Transcript Slide 1
Compression After Impact Load Prediction in Composites Eric v. K. Hill, Tom Ebert, Jonathan V. Kay, Gregory K. Lewis, Anthony M. Gunasekera, Michael W. Langdon, William D. Rice and Yi Zhao OBJECTIVES Instron Dynatup 9250 Calibrated Impactor 4 x 6 Inch Carbon/Epoxy Coupon with 15J BVID • 4 x 6 inch graphite/epoxy composite coupons clamped on edges and impacted from 8-20 joules with 0.50 inch diameter hemispherical tup to create barely visible impact damage (BVID) • Acoustic emission (AE) nondestructive testing combined with back-propagation neural network (BPNN) to predict compression after impact (CAI) loads of composite laminates • Goal: Predict CAI loads with a worst case error within ±15% APPROACH/TECHNICAL CHALLENGES • BPNN constructed to predict CAI loads of graphite/epoxy coupons from low proof load (≤25% avg Pcu) AE amplitude distribution data • Eliminate coupon/compression fixture rubbing noises from AE data ACCOMPLISHMENTS/RESULTS • Neural network predicted CAI ultimate loads with a worst case error of -11.53% (-2000 lbf) • Absolute value of worst case error fits B-basis allowables for composites: 95% confidence that 90% of all future compressive failure loads will fall within this tolerance interval CAI Test Fixture with AE Transducers Mounted on Graphite/Epoxy Coupon BPNN Prediction of CAI Loads in Graphite/Epoxy Coupons Subject to BVID Amplitude Histogram (≤25% avg Pcu) NeuralWorks Professional II/PLUS® Software Predicted CAI Failure Load Worst Case Error: -11.53% AE Waveform Parameter Data AE Nondestructive Testing BPNN Settings BPNN Output Compressive Failure Load versus Impact Energy in Graphite/Epoxy Coupons Graphite/Epoxy Coupons Clamped on Edges and Impacted with 0.50 inch Diameter Hemispherical Tup Best Fit Parabola Pcu = 39,877 – 2,657.0E + 82.003E2 X-Ray Image of 15J BVID Ultrasonic C-Scan Image of 15J BVID Absolute value of worst case error ±2,000 lbf fits within B-Basis allowables for composites: 95% confidence that 90% of all future compressive failure loads will fall within this tolerance interval