Neural NetworksWhat Are They?Neural Network DiagramTranslation – Input Stimuli to Output ResponseLearning CapabilityControl System ApplicationsAircraft – Flight Control SystemManufacturing Plant – Hot Dip Galvanized Steel StripPower Plant – Utility BoilersNeural NetworksSarah EzzellEngr. 315What Are They?Information processing system (non-algorithmic, non-digital)Inspired by the human brainMade of artificial neurons (neurodes)–Crude approximation of biological neurons–Connected by weights over which signals travel–Arranged in layersNeural Network DiagramInput LayerMiddle LayerOutput LayerInput SignalsOutput SignalsNeurodeConnection LinkTranslation – Input Stimuli to Output Response3 steps–Neurode computes net weighted input received–Converts net input into an activation level–Converts activation level into output signalLearning CapabilityLearns to solve problems, not just programmedLearning achieved by modifying weightsTraining methods–Supervised–Graded–UnsupervisedControl System ApplicationsAircraftManufacturing plantPower plantAircraft – Flight Control SystemNeural network like a human, but better–Adapts like human, only quicker and more accuratelyBack propagation vs. on line learningNetwork integrated with current flight control systems–Removes “gain scheduling”Manufacturing Plant – Hot Dip Galvanized Steel StripManually controlled coating thickness controlsNeural network control model–Developed by Siemens and Thyssen Stahl–Cost-effective, better quality product–On line learningEliminates need for hot measuring equipmentPower Plant – Utility BoilersAir pollutants from power plantsNeuSIGHT system–Used in coal-fired electric power plants–Provides real time closed-loop supervisory control–Improves operating efficiency & reduces
View Full Document