Date of Award

Spring 2003

Document Type

Thesis - Open Access

Degree Name

Master of Aerospace Engineering


Graduate Studies

Committee Chair

Dr. Yi Zhao

Committee Member

Dr. Eric v. K. Hill

Committee Member

Dr. Richard P. Anderson


Low energy impact damage to a composite structure is difficult to detect and can have profound effects on compressive strengths. Low energy impact damage is sometimes termed as barely visible impact damage (BVID). Detecting BVID is only possible by implementing nondestructive testing (NDT) techniques. Depending upon the support conditions, material system, laminate thickness, lay-up orientation, and impactor geometry, velocity, and hardness, the types of damage associated with BVID include delaminations, longitudinal and transverse matrix cracks, and in some cases, fiber breaks. Material properties such as the strengths of the matrix, fibers, fiber/matrix interface, and more important for BVID, ply interface properties in a multi-ply laminate, are all parameters that determine impact resistance. After the composite structure experiences BVID, the depletion of the structural strength is determined as result of compression after impact (CAI) material testing.

The primary emphasis of this research is to predict structural compressive strength after low energy/low velocity impact using neural networks. After the composite structure absorbs BVID, it is common to determine structural strength depletion based on impact energy. Because impact energy is seldom known in real world applications, it is more reasonable to determine ultimate strength based on amount of damage present. The technique used in this research to assess the damage and predict ultimate strength includes ultrasonic testing (UT), to generate an image representing the damage, and neural networks to predict future performance.

Using the pixel data from the ultrasonic C-scan image of the impact damage, in conjunction with CAI testing, and analyzing it with a backpropagation neural network, correlations on ultimate compressive strength can be made. This analysis demonstrates the ability of a neural network to predict the ultimate compressive strengths of impact damaged composite structures using UT data.