A rolling bearing signal model based on a correlation probability box

Abstract In this paper, the correlation between probability box (p-box) models was studied, and a method of constructing a correlation probability box (cp-box) model from raw bearing data was proposed. First, a copula function was used to construct the correlation model based on bearing data, and the inverse function of the cumulative distribution function was obtained by selecting the appropriate confidence intervals. Then, confidence bounds were obtained to establish the cp-box model. Next, the cp-box model was discretized into a series of focal elements by an average discretization method, and the measurement information of the cp-box model was obtained by an aggregated uncertainty measurement method. Finally, using a support vector machine as a pattern recognition tool, the effectiveness of the cp-box model was evaluated by a comparison with existing methods. The result shows that the copula function can be added to optimize the p-box model.