Performance comparison of artificial neural networks and expert systems applied to flow pattern identification in vertical ascendant gas-liquid flows

摘要: Abstract Instantaneous readouts of an electrical resistivity probe are taken in upward vertical air–water mixture. The signals further processed to render the statistical moments and probability density functions here used as objective flow pattern indicators. A series 73 experimental runs have its identified by visual inspection assisted analyses void fraction’s trace associated function. patterns classified into six groups labeled as: bubbly, spherical cap, slug, unstable semi-annular annular. This work compares analyzes performance artificial neural networks, ANN, expert systems identification. employed ANNs Multiple Layer Perceptrons, Radial Basis Functions Probabilistic Neural Network, with single multiple outputs. is gauged percentage right identifications based on observation. analysis extended clustering algorithms assist formation knowledge base during learning stages systems. following algorithms: self organized maps, K-means Fuzzy C-means also tested against data.