A methodology for building neural networks models from empirical engineering data

摘要: Abstract Neural networks (NN) are general tools for modeling functional relationships in engineering. They used to model the behavior of products and properties processes. Nevertheless, their use is often ad hoc. This paper provides a sound basis using NN as implicit empirical engineering data. First, clear definition task given, followed by reviewing theoretical capabilities estimation. Subsequently, procedure practice described illustrated with an example marine propeller behavior. Particular attention devoted better estimation quality, insight into influence measurement errors on advanced methods such stacked generalization ensemble further improve quality. Using new method SG(k-NN), one could quality models even if they close being optimal.