Process Monitoring Based on LTSA-Greedy-SVDD

There are some characteristics such as non linearity and non Gaussian in real industrial process data. Aiming at these problems, a process monitoring method based on the local tangent space alignment is proposed. Firstly, the local tangent space alignment algorithm is used to get the sub manifold of low dimension from the normalized normal sample data such that the dimension reduction can be achieved. Then, Greedy method is used to extract feature sample to establish the monitoring model by support vector data description. Finally, the corresponding statistic is used for process monitoring. The simulation is made on the TE model, whose results illustrate the effectiveness of the proposed method.