Making a Plant Live Faster Than Realtime
Exploring deterministic virtual time for simulated industrial environments, so operational behaviour can run faster than realtime without losing reproducibility.
PROJECTProcedural Industrial Environment Simulator →Industrial systems spend a surprising amount of time waiting.
Processes settle. Operators arrive. Maintenance windows open. Pumps run for hours. Equipment fails after days rather than seconds.
That is realistic, but not especially convenient when the environment exists to support experiments.
The problem
If a simulation simply runs against wall-clock time, even relatively small experiments can become painfully slow.
A twelve-hour operational sequence should not require twelve hours of waiting just to generate another dataset.
At the same time, simply making every timer shorter changes the behaviour being modelled.
The interesting question is therefore:
Can the simulated environment retain realistic temporal relationships while allowing the entire world to advance faster than realtime?
Deterministic virtual time
The current approach is to separate simulated time from wall-clock time.
Operational behaviour is scheduled against a deterministic simulation clock rather than directly against the host operating system’s clock.
That allows the environment to advance quickly while still preserving things such as:
- daily operational schedules;
- maintenance windows;
- process settling periods;
- operator activity;
- periodic polling;
- faults and recovery;
- reproducible event ordering.
The important constraint is determinism.
Given the same configuration and seed, the same events should occur in the same order at the same simulated times.
Why this matters
Faster-than-realtime execution is useful for more than convenience.
It makes it practical to generate long operational histories, repeat experiments, compare model behaviour and produce datasets representing weeks or months of simulated activity.
More importantly, it allows time itself to become part of the experimental apparatus without sacrificing reproducibility.
Next
The next step is connecting simulated activity to explicit ground-truth provenance.
A packet should not merely exist.
The simulator should be able to explain why it exists.