Heterogeneous recurrence T 2 charts for monitoring and control of nonlinear dynamic processes

摘要: Many real-world systems are evolving over time and exhibit dynamical behaviors. Real-time sensing brings the proliferation of big data (i.e., dynamic, nonlinear, nonstationary, high dimensional) that contains rich information on nonlinear dynamic processes. Nonetheless, limited work studying dynamics underlying for quality control has been reported. This paper presents a new approach heterogeneous recurrence T2 chart online monitoring anomaly detection in A partition scheme, named Q-tree indexing, is firstly introduced to delineate local regions multidimensional continuous state space. Further, we designed fractal representation transitions, among regions, then develop measures quantify patterns. Finally, developed multivariate Hotelling Chart on-line predictive process recurrences. Case studies show proposed not only captures patterns transformed space, but also provides an effective charts monitor detect transitions process.