Rolling-Element Bearings: Selection, Fits and Installation: Radial and Axial Clearance Considerations

Radial and Axial Clearance Considerations

Race fits absorb some of the original bearing clearance:

• Approximately 80% of the actual interference shows up as race diameter change
• Heavy, stiff housings or solid shafts may cause even greater race diameter change
• Light metal housings (aluminum, magnesium) and tubular shafts cause lesser change
• Thermal effects must be accounted for — allow clearance in the proper direction for operating temperature differentials

Ball bearings can run with moderate preloads (up to 0.0005 inch) without affecting life or temperature rise.

Roller bearings have much lesser tolerance for preloading and must be carefully controlled to avoid overheating and self-destruction.

Critical Clearance Verification

In all critical applications:

• Check axial and radial clearances with feeler gauges or dial indicators
• Verify mounted clearances fall within the design engineer's specified tolerances
• Watch for: chips, scores, race misalignment, shaft/housing denting, housing distortion, end cover off-squareness, and mismatch of rotor and housing axial dimensions — all of these rob the bearing of clearance

the practitioner's Bearing Installation Protocol

1. Before touching the bearing: Verify shaft seat roundness (three-point electronic indicator), measure housing bore diameter in multiple planes, check surface finish with a profilometer
2. Verify fit class: Confirm the tolerance symbol matches the load condition, rotation direction, and bearing type
3. Measure everything: Inner ring bore, outer ring OD, shaft diameter, housing bore — compare all measurements against AFBMA tables
4. Clean obsessively: Every surface that contacts the bearing must be free of nicks, burrs, contamination, and corrosion
5. Align precisely: Sweep races with dial indicator after seating — verify runout against commercial or precision tolerance requirements
6. Clamp correctly: Use calibrated torque tools on locknuts; never use impact wrenches on bearing assemblies
7. Verify clearance: Check running clearance with feeler gauges or dial indicators before closing the housing
8. Document: Record all measurements for baseline comparison at the next service interval

Quick Reference: The Bearing Installation Decision Matrix

Your Next Step

Take one bearing installation in your facility — the next one that comes up. Before mounting, measure the shaft seat roundness in three planes and compare it to the tolerances in this guide. If the shaft surface deviation exceeds 0.0001 inch per inch of diameter for a precision application, or 0.0003 inch for a drawn cup needle bearing raceway, stop the installation and correct the shaft geometry first.

That single habit — measuring before mounting — will eliminate more premature bearing failures than any other practice change you can make.

The bearing is already precision-made. Your job is to give it a precision home.

Reference Standards: ANSI/ABMA 7-1995, ANSI/ABMA 18.1-1982 (R1994), ANSI/ABMA 18.2-1982 (R1993), ANSI/ABMA 20-1987, ANSI/ABMA 24.1-1989, ANSI/ABMA 24.2-1998, AFBMA Standard 4 (1984), ANSI/AFBMA Standard 8.2-1991.

Context and scope

Updated for modern engineers worldwide

A conveyor belt screams to a halt at 2:47 AM. Three thousand units of product sit motionless. an illustrative engineering practitioner stares at the wreckage — a seized bearing, smaller than her fist, has just cost her factory an estimated six figures in downtime. She turns to her maintenance lead. "How did we miss this?"

Six months earlier, an illustrative engineering practitioner had been handed the bearing selection for that exact conveyor system. He'd picked a bearing from a catalog, checked that it "looked about right," and moved on to the next task.

That one decision — made in under ten minutes — had a six-figure price tag.

This is the story of how bearings work, why they fail, and exactly how you can select them with confidence — whether you're a student opening your first catalog, a working engineer refining your skills, or a decision-maker who needs to understand what your team is doing.

No fluff. No filler. Just the knowledge that separates catastrophic failures from machines that run for decades.

Current-state problem

Here's a truth that most engineering programs gloss over: bearing selection is not a "plug and chug" exercise. It's a multi-variable estimation problem where getting it wrong doesn't show up for months or years — and when it does, the consequences are brutal.

Consider the roles in the example:

• the practitioner — Junior mechanical engineer, 18 months out of university, eager but overconfident
• the practitioner — Plant operations manager, 15 years of experience, responsible for uptime targets
• the practitioner — Senior design engineer and the practitioner's mentor, the person who's seen every failure mode imaginable

When the practitioner first opened a bearing catalog, he saw a wall of numbers. Thousands of pages. Dozens of bearing types. Tables that seemed to stretch into infinity. So he did what most people do: he found a bearing that fit the shaft, checked the load rating, and moved on.

That's the equivalent of buying a car based solely on whether you fit in the driver's seat.

the practitioner had been trying to warn him. "A bearing isn't just a part," he'd said over coffee one morning. "It's a prediction. You're predicting how long a piece of steel will survive under punishment — and the punishment comes from directions you haven't even thought about yet."

the practitioner hadn't listened. Not really. Not until the call came at 2:47 AM.

The World of Rolling Element Bearings: What You Actually Need to Know

Before we follow the practitioner's journey to redemption, you need the foundation. Let's build it from the ground up.

What Is a Rolling Element Bearing?

At its simplest, a rolling element bearing is a device that allows a shaft to rotate with minimal friction inside a housing. Instead of the shaft rubbing directly against the housing (which would generate enormous heat and wear), small rolling elements — balls or rollers — carry the load between the inner and outer rings of the bearing.

Think of it this way: Imagine pushing a heavy box across a floor. Now imagine putting that box on top of a dozen marbles. Which is easier?

That's the fundamental principle. But the engineering is in the details.

The Four Bearing Types You Must Know

Out of dozens of bearing types available worldwide, four cover the vast majority of mechanical engineering applications. These are the ones the practitioner drilled into the practitioner after the incident:

. Deep Groove Ball Bearings

The workhorse. The most common bearing in the world.

"If you don't know what bearing to use," the practitioner told the practitioner, "start here. You'll be right more often than not."

. Self-Aligning Ball Bearings

The forgiving one. Built for imperfect installations.

"In a perfect world, shafts are perfectly aligned and housings are perfectly bored," the practitioner explained. "We don't live in a perfect world. When alignment is questionable, these bearings save you."

. Cylindrical Roller Bearings

The heavy lifter. When loads get serious.

. Spherical (Self-Aligning) Roller Bearings

The tank. Maximum load, maximum forgiveness.

Quick Comparison: Which Bearing Fits Your Situation?

★ = Low/Poor, ★★★★★ = High/Excellent

*Depends on bearing sub-type (NU, NJ, NUP)

The L₁₀ Life: A Statistical Bet

Here's a concept that surprises most beginners: bearing life is a probability, not a guarantee.

The industry standard measure is the L₁₀ life — defined as the number of revolutions (expressed in millions) at which 90% of a large group of identical bearings can be expected to survive under the given operating conditions.

Let that sink in.

• If you have 100 bearings operating under identical conditions, the L₁₀ life is the point where 10 of them will have failed
• 90 will still be running
• The average life is approximately 5 times the L₁₀ life

This is fundamentally a statistical prediction, not a certainty. And that's exactly where the practitioner went wrong — he treated it as a certainty.

"When the catalog says a bearing has an L₁₀ life of 500 million revolutions," the practitioner told the practitioner, "it's not promising you 500 million revolutions. It's saying that in a population of these bearings, 90% will make it to 500 million. Your specific bearing might fail at 100 million. Or it might last 2 billion."

Converting Between Hours and Revolutions

Since most engineers think in operating hours, not millions of revolutions, the conversion formula is essential:

Design Life Formula:

L = (60 × N × h) / 10⁶

Where:

Example: You need a bearing to last 20,000 hours at 1,500 rpm:

L = (60 × 1,500 × 20,000) / 1,000,000 L = 1,800,000,000 / 1,000,000 L = 1,800 million revolutions

Typical design life targets by application:

Hidden Killer #1: The Steel Itself

The ISO equations that underpin all bearing life calculations are based on standardized material specifications. In practice, most commercial bearings from reputable manufacturers meet or exceed these specifications. But the quality of the steel is foundational — it sets the upper limit of what's possible.

What this means for you: Unless you're ordering specialty bearings for extreme environments (aerospace, nuclear, deep-sea), standard commercial bearings from recognized brands will meet the ISO material assumptions. This factor typically works in your favor — most bearing steels today are better than what the equations assume.

Hidden Killer #2: Lubrication

This is the factor that destroyed the practitioner's conveyor bearing.

Lubrication isn't just "put some grease on it." It's a complex system of interacting variables:

The critical insight: Viscosity and temperature are intimately linked. For liquid lubricants, viscosity drops as temperature rises. The standard reference point used by major manufacturers is a viscosity of approximately 100 mm²/s at 40°C for pre-lubricated deep groove ball bearings.

the practitioner's mistake: The conveyor operated in a high-temperature environment near an oven line. The bearing's grease, selected for room-temperature operation, thinned out to the point where it could no longer maintain the protective film. Metal touched metal. The bearing was dead within months.

Hidden Killer #3: Cleanliness

Contamination is the silent assassin of bearings. Even microscopic particles can cause surface damage that initiates fatigue failure.

Sources of contamination include:

• Environmental dust and dirt
• Metallic particles from wear of adjacent components
• Water or other liquids that contaminate lubricant or cause corrosion
• Inadequate filtration of circulating lubricant
• Poor sealing that allows ingress of foreign material

The key relationship: Many of these factors are interconnected. A well-sealed bearing prevents contaminant entry. A good filtration system removes particles from circulating oil. Proper installation prevents introduction of debris during assembly.

"Cleanliness is the factor that engineers can control the most," the practitioner told the practitioner, "and it's the one they ignore the most."

The Three Methods: How to Actually Calculate Bearing Life

After the failure, the practitioner spent three weeks studying under the practitioner. He learned that there are three progressively more sophisticated methods for calculating bearing life. Each accounts for more real-world variables, and each produces a different answer.

Understanding all three is what separates competent engineers from exceptional ones.

Method 1: The ISO Basic Rating Life (L₁₀)

What it considers: Load only

When to use it: Quick estimates, preliminary sizing, or when detailed operating conditions are unknown

The core equation:

L₁₀ = (C / P)^p

Where:

The exponent matters. The fact that life is proportional to the cube (or higher) of the load ratio means small changes in load create enormous changes in life:

Read that table carefully. Doubling the C/P ratio multiplies bearing life by 8× for ball bearings and ~10× for roller bearings. This is the most powerful lever you have in design. If you can reduce the load on a bearing by even 20%, the life increase is dramatic.

Calculating the Equivalent Dynamic Bearing Load (P)

Real-world bearings rarely experience purely radial or purely axial loads. Most see a combination. The equivalent dynamic bearing load P converts the actual load combination into a single hypothetical load that would produce the same bearing life.

For deep groove ball bearings:

Step 1: Calculate Fa/Fr Step 2: Calculate Fa/C₀ (static load ratio) Step 3: From Fa/C₀, determine the value of 'e' (threshold factor) Step 4: Compare Fa/Fr to e

If Fa/Fr ≤ e → P = Fr (axial load is negligible) If Fa/Fr > e → P = X × Fr + Y × Fa where X = 0.56 (for normal clearance) and Y is determined from Fa/C₀ (read from manufacturer data)

For self-aligning ball bearings:

If Fa/Fr ≤ e → P = Fr + Y₁ × Fa If Fa/Fr > e → P = 0.65 × Fr + Y₂ × Fa

(Values of e, Y₁, Y₂ are specific to each bearing — found in bearing tables)

For cylindrical roller bearings:

Type NU (no flanges): P = Fr (pure radial only)

Flanged types: If Fa/Fr ≤ e → P = Fr If Fa/Fr > e → P = 0.92 × Fr + Y × Fa

where e and Y depend on the bearing series: Series 10, 2, 3, 4: e = 0.2, Y = 0.6 Series 22 and 23: e = 0.3, Y = 0.4

IMPORTANT: Fa/Fr must not exceed 0.5 for cylindrical roller bearings.

For spherical roller bearings:

If Fa/Fr ≤ e → P = Fr + Y₁ × Fa If Fa/Fr > e → P = 0.67 × Fr + Y₂ × Fa

(Values of e, Y₁, Y₂ from bearing tables)

Method 2: The ISO Adjusted Rating Life (Lₙₐ)

What it considers: Load + reliability + material + lubrication

When to use it: Standard engineering design with known operating conditions

After the industry recognized that load alone couldn't predict bearing life accurately, ISO introduced the adjusted rating life in 1977:

Lₙₐ = a₁ × a₂ × a₃ × L₁₀

Where:

Reliability factors (a₁):

Read this table. Going from 90% to 99% reliability reduces your calculated life by nearly 80%. This is why specifying reliability without understanding the consequences can massively oversize (and over-cost) your bearings.

The Viscosity Ratio (κ):

This is the single most important parameter in Method 2. It tells you whether your lubricant is doing its job:

κ = v / v₁

Where:

• v = actual kinematic viscosity of your lubricant at the bearing's operating temperature (mm²/s)
• v₁ = required kinematic viscosity for adequate lubrication (mm²/s), determined by bearing size and shaft speed

How to find v₁: Use the bearing's mean diameter (dₘ) and the shaft speed:

dₘ = 0.5 × (d + D)

Where d = bearing bore diameter and D = bearing outer diameter

Then look up v₁ from the manufacturer's chart using dₘ and rpm.

What κ values mean:

💡 the practitioner's realization: After seeing the numbers, she understood why the conveyor bearing failed. The lubricant viscosity at operating temperature was so low that κ was well below 0.4. The bearing was effectively running dry — and no amount of load capacity could compensate.

Method 3: The New Life Theory (Most Advanced)

What it considers: Everything in Method 2 + fatigue load limit + contamination

When to use it: Precision engineering, critical applications, or whenever contamination is a significant factor

This method introduces a revolutionary concept: the fatigue load limit (Pᵤ) — a threshold load below which fatigue failure will not occur in the bearing, provided lubrication and cleanliness are adequate.

Yes, you read that right. Below the fatigue load limit, with proper lubrication and cleanliness, the bearing can theoretically last indefinitely.

Lₙₐₐ = a₁ × a_the bearing supplier × L₁₀

For standard 90% reliability (a₁ = 1):

Lₙₐₐ = a_the bearing supplier × L₁₀

The value of a_the bearing supplier depends on two inputs:

1. The viscosity ratio κ (same as Method 2)
2. The contamination-adjusted fatigue ratio: η_c × (Pᵤ / P)

Where:

• Pᵤ = fatigue load limit (from bearing tables)
• P = equivalent dynamic bearing load
• η_c = contamination factor

Contamination Factor (η_c) Guide:

Problem Statement

A 45 mm diameter shaft running at 5,000 rpm carries a radial load of 8 kN with no axial load. The bearing will be oil-lubricated with oil having a viscosity of 20 mm²/s at the operating temperature. Reliability is standard (90%). Calculate the bearing life using all three methods under:

• (a) Clean conditions
• (b) Contaminated conditions (η_c = 0.2)

the practitioner selected: Deep groove ball bearing 6309 (45 mm bore)

From manufacturer data for bearing 6309:

Step-by-Step Solution

Since there is no axial load: P = Fr = 8 kN

Method 1: ISO L₁₀ Life

L₁₀ = (C / P)³ L₁₀ = (52.7 / 8)³ L₁₀ = (6.5875)³ L₁₀ = 286 million revolutions

Converting to hours:

h = (L₁₀ × 10⁶) / (60 × N) h = (286 × 10⁶) / (60 × 5,000) h = 286,000,000 / 300,000 h ≈ 953 hours

the practitioner's first reaction: "Only 953 hours? That's about 40 days of continuous operation!"

the practitioner: "Keep going."

Method 2: ISO Adjusted Life (Clean Conditions)

Step 1: Reliability factor for 90%: a₁ = 1

Step 2: Calculate mean diameter:

dₘ = 0.5 × (45 + 100) = 72.5 mm

Step 3: Determine required viscosity v₁

From the manufacturer chart at 5,000 rpm with dₘ = 72.5 mm: v₁ ≈ 7 mm²/s

Step 4: Calculate viscosity ratio:

κ = v / v₁ = 20 / 7 = 2.86 ≈ 2.9

Step 5: Determine a₂₃ factor

From the manufacturer diagram at κ = 2.9: a₂₃ = 2

Step 6: Calculate adjusted life:

Lₙₐ = a₁ × a₂₃ × L₁₀ Lₙₐ = 1 × 2 × 286 Lₙₐ = 572 million revolutions

Converting to hours: ≈ 1,907 hours

the practitioner: "See that? Just by accounting for the fact that our lubricant viscosity is nearly three times higher than the minimum required, the predicted life doubled. Lubrication matters."

Method 3: New Life Theory (Clean Conditions)

Step 1: a₁ = 1 (90% reliability)

Step 2: From bearing tables: Pᵤ = 1.34 kN

Step 3: For clean conditions: η_c = 0.8

Step 4: Calculate the contamination-adjusted fatigue ratio:

η_c × (Pᵤ / P) = 0.8 × (1.34 / 8) = 0.8 × 0.1675 = 0.134

Step 5: Using κ = 2.9 and η_c × (Pᵤ/P) = 0.134, from the manufacturer's diagram:

a_the bearing supplier = 8

Step 6: Calculate life:

Lₙₐₐ = a_the bearing supplier × L₁₀ Lₙₐₐ = 8 × 286 Lₙₐₐ = 2,288 million revolutions

Converting to hours: ≈ 7,627 hours

the practitioner: "The new life theory predicts a life four times higher than the adjusted method? How?"

the practitioner: "Because it accounts for the fatigue load limit. Our operating load is well above Pᵤ, but the ratio combined with good lubrication gives the bearing credit it deserves. This is closer to reality."

Method 3 Under Contaminated Conditions (η_c = 0.2)

η_c × (Pᵤ / P) = 0.2 × (1.34 / 8) = 0.2 × 0.1675 = 0.0335

From the diagram at κ = 2.9 and η_c × (Pᵤ/P) = 0.0335:

a_the bearing supplier = 1.2

Lₙₐₐ = 1.2 × 286 = 343 million revolutions

Converting to hours: ≈ 1,143 hours

The Results Summary: Three Methods, Three Answers

The lesson that hit the practitioner like a freight train:

Clean conditions with good lubrication → bearing life is 8 times the basic calculation

Contaminated conditions → life drops to barely above the basic calculation, despite the same good lubrication

Contamination erased nearly all of the life benefit from superior lubrication.

The Complete Selection Process: Your Step-by-Step Checklist

After his transformation, the practitioner created a checklist that he follows for every bearing selection. the practitioner approved it. the practitioner mandated it for the entire engineering team.

Here it is, applicable to all four major bearing types:

Phase 1: Define Requirements

• Determine the required design life in operating hours (see application table above)
• Convert to millions of revolutions using L = (60 × N × h) / 10⁶
• Determine average radial load (Fr) — usually from shaft reaction force calculations
• Determine average axial load (Fa) — from thrust forces, gear separating forces, etc.
• Apply shock/dynamic load factors if loads are not steady
• Calculate Fa/Fr ratio — this drives your bearing type and P calculation

Phase 2: Select Bearing Type and Size

• Choose bearing type based on load direction, speed, alignment, and environment (use the comparison table above)
• Knowing shaft diameter, look up available bearings in manufacturer tables
• Start with a mid-range bearing for that shaft size
• Record: C (dynamic rating), C₀ (static rating), and any calculation factors (e, Y₁, Y₂, etc.)

Phase 3: Calculate and Verify

• Calculate equivalent dynamic bearing load (P) using the appropriate formula for your bearing type
• Calculate L₁₀ life using L₁₀ = (C/P)^p
• Compare L₁₀ to your required life L

• • If L₁₀ < L → bearing is too small → try the next larger bearing
    • If L₁₀ ≈ L → acceptable
    • If L₁₀ >> L → bearing may be oversized → try a smaller bearing (saves cost and space)

• If lubrication and cleanliness data are available, apply Method 2 or 3 for more realistic life predictions

Phase 4: Safety Checks (Do Not Skip These)

These are the checks that the practitioner skipped the first time:

Check 1: Equivalent Static Load

P₀ = 0.6 × Fr + 0.5 × Fa (for deep groove ball bearings) P₀ = Fr + Y₀ × Fa (for self-aligning ball and spherical roller bearings) P₀ = Fr (for cylindrical roller bearings)

Rule: P₀ must be less than C₀ (or C₀/1.5 for roller bearings)

If P₀ < Fr, set P₀ = Fr

Check 2: Minimum Radial Load

For satisfactory operation, the bearing must carry a minimum radial load:

Why this matters: Under very light loads, the rolling elements can skid rather than roll, causing surface damage and premature failure. This is counter-intuitive — too little load can be just as destructive as too much.

Check 3: Maximum Speed

Verify that the maximum shaft speed does not exceed the bearing's speed rating (listed in manufacturer tables).

Check 4: Adapter Sleeve Load Limit (Self-Aligning Bearings Only)

If using a tapered bore bearing with an adapter sleeve:

Fa < 3 × B × d

Where:

• Fa = axial load (N)
• B = bearing width (mm)
• d = bearing internal diameter (mm)

Advanced Concepts: The Viscosity Deep Dive

For those of you who want to go deeper — the engineers, the specialists, the people who the practitioner calls at 2 AM — here's the full picture on lubricant viscosity and bearing life.

Temperature-Viscosity Relationship

Lubricant viscosity is not a fixed number. It changes with temperature, and this change is the single most common source of bearing life miscalculation.

General relationship for mineral oils and greases:

The engineering approach:

1. Estimate the bearing operating temperature — consider ambient conditions, heat generated by the load and speed, housing design, and cooling
2. Look up your lubricant's viscosity at that temperature using the lubricant manufacturer's viscosity-temperature chart
3. Calculate κ and apply it to your bearing life calculation