New Advances in Space SHM Project

Abstract

This paper is based on a consistent family of experimental data obtained in a national research project. More accurate, specimens representing spacecraft structures, accomplished from aluminum circular plates with PWAS bonded on them were subjected to extreme temperature variations and radiations, both specific to space applications. The structure itself is affected by mechanical damages caused by fatigue and aging. These mechanical damages were simulated by laser fabricated slit cuts. The signature of structure’s health is seen as the real part of electromechanical impedance (EMI) curves of a PWAS bonded on structure. Whatever the EMI signature (recorded via special devices) changes, it is important that it be signaled online. It is shown that a neural network (NN) has the willingness to “learn”, thus identifying a function more or less complicated, as it is the case with the real part of EMI characteristic. In view of NN preparation for in-situ installing, at least the following aspects must be elucidated: setting the number of iterations to learning; evaluation of common damages that can appear in the structure; investigations on their evolution time; investigations on the possible values of learning errors; the default value of error to stop the learning process.

Keywords

• spacecraft structures
• structural health monitoring
• piezoelectric wafer active sensors
• electromechanical impedance
• online identification