Mechanical Joint and Fastener Selection: Six Rules to Stop Bolt Fatigue Failure

Six Rules to Stop Bolt Fatigue Failure

the practitioner had these pinned to his workshop wall. the practitioner copied them:

1. Tighten the bolt properly. The bolt must be stretched beyond the external load to ensure the preload always exceeds the working load.
2. Bolt extension during tightening should be high. Use at least 1× bolt diameter of thread length under the nut.
3. Use high-strength bolts or small bolts — they have more elasticity relative to their size.
4. Shank bolts in clearance holes are better — use bolts where the shank (not the thread) sits in the hole, with maximum clearance of 2–3 mm.
5. Rolled threads beat cut threads. Rolling induces compressive surface stresses that resist fatigue cracking.
6. Under vibration, use locknuts or Nyloc nuts to prevent rotation. Non-axial loading and prising (bending) action must be avoided.

A New Kind of Problem

After the spring and bolt failures, the practitioner was assigned to a simpler project — or so she thought. A welded bracket to support a conveyor drive motor. The bracket experienced both a direct downward load and a bending moment from the motor's offset position.

She sized the weld using the conventional method: calculate the load, divide by the weld area, check against allowable stress. Easy.

The weld cracked in service after four months.

The problem wasn't the weld size. It was the weld stress distribution.

Two Methods, Two Mindsets

There are two fundamentally different approaches to designing fillet welds. Understanding both is what separates competent engineers from dangerous ones.

Method 1: Conventional Design (The "Area" Method)

Treat the weld as a separate component with a real cross-sectional area.

Weld Area: A = t × L
Where:
  t = throat thickness = 0.707 × leg size (s)
  L = total weld length

Stress: f_s = F / (t × L)

For a standard fillet weld, the throat thickness is always 0.707 times the leg length. So a 6 mm weld has a throat of about 4.2 mm.

This method works fine for simple, direct loads — welds that only see force in one direction.

Method 2: Weld-as-a-Line (The "Line Stress" Method)

This is the method the practitioner taught the practitioner — and it's the one professional designers use for anything involving bending or torsion.

Instead of treating the weld as an area, treat it as a line with no thickness. The stress becomes a "line stress" with units of force per unit length (N/mm):

Line Stress: f = F / L (units: N/mm)

To convert to real stress:
  f_s = f / t

To find required weld size:
  t = f / f_s
  s = t / 0.707

Why use this method? Because when bending or torsion is involved, the stress distribution across the weld group depends on the geometry of the weld pattern, not just the weld size. Using the line method, you can calculate the section modulus or polar moment of the weld pattern first, then size the weld afterward.

Bending in Welds: Where Most Failures Hide

When a load is applied at a distance from the weld group's centroid, it creates a bending moment. The bending line stress is:

f_b = M / Z

Where:
  M = bending moment = Force × distance to centroid
  Z = section modulus of the weld group (treated as a line)

The section modulus depends on the weld pattern. Here are common configurations:

For the L-shaped weld, you need to find the centroid first:

Centroid vertical distance: y_c = d² / [2(b + d)]
Centroid horizontal distance: x_c = b² / [2(b + d)]

Torsion in Welds: The Invisible Killer

When a load is offset from the weld group's shear centre, it creates a torsional moment. This is the failure mode that caught the practitioner.

Torsional Line Stress: f_t = T × r / J

Where:
  T = torque = Force × perpendicular distance to centroid
  r = distance from centroid to the farthest weld point
  J = polar moment of the weld group (treated as a line)

Common polar moment formulas:

Combining Stresses: The Vector Rule

Here's what killed the practitioner's bracket: when a weld experiences both direct stress and bending (or torsion) stress, you must combine them vectorially — not by simple addition.

If the direct and bending stresses act in different directions (which they almost always do), the resultant is:

f_r = √(f_b² + f²)

If they act in the same direction at the critical point, they add directly:

f_r = f_b + f

At every other point, you need to resolve them into components and combine using the cosine rule or vector addition.

The critical point is always the location where the stresses combine to give the maximum resultant. This is usually at the extreme fibre of the weld group — the point farthest from the centroid.

Allowable Weld Stress: The Number You Must Know

For static or steady loads with common welding rods:

Allowable shear stress = 0.3 × UTS of the weld rod

For E41xx rod (UTS = 410 MPa): Allowable = 123 MPa
For E48xx rod (UTS = 480 MPa): Allowable = 144 MPa

For dynamic or cyclic loads, apply a safety factor on top of this.

Worked Example: the practitioner's Bracket Redesign

The motor bracket carried a 30 kN load at 60° to the vertical, attached by an L-shaped weld (b = 120 mm horizontal, d = 100 mm vertical).

Resolve the load:

Vertical component:   F_v = 30 × cos(60°) = 15 kN
Horizontal component: F_h = 30 × sin(60°) = 25.98 kN

Calculate total weld length:

L = b + d = 120 + 100 = 220 mm (using 224 mm accounting for returns)

Direct line stresses:

f_v = 15,000 / 224 = 67 N/mm
f_h = 25,980 / 224 = 116 N/mm

Bending moment about weld centroid:

M = 25,980 × 80 = 2,078,000 N·mm

Section modulus (for the L-weld pattern):

Z = b × d + d²/3 = 120 × 100 + 100²/3 = 4,533 mm²

Bending line stress:

f_b = 2,078,000 / 4,533 = 458.4 N/mm (vertical direction)

Combine at critical point:

Total vertical: f_v(total) = 458.4 + 67 = 525.4 N/mm
Total horizontal: f_h = 116 N/mm

Resultant: f_r = √(525.4² + 116²) = 538 N/mm

Size the weld:

Allowable stress (E41xx, static): f_s = 123 MPa = 123 N/mm²
Throat: t = 538 / 123 = 4.37 mm
Leg size: s = 4.37 / 0.707 = 6.18 mm → Use 8 mm fillet weld

This time, the bracket held.

The Power Screw — Where Rotation Becomes Muscle

What a Power Screw Actually Does

A power screw converts rotational torque into linear force (or vice versa). It's the same principle behind every car jack, every vise, every CNC machine's lead screw, and every valve stem.

There are three mechanical arrangements:

1. Screw rotates, nut translates — most common in jacks and presses
2. Nut rotates, screw translates — used in some actuators
3. Screw rotates AND translates — rare, used with fixed nuts

Thread Forms: Shape Determines Everything

The shape of the thread isn't decorative — it controls friction, strength, and whether the screw can hold a load without power.

For most engineering applications, the modified square (5° face) or trapezoidal metric (15° face) thread is the standard choice. Pure square threads are theoretically optimal but impractical to manufacture.

The Critical Terminology

Before you can design a power screw, you need to speak its language:

Standard pitch-diameter combinations (no formal standard exists — these are engineering-recommended proportions):

The Torque Equations: Raising and Lowering Loads

This is the heart of power screw design. The torque required to raise or lower a load depends on the friction angle, the helix angle, and the load itself.

For a square or modified square thread face:

Raising Load:    T = F × (d/2) × tan(φ' + θ)

Lowering Load:   T = F × (d/2) × tan(φ' − θ)

Where:

• F = axial load (the weight being lifted)
• d = pitch diameter
• θ = helix angle
• φ' = effective friction angle = arctan(μ / cos α)
• α = thread face angle (0° for square, 15° for trapezoidal)
• μ = coefficient of friction (typically 0.1 to 0.15; average 0.125)

The Friction Reality

The coefficient of friction between metal-on-metal lubricated screw threads depends on:

• Lubrication quality — good oil with correct viscosity gives μ as low as 0.1
• Surface finish — accurately machined, well-run-in surfaces are smoothest
• Usage cycles — surfaces "bed in" over time, reducing friction

Worked Example: the practitioner's Lifting Platform

Problem: A trapezoidal single-start metric thread, 30 mm diameter, lifts an axial load of 2 kN. Calculate the torque to raise and lower the load, assuming average friction.

Solution:

From the pitch table:  p = 8 mm, so Lead L = 8 mm (single-start)
Pitch diameter:         d = D − 0.5p = 30 − 4 = 26 mm
Coefficient of friction: μ = 0.125
Face angle (trapezoidal): α = 15°

Effective friction angle:
  tan φ' = μ / cos α = 0.125 / cos 15° = 0.125 / 0.966 = 0.1294
  φ' = 7.374°

Helix angle:
  tan θ = L / (πd) = 8 / (π × 26) = 0.0979
  θ = 5.594°

RAISING the load:
  T = F × (d/2) × tan(φ' + θ)
  T = 2000 × 13 × tan(7.374° + 5.594°)
  T = 2000 × 13 × tan(12.968°)
  T = 2000 × 13 × 0.2305
  T = 5,993 N·mm = 5.99 Nm

LOWERING the load:
  T = F × (d/2) × tan(φ' − θ)
  T = 2000 × 13 × tan(7.374° − 5.594°)
  T = 2000 × 13 × tan(1.780°)
  T = 2000 × 13 × 0.0311
  T = 808 N·mm = 0.808 Nm

Key insight: It takes 5.99 Nm to raise the load but only 0.808 Nm to lower it. The screw does most of the "holding" through friction.

Self-Locking: The Built-In Safety Feature

Notice that the lowering torque is positive — you still need to apply torque to lower the load. This means even if you let go of the handle, the load stays put. This is called self-locking, and it's one of the most valuable properties of power screws.

A screw is self-locking when: φ' > θ (friction angle exceeds helix angle)

If the friction angle equals or drops below the helix angle, the load will overhaul — it will drive the screw backward under its own weight. This is dangerous in lifting applications.

To maintain self-locking:

• Keep helix angles low (use single-start threads when possible)
• Ensure adequate lubrication (paradoxically, too-smooth surfaces can reduce the self-locking margin)
• For high helix angles (multi-start threads), add a brake mechanism

Efficiency: The Uncomfortable Truth

Power screws are not efficient. the practitioner's lifting screw had an efficiency of:

η = (F × L) / (2π × T)
η = (2000 × 0.008) / (2π × 5.99)
η = 16 / 37.64
η = 0.425 = 42.5%

Less than half the input energy actually lifts the load. The rest is lost to friction.

With collar friction added (thrust bearing with mean radius 18.75 mm):

Collar friction torque: T_collar = μ × F × r_m
T_collar = 0.125 × 2000 × 0.01875 = 4.69 Nm

Total torque: T_total = 5.99 + 4.69 = 10.7 Nm

New efficiency: η = (2000 × 0.008) / (2π × 10.7)
η = 16 / 67.23 = 0.238 = 23.9%

Only 24% efficient. This is the trade-off: power screws give you self-locking, precise control, and massive mechanical advantage — at the cost of efficiency.

For you: If efficiency matters more than self-locking (like in a CNC lead screw), use ball screws — recirculating balls replace sliding friction with rolling friction, pushing efficiency above 90%. But you lose self-locking entirely and need a brake system.

Stress Analysis: Where Power Screws Really Get Tested

The stresses in a loaded power screw thread are surprisingly complex. Experiments show that the first one or two threads in contact carry the majority of the load due to deflection effects.

However, the simplified analysis (assuming uniform load distribution) is still useful because:

1. It provides a fundamental understanding of the stress types involved
2. Uncertainty can be handled through conservative safety factors
3. Full FEA analysis is often overkill for standard applications

Key design dimensions to check:

Engineering takeaway

What the practitioner Learned (And What You Should Take Away)

After two years of failures, redesigns, and late nights with engineering handbooks, the practitioner distilled everything into a framework she now teaches every junior engineer who walks through the door:

🔧 The Four-Connection Checklist

Before you finalise ANY mechanical design, verify all four connection types:

The Universal Design Formulas Card

Keep these formulas accessible. They cover 90% of connection design decisions:

Springs:

Spring Rate:     R = F / δ  (force per unit deflection)
Safe Operating:  Never compress past 80% of (Free Length − Solid Height)

Bolts:

Preload Force:   F = S × Applied Load
Tensile Area:    A = F / (0.65 × Yield Stress)
Assembly Torque: Always specify — never leave to the shop

Welds:

Throat Size:     t = 0.707 × leg size
Line Stress:     f = F / L
Combined Stress: f_r = √(f_bending² + f_direct²)
Weld Size:       s = (f_r / f_allowable) / 0.707

Power Screws:

Helix Angle:     tan θ = Lead / (π × pitch diameter)
Raising Torque:  T = F(d/2)tan(φ' + θ)
Efficiency:      η = F·L / (2π·T)
Self-Locking:    Requires φ' > θ

What To Do Next

You now understand the four critical connection methods better than most engineers learn in their first five years. But understanding and applying are different skills.

Here's your next step:

Pick a mechanical system near you — a door closer, a car jack, a bench vise, a bolted shelf bracket. Look at it through the lens of this article:

• What type of springs (if any) does it use? Are they operating near solid height?
• How are the bolts loaded — in tension, shear, or both? Are the nuts self-locking?
• If there are welds, where is the highest stress point? Is it the direct load or the bending/torsion that dominates?
• If there's a screw mechanism, is it self-locking? What's its approximate efficiency?

The engineers who build things that last aren't the ones who memorise formulas. They're the ones who see connections everywhere — and understand what happens when those connections fail.

the practitioner learned that the hard way. You don't have to.

What connection problem are you currently wrestling with? Drop a comment or question below — whether you're a first-year student or a 30-year veteran, the fundamentals apply equally.