Fault Detection and Diagnosis Based on Adam-ICA

The traditional FastICA method uses the Newton iteration method which often leads to local minimum solution, so that the suitable independent components are not obtained. To solve this problem, a modified ICA algorithm based on adaptive gradient algorithm (Adam) called Adam-ICA is proposed for the purpose of multivariate statistical processes monitoring (MSPM). The basic idea of this algorithm is to extract the main independent components from the original data by adaptive low-order moment estimation for fault diagnosis and analysis. The proposed monitoring method is applied to fault detection and diagnosis of the vehicle dataset. The results show that the Adam-ICA method is more efficient.