Abrasive Machining & Precision Grinding: The Complete Engineering Guide — 8 J/mm³ to Cut, 28 to Grind, 85% Into the Part

Grinding is the only machining process where the tool is deliberately dull, the chip is smaller than the grain, and most of the energy goes into the workpiece instead of the chip. This guide works the arithmetic that decides whether a hardened part survives: theoretical chip formation needs about 8 J/mm³ to heat steel to 1,250 °C, a real surface-grinding pass consumes 25–60 J/mm³, and with an alumina wheel 85% of that energy partitions into the part — a 3.5 kW cut spread over a 2.5 mm arc becomes 4.8 kW/cm² and an estimated 330 K surface rise. From that single fact follow tempering burn (ISO 14104 classes A/B/D/E), the residual-stress flip from −250 MPa compressive to +700 MPa tensile, the ISO 525 wheel code (A 60 K 5 V, FEPA F-grit, 62% grain at structure 0), the dressing ratios that make or break wheel topography, centerless feed at va = π·dr·Nr·sin ξ, and the economics that explain why a ₹8,000 wheel is 1% of a part's cost while a burned raceway is 100% of its failure.

·

Abrasive Machining & Precision Grinding: The Complete Engineering Guide

A bearing raceway is a surface that has been trained to within two micrometres of a perfect circle and polished to R_a 0.1 µm, and the tool that does it is a wheel of glued-together sand that costs less than the coolant it runs in. That is the first strange thing about grinding. The second is that the wheel is not sharp. An abrasive grain is a blunt, negatively raked crystal that ploughs more material than it cuts, and the chip it makes is often thinner than the grain itself.

The consequence is thermodynamic. Every machining process converts work into heat, but grinding is the only one where the majority of that heat is deliberately injected into the surface being finished. Turning sends most of its energy out with the chip. Grinding sends most of it into the part, through an arc a few millimetres long, in a contact time measured in milliseconds. The result is a surface that can be dimensionally perfect and metallurgically destroyed at the same time — a part that gauges good, passes a visual inspection, and fails 400 km into service because a 20 µm layer under the raceway was tempered back to 58 HRC and left in 700 MPa of tension.

This guide is about that layer. It runs in eleven sections:

1. Why Grinding Is a Different Process

1.1 The tool is supposed to be dull

A milling insert has a rake angle, a honed edge of 10–20 µm, and a coating that lets it slide through the shear plane. An alumina abrasive grain has no rake, an edge radius comparable to the depth of cut, and a fracture toughness low enough that it is designed to break down in service. It removes material by a combination of three mechanisms: cutting (a chip forms), ploughing (material is displaced sideways into a ridge), and rubbing (elastic contact, no removal at all). In fine grinding, cutting can be less than half of the energy budget — the rest is friction.

That is why the specific energy is so high.

1.2 The specific energy table

Specific energy u is the energy required to remove a unit volume of material:

where P is net grinding power (W), Q_w the removal rate (mm³/s), F_t the tangential force (N), v_s the wheel speed (m/s), a_e the depth of cut (mm), b the grinding width (mm), and v_w the work speed (m/s).

Process · Specific energy u (J/mm³) · Typical removal rate · Energy per cm³

Turning, plain carbon steel · 2–5 · 10–100 cm³/min · 0.6–1.4 Wh

Milling, steel · 3–7 · 20–200 cm³/min · 0.8–1.9 Wh

Surface grinding, hardened steel · 25–45 · 3–15 cm³/min · 7–12 Wh

Fine grinding, difficult alloy · 100–200 · < 2 cm³/min · 28–56 Wh

Dull/glazed wheel (same set-up) · 150–400+ · falling · 40–110 Wh

The numbers in the first two rows are the textbook range for conventional machining. The grinding rows come from measured spindle power in production and laboratory tests — a workhorse value of 25 J/mm³ for a free-cutting alumina wheel in hardened steel, rising to 200 J/mm³ for a hard-to-grind alloy with a fine-grain wheel. When a wheel glazes, specific energy has been measured climbing from 150 to over 400 J/mm³ on the same machine, same part, same nominal parameters.

1.3 The theoretical floor: 8 J/mm³

There is a physical lower bound. If a chip of steel is heated from ambient to the 1,250 °C that grinding chips typically reach, the enthalpy it carries away is roughly:

Published estimates for AISI 1095 using a mean specific heat to 1,500 °C give 8.1 J/mm³. That is the energy that must leave the zone in the chip if the chip is to carry all the heat. A real grinding pass uses three to seven times that, and the excess has nowhere to go but the wheel, the coolant, and the workpiece.

The gap between 8 and 28 J/mm³ is the entire surface-integrity problem in one sentence.

1.4 What this means operationally

Grinding is not a stock-removal process and should never be specified as one. At 7.5 cm³/min, a 3.5 kW surface grinder removes material about eight times slower than a 15 kW milling centre running a 50 mm face mill. If a drawing leaves 2 mm of stock on a hardened block "and we'll grind it", the shop will spend a day making the part smaller and a second day making it scrap. The correct allowance is 0.15–0.4 mm, applied after heat treatment, and the grinding operation's job is to remove the decarburised layer, correct the distortion, and generate the tolerance — not to create the geometry.

2. The Size Effect: Why Thinner Chips Cost More

2.1 Equivalent chip thickness

Grinding parameters collapse into one useful dimensionless-ish quantity, the equivalent chip thickness:

For a surface-grinding pass at a_e = 0.025 mm, v_w = 0.20 m/s (12 m/min), and v_s = 32 m/s:

A tenth of a micrometre. That is the scale at which grinding operates, and it explains why the process is sensitive to things a milling machine ignores: wheel topography, coolant chemistry, a 2 µm spindle runout.

2.2 Specific energy rises as chips get thinner

The relationship is empirical but robust:

The exponent is the size effect: at very small depths of cut, the material under the grit is confined and the shear-plane model breaks down. The grain ploughs, the workpiece work-hardens ahead of it, and the energy per unit volume rises. Practical consequences:

2.3 Power is the cheapest sensor in the shop

The single most useful instrumentation on a grinding machine is a kW meter on the spindle. At constant parameters, power is proportional to specific energy, and specific energy is a proxy for wheel sharpness. A practical rule used on production lines:

This is a better burn indicator than a thermal camera, because it responds before the surface discolours. A grinding burn begins roughly where power per unit width crosses a threshold that is characteristic of the material and coolant — and the operator can find that threshold once and then watch the meter.

3. Where the Heat Goes

3.1 The partition ratio

Of the energy entering the grinding zone, a fraction goes into the workpiece, a fraction into the wheel, and the rest leaves with the chip and the coolant. The workpiece fraction \varepsilon is the one that matters, and it is a strong function of wheel material:

Wheel · Partition to workpiece \varepsilon · Why

Aluminium oxide (vitrified) · 0.85 · Low conductivity, grains act as insulators

Cubic boron nitride (vitrified) · 0.46 · High-conductivity grains conduct heat into the wheel and chips

Diamond · 0.4–0.6 · Highest conductivity, but graphitises in steel

This single number explains why CBN "runs cooler" even when it removes material at the same rate. It is not that CBN generates less heat — although it often does, because it stays sharp longer — it is that CBN takes roughly half the heat away from the part that alumina puts in. An alumina wheel at 3.5 kW delivers about 2.98 kW into the workpiece; a CBN wheel at the same power delivers 1.61 kW, a 46% reduction in thermal load.

3.2 Contact length

The heat is injected through the arc of contact. For surface grinding:

where d_e is the equivalent diameter. For external cylindrical grinding:

and for internal grinding:

with d_s the wheel diameter and d_w the work diameter. Internal grinding has the longest contact arc of the family — the wheel wraps the bore — which is why bore grinding is the burn-prone operation and why small internal wheels need high wheel speed and low depth of cut.

For the surface example (a_e = 0.025 mm, 250 mm wheel): l_c = \sqrt{0.025 \times 250} = 2.5 mm. With a 25 mm width, the contact patch is 62.5 mm², and the heat flux is:

Four point eight kilowatts per square centimetre. For comparison, a domestic electric hob runs at about 0.02 kW/cm². The grinding arc is two orders of magnitude more intense, and it moves across the surface at 12 m/min.

3.3 A Jaeger estimate you can run at the machine

For a band heat source of half-width a = l_c/2 moving at work speed v_w over a semi-infinite body, the maximum surface temperature rise is approximately:

with k the workpiece conductivity (≈40 W/m·K for steel at 400 °C) and \kappa = k/(\rho c_p) \approx 1 \times 10^{-5} m²/s. Substituting the example:

A 330 K rise above a 25 °C bulk puts the surface at roughly 355 °C. Bearing steel 52100 is tempered at 150–200 °C. The pass is, by calculation, hot enough to soften the surface layer — before any discoloration is visible.

Two caveats keep this honest. First, the estimate assumes a smooth band; in reality contact is through discrete grits, and the flash temperature at a grain tip can exceed 1,000 °C for microseconds while the average band temperature stays near 350 °C. Second, the partition ratio itself changes with temperature and coolant; the 0.85 figure is a room-temperature-alumina value and falls when the coolant boils. Both caveats make the real surface worse, not better.

3.4 Why oil sometimes beats water

The counter-intuitive result from creep-feed research: an oil-based process at 190 J/mm³ measured a maximum temperature of 260 °C, while a water-based emulsion at 67 J/mm³ measured 40 °C. The oil process used less specific energy and a slower work speed, but the decisive factor was that oil wets the arc completely and suppresses boiling — no vapour blanket, no film-boiling spike. Water carries more heat per litre, but only while it stays liquid in contact. Once the arc exceeds the Leidenfrost point, water-based coolant can be worse than oil.

High-efficiency deep grinding (HEDG) pushes the other way: wheel speeds up to 200 m/s reduce specific energy to 11 J/mm³, because at high speed the chip gets thinner but the removal rate per unit width is enormous, and the heat is carried away by the chip in a shorter contact time.

4. Thermal Damage: What Actually Happens to the Surface

4.1 Two distinct mechanisms

Tempering burn. The surface is heated above the original tempering temperature but below the lower critical temperature A_{c1} (about 750 °C for 52100). The tempered martensite over-tempers; hardness drops; the microstructure coarsens. The part is softer exactly where the load is highest. This is the common burn, and it is invisible.

Rehardening burn. The surface exceeds A_{c1}, austenitises, and is quenched by the cold bulk underneath — a self-quench. The result is untempered, brittle, freshly formed martensite, often with retained austenite and a network of cracks. This is the severe burn. It is not a hardness loss; it can read harder than the base metal on a portable hardness tester, which is why hardness alone is a poor burn inspection.

4.2 The residual-stress signature

Grinding leaves residual stress in the surface layer, and its sign is the most reliable single indicator of process health:

A tensile skin on a bearing raceway is a crack waiting for a load cycle. Rolling contact fatigue initiates where the subsurface shear stress is highest, and a tensile residual field moves the initiation site toward the surface, where grinding defects already are.

4.3 Detection: nital, Barkhausen, or X-ray

Nital etch (ISO 14104). The reference method. A cleaned surface is etched in 3–5% nitric acid in alcohol, rinsed, bleached in 4–6% hydrochloric acid, neutralised, and inspected. The classes are:

Class · Meaning

A · No thermal damage

B · Light tempering — acceptable on many parts, reject on critical

D · Heavy tempering

E · Rehardening (with or without cracks)

Nital etch is reliable and standardised, but it is destructive in a subtle way: each inspection removes about 3 µm of material, so a part cannot be repeatedly inspected at the same location and a tight-tolerance surface must be masked. It also leaves discoloration that may need polishing.

Barkhausen noise analysis (BNA). Magnetise the surface and listen to the jumps in magnetisation as domain walls pin and unpin on defects and stress fields. Non-destructive, fast, and sensitive to residual stress before hardness drops. Its weakness is calibration: the signal depends on material, hardness, and surface finish, so it needs a master part and a recipe per product. On production lines it is the method of choice for 100% inspection of bearing rings and gear teeth.

X-ray diffraction. The quantitative referee — measures residual stress and FWHM directly. Slow, expensive, laboratory-scale, used for process qualification and failure analysis rather than production.

4.4 Cracks

Grinding cracks are fine, shallow, and typically perpendicular to the grinding direction. They are almost invisible to the eye and to a surface finish gauge. The inspection is magnetic particle (for ferromagnetic parts) or fluorescent penetrant after a careful clean. Any part that has seen a rehardening burn should be treated as cracked until MPI says otherwise — and on a safety-critical part, "treated as cracked" usually means scrapped.

4.5 What burn costs

A burned raceway does not fail immediately. It fails by fatigue, and fatigue life is hypersensitive to hardness and residual stress. A drop of a few HRC in the surface layer, combined with a tensile skin, can move the L10 life by a factor of two or more under the same load. The economic asymmetry is total: the grinding pass costs a few rupees of machine time, the part costs hundreds, and the field failure costs the customer relationship.

5. The Wheel, Decoded

5.1 The ISO 525 marking

A bonded wheel carries a compact code. The ISO 525 order is:

51   A    36   L     5     V     23   50
│    │    │    │     │     │     │    └── max speed, m/s
│    │    │    │     │     │     └─────── manufacturer's bond record
│    │    │    │     │     └───────────── bond type
│    │    │    │     └─────────────────── structure number
│    │    │    └───────────────────────── grade (hardness letter)
│    │    └────────────────────────────── grit size (FEPA F)
│    └─────────────────────────────────── abrasive type
└──────────────────────────────────────── manufacturer's grain record

The standard's own example is 51 A 36 L 5 V 23 50. The fields that matter to a buyer are the middle five.

5.2 Abrasive type

Code · Abrasive · Knoop hardness (kg/mm²) · Use

A / WA · Aluminium oxide (brown / white) · ~2,100 · Carbon and alloy steels, general purpose

PA / SA · Pink / monocrystalline alumina · ~2,200 · Heavy stock, less friable

SG / ceramic · Sol-gel microcrystalline alumina · ~2,200 · High-performance steels, 2–3× life, self-sharpening

C / GC · Silicon carbide (black / green) · ~2,500 · Cast iron, non-ferrous, carbide, stone

ZA · Zirconia alumina · ~1,400 but very tough · Snagging, foundry cleaning

B / cBN · Cubic boron nitride · ~4,700 · Hardened steel > 45 HRC

D / SD · Diamond · ~7,000–8,000 · Carbide, ceramics, glass, PCD, stone

The selection rule that matters most: diamond is the wrong abrasive for steel. Carbon dissolves into iron at grinding temperatures, the diamond graphitises, and wheel life collapses. Use CBN for hardened steel, diamond for everything else hard.

5.3 Grit size

FEPA F-grit is defined by sieving for macrogrits (ISO 8486) and by median particle size for microgrits. The coarse/fine divide matters because coated abrasives use a different scale (FEPA P, ISO 6344) and the numbers do not correspond — an F60 wheel and a P60 belt are not the same particle size.

FEPA F grit · Sieve opening, all pass (µm) · 40% pass (µm) · Typical use

F24 · 1,180 · 710 · Heavy stock, foundry

F36 · 850 · 500 · Rough surface and cylindrical

F46 · 600 · 355 · General surface grinding

F60 · 425 · 250 · General purpose, toolroom

F80 · 300 · 180 · Finishing

F120 · 180 · 106 · Fine finishing, tool sharpening

F220 · 106 · 53 · Fine finishing, honing

F320 · median 29.2 · — · Precision finishing

F600 · median 9.3 · — · Superfinishing, lapping

Coarse grits cut cooler and faster but leave deeper scratches; fine grits produce the finish but load and burn more readily. A useful starting point for hardened steel is F46–F60 for stock removal and F80–F120 for finish, with the grade adjusted around it.

5.4 Grade: the most misread letter on the wheel

The grade letter (A soft → Z hard) is not the hardness of the abrasive. It is how strongly the bond holds the grain — how much force is required to tear a dull grain out and expose a fresh one.

The rule is counter-intuitive and universal: hard workpiece → soft grade; soft workpiece → hard grade. A hard, brittle workpiece dulls grains quickly; a soft bond releases them before they rub. A soft, ductile workpiece does not dull grains quickly, so a hard bond can hold them for full life. Two corollaries:

The letters are also not comparable across makers or bond families. A "K" in a vitrified precision wheel and a "K" in a resin cut-off wheel are different animals.

5.5 Structure

Structure is the volumetric grain fraction. The convention used by many makers: structure 0 is 62% grain by volume, and every two percentage points of grain replaced by porosity adds one structure number:

So a structure 8 wheel is about 46% grain. Higher structure numbers mean more open, more porous wheels that clear chips, run cooler, and resist loading — at the cost of strength and form retention. Dense structures (0–4) hold form and produce fine finishes; open structures (8–14) are for soft, gummy, or burn-prone work and for creep-feed wheels that must pass large volumes of coolant.

5.6 Bond

Code · Bond · Character · Typical use

V · Vitrified · Rigid, heat-resistant, porous, dressable · Precision grinding, form holding

B · Resinoid · Tough, slight elasticity · Cut-off, high-speed, rough grinding

BF · Reinforced resinoid · Fibre-reinforced · Hand-held discs, EN 12413

R · Rubber · Elastic, dense · Regulating wheels, thin wheels, bearing work

E · Shellac · Very elastic, fine finish · Polishing, camshaft grinding

M · Metal · Strong, conductive · Diamond/CBN electroplated or brazed

PL · Plastic · Flexible · Special applications

MG · Magnesite · Cold-cured · Large wheels, no firing

Vitrified is the default for precision work because it can be dressed to a sharp, open topography and it holds form. Resinoid dislikes alkaline coolant — a wheel that wears too fast on a resin bond is often a coolant-pH problem before it is a wheel problem.

5.7 Speed, safety, and mounting

Wheel speed is marked on every bonded wheel, and the range in ISO 525 runs from 16 to 160 m/s. Common values: 35 m/s for standard vitrified bench and surface wheels, 45–60 m/s for high-speed precision wheels, 80–125 m/s for CBN, and up to 200 m/s in HEDG research machines.

Safety rules that are not negotiable:

6. Dressing: Ten Micrometres That Decide Two Hundred Parts

6.1 Truing versus dressing

Truing restores geometry — concentricity, profile, flatness — by cutting the wheel with a diamond. Dressing restores topography — sharpness, grit protrusion, openness — usually in the same pass. A new wheel must be trued; a working wheel is dressed at the end of every redress interval. Confusing the two is why some shops never get a wheel to run true.

6.2 Single-point diamond dressing

The classic tool is a single-crystal diamond, mounted in a holder, traversed across the wheel face. The variables:

Parameter · Rough dressing · Finish dressing

Depth of dress per pass · 20–50 µm · 5–15 µm

Lead (traverse per wheel revolution) · 0.10–0.25 mm/rev · 0.05–0.10 mm/rev

Overlap ratio · 3–5 · 15–20

Number of passes · 1–2 · 2–4 (spark-out)

A larger depth and lead produce a rougher, more open wheel that cuts cooler and faster. A smaller depth and lead produce a smoother wheel with more cutting edges, which gives a better finish but loads and burns more easily. The overlap ratio U = b_d/(f_d) — dresser contact width divided by lead — is the single number that predicts the resulting topography.

Dressing is a controlled crushing operation. The diamond does not sharpen grains; it fractures the bond and pulls whole grains out, leaving new ones with sharp edges. That is why a freshly dressed wheel cuts cooler for the first few parts and then settles into steady-state wear.

6.3 Rotary dressing

Production grinding uses a rotary diamond dresser — a disc with diamond-impregnated segments — because it dresses faster, lasts longer, and can dress at full wheel speed. Two parameters govern the result:

Measured behaviour: up-dressing produces a smoother wheel surface, makes the wheel act harder, and gives high G-ratios — use it for finish. Down-dressing produces an open, sharper topography, makes the wheel act softer, and allows higher removal rates at lower G-ratio — use it for roughing. A dressing increment below 10 µm and a speed ratio around 0.7 are common starting points; overlap of 3 for roughing and 20 for finishing is a standard recipe.

6.4 Redress life and compensation

Redress life — the number of parts between dressings — is set by the tolerance and finish budget, not by the wheel's ability to keep cutting. A wheel can keep removing material long after it has started producing taper, chatter, or burn. Typical production intervals run from 20 to 200 parts for conventional wheels, and 200 to 2,000 for CBN.

Every dressing removes wheel diameter, so the machine must compensate. The compensation is not just the dress depth: the effective wheel diameter changes the surface speed at constant spindle rpm, which changes the cut. A control that holds constant surface speed is worth more than a control that holds constant rpm.

7. The Process Family

7.1 Surface grinding

A reciprocating or rotary table, a horizontal or vertical spindle, and a workpiece held on a magnetic chuck or in a fixture. Achievable: IT5–IT7, R_a 0.1–0.8 µm, flatness 2–5 µm over 100 mm. The dominant error source is the machine's thermal growth, not the wheel.

7.2 Cylindrical grinding (OD and ID)

The workpiece rotates between centres or in a chuck; the wheel traverses or plunge-cuts. For OD work:

Achievable: IT5–IT6, R_a 0.1–0.4 µm, roundness 0.5–2 µm, size holding ±2–3 µm. For ID work the equivalent diameter uses the difference, the contact arc is longer, the wheel is smaller and runs at lower surface speed, and the burn risk is highest in the family. Internal grinding is where coolant delivery is hardest and where a 0.5 mm wheel wear allowance disappears fastest.

7.3 Centerless grinding

The workpiece is not held on centres. It sits on a workrest blade between a large grinding wheel and a smaller, slow regulating wheel, and it is driven by friction. The geometry does three things at once: it centres the part, it rotates it, and — when the regulating wheel is tilted — it feeds it axially.

The axial feed rate is:

where d_r is the regulating wheel diameter (mm), N_r its speed (rpm), and \xi the tilt angle. For d_r = 250 mm, N_r = 25 rpm, \xi = 2°:

At 80 mm part length plus 40 mm gap, that is about 5.7 parts per minute theoretical. A useful identity falls out of the same geometry: feed per work revolution = \pi d_w \sin\xi, which at 25 mm and 2° is 2.74 mm/rev.

Setup rules:

Parameter · Value

Regulating wheel tilt, through-feed · 1–3° typical; 2–6° roughing, 1–2° finishing, 0.5–1.5° long parts

Regulating wheel tilt, in-feed · 0–0.5°

Workrest blade top angle · 30° typical; 20–45° range; lower for heavy work

Work centre height · About half the part diameter above the wheel centreline; too high = chatter, too low = out-of-round

Blade thickness · 1.5–2 mm under the part diameter

Two details that separate a good centerless shop from a struggling one. First, the regulating wheel must be dressed as a hyperboloid, not a cylinder, when tilted — otherwise it contacts the part at a point and the part wanders, chatters, or grinds tapered. Second, the infeed ratio is 1:1, not 2:1: in conventional OD grinding, 0.001" of infeed removes 0.002" from the diameter because the part is held on centre; in centerless, the part floats against the regulating wheel and 0.001" of infeed removes 0.001". That 1:1 ratio is why centerless holds size so well in volume — and why operators trained on centres overshoot.

7.4 Creep-feed grinding

Full depth in one pass: a_e of 1–6 mm, work speed of 0.05–0.5 m/min, high-porosity wheels, high-pressure coolant or neat oil. Removal rates are high and specific energy is low (11–67 J/mm³ measured), but the arc is long and the thermal load is continuous. Creep feed is the process for turbine roots, broach-like profiles, and carbide form tools — and the one where coolant delivery is not optional.

7.5 Jig grinding

A precision vertical spindle machine with a planetary motion, used for holes, profiles, and coordinates in hardened tooling. Achievable position tolerance 1–3 µm, R_a 0.1–0.4 µm. Slow, expensive, and irreplaceable for a hardened die insert that must be corrected without annealing.

7.6 Honing, superfinishing, lapping

These are the processes after grinding, and they are defined by what they do not do: they do not change geometry much, they remove the damaged layer and the peaks.

Process · Stock removed · Achievable R_a · Notes

Honing · 20–100 µm · 0.05–0.4 µm · Cross-hatch, bore geometry correction

Superfinishing · 2–20 µm · 0.005–0.10 µm · Oscillating stone, 0.05–0.50 MPa, 500–3,000 strokes/min, 5–60 s

Lapping · 5–50 µm · 0.01–0.05 µm · Free abrasive, flatness to fractions of a µm

Superfinishing is the operation that turns a ground bearing raceway from R_a 0.1–0.2 µm into R_a < 0.05 µm, improves roundness by 50–80%, and raises the contact-area ratio t_p to 90–95%. A good rule for stock: three to four times the previous operation's R_a, which for a ground surface is 5–10 µm on diameter. Balls go further — lapped between cast-iron plates to R_a ~0.01 µm and roundness under 0.5 µm for G5–G10 grades.

7.7 The tolerance table

Process · IT grade · R_a (µm) · Size holding

Surface grinding · IT5–IT7 · 0.1–0.8 · ±3–5 µm

Cylindrical OD · IT5–IT6 · 0.1–0.4 · ±2–3 µm

Internal · IT6–IT7 · 0.2–0.8 · ±5 µm

Centerless · IT6–IT7 · 0.2–0.8 · ±3–5 µm

Creep feed · IT7–IT9 · 0.4–1.6 · ±10–20 µm

Jig grinding · IT5–IT6 · 0.1–0.4 · ±2–3 µm, position ±1–3 µm

Honing · IT4–IT6 · 0.05–0.4 · ±2–5 µm

Superfinishing · — · 0.005–0.10 · Roundness improved 50–80%

8. Coolant: The Part of the Process Nobody Specifies

8.1 The air barrier

A spinning wheel drags a boundary layer of air with it. At 35 m/s, that film is stiff enough to deflect a plain coolant stream away from the arc. The nozzle must penetrate it: aim at the contact point from close range (10–20 mm), with a coherent jet whose velocity is comparable to the wheel speed (velocity ratio 0.8–1.0). A floppy hose pointed at the wheel is not coolant delivery; it is coolant disposal.

8.2 Flow and pressure

Parameter · Conventional grinding · Creep feed

Flow per mm of wheel width · 0.5–1.5 L/min/mm · 2–4 L/min/mm

Nozzle pressure · 3–8 bar · 8–20 bar

Filtration · 20–50 µm · 10–20 µm

Emulsion concentration · 5–10% · 8–12% or neat oil

Through-spindle coolant is the only reliable way to reach an internal grinding arc. For small bores, a high-pressure jet through the wheel hub is not a luxury; it is the difference between a 40-part wheel life and a 400-part wheel life.

8.3 When to use oil, MQL, or cryogenic

8.4 Coolant is a chemistry problem

Concentration, pH, hardness, tramp oil, and bacterial load all change grinding behaviour. A resin-bond wheel that wears too fast, a part that rusts overnight, a sump that smells — these are coolant-management failures, not wheel failures. Test concentration weekly with a refractometer and pH monthly. Alkaline coolant attacks resin bonds and aluminium; acidic coolant attacks machine paint and cast iron.

9. The Economics: Stock, G-Ratio, and Cost per Part

9.1 Stock allowance is a contract

The grinding allowance must cover three things: heat-treatment distortion (typically 0.05–0.3 mm, more for long or thin parts), decarburisation (0.1–0.3 mm on parts heat-treated without atmosphere control), and the geometry correction needed to bring the part into tolerance. A practical allowance for a hardened shaft is 0.25–0.40 mm on diameter; for precision bearing rings, 0.10–0.20 mm. Anything more is paid for twice — once in grinding time, and once in the risk of burning the surface while removing it.

9.2 G-ratio

the volume of workpiece removed per unit volume of wheel worn. Typical steady-state values:

Wheel/workpiece · G-ratio

Alumina vitrified, hardened steel · 10–30

Alumina, mild steel (soft, gummy) · 5–15

Vitrified CBN, hardened steel · 300–2,000

Vitrified CBN, high-speed (120 m/s) · 850–2,480 measured

Ultrafine-crystalline CBN, M2 at 65 HRC · 13,045 vs 3,096 for conventional CBN

Plated CBN · 1,000–10,000+

The last two rows are the reason CBN exists. On the same machine, same part, vitrified CBN at 60 → 120 m/s changed specific energy from 15.2 to 22.0 J/mm³, R_a from 0.66 to 0.20 µm, roundness from 1.25 to 0.72 µm, size holding from 30 to 5 µm, and G-ratio from 850 to 2,480. Higher speed cost energy per unit volume and bought quality and wheel life.

9.3 A worked cost per part

Assumptions, stated so you can substitute your own: hardened 52100 sleeve, OD 40 mm × 30 mm long, 0.30 mm stock on diameter, 3.5 kW cut, 40 s floor-to-floor, shop rate ₹1,200/h, alumina wheel ₹8,000 with 1.74 × 10⁶ mm³ of usable abrasive, G = 30.

Part volume removed:

Wheel cost per part:

Machine time at 40 s: ₹13.33. Dressing at 50 s every 40 parts: ₹0.42. Coolant and filtration: ₹0.30. Total: about ₹14.2 per part, of which the wheel is 1.2%.

Now the CBN case: wheel ₹1,20,000 with 3.1 × 10⁵ mm³ of usable layer, G = 300, cycle 30% faster at 28 s.

Machine time at 28 s: ₹9.33. Dressing every 300 parts at 20 s: ₹0.02. Total: about ₹10.8 per part — a 24% saving, driven almost entirely by cycle time, not by wheel cost. CBN is justified by time and size holding, not by abrasive economy. That is why high-volume bearing plants run CBN and job shops run alumina: the crossover is a function of annual volume, not part complexity.

9.4 The hidden cost: scrap

The cost model above assumes the part is good. A burned part is not: it is scrap, or a rework that removes another 0.05 mm and risks the tolerance. On a ₹3,000 bearing ring, a 2% burn rate adds ₹60 per good part — four times the entire grinding cost. Process control is cheaper than inspection, and the cheapest control is the spindle power meter plus a dressing schedule that is actually followed.

10. The Failure Atlas

Symptom · Most likely cause · First fix

Burn/discolouration at the exit edge · Coolant starvation at the arc, wheel too hard, depth too high · Increase coolant pressure and flow, softer grade, reduce a_e

Burn in a band across the part · Wheel loading, unequal dressing, taper in the wheel face · Dress with more overlap, check coolant filtration, true the wheel

Chatter, regular marks · Wheel imbalance, spindle runout, worn bearings, wrong wheel speed · Balance to ISO 6103, check runout, change speed

Taper on a cylindrical part · Workhead misalignment, worn wheel, tailstock pressure · Realign, dress, reduce pressure

Out-of-round in centerless · Centre height wrong, blade angle wrong, regulating wheel not hyperbolically dressed · Set height, 30° blade, redress regulating wheel

Size drift over a batch · Thermal growth, wheel wear without compensation · Warm-up cycle, constant surface speed, gauge in-process

Loading (soft, gummy material) · Wheel too dense/hard, coarse grit needed, coolant poor · Open structure, softer grade, coarse grit, more flow

Glazing (shiny wheel, high power) · Wheel too hard, dress too fine, speed too high · Coarser dress, softer grade, reduce speed

Cracks after grinding · Rehardening burn, excessive depth, no spark-out · Reduce a_e, increase passes, inspect with MPI

Poor finish after a good wheel · Dressing lead too coarse, grit too coarse, wheel out of balance · Finer dress, finer grit, balance

11. Specifying Grinding on an RFQ

Grinding is the process most often under-specified on a purchase enquiry, and the gap is where quotes diverge by 3×. A grinding RFQ should carry:

  1. Material and condition — alloy, heat treatment, hardness in HRC, and whether the part is through-hardened, case-hardened, or nitrided.
  2. Stock allowance and its distribution — how much material is left on each surface, and whether heat treatment has already been done.
  3. Critical dimensions with tolerance — not just the nominal, but the tolerance band and the datum scheme.
  4. GD&T — roundness, cylindricity, flatness, parallelism, and runout, each with a value.
  5. Surface finish — R_a plus any lay requirement, and whether it applies to the whole surface or a band.
  6. Inspection — first article, CMM report, nital etch per ISO 14104, Barkhausen, MPI after grinding.
  7. Quantity and repeatability — a one-off toolroom part and a 10,000-piece production run are different processes, different wheels, and different machines.
  8. Any process constraints — "no visible burn", "no chatter", "maximum stock per pass", "grind after heat treatment only".

A shop that receives a 62 HRC part with 0.8 mm of stock and a 2 µm tolerance has two choices: burn it or spend four hours on it. Neither is the price that was quoted. Specify the allowance, and the quote becomes comparable.

The Discipline in One Page

Grinding is one energy balance with four consequences. The balance: chip formation needs about 8 J/mm³ to heat steel to 1,250 °C, a real alumina pass consumes 25–45 J/mm³, and 85% of that energy partitions into the workpiece — a 3.5 kW cut over a 2.5 mm arc is 4.8 kW/cm² and an estimated 330 K surface rise. The consequences: (1) burn — tempering above 150–200 °C or rehardening above A_{c1}, classified A/B/D/E by ISO 14104, with residual stress flipping from −250 MPa compressive to +700 MPa tensile; (2) the wheel — ISO 525's five fields (abrasive, FEPA grit, grade, structure, bond), where grade is bond retention and not grain hardness, structure 0 is 62% grain and each number costs 2 points of it, and diamond is disqualified on steel; (3) dressing — 5–50 µm of depth, 0.05–0.25 mm/rev of lead, overlap 3 for roughing and 20 for finishing, up-dressing for finish and down-dressing for removal; (4) the economics — leave 0.15–0.40 mm, not 2 mm, because at 7.5 cm³/min the grinder is eight times slower than the mill. Centerless adds its own arithmetic: v_a = \pi d_r N_r \sin\xi, a 30° blade, half-a-diameter of centre height, and a 1:1 infeed ratio. Superfinishing finishes the job that grinding starts — 2–20 µm of stock, R_a down to 0.005 µm, roundness improved 50–80%. And through all of it, the cheapest instrument is the spindle power meter: a 20% rise at constant parameters means the wheel is dull, and a dull wheel burns a part long before it measures one. The parts that last are not the ones ground to the tightest tolerance; they are the ones whose surface layer was never allowed to exceed the tempering temperature that the drawing's hardness implied.

Previous guides in this series: Springs & Spring Design · O-Rings, Gaskets & Mechanical Seals · Machine Safety & Guarding · Rolling-Element Bearings · Gears & Gearboxes · CNC Machining · Metrology & Dimensional Inspection · Fasteners, Threads & Bolted Joints.

[1] W. B. Rowe, "Energy and temperature analysis in grinding," Laser Metrology and Machine Performance VI, WIT Press — specific energy 25 J/mm³ for a typical steel grind, 200 J/mm³ for a hard-to-grind fine-grain wheel, 150 → 400+ J/mm³ as grains dull; workpiece partition 0.85 (alumina) vs 0.46 (CBN); creep grinding at 190 J/mm³ / 260 °C (oil) and 67 J/mm³ / 40 °C (water-based); HEDG at 11 J/mm³ and wheel speeds to 200 m/s; chip energy 8.1 J/mm³ for AISI 1095 at an assumed 1,250 °C chip temperature. [2] Malkin & Guo, Grinding Technology: Theory and Applications of Machining with Abrasives — specific energy, size effect, equivalent chip thickness, G-ratio, dressing mechanics, wheel wear regimes. [3] ISO 525:1999/2020 Bonded abrasive products — General requirements and ISO 8486-1 (FEPA F macrogrits) / ISO 6344 (FEPA P coated abrasives) — wheel marking order (51 A 36 L 5 V 23 50), grit designations F4–F2000, grade letters A–Z, structure numbers, bond codes (V, B, BF, R, E, M, PL, MG), marked speed range 16–160 m/s. [4] ISO 14104:2017 Gears — Surface temper etch inspection after grinding, chemical method — etching procedure, classes A (no damage), B (light temper), D (heavy temper), E (rehardening); used with type-2 etch for low-alloy steels. [5] The International Journal of Advanced Manufacturing Technology (2021), "In-process detection of grinding burn using machine learning" — nital etching and Barkhausen noise as the dominant industrial methods; ISO 14104 four-class scheme. [6] Production Engineering (2025), "In-process grinding burn detection by use of different Barkhausen noise measurement techniques" — residual stress −190 to −250 MPa compressive after low-energy grinding, rising to +700 MPa tensile and FWHM 5.2° → 3.3° in a thermo-mechanically damaged zone. [7] US Patent 8,353,739 B2 — nital etch chemistry (3–5% nitric, 4–6% hydrochloric), approximately 0.003 mm material removal per inspection, corrosion and discoloration limitations. [8] Liverpool John Moores University doctoral thesis, "Process requirements for precision grinding" — 60 → 120 m/s vitrified CBN: specific energy 15.2 → 22.0 J/mm³, R_a 0.66 → 0.20 µm, roundness 1.25 → 0.72 µm, size holding 30 → 5 µm, G-ratio 850 → 2,480; rotary dressing speed ratio ≈ 0.7, dressing increment < 10 µm, overlap 3 roughing / 20 finishing, up- vs down-dressing behaviour. [9] Materials Research (SciELO), "Analysis of the performance of superabrasive and alumina grinding wheels with different bonds and machining conditions" — CBN G-ratio up to 30× alumina; vitrified vs resin bond wear; tangential force and roughness comparison. [10] JSME International Journal, "Profile grinding of high-speed steels using ultrafine-crystalline cBN wheels" — steady G-ratio 13,045 (cBN-U) vs 3,096 (conventional cBN) on M2 at 65 HRC, 30 m/s, concentration 125. [11] Norton Abrasives / Cutting Tool Engineering, "Thrufeed centerless OD grinding: parameters and troubleshooting" — workrest blade top angle 30° typical (20–45° range), regulating wheel inclination +1 to +3°, centre-height relationship, guide alignment. [12] Journal of Manufacturing and Materials Processing (MDPI), "Dynamic rounding stability in through-feed centerless grinding" — axial feed v_a = \pi d_r N_r \sin\xi, hyperboloid regulating wheel dressing, centre-height angle variation along the grind length. [13] Nagel SG 50/SG 75 superfinishing technical data — stock removal 3–4× previous R_a, 5–10 µm on diameter for ground parts, R_a to 0.02 µm, roundness improvement 50–80%, contact area t_p 90–95%. [14] Jain Precision Tools superfinishing stone data — R_a 0.005–0.10 µm, stone pressure 0.05–0.50 MPa, oscillation 500–3,000 strokes/min, stock 2–20 µm, cycle 5–60 s. [15] CNC57 technical library, grinding-wheel marking and centerless guides — ISO 525 field order, structure convention (62% grain at 0, −2% per number), grade semantics, centerless tilt and blade tables. [16] High Speed Training, "Understanding the marking system for abrasive wheels" — abrasive/grit/grade/structure/bond codes, organic-bond three-year expiry, ISO type numbers. [17] CarTech 52100 alloy datasheet — 1.0% C, 1.5% Cr bearing steel; annealed 223 HB max; hardening 832–843 °C oil quench; tempering practice; ASTM A295/AMS 6440 scope. [18] Kalpakjian & Schmid, Manufacturing Engineering and Technology — specific energy ranges for turning, milling, and grinding; abrasive process classifications. [19] EN 12413 Safety requirements for bonded abrasive products — marked maximum speed, burst margin, mounting and guarding requirements; ISO 6103 permissible unbalance.

Note on sources: the governing references for the equations in this guide are Malkin & Guo for grinding mechanics and Rowe for the thermal analysis. Where sources disagree — partition ratios by a few points, G-ratio bands by an order of magnitude depending on wheel and coolant, specific energy by 20% between labs — the guide prints the working value and the conservative alternative beside it. Grinding is a process where the machine, the wheel, the coolant and the part hardness form a single system; no table survives contact with a specific machine. Measure power, measure burn on a first article, and set the dressing schedule from the part that passed, not from the catalogue.

FabFlow connects buyers with grinding shops that can hold the tolerance and prove the surface integrity — surface, cylindrical, centerless, internal, jig, and superfinishing work, with first-article inspection and nital-etch documentation on request. Post a grinding job at fabflow.app and compare quotes from verified precision fabricators.

More FabFlow blog posts