# Distortion after machining

Shape changes after material removal: bulk and machining-induced stress, clamping, temperature and a controlled measurement handoff.

Reviewed: 2026-11-02

Tags: production, machining, distortion, residual-stress

In the worked comparison, five points along a **200 mm** specimen have a **0.02 mm** sampled height range while restrained and **0.18 mm** after release on supports. At the same temperature and measurement procedure, each row uses its own reference line. The ranges differ by **0.16 mm**. Five points on one line do not establish full-surface flatness, and the difference does not isolate clamping from other mechanisms or measure residual stress.

UC Davis distinguishes redistribution of existing bulk stress from stress introduced by machining, both relevant to thin-walled parts. [UC Davis, machining-distortion research project](https://research.engineering.ucdavis.edu/master/research/understanding-machining-distortion/)

Compare shapes under restraint and in the required inspection condition. A fixture can temporarily hold a part flat.

## Separate the mechanisms

| Candidate mechanism | Evidence to investigate |
| --- | --- |
| Redistribution of bulk residual stress | Stock manufacturing history, location/orientation and material-removal sequence |
| Machining-induced residual stress | Near-surface condition, tooling, cutting parameters and repeatability |
| Elastic fixture or cutting-force deflection | Support, contact forces, restraint and changes on release |
| Temperature effects | Part/fixture temperatures, gradients and time to stabilization |
| Measurement/setup difference | Support and datum conventions, instrument performance and repeatability |

Clamping, bulk stress, machining-induced stress and heat can act together. Record stock and process history alongside shape and temperature.

UC Davis's project studies the interaction of initial bulk and machining-induced stresses. Removing material changes the remaining body's stress equilibrium; a machining strategy must therefore be considered with the stock and final geometry, rather than as an isolated feed or speed setting. [UC Davis, project scope and mechanism](https://research.engineering.ucdavis.edu/master/research/understanding-machining-distortion/)

## What research does and does not establish

Weber and coauthors studied milled aluminum workpieces, machining-induced residual-stress repeatability and the distortion of thin wafers removed from the machined surface. Their experiments and numerical analysis connected the measured stress profiles with distortion; more or deeper machining-induced stresses produced more distortion in the investigated conditions. This is evidence for a mechanism in that study, not a universal stress-to-deflection conversion for arbitrary parts. [Weber et al., published May 22, 2021](https://link.springer.com/article/10.1007/s00170-021-07171-7)

Cherif and coauthors studied **AISI 316L plates** with an instrumented clamping system and a numerical model. Their results indicate that clamping and machining sequence can influence distortion. Keep that material, geometry and method scope when using the work to motivate a trial on another component. [Cherif et al., 2019, institutional repository record and abstract](https://sam.ensam.eu/handle/10985/17656)

Fixture deformation is also a separate modeling problem. Yadav and Mohite's study of thin-walled **Al 6061-T6** workpieces compared numerical contact/structural deformation with measurements in a specified milling fixture. It supports examining force, support and geometry together; it does not establish one clamp pressure for all thin walls. [Yadav and Mohite, Procedia Manufacturing 20, 2018, pp. 509–516](https://www.sciencedirect.com/science/article/pii/S2351978918301112)

## Original example: compare restraint conditions explicitly

Assume an invented 200 mm-long specimen is probed at five longitudinal positions. The same temperature and measurement procedure are used, and each row is aligned to its own stated reference line. The values are hypothetical measured heights, not model predictions:

| Position along specimen | 0 mm | 50 mm | 100 mm | 150 mm | 200 mm |
| --- | ---: | ---: | ---: | ---: | ---: |
| Restrained height | 0.00 | 0.01 | 0.02 | 0.01 | 0.00 |
| Released, supported height | 0.00 | 0.08 | 0.18 | 0.08 | 0.00 |

All heights are millimeters. The sampled height ranges are **0.02 mm restrained** and **0.18 mm released**. Their difference, **0.16 mm**, describes the difference between these two sampled ranges. It is not a residual-stress value or a proof that clamping alone created the change.

Five points along one line do not establish the full surface's flatness or compliance with a drawing. The reference alignment, support, unmeasured regions and measurement uncertainty still matter. Nor may a restrained result substitute for a free-state requirement, or the reverse, without the applicable drawing and inspection convention.

To investigate, retain the full measurement maps and restraint definitions, repeat measurements where needed, and compare the states after temperature stabilization. Establish whether the observed difference is repeatable and larger than the relevant measurement uncertainty before attributing it to a process change.

## Separate temperature correction from permanent shape change

NIST's dimensional-calibration treatment explicitly accounts for temperature and uncertainty in the thermal-expansion coefficient. Its simple length correction concerns uniform dimensional expansion relative to a reference temperature; a temperature gradient or bowed part requires additional treatment. [Doiron and Stoup, Uncertainty and Dimensional Calibrations, November–December 1997, section 3.3](https://pmc.ncbi.nlm.nih.gov/articles/PMC4894587/)

As a separate original illustration, assume a uniform **200 mm** length, an explicitly hypothetical expansion coefficient of **23 × 10^-6/K**, and a uniform **10 K** temperature difference. First-order length change is:

`ΔL = α × L × ΔT = 23 × 10^-6 × 200 × 10 = 0.046 mm`.

This assumed coefficient is not a certified property of an identified alloy. Uniform expansion of the length does not explain the earlier 0.18 mm transverse height range or calculate thermal bow. Record the actual material, temperature and reference convention before applying a correction.

## Evaluate process changes against the actual part

Candidate trials may examine stock selection and orientation, roughing/finishing sequence, material-removal balance, support and restraint, tooling and thermal control. Select trials from the diagnosed mechanisms and qualified process requirements. None is a universal instruction to alternate faces, increase clamp force or add a heat treatment.

If stress relief or another thermal operation is proposed, evaluate its effect on the specified material condition and downstream requirements with the responsible process authority. Do not assume a treatment that changes residual stress preserves temper, hardness, dimensions or surface requirements. Likewise, extra finish allowance is useful only if the remaining stock and revised sequence can produce the required released geometry.

Keep the trial's stock identity, location within the parent stock, orientation, fixture/support, removal stages, tool and cutting data, coolant, temperatures and measurement condition. Compare final accepted geometry as well as cycle time. A better in-fixture result that disappears on release does not demonstrate improvement at a released inspection boundary.

## Handoff and source versions

For a useful handoff, record the observed feature and quantity, process stage at which it appears, restrained/released state, temperature/time, datum/support scheme, measurement method and uncertainty, material history and trial results. Keep the diagnosis, proposed change and acceptance decision as separate statements.

Reviewed **2026-11-02 PST**. Weber et al.'s publisher dates the article **May 22, 2021**. Cherif et al.'s repository identifies **2019**; Yadav and Mohite's publisher identifies **2018**. The UC Davis page gives project duration **April 15, 2017–September 30, 2021**, not a separate page publication date. The NIST-authored calibration paper is from **November–December 1997**. Numerical inputs and diagnostic recommendations above are original, bounded guidance; the cited studies do not supply a universal machining recipe.

## Related reference

[Manufacturing process references](../../processes/index.md)