Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

Neural networks in engineering diagnostics

Journal Article · · SAE Special Publication
DOI:https://doi.org/10.4271/941116· OSTI ID:81260
;  [1]
  1. Univ. of Illinois, Urbana-Champaign, IL (United States)
Neural networks are massively parallel computational models for knowledge representation and information processing. The capabilities of neural networks, namely learning, noise tolerance, adaptivity, and parallel structure make them good candidates for application to a wide range of engineering problems including diagnostics problems. The general approach in developing neural network based diagnostic methods is described through a case study. The development of an acoustic wayside train inspection system using neural networks is described. The study is aimed at developing a neural network based method for detection defective wheels from acoustic measurements. The actual signals recorded when a train passes a wayside station are used to develop a neural network based wheel defect detector and to study its performance. Signal averaging and scoring techniques are developed to improve the performance of the constructed neural inspection system. 12 refs., 13 figs., 2 tabs.
Sponsoring Organization:
USDOE
OSTI ID:
81260
Report Number(s):
CONF-9409289--; CNN: Grant MSS-9214910; Grant BCS-9201437
Journal Information:
SAE Special Publication, Journal Name: SAE Special Publication Journal Issue: 1048; ISSN SAESA2
Country of Publication:
United States
Language:
English