Loads

Every load has the same system-independent interface. Pass axes directly for an explicit, reusable call:

history = load.generate(time=time, positions=positions, seed=42)

history.time  # caller-supplied time axis
history.positions  # caller-supplied position axis
history.values  # shape: (n_positions, n_times)

Fixed loads use one position argument for both MDOF systems and beams:

load = loads.Harmonic(position=0, amplitude=10.0, frequency=2.0)

history.values places each point force at the nearest point on positions for plotting and storage. The exact coordinates are retained in history.point_forces; simulation uses those coordinates instead of the nearest-grid approximation. Consequently, a two-wheel diagnostic field contains at most two nonzero spatial samples in each time column without making the physical response depend on grid resolution.

vehicle = loads.TwoWheelLoad(
    front_load=-4000.0,
    rear_load=-3500.0,
    speed=2.5,
    wheel_spacing=1.25,
)

time = np.arange(600) / 200
positions = np.linspace(0.0, 5.0, 101)
history = vehicle.generate(time, positions)

Moving, two-wheel, and runner trajectories require a system with a continuous position domain, currently Beam. MDOF systems use discrete integer DOF indices; apply fixed-position loads to those indices instead of interpolating a moving trajectory between them.

Available loads are WhiteNoise, Harmonic, Impulse, Chirp, TimeSeries, MovingLoad, TwoWheelLoad, RunnerLoad, and FunctionLoad. Loads can be added; their fields are summed on the same axes.

Custom functions receive the axes and a random generator:

def force(inputs: loads.ForceInput) -> np.ndarray:
    spatial = np.sin(np.pi * inputs.positions / inputs.positions[-1])
    temporal = np.sin(2 * np.pi * 2.0 * inputs.time)
    return spatial[:, None] * temporal[None, :]


custom = loads.FunctionLoad(force)

Function loads are distributed grid fields and are projected from the supplied positions. Custom exact forces can instead return a LoadHistory with PointForceComponent entries. Set distributed_values to the independently projectable grid field, or to a zero field when the history contains point forces only. values is derived as the complete nearest-grid diagnostic shown to users and stored in datasets.