Kurtosis forecasting of bearing vibration signal based on the hybrid model of empirical mode decomposition and RVM with artificial bee colony algorithm

• EMD-ABCRVM is presented for kurtosis forecasting of bearing vibration signal.
• Establish IMFs’ prediction models by RVM with each appropriate embedding dimension.
• Select the appropriate kernel parameters of IMFs’ RVM models by artificial bee colony algorithm.

Accurate prediction for kurtosis of bearing vibration signal is helpful to find out the fault of bearing as soon as possible. As it is difficult to obtain an appropriate embedding dimension in creating directly the prediction model of kurtosis of bearing vibration signal by relevance vector machine (RVM), the hybrid model of empirical mode decomposition and RVM with artificial bee colony algorithm (EMD-ABCRVM) is proposed for kurtosis forecasting of bearing vibration signal. The seven decomposed signals with different frequency range can be obtained by empirical mode decomposition for kurtosis of bearing vibration signal. The prediction models of the seven decomposed signals can be established by RVM with their each appropriate embedding dimension, and artificial bee colony algorithm (ABC) is used to select the appropriate kernel parameters of their RVM models. Thus, each RVM model of the seven decomposed signals has appropriate embedding dimension and kernel parameter. In order to show the superiority of the proposed EMD-ABCRVM method, the RVM models with several different embedding dimensions and Gaussian RBF kernel parameters are used to compare with the proposed EMD-ABCRVM method. The experimental results show that it is feasible for the proposed combination scheme to improve the prediction accuracy of RVM for kurtosis of bearing vibration signal.