Previous: LIGO Detector Next: EUV Scanner

The Self-Carving Piston — Generative Design

Generative design · c. 2020

The Self-Carving Piston — Generative Design

Preparing exhibit…

Operate

Step computes one optimizer iteration: a full finite-element solve, which can take tens of seconds. Reset starts a new run from the uniform field and clears the history; the target volume is kept.

TARGET MATERIAL VOLUME

The target applies from the next iteration. Lower targets remove more material and make the part less stiff; the peak-stress warning threshold is illustrative.

Inspect

TRAINING HISTORY

Drag to inspect earlier density fields in the optimization history.

How it works

This SIMP optimizer minimizes structural compliance under a material-volume constraint. The density field and stress post-processing are educational numerical results, not a certified component design.

SIMP step	Machine-learning analogy
density variables x[e]	trainable parameters
FEA stress solve	forward pass
sensitivities dc/dx	gradient / backprop
OC density update	gradient-descent step
constraint / stress warning	instability / constraint failure

OPTIMIZER STATE

Iteration / epochVolume / materialCompliance / lossPeak von Mises (model units)

CONVERGENCE

compliance (log scale)volume fraction, target dashed

MATERIAL ⇄ STIFFNESS PROXY

Material (volume fraction) against relative stiffness (first analysed compliance divided by current compliance).

STATE

.museum-body--immersive { overflow: auto !important; } .museum-main--fill, .museum-runtime { height: auto !important; min-height: 100vh; overflow: visible !important; } .museum-runtime > .exhibit-root, .museum-runtime > :not(#mw-about):not(noscript):not(script) { display: none !important; } .museum-runtime #mw-about { display: block !important; position: static; width: min(100%, 800px); margin: 24px auto; padding: 24px; box-sizing: border-box; } .museum-runtime [data-about-close], .museum-runtime #mw-about > h2 { display: none !important; }

c. 2020 · Generative design

The Self-Carving Piston — Generative Design

A SIMP topology optimizer redistributes material in a piston-shaped design domain under modeled loads and supports.

Topology optimization seeks a material distribution that improves a stated objective while satisfying constraints. This exhibit minimizes structural compliance at a selected material-volume fraction for a piston-shaped design domain with prescribed supports and a unit loading case.

Each iteration solves a finite-element model, computes compliance sensitivities, filters them, and applies an optimality-criteria SIMP update. The von Mises color field is post-processing from the current numerical solution. The process is analogous to iterative optimization in machine learning, but it is not a neural network and its stress warning is an illustrative threshold rather than a fracture prediction.

“Show concept surface” displays a separately authored, smoothed piston concept for visual comparison. It is not reconstructed from the live voxel field, manufacturing-ready, or validated for a real engine. The live result remains an educational numerical model whose interpretation depends on its mesh, boundary conditions, material assumptions, objective, and constraints.

Specifications

Method: SIMP density-based topology optimization
Physics: Finite-element displacement solve in the optimization loop
Update: Filtered sensitivities and an optimality-criteria SIMP update
Displayed: Density field and post-processed von Mises stress

All exhibits

Operate

Inspect

In the current issue