Rolling-Element Bearing Life in Travel and Swing Drives

2. Basic rating life and what it assumes

2.1 The model

The basic rating life is defined in ISO 281¹ as

L₁₀ = (C / P)^p   [10⁶ revolutions]

with p = 3 for ball bearings and p = 10/3 for roller bearings. It derives from the Lundberg–Palmgren treatment of subsurface-initiated rolling contact fatigue, in which the probability of survival is a Weibull function of the orthogonal shear stress amplitude, the depth at which it acts, the stressed volume and the number of stress cycles. The load–life exponent is not a free fit to life data, but nor does it follow from geometry alone: it comes from how Hertzian contact dimensions scale with load, combined with the exponents Lundberg and Palmgren gave to stress, depth and number of cycles — values they set to match bearing tests. In point contact (ball on raceway) the contact ellipse grows as Q^1/3 in each semi-axis and stress as Q^1/3; in line contact (roller on raceway) the contact half-width grows as Q^1/2 and stress as Q^1/2. Propagating those through the stressed-volume integral yields exponents of 3 and 4 respectively in the idealized case. The 10/3 adopted for roller bearings is a deliberate, documented compromise: real rollers are crowned, so the contact is neither purely line nor purely point, and edge effects prevent the theoretical exponent of 4 from being realized. This is one of the places where genuine disagreement persists in the literature — some authorities argue that a well-profiled roller under good lubrication behaves closer to the theoretical exponent, others that 10/3 is already optimistic for imperfectly aligned installations.

2.2 What L₁₀ is not

L₁₀ is the life that 90 % of a large population of apparently identical bearings will reach or exceed before the first evidence of subsurface fatigue spalling, under the stated load and speed. Three consequences follow, and all three are routinely lost in translation:

• It is a fatigue life. It predicts nothing about wear, corrosion, electrical erosion, cage failure, seal failure, lubricant degradation or contamination-driven surface-initiated damage — which, in field service on mobile equipment, together account for the large majority of removals. ISO 15243⁶ catalogs these mechanisms and is the correct companion document when diagnosing a bearing that failed early.
• It is a statistical statement about a population, not a prediction about one bearing. Individual lives in a homogeneous test group commonly spread over more than an order of magnitude.
• It is a lower decile, not a mean. Under the Weibull dispersion assumed in the standard, the mean life of the population is several times L₁₀ — a fact that makes L₁₀ conservative in the aggregate and useless as a maintenance interval for an individual machine.

2.3 The assumptions embedded in C

The catalog value of C is not a property of the bearing alone; it is the load that would produce L₁₀ = 1 million revolutions under a specific idealized set of conditions: contemporary vacuum-degassed through-hardened or case-carburized bearing steel of stated cleanliness, ring support that is rigid and round, negligible misalignment, a constant load of constant direction, an adequate lubricant film, a temperature within the dimensional stabilization range of the rings, and no contamination. Every departure from that list is handled outside C, in the modification factors of Section 4. A calculation that stops at (C/P)^p has silently asserted that a travel drive running in abrasive slurry at 95 °C meets laboratory conditions.

3. Equivalent dynamic load

3.1 Combining radial and axial components

Bearings are rated radially or axially, but loaded in both directions at once. The standard reduces a combined load to a single equivalent radial load that produces the same fatigue damage:

P = X·F_r + Y·F_a

X and Y are tabulated per bearing type and depend on the ratio F_a/F_r relative to a limit e, and — for angular contact and tapered roller bearings — on the contact angle. Two points repay attention in drive applications. First, in a tapered roller pair the axial reaction is not independent of the radial load: a radial load on a tapered bearing generates an internal axial component of approximately 0.5·F_r/Y, which is reacted by the opposing bearing. The pair must be solved together; treating them as two independent bearings understates the load on one of them. Second, some conventions (notably the ABMA formulation) include a rotation factor V, taken as 1 for inner-ring rotation and 1.2 for outer-ring rotation, reflecting the shorter life of the inner raceway, usually the more highly stressed of the two, when it is stationary relative to the load: one arc of it then takes the peak element load at every element pass rather than once per revolution. Planetary gear bearings — where the load direction is fixed relative to the rotating planet pin — are precisely the case where this distinction matters.

3.2 Aggregating a duty cycle

Real travel and swing duty is a sequence of discrete regimes: tracking loaded, tracking empty, grade climbing, counter-rotation, slewing loaded, slewing empty, idle. Because life is non-linear in load, the correct aggregation is a p-th power mean weighted by revolutions, not by time:

P_m = [ Σ ( U_i · P_i^p ) / Σ U_i ]^1/p,   U_i = n_i · t_i

Equivalently, and more transparently for review, one may apply Palmgren–Miner linear damage accumulation directly: 1/L = Σ (u_i / L_i), where u_i is the fraction of total revolutions spent in regime i and L_i the rating life at that regime's load. The two are algebraically identical when speed is included correctly, and the second makes it obvious which regime dominates. Almost always, one short high-load regime does. A duty cycle in which 5 % of revolutions occur at three times the load of the other 95 % does more damage in that 5 % than in the other 95 % combined: about 1.4 times as much for a ball bearing (p = 3) and about twice as much for a roller bearing (p = 10/3).

3.3 Why modest overload is disproportionate

The cubic exponent is the single most important practical consequence of the model, and the one most often underestimated in specification meetings. Table 2 gives the relative life for a load increase over the design value.

A 25 % load increase — a heavier bucket, a wider track shoe, an attachment that moves the center of gravity outboard, an operator who counter-rotates on rock — halves bearing fatigue life. Doubling load leaves one eighth. The corollary is more useful than the warning: because the relationship is symmetric, modest reductions in equivalent load buy large life gains, and effort spent on load path and alignment is repaid at the third power.

4. Modified rating life

ISO 281¹ extends the basic model to

L_nm = a₁ · a_ISO · L₁₀

4.1 Reliability factor a₁

a₁ converts the 90 % survival probability to another. Representative values from the standard: 1.00 at 90 %, 0.64 at 95 %, 0.37 at 98 %, 0.25 at 99 %, and 0.093 at 99.9 %. The cost of high reliability is severe and it is worth stating plainly to a fleet engineer: specifying 99 % survival rather than 90 % reduces the usable calculated life to a quarter. Note that these values reflect the modified Weibull dispersion adopted in the 2007 revision of ISO 281¹ and differ from figures published under earlier editions — a₁ tables from different vintages are not interchangeable.

4.2 Life modification factor a_ISO

a_ISO is where the lubrication and contamination physics enters. It is presented in the standard as a family of curves — one set per bearing type — of the form

a_ISO = f ( e_C · C_u / P , κ )

The fatigue load limit C_u is the load below which, in clean and well-lubricated conditions, fatigue does not initiate — the rolling-contact analog of an endurance limit. It follows from the Ioannides–Harris introduction of a stress-fatigue threshold into the Lundberg–Palmgren integral. ISO 281¹ approximates C_u as the load producing a maximum Hertzian contact stress of about 1 500 MPa in conventional through-hardened bearing steel; the value is published per part number by bearing manufacturers. Where the ratio e_C·C_u/P is high and κ is good, a_ISO can exceed 10 and the standard truncates it at 50. Where the ratio is low, a_ISO falls towards 0.1, the lowest value the standard's curves reach — up to a tenfold life penalty relative to the catalog figure, arising entirely from oil condition. It is not unusual for a_ISO to move the answer further than any plausible change of bearing size.

The contamination factor e_C spans roughly 1.0 for extreme cleanliness down to values approaching zero for severe contamination; for a given cleanliness level it also depends on bearing mean diameter, because a large bearing tolerates a given particle size better than a small one. Published guidance ties e_C bands to oil cleanliness expressed as an ISO 4406⁸ code (particle counts per millilitre above 4, 6 and 14 µm(c), calibrated per ISO 11171⁹). Two practical consequences for splash-lubricated planetary final drives: the enclosure is a closed, unfiltered system in which wear debris accumulates monotonically between oil changes, so e_C degrades over the service interval rather than sitting at a constant value; and any seal breach admitting site abrasives collapses e_C immediately. Contamination-driven damage is also surface-initiated — dent-raised stress risers on the raceway — so it does not merely shorten fatigue life, it changes the failure mode to one the fatigue model does not describe.

5. Lubrication regime and viscosity ratio

κ is the ratio of the actual kinematic viscosity of the lubricant at operating temperature to a reference viscosity ν₁ required to form an adequate elastohydrodynamic film. ν₁ is read from a chart in ISO 281¹ as a function of bearing mean diameter d_m and speed n: large, slow bearings need high viscosity; small, fast ones need less. Viscosity grade is classified per ISO 3448¹⁰ and measured per ASTM D445¹¹, but the grade is quoted at 40 °C — what matters is the viscosity at the actual sump temperature, which in a hard-working travel drive may be 40–60 °C above ambient. Because viscosity falls steeply with temperature, a lubricant selected on its grade number alone can be well below the reference viscosity in service while appearing correct on paper.

κ relates to the more fundamental film parameter Λ = h_min/σ, the ratio of minimum EHL film thickness (Hamrock–Dowson) to composite surface roughness of the two contacting bodies. The regimes are conventionally: Λ > 3, full film separation and fatigue-limited life; 1 < Λ < 3, mixed film with intermittent asperity contact; Λ < 1, boundary lubrication where the additive package, not the base oil, is doing the work. Note the dependence on σ as well as h_min: a smoother raceway raises Λ at unchanged oil and speed, which is why surface finish specification is a lubrication decision as much as a manufacturing one.

Two caveats. First, a high κ obtained with a very viscous oil raises churning losses and sump temperature, which lowers viscosity again — the system finds its own equilibrium and the naive calculation can be self-defeating. Second, κ and e_C are not independent in the a_ISO chart: a thicker film raises the particle size that can pass through a contact without denting it, so good lubrication partially mitigates moderate contamination, and poor lubrication amplifies it.

6. Internal geometry: load zone, clearance and setting

6.1 Load distribution

Only a fraction of the rolling elements carries load at any instant. For a radially loaded bearing with zero operating clearance, the load zone spans 180°, and for point contact (ball bearings) the Stribeck result gives

Q_max ≈ 4.37 · F_r / (Z · cos α)

with the coefficient conventionally rounded to 5 for ball bearings to allow for normal operating clearance. For roller bearings (line contact) the same equilibrium, taken over the distribution below, gives a coefficient of about 4.08. The elemental load distribution across the zone follows Q(ψ) = Q_max[1 − (1 − cos ψ)/(2ε)]ⁿ, with n = 3/2 for point contact and 10/9 for line contact, and ε the load distribution factor: ε = 0.5 at zero clearance, ε < 0.5 with clearance (narrower zone, higher peak), ε > 0.5 under preload (wider zone, and a lower peak only while the preload is light: heavier preload raises the peak above its zero-clearance value). Because life scales as roughly Q_max⁻³, the load zone geometry is not a second-order effect. Excessive operating clearance concentrates the whole radial load onto two or three elements and can cost more life than a substantial increase in external load.

6.2 Operating clearance is not fitted clearance

The clearance that matters is the one present at temperature and under load, not the value in the box. Radial internal clearance classes are specified per ISO 5753-1⁷. It is reduced by interference fits on the shaft and in the housing, and modified further by the differential thermal expansion between an inner ring running hot on a shaft and an outer ring in a comparatively cool aluminum or cast-iron housing. In a travel drive the inner ring commonly runs 10–20 °C hotter than the outer, closing clearance in service. The correct design sequence is therefore: select the clearance class such that after fit-up and thermal equilibrium the operating clearance is near zero to slightly negative — not such that the mounted cold clearance looks tidy.

6.3 Tapered roller bearing setting

Tapered roller bearings are mounted in opposed pairs and their axial setting — end play, line-to-line, or preload — is an installation variable, not a bearing property. The published relationship between setting and fatigue life is asymmetric: life rises as end play is removed, peaks at a small preload, then falls very steeply as preload increases, because preload adds directly to Q_max on every element and raises operating temperature, which increases preload further in a thermal runaway loop. End play, by contrast, degrades life gradually. Standard practice is therefore to set a cold end play calculated so that thermal growth carries the assembly to light preload at operating temperature, and to accept that the safe side of the optimum is the end-play side. Setting is verified by rolling torque, by axial displacement measurement, or by a calibrated collapsible spacer; each method has its own tolerance stack and the method must be stated with the specification.

6.4 Misalignment and edge loading

Angular misalignment between inner and outer ring tilts the load distribution along the roller length, concentrating stress at one end of the contact. In an uncrowned line contact this produces a theoretically singular edge stress; real rollers use a crowned or logarithmic profile specifically to bound it, and ISO/TS 16281³ provides the slice-based method for computing life under such non-uniform contact conditions. Published permissible misalignment is very small for the non-aligning types — of the order of a few arc-minutes for cylindrical and tapered roller bearings, somewhat more for deep-groove ball bearings — against one to several degrees for self-aligning ball and spherical roller bearings. Since housing bore alignment, shaft deflection under load and carrier plate distortion all contribute, misalignment should be assessed as a loaded, not an unloaded, quantity. Where geometry cannot guarantee alignment, the correct response is a self-aligning bearing type, not a larger rigid one.

7. Where the fatigue model stops applying

7.1 Static capacity and brinelling

The basic static load rating C₀ of ISO 76² is defined as the load producing a calculated contact stress, at the center of the most heavily loaded rolling element/raceway contact, of 4 200 MPa for ball bearings other than self-aligning types and 4 000 MPa for roller bearings — a stress at which the total permanent deformation of rolling element and raceway is approximately 0.0001 of the rolling element diameter in conventional bearing steel. This is the governing criterion for shock events: a machine dropped onto its tracks, a swing arrested against a stop, a boom impact. A true brinell dent is a permanent stress raiser and initiates surface fatigue immediately thereafter, so a single overload event can invalidate an otherwise sound fatigue calculation for the whole remaining life.

7.2 Oscillating and slow-moving applications

Swing drives, and slewing rings in particular, sit awkwardly in a model whose unit is the revolution. An equivalent rotational speed can be formed by scaling the oscillation frequency by the ratio of swept angle to a full revolution, but conventions differ over whether the stated angle is the total swept angle or the half amplitude and whether a cycle includes the return stroke. Each convention is worth a factor of two, and the two compound: the same stated angle and frequency can imply equivalent speeds up to four times apart. That is large enough to matter, so the convention must always be stated alongside the number.

More important is a threshold effect. When the swept angle is smaller than roughly twice the angular pitch between rolling elements, no element travels as far as the position its neighbor started from: each works back and forth over its own short track, and the raceway between those tracks is never rolled over. Fresh lubricant cannot be entrained, no hydrodynamic film forms, and the dominant mechanism becomes false brinelling and fretting corrosion — a wear process producing raceway markings superficially resembling brinell dents but of entirely different origin, cataloged as such in ISO 15243⁶. No fatigue calculation predicts it, and remedies are different in kind: grease selection with appropriate anti-fretting additives, periodic full-rotation exercise of the mechanism, and attention to vibration transmitted while parked. Large slewing bearings are conventionally rated by the manufacturer's static capacity curve — combining axial load, radial load and tilting moment — rather than by ISO 281¹ at all, and mixing the two rating bases in one comparison is a common and serious error.

7.3 Reading a quoted life figure critically

Given the structure above, a defensible bearing life statement should disclose: which life is quoted (L₁₀ or L_nm, and at what reliability); the equivalent load and how a duty cycle was aggregated into it; the assumed κ, and at what oil temperature; the assumed e_C and the cleanliness level it corresponds to; the assumed operating clearance or setting; and the assumed misalignment. A figure quoted without those is a statement about a catalog, not about a machine. Conversely, where all six are disclosed, the calculation is auditable, and disagreements become disagreements about inputs — which can be resolved by measurement — rather than about arithmetic.