Abstract:
Motivation for the shrinking horizon model predictive control (SHMPC) strategy is developed, as is the need for SHMPC experimental verification. A miniature autoclave is described, as are the 112 autoclave curing runs performed in it. Nondestructive tests such as the displacement method, the buoyancy method (with different fluids), and the resin balance method--all described herein-are applied to the 112 cured laminates. Ignition loss testing is also applied to pieces of these laminates. Thickness and void fraction data from the 112 laminates are used to train off-line and on-line artificial neural network (ANN) models, for use in SHMPC. SHMPC is then used to control the thickness (to within .002 inch--on average--of target thickness) and void fraction (to less than .01 of laminate volume on average) of ten composite laminates. On-line control results in greater thickness accuracy than its off-line counterpart.
Source: Proceedings of the 'NDE applied to Process Control of Composite Fabrication' - Conference, 4-5 Oct 1994 St. Louis, Missouri. Publisher and Organizer: Nondestructive Testing Information Analysis Center (NTIAC) Texas Research Institute Austin, Inc. Austin, Texas [http://www.ntiac.com] [Buying the Proceedings]
| |UTonline| |Abstract Database| |Access to all Abstracts of the Conference | |