Skip to content

Cycles and Initial Conditions

A cycle is the sequence of operating steps a bed repeats — feed, blowdown, purge, pressurisation, and so on. In Skarstrom a cycle is a list of steps, each with its own duration and its own pair of boundary conditions, executed in order and then repeated. A deck's cycle also holds the connections between steps for multi-bed processes and the KPI configuration for the performance metrics.

Start from a template. data/cycles/ holds ready-made cycles: breakthrough.json, 4_step_vpsa.json, 4_step_vpsa_lpp.json (with light-product pressurisation), skarstrom_cycle.json and 6_step_vpsa_rinse.json. They are written by core/test/unit/test_cycle_designs.cpp, which is also the clearest example of building a cycle in C++, and load with adsim::LoadCycleFromJSON. The example decks are complete processes to adapt.

Steps

Each step has a name and two boundary conditions:

  • the Top boundary at \(z = L\) (the far end of the column), and
  • the Bottom boundary at \(z = 0\) (the near end).

A step can also apply an energy source to a layer (heating or cooling). In C++ a cycle is edited with Cycle::AddStep, Cycle::GetStep, DuplicateStep (which copies a step with its settings and clears connections on the copy) and the boundary-condition factories; from Python the existing steps' values are parameters:

sim.steps                          # ['Feed', 'Blowdown', 'Evacuation', 'Pressurisation']
sim["step[Feed].time"] = 60.0      # by name (alias) ...
sim["step[1].time"] = 60.0         # ... or by the step's stable ID

A few rules are enforced by the deck checks before a run:

  • A cycle must contain at least one step, and step names must be unique.
  • The two boundaries of a step cannot both be inlets — one must be an outlet (or a closed end).
  • The Top and Bottom boundaries must be at different ends of the column.

Number of cycles

A PSA/TSA process must run many cycles before it settles into a repeating pattern. The deck's number of cycles (cycle.number_of_cycles) is how many times the whole step sequence repeats; the total simulated time is the cycle time multiplied by the number of cycles. sim.run(cycles=N) and skarstrom run --cycles N override it for one run.

Rather than guessing the number of cycles, let the solver stop automatically once the bed stops changing from one cycle to the next — see cyclic steady state. The number of cycles is then an upper limit.

Step timing

The step time mode controls how step durations are given:

  • Absolute — each step's duration in seconds (step[...].time). The cycle time is the sum of the step times.
  • Fraction — a single total cycle time (cycle.total_time) and a fraction of it per step (step[...].fraction). The fractions must sum to 1. This makes it easy to rescale a whole cycle by changing one number.

Step times are expressions, so two steps can share a global variable (sim["step[Rinse].time"] = "t_feed").

Initial conditions

The initial conditions set the state of the bed everywhere at \(t = 0\), before the first step begins:

  • Pressure
  • Bed temperature (the initial gas and solid temperature)
  • Wall temperature
  • Velocity
  • Initial mole fraction for each component — these must sum to 1.

Choose initial conditions that are physically consistent with the first step's boundary conditions; for example, if the first step feeds gas at high pressure into an evacuated bed, start from a low pressure. A poor initial guess can make the first solve harder to converge. (If the model reaches cyclic steady state, the initial condition only affects how many cycles that takes, not the final answer.)

Next: define what happens at each end of the column in Boundary Conditions.