APPLICATION OF NONPARAMETRIC REGRESSION IN LOCATORS OF ACOUSTIC-EMISSION SOURCES

This paper describes a locator of acoustic emission sources based on a non-parametric regression procedure. It includes a network of AE sensors, a triggered clock and PC. A learning and an analysis phases are characteristic of the operation of a locator. During learning, AE bursts in the specimen are detected by sensors, while the clock measures the time delays between them. In the computer, a particular AE burst is represented by a vector comprised of time delays and corresponding source coordinates. In this phase, a set of experiments provides prototype data which are stored in the computer and represent an empirical model of the AE phenomenon. After learning, the locator is capable of determining the source position from measured time delays throughout the procedure of non-parametric regression. It Is shown in the paper that the same locator is applicable on samples of various geometrical properties and with various sensor network configurations.