The process historian is not the plant

Direct answer

A historian stores values and timestamps. It does not necessarily store the operating regime, instrument condition, disturbances, manual actions or constraints that explain those values.

Data volume does not guarantee that the history can distinguish a physical phenomenon, noise, a regime change and the operating response. Before applying an algorithm, determine which decisions the dataset can reconstruct.

What must accompany the data

• operating regime during the observed interval;
• the instrument that generated the signal and its service condition;
• availability, resolution and synchronisation;
• changes in ore, water, reagent or configuration;
• feed disturbances;
• operator and control-system actions;
• alarms, maintenance and periods of restricted operation.

Without this context, a model can find consistent relationships within a mixture of different physical states. That relationship may no longer hold when the regime changes.

Implications for modelling

Historical-data preparation should separate, where possible, identification data, validation data and periods requiring review. It must also retain the decisions made for cleaning, exclusion or imputation.

The output is not just a database with fewer missing values. It should explain which part of the process is observed, which part remains hidden and which uncertainty remains.

Sources and evidence status

Discovery source LinkedIn post dated 4 Aug 26.

Use of the source The post provides the distinction between records and operational context.

Publication status PUBLISHED. No specific installation or modelling result is presented.

Editorial scope The metadata list must be checked against each plant's historian and control architecture.