Recast layers

An ideal 10.000 mm EDM slot with a 10.010 mm upper limit grows to 10.016 mm after removing 0.008 mm normal to each wall, exceeding the limit by 0.006 mm. If local recast depth is 0.009 mm, that same removal leaves 0.001 mm at that location. The model assumes parallel walls and uniform removal. Verify finished dimensions and actual layer depth independently.

Recast material melts and resolidifies on the part. POCO distinguishes EDM’s recast or white layer from underlying thermally affected material. Jerry Mercer, POCO, EDM Effect on Surface Integrity, pp. 1–2.

Apply thickness and property observations to the actual alloy and EDM conditions.

Recast, redeposition and heat effects

Terminology varies across publications. POCO associates carbon enrichment with hydrocarbons in its mold-cavity electrode/dielectric context; this does not establish the composition or transformation behavior of every EDM layer. POCO, thermally altered layers.

The distinction between melting and heating matters because removal of visible surface deposits need not remove the underlying altered material. Likewise, a nominal recast thickness does not automatically include every microstructural or residual-stress change beneath it.

Process and alloy determine the result

POCO discusses electrical conditions, workpiece thermal characteristics and lower-energy finishing in moldmaking. These relationships distinguish surface finish from subsurface integrity without supplying a universal pulse program, skim count or polishing allowance. POCO, microcracking and surface-integrity discussion.

NASA’s January 2023 study examined wire-EDM rough, semi-finish and finish condition sets on superalloy 718, with brief additional assessments of ME209 and LSHR. Although roughness decreased toward finishing, tensile residual stresses increased in the investigated conditions. A smoother finish therefore did not establish an improvement in every surface-integrity quantity. Gabb et al., NASA/TM-20220017061, report and abstract.

For 718, the report’s Table IV gives:

These are measurements from that experiment, not guaranteed machine capabilities or statistical limits for all production. The underlying heat-affected zone was reported separately. NASA/TM-20220017061, Table IV, printed p. 11.

Removal can change more than the layer thickness

Holmberg, Wretland and Berglund studied grit blasting after wire EDM of an Inconel 718 aerospace shaft. Their original work found that blasting conditions strongly affected recast removal and residual stress. Longer exposure increased the risk of blasting medium becoming embedded in the surface. Removal, roughness, contamination and stress state therefore represented different outcomes, not one interchangeable “clean surface” result. Holmberg et al., published October 26, 2016.

That experiment does not establish a generic blasting recipe for a mold, thin wall or another alloy. Mechanical finishing can consume dimensional allowance, round edges or alter texture. A layer-removal claim consequently has both a metallurgical boundary and a geometric boundary.

Check local recast and final size separately. Remaining recast and stock allowance depend on the specified region and removal process.

Original example: removal allowance on opposing walls

Assume a hypothetical parallel-sided EDM slot is already 10.000 mm wide, with a permitted final width of 10.000 ± 0.010 mm. A proposed finishing operation removes 0.008 mm from each internal wall, measured normal to those walls. Ignore tolerances and nonuniform removal for this first calculation.

final width = 10.000 + 2 × 0.008 = 10.016 mm.

The result exceeds the 10.010 mm upper limit by 0.006 mm. Removing material from internal walls enlarges the opening; it does not reduce the slot width. With the same ideal removal, a prefinish width of 9.984 mm would finish at 10.000 mm.

Now assume a separate metallographic observation finds a local recast depth of 0.009 mm. Removing 0.008 mm normally from that location would leave 0.001 mm under the idealized model. A visually improved surface or correct final slot width would not establish complete removal. These invented dimensions are not recommended stock allowances; the calculation excludes sectioning uncertainty, variable layer depth, wall geometry and any deeper heat-affected material.

Measurement describes a defined region and quantity

A cross-section can reveal layer thickness and continuity at the sectioned location, while surface topography measures another property. Sample preparation, boundary interpretation and section orientation affect the measured thickness. A mean from several locations does not establish the maximum over an entire cavity or passage.

For a separate original illustration, three local depths of 3, 5 and 9 µm have a mean of 5.67 µm and an observed maximum of 9 µm. Under an invented 7 µm per-location maximum, the 9 µm observation exceeds the limit even though the mean is lower. Three points do not establish complete surface coverage or a validated sampling plan.

In the 2016 shaft study, handheld X-ray fluorescence was evaluated as a nondestructive indicator by detecting residual elements from the EDM wire. Its interpretation relied on that wire/material system and correlation with the examined surface. A chemistry signal is not automatically a direct thickness measurement or a universally applicable demonstration of recast absence. Holmberg et al., XRF and surface-integrity evaluation.

Service significance and evidence limits

Assess recast depth, cracks, stress, hardness, chemistry and roughness separately for the service load, environment and geometry.

A finishing-pass count describes an operation, not the residual layer itself. Similarly, a broad claim about modern generator control does not establish the result on a specified material and feature. The NASA and shaft-removal studies illustrate why measured results remain tied to particular materials, conditions and examination methods rather than becoming universal manufacturing limits.

Source versions

Reviewed 2026-11-02 PST. The NASA PDF cover explicitly reads January 2023, despite the report identifier beginning 2022; Table IV appears on printed p. 11. POCO’s final page prints ©2010–2016 and document code EDM-117919-0816; the code is reproduced without assigning an unverified issue day. The original grit-blasting paper’s publisher gives October 26, 2016, volume 25, pp. 5540–5550. Numerical removal and sampling examples are original, bounded calculations.