Time Interval and Time Delays

You can model quasi-steady-state processes having significant time delays among process variable interactions by specifying a time delay value for each variable. The time delay value adjusts the temporal relationship of a process variable with respect to the other variables in the process. A positive time delay value advances the variable in time; a negative value moves it back in time. One-time delay unit represents the time interval between rows in the after-transforms dataset (see Delay Filter or Dynamic Filter, but in general, you want to perform a time merge followed by a dynamic alignment).

Before executing a time merge, you must determine the time interval that you will want between rows in the resulting dataset. Time statistics information on the time intervals currently in the data may help you to decide on an appropriate interval. As a rule, the time interval should simply reflect the rate at which you want to obtain output predictions. However, the time interval should not be smaller than the sampling interval of the most frequently sampled process input variable.

In the special case of a process that does not include time delays, the only reason to time merge would be if the dataset were not row synchronous. In this case, any convenient time interval could be used.