Process Performance Monitoring

Pattern recognition contributes in several areas of the scientific and analytic process. It underlies a substantial portion of cognitive theory regarding spatial thought, and can be a fundamental step in the development of scientific theory as well. Our ASOM (Adaptive Self-Organizing Maps) overcomes the topological constraints. Through pattern evolving process, it can discover new clusters, which is practically useful in unsupervised pattern recognition without a prior knowledge of cluster numbers. We focus on a trending based mechanism for real-time abnormal event detection. This method has shown to be sensitive to small signal variation and quite effective in detecting abnormal signals embedded in the normal signals. Test results have shown that our method can detect the abnormal events by analyzing only the raw sensor measurements.