No abstract

No abstract

A novel monitoring strategy is proposed for multimode process in which mode clustering and fault detection based on geodesic distance (GD) are integrated. To start with, the empowered adjacency matrix of normalized training dataset is obtained and improved Dijkstra algorithm (IDA) is utilized to calculate the geodesic distance between each sample data so as to characterize the shortest distance of the nonlinear data within local areas accurately. Next, GD matrix algorithm is presented as an optimal clustering solution for a multimode process dataset. Then, the GDS model is established in each operating mode. Monitoring statistics based on the power of geodesic distance are structured based on square sum of Euclidean distances. Once the test data is detected as fault data, mode location based on deviation coefficient is conducted to narrow the scope of the inspection fault. Finally, the validity and usefulness of the proposed GDMPM monitoring method are demonstrated through the Tennessee Eastman (TE) benchmark process.

No abstract

No abstract