Artificial Neural Network Modeling of the Automated Thermoplastic Composite Tow-Placement System

摘要

The Automated Thermoplastic Composite TowPlacement process is being investigated as a rapid affordabletechnology for the fabrication composite parts. The componentsare manufactured by applying adequate heat and pressure to athermoplastic prepreg tape so that bonding is achieved at theinterface and consolidation is obtained within the composite tow.Process models, including heat transfer, consolidation andbonding, have been validated and can accurately predict partquality as a function of process parameters. Unfortunately, theslow processing times of the first principle models limit theiruse, i.e., for extensive parametric studies, for on-line processsimulation or on-line process optimization. Therefore, this studyis investigating two different classes of artificial neural networks,conventional feed-forward neural networks and CerebellarModel Arithmetic Controller, to substrate the first principlemodels and subsequently decrease computational time several orders of magnitude.