Methodology for implementing Virtual Sensors using Neural Networks

摘要

In this work a Methodology framework for implanting Virtual Sensors using Neural Networks will be presented, including the statistical analysis techniques that can be used for studying and processing the data. The proposed Methodology is based upon Software Engineering, Knowledge-based systems and neural networks methodologies. This methodological framework includes both technical and economical feasibility to build the virtual sensors and considers important aspects as the available computational platform, historical data files, data processing requirements such as filtering, pruning, set of variables that must be selected for the best performance of the virtual sensor, etc. There are also presented the statistical consideration and the corresponding techniques for data analysis and processing. The methodology includes techniques as principal components, cluster analysis, factorial analysis, etc.