Utilities

AbstractPenalty{T}

Abstract type for SIMP-style penalties applied to density variables. T is the numeric type. Concrete subtypes: PowerPenaltyFun, RationalPenaltyFun, SinhPenaltyFun, ProjectedPenaltyFun.

PowerPenaltyFun(p)

Classic SIMP power penalty: ρ^p. p > 1 penalises intermediate densities. The most common choice in topology optimization.

RationalPenaltyFun(p)

Rational SIMP penalty: x / (1 + p * (1 - x)). Produces a smoother penalty than PowerPenaltyFun for the same exponent.

SinhPenaltyFun(p)

Hyperbolic-sine penalty: sinh(p*x) / sinh(p). An alternative smooth penalty.

ProjectedPenaltyFun(penalty, proj)

Composite penalty that applies a projection (default HeavisideProjectionFun(10)) before the penalty. Pushes the design toward 0/1 and then penalises.

AbstractProjection

Abstract type for projection functions that push densities toward 0/1. Subtypes: HeavisideProjectionFun, SigmoidProjectionFun.

HeavisideProjectionFun(β)

Heaviside projection with steepness β. Larger β produces sharper 0/1 transitions. y = 1 - exp(-β*x) + x*exp(-β).

SigmoidProjectionFun(β)

Sigmoid projection with steepness β. y = 1 / (1 + exp((β+1)*(-x+0.5))).

getpenalty(solver)

Return the current penalty object of solver (or of a function wrapping a solver).

getprevpenalty(solver)

Return the previous penalty object (before the last setpenalty! call).

setpenalty!(solver, p)

Update the penalty of solver to p (a number or an AbstractPenalty). Stashes the old penalty in getprevpenalty.

density(var, xmin)

Map a design variable var ∈ [0, 1] to a physical density in [xmin, 1]: ρ = var * (1 - xmin) + xmin. This is the interpolation step that gives void (xmin) a small but nonzero stiffness to keep the system non-singular.

RaggedArray(vv::Vector{Vector{T}})

Store a vector of variable-length vectors in a single flat values array with an offsets array marking where each row begins. Used to pack the per-cell DOF and cell connectivity tables of the FEA metadata.

compliance(Ke, u, dofs)

Compute the local compliance uᵀ Ke u over the degrees of freedom dofs.

sumdiag(K::Union{StaticMatrix, Symmetric{<:Any, <:StaticMatrix}})

Computes the sum of the diagonal of the static matrix K.

meandiag(K::AbstractMatrix)

Compute the mean of the absolute diagonal entries of K.

getsolver(f)

Return the solver wrapped by f (a function or algorithm that stores its solver in the solver field).

@debug expr

Evaluate expr only when the package-level DEBUG flag is enabled.

@forward_property T field

Forward any getproperty/setproperty! access on T to its field, so T transparently exposes that field's properties.