novel four-step soft computing knowledge identification model called IKBIS to perform thermal insulation failure detection. It proposes the use of Exploratory Projection Pursuit methods to study the relation between input and output variables and data dimensionality reduction. It also applies system identification theory and neural networks for modeling the thermal dynamics of the building. Finally, the novel model is used to predict dynamic thermal biases, and two real cases of study as part of its empirical validation.

This interdisciplinary research is based on the application of unsupervized connectionist architectures in conjunction with modelling systems and on the determining of the optimal operating conditions of a new high precision industrial process known as laser milling. Laser milling is a relatively new micro-manufacturing technique in the production of high-value industrial components. The industrial problem is defined by a data set relayed through standard sensors situated on a laser-milling centre, which is a machine tool for manufacturing high-value micro-moulds, micro-dies and micro-tools. The new three-phase industrial system presented in this study is capable of identifying a model for the laser-milling process based on low-order models. The first two steps are based on the use of unsupervized connectionist models. The first step involves the analysis of the data sets that define each case study to identify if they are informative enough or if the experiments have to be performed again. In the second step, a feature selection phase is performed to determine the main variables to be processed in the third step. In this last step, the results of the study provide a model for a laser-milling procedure based on low-order models, such as black-box, in order to approximate the optimal form of the laser-milling process. The three-step model has been tested with real data obtained for three different materials: aluminium, copper and hardened steel. These three materials are used in the manufacture of micro-moulds, micro-coolers and micro-dies, high-value tools for the medical and automotive industries among others. As the model inputs are standard data provided by the laser-milling centre, the industrial implementation of the model is immediate. Thus, this study demonstrates how a high precision industrial process can be improved using a combination of artificial intelligence and identification techniques.

Abstract. Improving the detection of thermal insulation in buildings -which includes the development of models for heating and ventilation processes and fabric gain -could significantly increase building energy efficiency and substantially contribute to reductions in energy consumption and in the carbon footprints of domestic heating systems. Thermal insulation standards are now contractual obligations in new buildings, although poor energy efficiency is often a defining characteristic of buildings built before the introduction of those standards. Lighting, occupancy, set point temperature profiles, air conditioning and ventilation services all increase the complexity of measuring insulation efficiency. The identification of thermal insulation failure can help to reduce energy consumption in heating systems. Conventional methods can be greatly improved through the application of hybridized machine learning techniques to detect thermal insulation failures when a building is in operation. A three-step procedure is proposed in this paper that begins by considering the local building and heating system regulations as well as the specific features of the climate zone. Firstly, the dynamic thermal performance of different variables is specifically modelled, for each building type and climate zone. Secondly, Cooperative Maximum-Likelihood Hebbian Learning is used to extract the relevant features. Finally, neural projections and identification techniques are applied, in order to detect fluctuations in room temperatures and, in consequence, thermal insulation failures. The reliability of the proposed method is validated in three winter zone C cities in Spain. Although a great deal of further research remains to be done in this field, the proposed system is expected to outperform conventional methods described in Spanish building codes that are used to calculate energetic profiles in domestic and residential buildings.

A novel procedure for learning Fuzzy Controllers (FC) is proposed that processes energy efficiency issues and uses them to distribute electrical energy to heaters in an electrical energy heating system. Energy rationalisation together with temperature control can significantly improve energy efficiency, by efficiently controlling electrical heating systems and electrical energy consumption. The novel procedure, which improves the training process, is designed to train the FC, as well as to run the control algorithm and to carry out energy distribution. Firstly, the dynamic thermal performance of different variables is mathematically modelled for each specific building type and climate zone. Secondly, an exploratory projection pursuit method is used to extract the relevant features. Finally, a supervised dynamic neural network model and identification techniques are applied to FC learning and training. The FC rule-set and parameter-set learning process is a multi-objective problem that minimises both the indoor temperature error and the energy deficit in the house. The reliability of the proposed procedure is validated for a city in a winter zone in Spain.

No abstract

This ongoing interdisciplinary research is based on the application of genetic algorithms to simplify the process of predicting the mortality of a critical illness called endocarditis. The goal is to determine the most relevant features (symptoms) of patients (samples) observed by doctors to predict the possible mortality once the patient is in treatment of bacterial endocarditis. This can help doctors to prognose the illness in early stages; by helping them to identify in advance possible solutions in order to aid the patient recover faster. The results obtained using a real data set, show that using only the features selected by employing a genetic algorithm from each patient's case can predict with a quite high accuracy the most probable evolution of the patient.

In this paper we present a soft computing system developed to optimize the laser milling manufacture of high value steel components, a relatively new and interesting industrial technique. This multidisciplinary study is based on the application of neural projection models in conjunction with identification systems, in order to find the optimal operating conditions in this industrial issue. Sensors on a laser milling centre capture the data used in this industrial case study defined under the frame of a machine-tool that manufactures steel components like high value molds and dies. The presented model is based on a two-phase application. The first phase uses a neural projection model capable of determine if the data collected is informative enough based on the existence of internal patterns. The second phase is focus on identifying a model for the laser-milling process based on low-order models such as Black Box ones. The whole system is capable of approximating the optimal form of the model. Finally, it is shown that the Box-Jenkins algorithm, which calculates the function of a linear system from its input and output samples, is the most appropriate model to control such industrial task for the case of steel components.

No abstract