Friction Coefficient µ: Why 90% of Your Torque Never Becomes Clamping Force

You torque a fastener to the specified value and assume the joint is secure. In reality, most of that energy was already gone the moment you stopped turning - lost to friction. Understanding how tightening torque, friction coefficient, and clamp force relate to one another leads to better design and process decisions. This article explains the physics precisely, without oversimplification, with direct reference to VDI 2230.

The 90/10 Rule: What Your Torque Actually Does

Imagine applying 100 Nm to a bolt. How much of that actually generates clamping force?

At a typical overall friction coefficient of µ_ges = 0.12, roughly 85-90% of the tightening torque is consumed by friction - only the remaining portion generates actual preload in the fastener. Breaking it down further: Without lubrication, approximately 50% of the torque is lost to underhead friction and 40% to thread friction, leaving only around 10% to produce clamp force.

This is not an inefficiency that can be engineered away - it is physics. But it is physics that can be managed, once you know which variables to control.

The VDI 2230 Formula, Plainly Stated

VDI 2230 describes the relationship between tightening torque M_A and preload F_V with the following complete formula:

M_A = F_V × (0.16 × P + 0.58 × µ_G × d₂ + µ_K × D_Km / 2)

Where:

• P - Thread pitch
• µ_G - Thread friction coefficient (friction between bolt and nut thread flanks)
• d₂ - Pitch diameter of the thread
• µ_K - Underhead friction coefficient (friction beneath the bolt head or nut bearing face)
• D_Km - Mean friction radius under the head

The tightening torque therefore consists of two components: the thread torque M_G, which overcomes thread friction, and the underhead friction torque M_K, which accounts for friction at the bearing surface. Preload F_V appears in the formula as a factor - it is the result, not the input.

What does this mean in practice? If you specify a tightening torque without knowing µ_G and µ_K, you cannot reliably predict F_V. The friction coefficient is the decisive - yet difficult to control - variable in the system.

Why µ Is So Hard to Control

Typical Scatter Ranges in Practice

The friction coefficient is not a material constant - it is a system parameter that depends on coating, lubrication, material pairing, surface roughness, and even tightening speed.

VDI 2230 distinguishes two friction coefficient classes: Class A (narrow scatter, µ = 0.08 to 0.16, under defined lubrication conditions) and Class B (wide scatter, µ = 0.10 to 0.23, under undefined conditions).

Typical reference values from practice:

Particularly critical: Trivalent zinc passivation surfaces (Cr(III) passivation, now widespread following the RoHS transition) exhibit a significantly wider friction coefficient window of µ_ges = 0.18-0.38, compared to µ_ges = 0.21-0.28 for the earlier hexavalent chromate coatings. Anyone who set their tightening torque based on older coating data and then switched to Cr(III) without re-characterizing the friction coefficient is systematically risking insufficient preload.

How Friction Variation Shifts Preload

The direct impact is substantial: When the underhead friction coefficient rises from 0.10 to 0.16 at constant tightening torque, the assembly preload drops by 22%. This is not an edge case - it is the difference between a secure joint and an under-tensioned one.

There is a further consideration: Every individual bolted joint will exhibit a specific assembly preload, even when an entire production batch of identical fastener assemblies is tightened in exactly the same way - the root cause is the unavoidable scatter in friction coefficients at the thread and under the bolt head.

The Tightening Factor α_A: How VDI 2230 Addresses Scatter

VDI 2230 quantifies scatter through the tightening factor α_A - the ratio of the maximum to the minimum achievable preload for a given tightening method.

With torque-controlled tightening, α_A typically falls between 1.4 and 2.0 - meaning the maximum preload can be twice as large as the minimum. Under undefined friction conditions (Class B), α_A can rise to 2.5-4.0.

The design implication: the larger α_A, the more the fastener must withstand in the worst case - and the more preload is lost in the best case. The window between "insufficient clamping force" and "overloaded fastener" becomes dangerously narrow.

Key takeaway: Angle-controlled tightening reduces preload scatter to ±15%, because the rotation angle correlates more directly with bolt elongation than torque does. Pure torque control is structurally imprecise when friction is unknown or variable - not because of the tool, but because of the physics.

Embedment Relaxation: The Silent Preload Loss After Assembly

Even a correctly tightened joint loses preload after assembly. Typical embedment losses range from 5 to 25% of the assembly preload, depending on the joint configuration. The cause: Micro-asperities at the bearing surface, thread flanks, and interfaces smooth out plastically under load - the joint "settles."

VDI 2230 accounts for this effect through the embedment allowance f_Z. Embedment is most pronounced in joints with:

• Multiple interfaces (each interface contributes to the total embedment)
• Soft materials such as aluminum or plastics
• Short grip lengths (limited elastic reserve)
• High surface pressures under the bolt head

Anyone who wants to observe embedment behavior in the torque-angle curve needs more than a single torque reading at the end of tightening - they need the complete torque-angle trace.

The Torque-Angle Curve: The Fingerprint of a Joint

What does an engineer see when they only know the final torque? A single data point. What do they see with the complete torque-angle curve? The entire history of the joint.

A complete torque-angle curve reveals four characteristic phases:

The decisive advantage of curve analysis over pure torque monitoring: the slope in the elastic region is friction-independent. It depends on system stiffness - bolt elongation and joint compliance. This means the curve can reveal whether the joint actually reached the desired preload - regardless of whether µ was 0.10 or 0.16 on that particular assembly.

Measuring µ Instead of Guessing It: The Practical Approach

How do you reliably determine µ_G and µ_K for a specific joint?

Friction coefficients cannot be measured directly - they must be calculated from measured torques and preloads. This requires at least two measured quantities: the total tightening torque and - to separate thread friction from underhead friction - the partial torque.

In practice, this means an analysis tool must capture the complete torque-angle curve at sufficient resolution, without requiring a fixed reference point on the component. This is precisely where the difference lies between a simple torque wrench and an analysis instrument like the QUANTEC MCS® from GWK.

The QUANTEC MCS®'s reference-point-free angle measurement captures rotation angle independently of a defined starting point - a critical advantage for joints where the tightening zone does not begin at a reproducible position, as is frequently the case with variable friction behavior. With an accuracy of ±1% between 10 and 100% of the rated range, the instrument provides the data foundation needed to determine µ_G and µ_K for a specific joint and to verify the tightening factor α_A by measurement - rather than estimating it from tables.

Interactive Friction Influence Calculator

How much does a change in friction coefficient shift your preload at constant tightening torque? Try it yourself:

What This Means for Design Engineers and Process Planners

Three conclusions that follow directly from the physics:

1. µ must be defined, not assumed. Anyone who selects a tightening torque from a table without knowing the friction conditions of their specific joint is working with an unknown control variable. Safe assembly requires precisely defined friction conditions and the tightest possible control of their scatter. This is especially important after changes in coating or supplier.

2. Lubrication is a process parameter, not a convenience. Lubrication reduces µ and narrows scatter - but only when applied reproducibly. Adding lubrication to threads significantly changes the friction coefficient and leads to undefined tightening conditions if it was not accounted for in the design. A specified lubricant with a documented µ window is always preferable to no lubricant with a wide scatter band.

3. Torque control alone is not sufficient for critical joints. Torque is the most widely used control variable in fastening technology - but it is an indirect measurement. Preload always depends on µ. Anyone who must demonstrate process capability (VDI/VDE 2862, Class A joints) needs torque-angle analysis as a foundation - not as an option.

Conclusion: See µ, Don't Guess It

The friction coefficient µ is the invisible control variable in every bolted joint. It determines how much of your tightening torque arrives as preload - and how much that preload varies from joint to joint. Knowing and controlling µ allows you to minimize α_A, precisely specify tightening torques, and demonstrably reduce process scatter.

Torque-angle analysis makes µ visible. It shows not just the final value, but the complete torque-angle curve - with full information about system stiffness, embedment behavior, and friction state. The QUANTEC MCS® from GWK was built for exactly this purpose: a compact joint analysis lab that can be deployed wherever joints need to be designed, validated, or monitored.