PEM

pem.py

Module for full multidisciplinary Hall thruster predictive engineering model(s).

Note

Multidisciplinary systems are specified using the SystemSurrogate object from the amisc package. This data structure allows feedforward and feedback connections between a set of component models. It can be used to build a surrogate for the MD system, or to evaluate the system directly using the underlying component models.

Includes

• pem_v0() - The v0 cathode-thruster-plume feedforward multidisciplinary system.

pem_v0(save_dir=None, executor=None, init=True, hf_override=False, var_file=CONFIG_DIR / 'variables_v0.json', from_file=None)

Return a SystemSurrogate object for the feedforward v0 PEM system.

PARAMETER	DESCRIPTION
save_dir	where to save surrogate and model outputs TYPE: str | Path DEFAULT: None
executor	the parallel execution manager TYPE: Executor DEFAULT: None
init	whether to initialize the surrogate (will evaluate all component models) TYPE: bool DEFAULT: True
hf_override	whether to use only highest-fidelity for all models TYPE: bool DEFAULT: False
var_file	the path to the .json config file storing information about all variables TYPE: str | Path DEFAULT: CONFIG_DIR / 'variables_v0.json'
from_file	the .pkl save file to load the surrogate from, (instead of building from scratch) TYPE: str | Path DEFAULT: None

RETURNS	DESCRIPTION
SystemSurrogate	the SystemSurrogate object

Source code in src/hallmd/models/pem.py

def pem_v0(save_dir: str | Path = None, executor: Executor = None, init: bool = True,
           hf_override: bool = False, var_file: str | Path = CONFIG_DIR / 'variables_v0.json',
           from_file: str | Path = None) -> SystemSurrogate:
    """Return a `SystemSurrogate` object for the feedforward v0 PEM system.

    :param save_dir: where to save surrogate and model outputs
    :param executor: the parallel execution manager
    :param init: whether to initialize the surrogate (will evaluate all component models)
    :param hf_override: whether to use only highest-fidelity for all models
    :param var_file: the path to the `.json` config file storing information about all variables
    :param from_file: the `.pkl` save file to load the surrogate from, (instead of building from scratch)
    :returns: the `SystemSurrogate` object
    """
    exo_vars = load_variables(['PB', 'Va', 'mdot_a', 'T_ec', 'V_vac', 'P*', 'PT', 'u_n', 'l_t', 'vAN1', 'vAN2',
                               'delta_z', 'z0', 'p0', 'c0', 'c1', 'c2', 'c3', 'c4', 'c5', 'sigma_cex', 'r_m'],
                              var_file)
    coupling_vars = load_variables(['V_cc', 'I_B0', 'I_D', 'T', 'eta_v', 'eta_c', 'eta_m', 'ui_avg', 'theta_d'], var_file)

    if from_file is not None:
        surr = SystemSurrogate.load_from_file(Path(from_file), stdout=False, executor=executor)
        for v in exo_vars:
            # Make sure nominal values are up to date with the current config file
            j = surr.exo_vars.index(v)
            surr.exo_vars[j] = v
            for node, node_obj in surr.graph.nodes.items():
                try:
                    j = surr[node].x_vars.index(v)
                    surr[node].x_vars[j] = v
                except:
                    pass

        for v in coupling_vars:
            j = surr.coupling_vars.index(v)
            surr.coupling_vars[j].tex = v.tex
            surr.coupling_vars[j].nominal = v.nominal

        return surr

    # Get number of reconstruction coefficients for ion velocity and ion current density profiles
    try:
        with open(CONFIG_DIR / 'thruster_svd.pkl', 'rb') as fd, open(CONFIG_DIR / 'plume_svd.pkl', 'rb') as fd2:
            d = pickle.load(fd)
            r1 = d['vtr'].shape[0]
            d2 = pickle.load(fd2)
            r2 = d2['vtr'].shape[0]
    except FileNotFoundError:
        r1 = 4
        r2 = 2

    coupling_vars.extend([UniformRV(-20, 20, id=f'uion{i}', tex=f"$\\tilde{{u}}_{{ion,{i}}}$",
                                    description=f'Ion velocity latent coefficient {i}',
                                    param_type='coupling') for i in range(r1)])
    coupling_vars.extend([UniformRV(-20, 20, id=f'jion{i}', tex=f"$\\tilde{{j}}_{{ion,{i}}}$",
                                    description=f'Current density latent coefficient {i}',
                                    param_type='coupling') for i in range(r2)])

    # Component inputs
    cathode_exo = ['PB', 'Va', 'T_ec', 'V_vac', 'P*', 'PT']
    thruster_exo = ['PB', 'Va', 'mdot_a', 'T_ec', 'u_n', 'l_t', 'vAN1', 'vAN2', 'delta_z', 'z0', 'p0']
    plume_exo = ['PB', 'c0', 'c1', 'c2', 'c3', 'c4', 'c5', 'sigma_cex', 'r_m']

    # Models should be specified at the global scope for pickling
    cathode = ComponentSpec(cc_feedforward, name='Cathode', exo_in=cathode_exo, coupling_out='V_cc',
                            surrogate='analytical')
    thruster = ComponentSpec(hallthruster_jl_wrapper, name='Thruster', exo_in=thruster_exo, truth_alpha=(2, 2),
                             max_alpha=(2, 2), coupling_in='V_cc', max_beta=(2,)*(len(thruster_exo)+1), save_output=True,
                             coupling_out=['I_B0', 'I_D', 'T', 'eta_v', 'eta_c', 'eta_m', 'ui_avg'] + [f'uion{i}' for i in range(r1)],
                             model_kwargs=dict(n_jobs=-1, compress=True, hf_override=hf_override))
    plume = ComponentSpec(plume_feedforward, name='Plume', exo_in=plume_exo, coupling_in='I_B0', surrogate='analytical',
                          coupling_out=['theta_d'] + [f'jion{i}' for i in range(r2)], model_kwargs={'compress': True})

    logger_name = 'SF-surrogate' if hf_override else 'MF-surrogate'
    surr = SystemSurrogate([cathode, thruster, plume], exo_vars, coupling_vars, executor=executor, init_surr=init,
                           stdout=False, save_dir=save_dir, logger_name=logger_name)

    return surr