《Neural network approach of active ultrasonic signals for structural health monitoring analysis》.pdf

Neural Network Approach of Active Ultrasonic Signals for Structural Health Monitoring Analysis Zachary Kral, Walter Horn, and James Steck Dept. of Aerospace Engineering, Wichita State University, Wichita, KS 67260 ABSTRACT Maintenance is an important issue for aerospace systems, since they are in service beyond their designed lifetime. This requires scheduled inspections and damage repair before failure. Research is in progress to develop a structural health monitoring system (SHMS) to improve this maintenance routine. Ultrasonic testing, utilizing a system of piezoelectric actuators and sensors, is a promising concept Measured wave signals are compared with signals for previously scanned states. Changes to the signal could be the result of damage to the component. This paper focuses on analyzing the differences of states, using artificial neural networks. Neural network analysis has the potential of creating a SHMS of greater ability and processing. Experiments were performed on a thin, flat aluminum panel. Ultrasonic actuators and sensors were installed and a baseline scan was performed on the undamaged panel. Simulated damage was introduced in specific areas, and scans were conducted for several damaged states. Neural networks were created to assess the changing conditions of the panel. The system was later tested on a lap joint specimen to confirm the abilities of the neural network. This form of analysis performed well at locating and quantifying areas of change within the structure. The neural network performance indicated that it has a role in the SHMS of aerospace structures. Keywords: Structural health monitoring, artificial neural network, ultrasonic testing, damage det