A process historian is a database that records every reading in a plant, with its time, and keeps it for years. A reading is one measured value, such as the temperature of a furnace or the current drawn by a motor. The historian is built to store a very large number of such readings every second.
The data it stores is time-series data: a long list of values, each with the time at which it was measured. Each measuring point in the plant is called a tag, for example the outlet pressure of pump 3. A plant with 20,000 tags recording one value every second stores 20,000 readings every second. That is about 1.7 billion readings a day. A historian compresses these readings, so that it can keep them for years in a reasonable amount of storage.
For instance, a pump in a fertiliser plant stops at 2:14 a.m. and the night shift restarts it. The next morning, the maintenance engineer opens the historian and plots the pump's pressure, current and temperature for the previous hour. The curve shows that the pressure began to fall at 2:05 a.m., nine minutes before the stop. The engineer now knows that the cause was a blocked suction line and not a motor fault.
Historians help in three ways. First, engineers find the cause of a fault by looking back at what happened. Second, they compare good days with bad days to run the plant better. Third, other software uses the same data. Asset performance management studies it to predict failures, an industrial data platform combines it with other data, and a plant digital twin uses it to keep a software copy of the plant up to date. In an oil field, the same data supports production accounting.
An ordinary database could not do this job. A business database is built for records that change, such as a customer's address. In normal use, a historian does not change a reading once it is written. It only adds new ones, millions of them every minute, and it can answer a question such as "show the pressure of pump 3 for the last week" in seconds. To save space, it often stores a new reading only when the value has moved by more than a small amount. A flat line then takes almost no storage.
The historian receives its readings from SCADA and the distributed control system. It is not used to control the plant. It only records. The users are the plant's process engineers, the maintenance team and, in regulated industries, the auditors who want proof of what the plant did on a given day.