| Abstract: | In order to make trend analysis and predic tion to acquisition data in a mechanical equipment condition monitoring system, a new method of trend feature extraction and prediction of acquisition data is p roposed which constructs an adaptive wavelet on the acquisition data by means of second generation wavelet transform (SGWT). Firstly, taking the vanishing momen t number of the predictor as a constraint, the linear predictor and updater are designed according to the acquisition data by using symmetrical interpolating sc heme. Then the trend of the data is obtained through doing SGWT decomposition, t hreshold processing and SGWT reconstruction. Secondly, under the constraint of t he vanishing moment number of the predictor, another predictor based on the acqu isition data is devised to predict the future trend of the data using a non-sym metrical interpolating scheme. A one-step prediction algorithm is presented to predict the future evolution trend with historical data. The proposed method obt ained a desirable effect in peak-to-peak value trend analysis for a machine se t in an oil refinery. |
