Batch data analysis

Three case studies on batch process data. In each one the trajectories of every batch are unfolded batchwise, one row per batch, so that PCA and PLS can describe how the batches differ from each other, flag the abnormal ones, and trace an event back to the variables involved and the time at which it happened.

Each case study is a plain Python script that sits next to its page. The page quotes the script section by section and states what each section prints; run the script to reproduce the numbers and to get every figure as an HTML file. The data are downloaded from openmv.net when the script runs.

• Batch PCA outlier diagnosis: the DuPont polymerization reactor
• Batch PLS fault diagnosis: the simulated SBR reactor
• Multiblock batch PLS: the FMC batch dryer