Operation, monitoring, diagnostics, maintenance, planning

Since process monitoring and diagnostics are closely tied to data from production units, these fields began using data analysis methods early on to achieve specific objectives. Since the 1980s, numerous demonstrations have been conducted, and effective, comprehensive approaches are now available. These approaches are unique in that they seamlessly combine several learning methods into a single framework, such as clustering, dimensionality reduction, and classification methods.

Figure 12 provides an overview of the development and operational phases of a method for automatically diagnosing process faults or deviations. During the modeling phase, data is collected from historical process data and previous diagnostics, enabling the association...