Quick Summary
- What this covers: The four connection methods that hold almost every machine together — helical springs, bolted joints, fillet welds, and power screws — and the specific failure modes that destroy each one.
- Why it matters: Most mechanical failures aren't caused by choosing the wrong part. They're caused by misunderstanding how the part carries load. A correct spring in the wrong operating window still shatters. A correct bolt with no specified preload still loosens.
- The key insight: A connection isn't a piece of hardware. It's a managed force path with a failure mode attached. Master the force path, and the hardware almost selects itself.
- Who this is for: Mechanical engineering students, junior design engineers, maintenance and reliability technicians, makers, and anyone who has ever wondered why a "correctly sized" component failed anyway.
Introduction
Maya Chen graduated top of her class — mechanical engineering, honours, an internship at a premier fabrication firm. She walked into her first real project, the frame for a hydraulic press, with the quiet confidence that comes from four years of perfect grades.
Six months later she was standing in front of a shattered bolted bracket, a fractured weld line, and a return spring that had buckled at half its rated load. The press was offline. Production was halted. Her boss didn't raise his voice — which was worse. He looked at the wreckage and said:
"You knew the formulas. But you didn't understand the connections."
That sentence reorganised her entire career. Over the next two years — through mentors, failures, and relentless hands-on study — Maya learned the difference between passing an exam and building something that survives the field. This article distills what she learned: not just the "what," but the "why it fails" and the "how to get it right."
There are four connection methods that hold the mechanical world together. Get any one of them wrong, and your design becomes a ticking clock. We'll take them one at a time, the way Maya was forced to — each through the lens of a real failure and the framework that prevents it.
Core Concepts: What a "Connection" Actually Is
Before the four methods, internalise the idea that unifies them.
Engineers are taught to think of springs, bolts, welds, and screws as components — things you pull from a catalogue and bolt into an assembly. That mental model is exactly why they fail. The professional model is different:
A connection is a force path with a failure mode waiting at the weakest point along it.
Every one of the four methods is just a different strategy for routing force from one part to another:
HOW EACH CONNECTION ROUTES FORCE
Spring → stores and returns force → fails by fatigue / buckling
Bolt → clamps via stored stretch → fails by losing preload
Weld → fuses, carries combined stress → fails by stress concentration
Power Screw → converts torque ↔ thrust → fails by overhauling / seizing
Notice the pattern: in all four, the failure is rarely a simple overload of the headline property. The spring doesn't fail because it can't hold the load — it fails because it was cycled near its solid height. The bolt doesn't fail because it's too weak in shear — it fails because nobody specified its preload. The headline number is the trap. The force path is the truth.
Three questions cut through every connection decision you will ever make:
- What is the real force path? (Tension? Shear? Bending? Torsion? Cyclic? Shock?)
- Where is the weakest point along that path? (It is almost never where the load is applied.)
- What happens at that point over time? (Fatigue, loosening, creep, wear, corrosion.)
Hold those three questions in mind as we work through each method.
Connection 1 — Springs: The Battery That Betrayed Her
The failure
Maya's first mistake was spring selection. The press needed helical compression springs to push the ram back up after each cycle. She did what most juniors do: found the load, found the length, ordered the part. She chose a C0600-049-2000 spring — 0.600 in outside diameter, 0.049 in wire, 2.00 in free length, rated at 8.40 lb at the working length.
It failed in three weeks.
Why it failed (and what the catalogue doesn't tell you)
The catalogue gives you data. It does not give you judgment. Here is what every line on a spring spec sheet actually means — and which one killed Maya's design:
| Parameter | What it means | Why it matters |
|---|---|---|
| Outside Diameter | The coil's outer boundary | Must fit the housing with clearance |
| Wire Diameter | Thickness of the spring wire | Drives fatigue life and load capacity |
| Free Length | Length under zero load | Your datum — everything is measured from here |
| Load at Length | Force at a specific compression | NOT the maximum safe load |
| Solid Height | Fully compressed length, coils touching | Never operate here — this kills springs |
| Spring Rate | Force per unit of deflection | The spring's "personality" — stiff or soft |
Maya's fatal error: she operated the spring too close to its solid height. When the ram bottomed out on heavy cycles, the coils were slamming into each other. Each impact hammered the wire, created stress concentrations, and nucleated the cracks that finished it.
The mentor's framework
Maya's mentor — a retired toolmaker named Haruto — put it best:
"A spring is a battery. It stores energy. And like a battery, if you overcharge it or drain it past its limit, it dies."
The framework he taught her, in four steps:
Step 1 — Define the operating window first. Before you open a catalogue, you must know three numbers: the load you need, the space you have (diameter and length), and the motion range (how far the spring travels in service).
Step 2 — Keep a survival margin above solid height. Never use the last slice of available deflection. If free length is 50 mm and solid height is 30 mm, your available deflection is 20 mm — but design for roughly 16 mm of it. The remaining 4 mm is your survival margin, the buffer that keeps the coils from clashing under overload or dynamic overshoot.
Step 3 — Check the spring rate, not just the load. Two springs can deliver identical force at identical length and behave completely differently. A high spring rate means force changes rapidly with small movement — ideal for a valve, terrible for cushioning.
Step 4 — Match the spring to the duty cycle. A spring that's fine for 10,000 cycles can shatter at 100,000. A continuously running machine needs springs rated for infinite life, which means operating well inside the endurance limit of the wire.
Spring rate itself is simple:
Spring Rate: R = F / δ (force ÷ deflection)
Safe stroke: never compress past ~80% of (Free Length − Solid Height)
The decision logic, end to end:
SPRING SELECTION DECISION FLOW
Required Force (F)
↓
Available Space (OD, Length) → narrows catalogue options
↓
Deflection Range (Free − Working) → check solid-height margin
↓
Duty Cycle → fatigue / infinite-life check
↓
Environment → corrosion, temperature, vibration
↓
Verify Spring Rate (R = F ÷ δ) → confirm force is consistent across stroke
Where engineers get it wrong: The most common spring failures aren't from choosing the wrong spring. They're from running a correct spring in the wrong conditions — too close to solid height, or past its fatigue limit.
Connection 2 — Bolts: The Clamp Everyone Mistakes for a Pin
The failure
Two months later, Maya redesigned the frame. The springs were now perfect. But the bolted bracket holding the hydraulic cylinder failed under repeated loading — and it failed in a way that confuses most juniors: the bolts didn't snap. They loosened. Cycle after cycle the nuts backed off, the bracket shifted, the cylinder misaligned, and the assembly seized.
Why bolts fail (and why it's almost never the bolt)
Here is the single idea that prevents most bolted-joint failures:
A bolt is not a pin. A bolt is a clamp.
When you tighten a bolt, you stretch it. That stretch creates a clamping force called preload. Preload is what holds the joint together — not the bolt's shear strength, not thread friction, not the nut. The preload. If preload is insufficient, the joint slips, loosens, or fatigues under cyclic load, no matter how "strong" the bolt is.
The 5-step bolt framework
Step 1 — Identify the load and its nature. Static, dynamic, or shock loading sets your safety factor:
| Nature of loading | Safety factor |
|---|---|
| Steady stress | 1.5 – 2 |
| Repeated stress, gradually applied | 2 – 3.5 |
| Repeated stress with shock | 4.5 – 6 |
A safety factor of 3 on a 10 kN load means you design for 30 kN.
Step 2 — Calculate required preload.
Total required preload F = Safety Factor (S) × Applied Load (L)
For flexible gasket joints, add 10% to the design pressure load:
Total preload F = 1.1 × Q
Design load W = F + Q
Step 3 — Select bolt grade. Two grades dominate general engineering:
| Property | Grade 4.6 (commercial) | Grade 8.8 (high tensile) |
|---|---|---|
| Tensile strength | 400 MPa min | 830 MPa min |
| Yield stress | 240 MPa | 640 MPa |
| Proof load stress | 225 MPa | 600 MPa |
| Best for | General purpose, low cost | High-load, precision |
| Working preload stress | 0.65 × yield = 156 MPa | 0.65 × yield = 416 MPa |
Use 65% of yield as your working preload stress. Higher risks permanent deformation; lower wastes the bolt's capacity.
Step 4 — Size the tensile area.
Tensile Area A = Design Load (W) ÷ Allowable Stress (f)
Then round up to the next standard bolt size. Worked example — a repeated tensile load of 10 kN on commercial-grade bolts:
Safety factor S = 3 (repeated loading)
Design preload F = 3 × 10 kN = 30 kN
Preload stress f = 0.65 × 240 = 156 MPa
Tensile area A = 30,000 ÷ 156 = 192 mm²
→ Select M18 bolt (tensile stress area ≈ 192 mm²)
→ Assembly torque ≈ 100 N·m (specify per the fastener standard)
Step 5 — Specify the tightening torque and verify it. This is the step 90% of juniors skip. Every size-and-grade combination has a recommended assembly torque. Leave it unspecified and the shop will either under-tighten (joint loosens) or over-tighten (bolt yields, or snaps on assembly).
Shear vs. tension: two different worlds
Load path determines bolt mode — and which area you design against:
| Load type | How the bolt resists | Design against |
|---|---|---|
| Tension | Stretches along its axis | Tensile stress area of the thread |
| Shear | Resists sideways sliding | Shank cross-sectional area (not thread) |
| Combined | Both at once | Vector-sum the stresses |
For shear joints, load passes through bearing on the bolt shank, not the threads:
Shear Area = π × d² ÷ 4 (d = bolt shank diameter)
A critical and counter-intuitive point: even shear joints should be preloaded. Preload generates friction between the clamped faces. Ignore that friction in the calculation (the conservative approach) and it still buys you a safety margin in service.
Six rules to stop bolt fatigue
Haruto kept these pinned to his workshop wall:
- Tighten properly. The bolt must be stretched so preload always exceeds the working load.
- Maximise elastic length. Use at least one bolt diameter of thread length under the nut — a longer, more elastic bolt rides cyclic loads better.
- Prefer high-strength or small bolts. They have more elasticity relative to their size.
- Put the shank in the hole. Shank bolts in clearance holes (2–3 mm max clearance) beat threaded shanks in bearing.
- Roll, don't cut, the threads. Rolling induces compressive surface stresses that resist fatigue cracking.
- Lock under vibration. Use locknuts or Nyloc nuts; avoid non-axial loading and any prising (bending) action on the bolt.
Where engineers get it wrong: They size the bolt for strength and stop there. The number that actually keeps the joint alive — preload, expressed as a tightening torque — never makes it onto the drawing.
Connection 3 — Welds: The Combined Stress That Hides in Plain Sight
The failure
After the spring and bolt incidents, Maya was handed a "simpler" job: a welded bracket to support a conveyor drive motor. The bracket saw a direct downward load and a bending moment from the motor's offset weight. She sized the weld the easy way — load ÷ weld area, check against allowable stress.
It cracked in service after four months. The problem wasn't the weld size. It was the stress distribution.
Two methods, two mindsets
There are two fundamentally different ways to design a fillet weld, and the gap between them separates competent engineers from dangerous ones.
Method 1 — Conventional ("Area") method. Treat the weld as a real cross-section.
Weld Area A = t × L
t = throat thickness = 0.707 × leg size (s)
L = total weld length
Stress f = F ÷ (t × L)
For a standard equal-leg fillet, the throat is always 0.707 × the leg (because cos 45° = 0.707). So a 6 mm weld has a ~4.2 mm throat. This works fine for simple, single-direction loads.
Method 2 — Weld-as-a-Line ("Line Stress") method. This is the professional approach for anything involving bending or torsion. Treat the weld as a line with no thickness, and express stress as force per unit length (N/mm):
Line Stress f = F ÷ L (units: N/mm)
Convert to real stress: f_s = f ÷ t
Find required weld size: t = f ÷ f_s , then s = t ÷ 0.707
Why bother? Because under bending or torsion, the stress distribution depends on the geometry of the weld pattern, not just its size. The line method lets you calculate the pattern's section modulus or polar moment first, then size the weld afterward.
Bending in welds
A load applied away from the weld group's centroid creates a bending moment:
f_b = M ÷ Z
M = bending moment = Force × distance to centroid
Z = section modulus of the weld group (treated as a line)
Section modulus depends on the pattern. The three you can rely on without ambiguity:
| Weld pattern (bending about horizontal axis) | Section modulus Z |
|---|---|
| Single vertical weld, length d | Z = d² ÷ 6 |
| Two horizontal welds (top + bottom), length b, spacing d | Z = b × d |
| Rectangle (b wide × d tall), welded all round | Z = b × d + d² ÷ 3 |
For L-shaped or single-sided patterns, you must locate the centroid first before computing Z — never apply a rectangular-pattern formula to an L-shape.
Torsion in welds: the invisible killer
A load offset from the weld group's shear centre creates a torsional moment — the failure mode that caught Maya's first bracket:
f_t = T × r ÷ J
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)
| Weld pattern | Polar moment J |
|---|---|
| Single straight weld, length d | J = d³ ÷ 12 |
| Two parallel welds, length d, spacing b | J = d(3b² + d²) ÷ 6 |
| Rectangle (b × d) | J = (2b + d)³ ÷ 12 − b²(b + d)² ÷ (2b + d) |
Combining stresses: the vector rule
This is what actually killed the bracket. When a weld sees both direct stress and bending (or torsion) stress, you combine them vectorially — not by simple addition.
Perpendicular components: f_r = √(f_b² + f²)
Same direction (worst case): f_r = f_b + f
The critical point is wherever the components combine to the maximum resultant — usually the extreme fibre, the point farthest from the centroid.
Allowable weld stress
Allowable shear stress ≈ 0.3 × UTS of the weld rod
E41xx rod (UTS 410 MPa) → ≈ 123 MPa
E48xx rod (UTS 480 MPa) → ≈ 144 MPa
For dynamic or cyclic loading, apply a further safety factor on top.
Worked example: the bracket, done right
Maya's redesigned motor bracket is welded to the machine frame with a fillet weld around a rectangular footprint, b = 120 mm wide × d = 100 mm tall. The motor hangs from it, applying a downward load P = 30 kN at a horizontal eccentricity e = 250 mm from the weld centroid.
Step 1 — Direct (primary) shear line stress. Treat the weld as a line; total length L = 2(b + d).
L = 2 × (120 + 100) = 440 mm
f_direct = P ÷ L = 30,000 ÷ 440 = 68.2 N/mm (vertical, with the load)
Step 2 — Bending moment.
M = P × e = 30,000 × 250 = 7,500,000 N·mm
Step 3 — Section modulus of the rectangular weld group.
Z = b × d + d² ÷ 3
Z = (120 × 100) + (100² ÷ 3)
Z = 12,000 + 3,333 = 15,333 mm²
Step 4 — Bending (secondary) line stress at the extreme fibre.
f_bending = M ÷ Z = 7,500,000 ÷ 15,333 = 489.1 N/mm (perpendicular to the direct shear)
Step 5 — Combine vectorially at the critical point. Direct shear and bending act at right angles, so:
f_r = √(f_direct² + f_bending²)
f_r = √(68.2² + 489.1²)
f_r = √(4,651 + 239,219) = √243,870 = 493.8 N/mm
Step 6 — Size the weld. Using E41xx rod, static allowable f_s = 0.3 × 410 = 123 N/mm²:
Throat t = f_r ÷ f_s = 493.8 ÷ 123 = 4.01 mm
Leg s = t ÷ 0.707 = 4.01 ÷ 0.707 = 5.67 mm
→ Use a 6 mm fillet weld
This time, the bracket held — because the design accounted for the bending stress that dwarfed the direct load (489 N/mm vs. 68 N/mm). The conventional "area method" would have missed it entirely.
Where engineers get it wrong: They size welds for the direct load and ignore bending and torsion. In offset and bracket joints, those secondary stresses are usually the dominant ones — and they concentrate at the weld toe, exactly where cracks start.
Connection 4 — Power Screws: Where Rotation Becomes Muscle
The transformation
By her second year, Maya was the engineer other juniors came to. The project that cemented her reputation was the lifting mechanism for a custom hydraulic platform — driven by a power screw. Turn a handle, raise a load. Simple on the surface; the physics underneath is where the design lives.
What a power screw does
A power screw converts rotational torque into linear force (or the reverse). It's the same principle in every car jack, vice, CNC lead screw, and valve stem. Three arrangements exist:
- Screw rotates, nut translates — most jacks and presses.
- Nut rotates, screw translates — some linear actuators.
- Screw rotates and translates — rare, used with a fixed nut.
Thread form determines everything
The thread shape isn't cosmetic — it controls friction, strength, and whether the screw can hold a load with no power applied:
| Thread form | Face angle | Best for | Machinability |
|---|---|---|---|
| Square | 0° | Maximum efficiency, low friction | Lathe only — can't mill or grind |
| Modified Square (Acme) | 5° | General power transmission | All methods — widely preferred |
| Trapezoidal Metric | 15° (30° included) | Metric equivalent of Acme | Standard in metric regions |
| Buttress | 45° (one side) | Uni-directional heavy loads | Special applications |
Pure square threads are theoretically optimal but impractical to manufacture; in practice the 5° (Acme) or 15° (trapezoidal) form is standard.
The vocabulary
| Term | Symbol | Definition | Formula |
|---|---|---|---|
| Pitch | p | Distance between adjacent threads | — |
| Lead | L | Distance advanced per revolution | L = p (single-start) |
| Nominal diameter | D | Outside diameter of thread | — |
| Root diameter | dᵢ | Inside diameter (thread bottom) | dᵢ = D − p |
| Pitch diameter | d | Mean diameter (thread midpoint) | d = D − 0.5p |
| Helix angle | θ | Thread's slope angle | tan θ = L ÷ (π × d) |
| Friction angle | φ | Angle where a block just slides | tan φ = μ |
| Thread depth | t | Radial depth of thread | t = 0.5p |
Recommended pitch-diameter proportions (no formal standard — these are engineering conventions):
| Nominal diameter D (mm) | Recommended pitch p (mm) |
|---|---|
| 10, 12 | 3 |
| 15 | 4 |
| 20 | 5 |
| 25 | 6 |
| 30, 35 | 8 |
| 40, 45 | 10 |
| 50, 55 | 12 |
| 60, 65 | 13 |
| 70, 75 | 14 |
| 80, 85 | 15 |
| 90, 95 | 16 |
| 100 | 17 |
The torque equations
The heart of power-screw design. For a square or modified-square thread:
Raising load: T = F × (d ÷ 2) × tan(φ' + θ)
Lowering load: T = F × (d ÷ 2) × tan(φ' − θ)
F = axial load (the weight being moved)
d = pitch diameter
θ = helix angle
φ' = effective friction angle = arctan(μ ÷ cos α)
α = thread face angle (0° square, 15° trapezoidal)
μ = coefficient of friction (0.1–0.15; use 0.125 as an average)
The friction reality
Friction between lubricated metal threads depends on lubrication, surface finish, and how "run-in" the surfaces are:
| Condition | μ (approx.) |
|---|---|
| Well-lubricated, run-in, precision machined | 0.10 |
| Average conditions | 0.125 |
| Poor lubrication, new or rough surfaces | 0.15 |
| Start-up (static) friction | multiply operating μ by 4/3 |
Worked example: Maya's lifting platform
Problem: A trapezoidal single-start metric thread, 30 mm diameter, lifts an axial load of 2 kN. Find the torque to raise and lower the load, at average friction.
From the pitch table: p = 8 mm → Lead L = 8 mm (single-start)
Pitch diameter: d = D − 0.5p = 30 − 4 = 26 mm
Coefficient of friction: μ = 0.125
Face angle: 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 ÷ (π × d) = 8 ÷ (π × 26) = 0.0979
θ = 5.594°
RAISING the load:
T = F × (d ÷ 2) × tan(φ' + θ)
T = 2000 × 13 × tan(12.968°)
T = 2000 × 13 × 0.2305 = 5,993 N·mm = 5.99 N·m
LOWERING the load:
T = F × (d ÷ 2) × tan(φ' − θ)
T = 2000 × 13 × tan(1.780°)
T = 2000 × 13 × 0.0311 = 808 N·mm = 0.808 N·m
Key insight: It takes 5.99 N·m to raise the load but only 0.808 N·m to lower it. The screw does most of the holding through friction.
Self-locking: the built-in safety feature
The lowering torque is positive — you still need to apply torque to lower the load. So if you release the handle, the load stays put. This is self-locking, one of the most valuable properties of power screws.
A screw is self-locking when φ' > θ — the friction angle exceeds the helix angle.
If the friction angle drops to or below the helix angle, the load will overhaul — drive the screw backward under its own weight. In a lifting application, that's a falling load.
To preserve self-locking: keep helix angles low (favour single-start threads), maintain adequate lubrication (paradoxically, too smooth a surface can erode the self-locking margin), and for high-helix multi-start threads, add a brake mechanism.
Efficiency: the uncomfortable truth
Power screws are not efficient. Maya's lifting screw, thread friction only:
η = (F × L) ÷ (2π × T)
η = (2000 × 0.008) ÷ (2π × 5.99)
η = 16 ÷ 37.64 = 0.425 = 42.5%
Less than half the input energy actually lifts the load. Now add collar friction (a thrust bearing, mean radius 18.75 mm):
Collar torque: T_collar = μ × F × r_m = 0.125 × 2000 × 0.01875 = 4.69 N·m
Total torque: T_total = 5.99 + 4.69 = 10.7 N·m
New efficiency: η = 16 ÷ (2π × 10.7) = 16 ÷ 67.23 = 0.238 = 23.9%
| Configuration | Torque required | Efficiency |
|---|---|---|
| Thread friction only | 5.99 N·m | 42.5% |
| Thread + collar friction | 10.7 N·m | 23.9% |
Only ~24% efficient. That's the trade-off: power screws give you self-locking, precise control, and enormous mechanical advantage — at the cost of efficiency. When efficiency matters more than self-locking (a CNC lead screw, for example), switch to a ball screw: recirculating balls replace sliding friction with rolling friction, pushing efficiency above 90%. But you lose self-locking entirely and must add a brake.
Stress in the threads
The stress distribution in a loaded screw is genuinely complex — experiments show the first one or two engaged threads carry most of the load due to deflection. The simplified, uniform-load analysis is still useful because it builds intuition, the uncertainty is absorbed by conservative safety factors, and full FEA is overkill for standard applications. Key dimensions to check:
| Dimension | Typical range | Purpose |
|---|---|---|
| Nut thickness (a) | 0.75D to 1.5D | Adequate thread engagement |
| Thread depth (t) | 0.5p standard, 0.75p buttress | Load-bearing surface |
| Engagement length (b) | a multiple of pitch | Distributes load across threads |
Where engineers get it wrong: They design the screw to lift the load and forget to check self-locking and collar friction. A screw that lifts beautifully can drop its load the instant the handle is released — and the "spare" torque capacity they thought they had is eaten alive by the collar.
The Cross-Cutting Comparison
Before you finalise any mechanical design, run the four-connection checklist:
| Connection | The critical question | Failure mode if ignored |
|---|---|---|
| Springs | Am I operating within the safe deflection range, away from solid height? | Fatigue cracking, buckling, premature failure |
| Bolts | Have I specified preload, torque, and grade — not just bolt size? | Loosening, fatigue, joint separation |
| Welds | Have I accounted for bending and torsion, not just direct load? | Crack initiation at the weld toe, fracture |
| Power Screws | Is the screw self-locking? Have I accounted for collar friction? | Overhauling load drop, seized mechanism |
The universal formulas card
Keep these accessible — they cover roughly 90% of connection-design decisions.
SPRINGS
Spring rate: R = F ÷ δ
Safe stroke: 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 it 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
The Most Common Mistakes (And How to Prevent Them)
Across thousands of failed joints, the same handful of mistakes recur. Each has the same root cause — designing to the headline number instead of the force path — and each is preventable.
1. Operating a spring near its solid height. Why it happens: the catalogue load looks fine, so the deflection margin gets ignored. Consequence: coil clash, stress spikes, fatigue cracks. Prevention: design to ~80% of available deflection and reserve a survival margin.
2. Treating a bolt as a pin instead of a clamp. Why it happens: "shear strength" is the number students memorise. Consequence: under-preloaded joints loosen and fatigue. Prevention: design the preload, then convert it to a specified tightening torque on the drawing.
3. Leaving assembly torque unspecified. Why it happens: it feels like a shop-floor detail. Consequence: under- or over-tightening — loosening or bolt yield. Prevention: every fastener on the drawing gets a torque value.
4. Sizing a weld for direct load only. Why it happens: the area method is fast and feels complete. Consequence: bending/torsion stresses (often the larger ones) are missed, and cracks start at the toe. Prevention: use the weld-as-a-line method whenever the load is offset.
5. Adding weld stresses instead of combining them vectorially. Why it happens: arithmetic is easier than vectors. Consequence: the true resultant at the critical point is underestimated. Prevention: f_r = √(f_b² + f²) for perpendicular components; only add directly when they're collinear.
6. Forgetting to check power-screw self-locking. Why it happens: the raising torque "works," so the design feels done. Consequence: the load overhauls and drops when power is removed. Prevention: verify φ' > θ, and add a brake if you're running multi-start threads.
7. Ignoring collar/thrust friction in screws. Why it happens: the thread torque is the "main" calculation. Consequence: real torque demand can double; the drive is undersized. Prevention: always add the collar torque term before sizing the motor or handle.
Expert Insights
A few heuristics that separate engineers who pass calculations from engineers who build things that last:
- Design the failure mode, not the part. Decide how the joint is allowed to fail (and how it must never fail) before you pick a size. The size then follows from the force path.
- Simplified models are a feature, not a flaw. Hand calculations with honest safety factors beat false-precision FEA on a standard part. Reserve FEA for the joints where the load path is genuinely unclear or the consequence of failure is severe.
- Preload is a system property, not a bolt property. The same bolt at the same torque gives different preload depending on lubrication, surface finish, and joint stiffness. For critical joints, control the clamp force (torque-angle, or direct tension measurement), not just the torque.
- The critical point is rarely where the load is applied. It's at the extreme fibre of a weld group, the most-engaged screw thread, or the coil nearest solid height. Train yourself to look there first.
- Self-locking and efficiency are a trade, not a free lunch. Want a screw that holds with no power? Accept low efficiency. Want >90% efficiency? Accept that you need a brake. There is no thread form that gives you both.
- Match the connection to the duty cycle, not the peak load. A joint that survives one overload can still die from a million ordinary cycles. Fatigue, not strength, is what kills most well-sized connections.
Frequently Asked Questions
What are the four main types of mechanical connections?
Helical springs (which store and return force), bolted joints (which clamp parts together through stored stretch, or preload), fillet welds (which fuse parts and must carry combined direct, bending, and torsional stress), and power screws (which convert rotary torque into linear thrust). Each routes force differently and fails differently.
Why do bolted joints loosen over time?
Almost always because of insufficient preload. A bolt holds a joint by being stretched, creating a clamping force. Under cyclic or vibrating loads, an under-preloaded bolt allows micro-movement between the clamped faces, the nut backs off, and the joint loses clamp force entirely. The fix is adequate, specified preload — plus locking features (Nyloc or locknuts) under vibration.
What is spring "solid height," and why does it matter?
Solid height is the length of a compression spring when all its coils are touching — fully compressed. Operating a spring at or near solid height makes the coils clash on each cycle, hammering the wire and creating stress concentrations that nucleate fatigue cracks. As a rule, never use more than about 80% of the available deflection between free length and solid height.
What does "weld throat thickness" mean?
The throat is the shortest distance through a fillet weld's cross-section — the plane on which it actually shears. For a standard equal-leg fillet, throat = 0.707 × leg size (because cos 45° = 0.707). So a 6 mm fillet weld has a throat of roughly 4.2 mm, and that throat — not the visible leg — carries the load.
What makes a power screw self-locking?
A power screw is self-locking when its effective friction angle exceeds its helix angle (φ' > θ). Practically, that means it holds its load without any applied torque — release the handle and nothing moves. If the helix angle grows larger than the friction angle (common with steep, multi-start threads), the load "overhauls" and drives the screw backward, which is dangerous in lifting applications and requires a brake.
Is a higher safety factor always better?
No. A safety factor matched to the loading (around 1.5–2 for steady loads, 2–3.5 for repeated, 4.5–6 for shock) protects against uncertainty. An arbitrarily large factor wastes material, adds weight and cost, and can mask the real problem — a misunderstood force path. Pick the factor that fits the load's nature, then design honestly.
What's the difference between a power screw and a ball screw?
A conventional power screw uses sliding friction between threads — giving self-locking and low cost, but efficiency typically below 50%. A ball screw replaces that sliding contact with recirculating balls (rolling friction), pushing efficiency above 90%. The trade-off: ball screws are not self-locking, so they need a brake to hold a load. Use power screws for clamping and holding; ball screws for fast, efficient positioning.
How do I choose between Grade 4.6 and Grade 8.8 bolts?
Grade 4.6 (≈400 MPa tensile, 240 MPa yield) is the low-cost, general-purpose choice for lightly loaded, non-critical joints. Grade 8.8 (≈830 MPa tensile, 640 MPa yield) is for high-load, fatigue-prone, or precision joints where you need higher preload capacity. Both should be preloaded to about 65% of their yield stress.
Why combine weld stresses vectorially instead of just adding them?
Because direct stress and bending or torsional stress usually act in different directions. Adding their magnitudes assumes they point the same way, which overstates the result in some places and understates it at the true critical point. When the components are perpendicular, the correct resultant is f_r = √(f_b² + f²); you only add directly when the components genuinely act along the same line.
What coefficient of friction should I use for a power screw?
For lubricated metal-on-metal threads, use about 0.125 as a working average. Drop toward 0.10 for precision-machined, well-lubricated, run-in surfaces, and up toward 0.15 for rough or poorly lubricated ones. For start-up (static) friction, multiply the operating value by roughly 4/3.
Final Takeaways
- A connection is a force path with a failure mode, not a part on a shelf. Master the path and the part nearly selects itself.
- The headline number is the trap. Springs fail near solid height, not at their rated load. Bolts fail from lost preload, not weak shear. Welds fail from missed bending, not insufficient area. Screws fail from overhauling, not inadequate lift.
- Specify what actually keeps the joint alive. Deflection margin for springs, preload and torque for bolts, combined stress for welds, self-locking for screws.
- Design for the duty cycle, not the single peak. Fatigue kills more well-sized connections than overload ever will.
- Use simplified models with honest safety factors. They beat false precision, and they keep you focused on the force path.
What to do next
You now understand these four connection methods better than most engineers do in their first five years — but understanding and applying are different skills. So apply it: pick a mechanical system near you — a door closer, a car jack, a bench vice, a bolted shelf bracket — and read it through this lens.
- What springs does it use, and are any operating near solid height?
- How are the bolts loaded — tension, shear, or both — and are the nuts locked?
- If there are welds, where is the highest stress, and is it the direct load or the bending/torsion that dominates?
- If there's a screw mechanism, is it self-locking, and roughly how efficient is it?
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.
Maya learned that the hard way. You don't have to.
What connection problem are you wrestling with right now? Whether you're a first-year student or a thirty-year veteran, the fundamentals apply equally — drop your question below.
Further Reading & References
Suggested internal links (point these at your own related posts):
- Anchor: "how fatigue failure actually works in metals" →
/fatigue-failure-explained - Anchor: "reading an engineering drawing without missing the critical callouts" →
/engineering-drawing-basics - Anchor: "selecting the right bearing for the load" →
/bearing-selection-guide - Anchor: "material selection for machine elements" →
/material-selection-machine-design - Anchor: "when to trust a hand calculation over FEA" →
/hand-calculations-vs-fea - Anchor: "torque wrenches and how to actually hit a preload target" →
/torque-wrench-guide
Suggested external references (authoritative, for citations and deeper study):
- Shigley's Mechanical Engineering Design — Budynas & Nisbett (McGraw-Hill)
- Machine Design — Robert L. Norton (Pearson)
- Roark's Formulas for Stress and Strain — Young, Budynas & Sadegh (McGraw-Hill)
- Design of Welded Structures — Omer W. Blodgett (Lincoln Electric)
- ISO 898-1 — Mechanical properties of fasteners (bolt property classes such as 4.6 and 8.8)
- ISO 2901–2904 / DIN 103 — ISO metric trapezoidal screw threads
- AWS D1.1 — Structural Welding Code, Steel
- Spring Manufacturers Institute (SMI) — Handbook of Spring Design
A note on scope: The models above are simplified first-pass design tools — excellent for sizing and intuition. For safety-critical, regulated, or high-consequence applications, verify against the governing standards and, where appropriate, detailed analysis or testing.
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
SEO & Publishing Metadata
(For Ghost post settings — copy into the relevant fields, then delete this block before publishing.) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
SEO Title: Mechanical Connections: Springs, Bolts, Welds & Screws
Meta Description: An engineer's guide to the four mechanical connections — spring selection, bolt preload, fillet weld design, and power-screw torque — with worked examples.
URL Slug: mechanical-connections-springs-bolts-welds-screws
Focus Keyword: mechanical connections
Related Keywords: spring selection guide · bolted joint design · bolt preload · fillet weld design · weld-as-a-line method · power screw torque · self-locking screw · fastener safety factor · machine design fundamentals · throat thickness
Semantic Keywords: solid height · spring rate · clamping force · tensile stress area · Grade 4.6 / Grade 8.8 bolts · section modulus of weld · polar moment of weld · helix angle · friction angle · effective friction angle · Acme thread · trapezoidal thread · buttress thread · ball screw · lead screw efficiency · collar friction · fatigue failure · overhauling load · weld toe
Suggested Featured-Snippet Target: "What are the four main types of mechanical connections?" (answered in the FAQ in a single concise paragraph).
Suggested FAQ Schema: Mark up the ten FAQ questions above as FAQPage structured data to compete for rich results.