The rotor fault prediction based on support vector regression and phase space reconstruction

Support vector regression (SVR) is a popular machine learning method that develops these years and has been widely used in the prediction field. But the input feature vectors largely affect the accuracy of the forecast error, so the feature vector choice has been the hot issues of attention and research scholars. For these problems, some scholars have proposed a characteristics selection method of support vector regression machine based on the phase space reconstruction, but the value of the time delay and embedding dimension became discussion hotspot. For these problems, some scholars have proposed characteristics selection method of the support vector regression machine based on phase space reconstruction, but the value of the time delay and embedding dimension became discussion hotspot. So the optimization method of particle swarm optimization (PSO) is proposed. This method is able to quickly identify the best combination of parameters (tau, m, C, sigma) and improve forecast accuracy. This method is applied to the prediction of the rotor misalignment of rotating machinery fault data. The experiment proved that the method is feasible.