The progressive degradation of presently operating electro-mechanical systems is a certain future fact. To minimize losses, maintenance costs and eventual replacements, condition monitoring should be applied to critical equipment (Condition Based Maintenance - CBM). The state of equipment can be predicted at any moment using statistical methods to analyze condition monitoring data. In this paper, collected data are vibration values, obtained at p points (p = 4 for instance) of an experimental equipment, forming p variables. When independence condition does not hold, it is suggested modeling data with Auto-Regressive Integrated Moving Average (ARIMA) models, and using the residues of the estimated model for Phase I. In Phase I, the estimation of parameters is achieved using the Hotelling T2 control chart; only after applying the defined ARIMA model, the p variables are treated. In Phase II, equipment state is artificially degraded through induced failures and failure prediction obtained using special multivariate control charts for data statistical treatment. Assuming data independence and normality, Multivariate ExponentiallyWeighted Moving Average Modified (MEWMAM) control charts are applied in Phase II to data collected from an electric pump, controlling the behavior of data using this procedure. In Phase II, for non-independent data the prediction errors from the adjusted model are used instead of original data. To show that the suggested methodology can be applied to propulsion systems, simulated data from a gas turbine are used. Using these methodologies it is possible to run online condition monitoring, and act in time, to minimize maintenance costs and maximize equipment performance.
|Number of pages||11|
|Journal||Journal of Vibrational Engineering and Technologies|
|Publication status||Published - 1 Dec 2015|
- Condition monitoring
- Multivariate Exponential Weighted Moving Average (MEWMA) control chart
- Statistical process control
- Vibration detection and analysis