Update ic_engine.py example

[ischoegl]

Cantera version

2.4 / 2.5a3

Operating System

all

Python/MATLAB version

python

Expected Behavior

The example should run simulations in an efficient manner (i.e. use Func1 routines for valve train and injection timing)

Actual Behavior

Implementing valve train and injection in the main simulation loop (between sim.advance steps) requires a time step that is set much smaller than necessary, making the example highly inefficient.

Steps to reproduce

1. run stock example ic_engine.py

[bryanwweber]

Could you be more specific about where Func1 routines should be used? When Python functions are passed to methods like set_velocity() it is equivalent to directly creating a Func1 class instance (to my understanding). I think the example could certainly be improved (and see the comments on #587 as well) but I'm not sure what you mean here.

[ischoegl]

The Func1 approach should be used for inlet_valve.set_valve_coeff, outlet_valve.set_valve_coeff and injector_mfc.set_mass_flow_rate. I haven't run any tests yet, but there is potential to simplify (eliminate?) the refinement machinery also (using the newly implemented r.set_advance_limit capabilies). I will submit a PR before long.

[ischoegl]

@bryanwweber ... I ran some tests and realized that Func1 for valves does not accept time as an input, which effectively blocks this approach. I consider this a bug, as the set_valve_coeff should replace the valve coefficient, not replace the valve equation, see:

void Valve::updateMassFlowRate(double time)
{
    if (!ready()) {
        throw CanteraError("Valve::updateMassFlowRate",
                           "Device is not ready; some parameters have not been set.");
    }
    double delta_P = in().pressure() - out().pressure();
    if (m_func) {
        m_mdot = m_func->eval(delta_P);
    } else {
        m_mdot = m_coeffs[0]*delta_P;
    }
    m_mdot = std::max(m_mdot, 0.0);
}

Imho, the function should be implemented as:

void Valve::updateMassFlowRate(double time)
{
    if (!ready()) {
        throw CanteraError("Valve::updateMassFlowRate",
                           "Device is not ready; some parameters have not been set.");
    }
    double delta_P = in().pressure() - out().pressure();
    if (m_func) {
        m_mdot = m_func->eval(time)*delta_P;
    } else {
        m_mdot = m_coeffs[0]*delta_P;
    }
    m_mdot = std::max(m_mdot, 0.0);
}

PS: the python documentation does describe the implemented behavior, which, however, does not correspond to the expected function name.

[bryanwweber]

Thanks @ischoegl I think this is probably what @speth was referring to on #587

[ischoegl]

Correct. As @speth commented

The other is to set the parameters from a function that is called as part of the ODE evaluation, e.g. the way we allow MassFlowController to have a mass flow rate as a function of time.

I opened a new issue (#666) that seeks to clarify some of the inconsistencies of the interface.