DMMD develops algorithms for preventive maintenance applications. Sub-periodic signals are extracted from vibration data that has been recorded using an accelerometer. These sub-periodic signals are then analyzed using time-frequency analysis tools.
The analysis of machine vibrations has proven to be a valuable application of signal processing. A variety of well-known techniques used in this area, require that good period and periodic sub-signal estimates can be made. Other applications of period estimation techniques are separation of periodic waveforms with overlapping spectra, finding musical rhythms and generally finding patterns in a wide variety of data sources. Our aim is to develop a method for periodic sub-signal estimation.
In the specific case, of dealing with the problem of recovery and detection of multiple sinusoidal signals there is an entire class of estimators, including: periodigram, Prony’s Method, Piseranko Harmonic Decomposition (PHD), the MUSIC, ESPRIT and IQML. For more references on Period Estimation techniques, the reader is encouraged to download our conference and/or journal papers.
Our work deals with the detection and estimation of machine vibration multi-periodic signals of unknown frequencies in white Gaussian noise. New estimates for the sub-signals (signals making up the received signal) and their periods are derived using an orthogonal subspace decomposition approach. We introduce the concept of exactly periodic signals. This in turn simplifies and enhances the understanding of periodic signals.