Enum ConvergenceTest 

pub enum ConvergenceTest {
    NormUnbalance {
        tol: f64,
        max_iter: usize,
    },
    NormDispIncr {
        tol: f64,
        max_iter: usize,
    },
    EnergyIncr {
        tol: f64,
        max_iter: usize,
    },
    Combined {
        force_tol: f64,
        moment_tol: f64,
        relative_tol: f64,
        displacement_tol: Option<f64>,
        max_iter: usize,
    },
}

Convergence criterion for Algorithm::Newton/KrylovNewton. Closed enum . Unused by Algorithm::Linear (which never iterates).

Every variant is evaluated against the accepted state of an iteration (IterationContext): the residual reassembled after the iteration’s correction (and any line-search rescaling) was applied, not the residual that correction was solved against. A force test can therefore never pass on a stale pre-correction residual.

Variants

NormUnbalance

Converged once the Euclidean norm of the post-update unbalanced force residual drops below tol.

Fields

tol: f64

max_iter: usize

NormDispIncr

Converged once the displacement correction’s norm drops below tol. Does not certify force equilibrium on its own (a stiff or stalled iteration can take a tiny step at a large residual).

Fields

tol: f64

max_iter: usize

EnergyIncr

Converged once the (absolute) incremental work |du . residual| (accepted correction against the accepted residual) drops below tol. Pairs force with translation and moment with rotation consistently, but its value changes with the work unit and can be small through cancellation, so it does not certify force equilibrium on its own either.

Fields

tol: f64

max_iter: usize

Combined

Every configured criterion must pass:

• Force equilibrium (always): on each reduced equation i, |r_i| <= abs_i + relative_tol * F_ref_i, where abs_i is force_tol for translational equations and moment_tol for rotational ones, and F_ref_i is the largest of the committed and predicted external force and the committed internal force on that equation, frozen at the start of the step. A per-equation maximum is used rather than a norm so many quiet DOFs can’t dilute one unbalanced one, and the reference is never the initial or current residual (either can be zero) or the current load alone (which can pass through zero during continuation).
• Displacement correction (when displacement_tol is set): the largest component of the last correction is at most displacement_tol. Under arc length this is the last undamped correction, so a tiny accepted backtracking step can’t fake it.

Fields

force_tol: f64

moment_tol: f64

relative_tol: f64

displacement_tol: Option<f64>

max_iter: usize

Implementations

impl ConvergenceTest

pub fn validate(&self) -> Result<(), AnalysisError>

Rejects nonfinite or nonpositive tolerances and a zero iteration budget. relative_tol may be zero; the absolute tolerances may not. Every Newton-family step runs this; it is public so wire decoders can reject bad input up front.

Trait Implementations

impl Clone for ConvergenceTest

fn clone(&self) -> ConvergenceTest

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Copy for ConvergenceTest

impl Debug for ConvergenceTest

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl PartialEq for ConvergenceTest

fn eq(&self, other: &ConvergenceTest) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for ConvergenceTest