With the advancement of sensing and information technology, high-dimensional spatial-temporal (ST) data are readily available in many engineering applications. Characterization of the patterns imbedded in such ST data is of critical importance for understanding the underlying engineering processes and support decision-making for process monitoring and control. Conventional statistical analysis and process control methodologies fall short in explicit consideration of spatial and temporal correlations. In this research, a generic approach to modeling the patterns in ST data is developed based on functional regression. Regularizations are incorporated in the model estimation to achieve appropriate model selection and thus achieve desirable analytical results. The effectiveness of the proposed approach is demonstrated in a water distribution system burst detection and a high-speed video processing.
