Incipient Fault Detection Using Higher-Order Statistics
Abstract
This study balances the development of theory and its application to real an simulated incipient fault data from systems which have cyclostationary properties. The study's theoretical contribution reveals the advantages of approaching estimation of time series in a general framework where estimation of the cumulant spectrum can reveal implications for three classes of stochastic processes: stationary, cyclostationary, and nonstationary. The developed cumulant spectrum estimation capability provides for feature construction in addition to bispectrum and power spectrum estimates of stochastic process data. Actual experimental data is obtained to study the incipient wear process of manufacturing drill bits cutting through epoxy glass composite material used for construction of electronic semiconductor panels. The fluctuating vibrations caused by the drill bits cutting through the epoxy-glass composite material used for construction of electronic semiconductor panels. The fluctuating vibrations caused by the drill bits cutting through the epoxy-glass composite are not subject to precise prediction, nor are the external noise, measurement errors, and other disturbances in the transmission of the vibration signal to three accelerometers mounted on the drilling machine considered to have the same characteristic of unpredictability.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 1991
- Accession Number
- ADA243493
Entities
People
- Richard W. Barker
Organizations
- Air Force Institute of Technology