MINIMISATION OF DECISION ERRORS IN A PROBABILISTIC NEURAL NETWORK FOR CHANGE POINT DETECTION IN MECHANICAL SYSTEMS
✍ Scribed by G.M. LLOYD; M.L. WANG; T.L. PAEZ
- Publisher
- Elsevier Science
- Year
- 1999
- Tongue
- English
- Weight
- 298 KB
- Volume
- 13
- Category
- Article
- ISSN
- 0888-3270
No coin nor oath required. For personal study only.
✦ Synopsis
Probabilistic neural nets have been applied in the detection of structural damage. These networks, which rely upon approximating the multivariate density of the training data, have been shown to be e!ective in some applications. However, quanti"cation of decision errors, which must ultimately be used to rank their e!ectiveness, has received little attention. In this paper, a two-dimensional probabilistic pattern classi"er (PPC) based on an ¸ detector is studied. Each dimension is modelled as a Gaussian random variable in order to directly study the e!ects introduced by a commonly used density estimator. The region of acceptance is examined for di!erent parameters of the kernel density estimator. The detector behaviour of the original PPC is compared to theory. It is demonstrated that performance is a!ected by errors introduced by a kernel constructed using a "nite set of data, and also from theoretical limitations inherent in the test statistic. A modi"cation of the test statistic is suggested to improve the sensitivity for structural damage detection.