Every plate, weld and bolt in the hull is chosen against two questions: will it stay tough enough not to crack, and will it stay intact long enough not to waste away. This chapter works through both, then turns weld and bolt sizing into arithmetic you can check under exam pressure.
Two plates can carry the identical yield strength on the mill certificate and behave completely differently in cold water. That is the point of the letter after a shipbuilding steel's grade. Grades A, B, D and E are the normal-strength series; AH, DH and EH are the matching higher-strength series. The letter has nothing to do with the tensile or yield number — it records the lowest temperature at which the plate still absorbs a specified amount of energy in a Charpy V-notch impact test, rather than snapping in a brittle, low-energy fracture. A grade D plate is simply a plate proven tough down to a colder test temperature than a grade A plate of the same strength class.
The reason this matters is that a ship's steel does not fail the way a tensile test coupon fails. A crack that starts small — at a weld defect, a notch, a hatch corner — can run at almost the speed of sound through steel that has gone brittle, well below the load that would ever yield the plate in a slow pull test. Thicker sections, colder service temperatures, and structure exposed to dynamic or impact loading (bow flare, ice belts, areas near stress concentrations) are the places where the specification calls for a tougher grade, even when the strength requirement alone would allow a cheaper one.
The grade letter is a temperature rating for toughness, not a strength class. Picking a grade on yield strength alone answers the wrong question — the plate can still shatter rather than yield.
Marine aluminium is almost always a 5000-series alloy — aluminium alloyed mainly with magnesium, chosen because it is not heat-treatable and so keeps its properties through welding without a post-weld heat treatment, and because its own oxide layer gives it good corrosion resistance in seawater on its own. Its strength and stiffness are both roughly a third of steel's, so an aluminium structure needs more section to do the same job, and it loses a much larger fraction of its strength at a much lower temperature in a fire — a real driver of how superstructure fire zones and escape routes get arranged on an aluminium-topped hull.
Composites (glass or carbon fibre in a resin matrix) sidestep galvanic corrosion entirely — there is no metal to corrode — but they bring their own design questions: strength and stiffness are directional, following the fibre layup, rather than the same in every direction the way rolled steel plate is; fatigue and impact damage can be invisible from the surface; and repair generally means re-laying material rather than welding a patch.
Put an aluminium superstructure directly on a steel hull, wet with seawater, and the two metals plus the electrolyte plus the metallic path through the structure is exactly a battery. Aluminium sits well below steel in the practical galvanic ordering of common shipbuilding metals, so it becomes the anode and corrodes to protect the steel. Worse, the area ratio works against it: a relatively small aluminium contact area wired to a large steel cathode concentrates the corrosion current into that small area, and the aluminium can waste away at a rate far faster than either metal would corrode alone. The standard fix is an insulating joint — a non-conducting gasket, coating or strip between the two metals — that breaks the metallic path the cell needs to work.
A small anode wired to a large cathode is the worst-case galvanic geometry — the fix is to break the electrical connection, not just to paint one of the two metals.
Ordinary steel corrosion in seawater is an electrochemical cell running on the hull plate itself: iron goes into solution at anodic sites, oxygen is reduced at cathodic sites, and the seawater carries the ionic current between them while the steel itself carries the electron flow back. Because the reaction needs dissolved oxygen at the cathode, corrosion rate tracks oxygen availability as much as it tracks anything else — which is why a well-aerated splash zone or a flowing ballast line often corrodes faster than a stagnant, oxygen-depleted tank bottom, even though the stagnant location looks like the more hostile environment.
Uniform corrosion, thinning the plate more or less evenly, is the easy case to allow for with a corrosion margin. The harder cases are localised: pitting, where a small breakdown in a protective film concentrates attack into a deep, narrow hole out of proportion to the metal lost; crevice corrosion, in the stagnant, oxygen-starved gap under a gasket, a doubler plate or a bolted lap joint, where the local chemistry inside the crevice drifts more acidic and self-accelerates; and erosion-corrosion, where high-velocity flow or cavitation — typical near a propeller, a pump impeller or a valve throttling point — physically strips the protective film as fast as it can re-form.
Prevention leans on three layers together, not any one of them alone. A coating is the primary barrier and does most of the work while it is intact. Cathodic protection — sacrificial anodes or an impressed current system — supplements the coating by protecting whatever bare steel shows through a holiday or a scratch; it is not sized to protect large areas of bare, uncoated plate economically. And design detail matters as much as either: avoiding faying surfaces that trap water, fitting drain holes so structure does not hold a standing puddle, and keeping dissimilar metals apart or insulated all remove the conditions corrosion needs before it can start.
Coating and cathodic protection are not alternatives — CP only protects where the coating has already failed, so a system relying on CP to protect bare steel over a large area is not really designed at all.
A butt weld, properly fused through the full thickness, is generally taken as being at least as strong as the plate either side of it, so butt joints are rarely the calculation that decides a connection. Fillet welds are different: the weld metal sits as a roughly triangular bead across the corner between two plates rather than replacing plate thickness, and its strength has to be established from its own geometry rather than assumed equal to the parent plate.
The dimension that carries the load is the throat, not the leg. For an equal-leg fillet the throat — the shortest distance from the weld root to the face, which is the plane a shear failure actually runs along — works out to about 0.7 times the leg length. Because the weld's cross-sectional area (and so its metal volume, consumable use and arc time) scales with the leg squared while its strength scales roughly with the leg, doubling the leg to gain margin costs four times the metal for twice the capacity — oversizing is a real, and often unnecessary, cost.
Heat input governs the thermal cycle the plate sees, and through that, the mechanical properties of the heat-affected zone. Too little heat input, on a steel with a high carbon equivalent, cools too fast and can leave a hard, crack-susceptible microstructure right where residual stress and any hydrogen picked up during welding are both at their worst. Too much heat input coarsens the grain and can drag toughness down even on an easy-to-weld steel. Preheat works alongside heat input rather than instead of it — raising the plate's starting temperature slows the cooling rate without changing the arc energy put into the joint, which is exactly the lever needed when the heat input is already sitting where the procedure wants it.
A bolted connection can fail four distinct ways, and a proper check runs all four rather than stopping at the first comfortable pass. Bolt shear is the bolt shank parting across the plane between the plates. Bearing is the plate hole ovalling or crushing ahead of the bolt shank. Net-section tension is the plate yielding or fracturing across whatever reduced cross-section is left once every hole on the critical line is deducted. And tear-out is a wedge of plate shearing free between a bolt hole and the nearest free edge. Each is an independent calculation with its own allowable, and the smallest of the four — not the first one checked — is what actually governs the joint.
It is easy to treat bolt shear as the whole check, because it is usually the first one done and the one most engineers are fastest with. But net section is driven by geometry — hole size relative to plate width, and how many holes land on the same transverse line — and it can easily be the weaker mode even in a joint with generously rated bolts. Staggering a bolt row so that not every hole sits on one cross-section, using the smallest hole clearance the assembly tolerance allows, or simply widening the plate are the usual fixes once net section is found to govern.
Once the joint (or the weld) is made, non-destructive testing is how its quality gets confirmed without cutting it open. Visual inspection is always the first pass and catches surface issues an instrument would otherwise be sent to find. Dye penetrant and magnetic particle testing both find defects that break the surface — penetrant by capillary action into a crack too fine to see unaided, magnetic particle by the leakage field a surface-breaking flaw creates in a magnetised part — but neither sees anything below the surface. Ultrasonic testing sends sound into the material and reads the echo from an internal discontinuity, which makes it good at finding planar flaws like lack of fusion, particularly when the beam angle is chosen to catch the flaw's likely orientation. Radiography images the volume the way an X-ray images a bone, which makes it well suited to rounded, volumetric defects like porosity or slag, but weaker at detecting a tight planar crack lying across the beam at an unhelpful angle.
Every bolted joint has four failure modes and the smallest capacity governs; every NDT method is good at some defect shapes and blind to others, so the method has to match the defect you actually expect, not just “internal versus external”.
Three fully stepped problems: sizing a fillet weld on the throat and pricing the cost of oversizing it, running all four bolted-joint checks to find which one actually governs, and combining a carbon-equivalent calculation with a heat-input check on the same welding procedure.
A steel bracket transmits a factored load of 168 kN into a bulkhead through two identical fillet welds, one on each side of the bracket, each 100 mm long. The yard's first attempt uses an 8 mm leg. The weld metal's allowable shear stress on the throat is 110 N/mm². Check the 8 mm weld, re-size it if it fails, and state how much more weld metal the fix costs.
F = 168 kN (168 000 N) Two fillet welds, each L = 100 mm long Initial leg size z₁ = 8 mm Allowable weld shear stress on the throat, τ_allow = 110 N/mm²
Check the 8 mm weld, re-size it if it fails, and state how much more weld metal the fix costs
Find the throat, not the leg.
An equal-leg fillet weld fails on a diagonal plane through the throat — the shortest distance across the weld cross-section — which for an equal-leg fillet works out to about 0.7 times the leg length.
Total effective throat area is the throat thickness times the combined weld length on both sides of the bracket, and the applied stress on that area is what has to clear the allowable.
150 N/mm² exceeds the 110 N/mm² allowable.
The 8 mm weld is undersized. Step the leg up to 12 mm and re-check before the joint is accepted.
That fix is not free.
Weld metal cross-section — and with it consumable, arc time and shrinkage — scales with the leg squared, not linearly with the leg.
Answer12 mm leg required (throat stress 100 N/mm² against a 110 N/mm² allowable); the resize costs roughly 125% more weld metal than the 8 mm attempt did.
The trap: using the 8 mm leg itself as the throat gives 168 000/(8×100×2) = 105 N/mm², a false pass against the 110 N/mm² allowable — the true throat stress is 150 N/mm², understated by 30%, on a weld that is actually overstressed.
A bracket is bolted to a bulkhead with four M20 bolts in a single row set perpendicular to the load, transmitting a tensile load of 240 kN. Plate thickness is 12 mm, gross plate width 200 mm, and each bolt hole (with clearance) is 22 mm diameter. The bolt supplier rates each bolt's allowable single-shear capacity at 70 kN. The plate's allowable bearing stress is 400 N/mm², its allowable net-section tensile stress is 150 N/mm², and its allowable tear-out (edge shear) stress is 170 N/mm² over an edge distance of 40 mm from each bolt centre to the plate edge. Check all four failure modes and state which one governs.
Applied load P = 240 kN 4 bolts, M20, single shear, rated 70 kN/bolt Plate thickness t = 12 mm, gross width w = 200 mm Hole diameter (with clearance) d_h = 22 mm, all 4 holes on the critical section Allowable bearing stress σ_br = 400 N/mm² Allowable net-section tensile stress σ_t = 150 N/mm² Allowable tear-out shear stress τ_to = 170 N/mm², edge distance e = 40 mm
Check all four failure modes and state which one governs
Bolt shear first.
The rated capacity already accounts for the bolt's shear area, so the group capacity is a straight multiplication by the bolt count.
Bearing checks whether the plate crushes and ovals ahead of the hole.
The bearing area per bolt is the bolt diameter times the plate thickness.
Net section is the tension check on whatever plate cross-section actually remains once every hole on that section is deducted — all four holes sit on the same line here, so all four come out together.
Tear-out checks a wedge of plate shearing away from each bolt toward the free edge.
On two shear planes per bolt.
Compare the four.
| Mode | Capacity (kN) | vs 240 kN demand |
|---|---|---|
| Bolt shear | 280 | OK |
| Bearing | 384 | OK |
| Net section | 201.6 | FAILS |
| Tear-out | 652.8 | OK |
Shear, bearing and tear-out all clear 240 kN comfortably — net section does not.
AnswerNet section governs at 201.6 kN against a 240 kN demand — the joint as drawn is undersized even though bolt shear and bearing both pass. The fix is a wider plate, smaller holes, or a staggered row so no single section loses all four holes at once.
The trap: stopping after bolt shear (280 kN, a comfortable pass) and bearing (384 kN, also comfortable) and signing the joint off — net section is the actual governing mode here, at barely 84% of the applied load, and skipping it passes a plate that will yield straight through the bolt line.
A 25 mm plate with a measured composition of C 0.18%, Mn 1.20%, Cr 0.30%, Mo 0.05%, Ni 0.40% and Cu 0.20% is to be welded by SMAW. The proposed procedure runs at 24 V and 180 A, travel speed 240 mm/min, with an arc efficiency of 0.8. Find the carbon equivalent, decide whether preheat is warranted, and check whether the proposed heat input sits inside the procedure's target band of 0.7–1.2 kJ/mm.
Composition: C 0.18%, Mn 1.20%, Cr 0.30%, Mo 0.05%, Ni 0.40%, Cu 0.20% Arc voltage V = 24 V, current I = 180 A Travel speed v = 240 mm/min Arc efficiency η = 0.8 Target heat-input band: 0.7–1.2 kJ/mm
Find the carbon equivalent, decide whether preheat is warranted, and check whether the proposed heat input sits inside the procedure's target band of 0.7–1.2 kJ/mm
Carbon equivalent first.
Composition alone doesn't say whether a steel is crack-sensitive in the heat-affected zone; the carbon-equivalent formula rolls all the alloying content into one number that tracks hardenability.
A CEV around 0.49 sits above the band where shipyard procedures typically start calling for preheat.
This plate needs it — skipping preheat raises the risk of hydrogen-assisted cracking in the HAZ once the joint cools.
Convert the travel speed before touching the heat-input formula.
It is specified per minute; the formula wants a consistent time base, here mm/s.
Now the heat input itself.
From the arc power and the travel speed.
Check against the target band.
0.864 kJ/mm sits inside 0.7–1.2 kJ/mm, so the procedure's energy is reasonable — but that is a separate control from preheat, not a substitute for it; both act on the same cooling curve and this plate needs both addressed.
AnswerCEV ≈ 0.49 (elevated) ⇒ preheat required; proposed heat input ≈ 0.864 kJ/mm falls inside the 0.7–1.2 kJ/mm target band, which satisfies the heat-input control only — not the preheat requirement.
The trap: forgetting to convert 240 mm/min to mm/s before dividing gives Q = 0.8×24×180/240 = 14.4 J/mm — two orders of magnitude too low — which would wrongly flag an adequately-run procedure as dangerously heat-starved.
Grades A, B, D, E / AH, DH, EHLetter denotes notch-toughness temperature rating, not strengthCEV = C + Mn/6 + (Cr+Mo+V)/5 + (Ni+Cu)/15Higher CEV needs more preheatGalvanic seriesMore anodic member corrodes; small anode/large cathode area ratio accelerates itFillet: throat a = 0.7 zStrength on the throat; weld metal volume ∝ z²Weld stress = F/(a·L)Total effective throat area over the weld lengthHeat input Q = ηVI/vControls the HAZ peak temperature and cooling rateBolted joint checksShear, bearing, net section, tear-out — smallest governsNet area = (w − n×d_hole) × tDeduct every hole on the critical sectionCathodic protectionSacrificial anode or impressed current; supplements a coating, doesn't replace itNDTVT, PT, MT surface; UT, RT internal/volumetric