A casting does not freeze everywhere at once. It freezes from the surfaces inward, and the last place to freeze is decided by the thickness of the sections feeding it and by how those sections are arranged. This is not a subtle effect to be corrected later by a process adjustment. It is the governing fact of the whole solidification pattern, and it determines where shrinkage ends up.
For an exhaust or intake component, section thickness is rarely uniform, and it should not be. A flange needs more material than a wall. A boss needs more than the web around it. The problem arises when the differences are accidental rather than deliberate, because an accidental thick section acts as a reservoir that has to be fed from somewhere, and the somewhere is usually a thinner section that solidifies first and cuts off the supply.
The rule that matters on the drawing board is that a section which is heavier than its surroundings will draw liquid metal from them as it solidifies. If the surrounding section is thin, it freezes early and the heavy section is left to shrink on its own. The result is a cavity in the middle of the heaviest part of the casting, which is often exactly where a machined face or a pressure boundary will be.
This is why section thickness is examined as a network rather than as a set of independent dimensions. The question is not whether a wall is thin, but whether the path from the feeder to the last point to freeze is continuous and progressively heavier. A casting that satisfies that condition is far more forgiving than one that does not, regardless of how carefully the individual dimensions are held.
Isolated heavy sections are the classic failure. A pad, a lug or a local boss set into a thin wall has no feeder path and will show internal shrinkage at the point where the customer most wants sound metal. The remedy is either to feed the section directly or to reduce its mass so that it no longer outruns its surroundings.
There is a temptation to treat thinner as better, on the reasoning that less metal means less shrinkage. That holds only while the thin section can still be filled. Below a certain thickness the metal chills before it reaches the end of the flow path, and the casting shows a cold shut or a surface that never formed.
Thin sections also lose the ability to feed anything else. A wall that freezes almost immediately acts as a barrier in the feeding network, isolating whatever lies beyond it. So a design that is uniformly thin may be free of internal shrinkage in the walls and full of defects at the far end of the flow, which is a different problem with the same root cause.
The practical answer is a graduated section, heavier at the feeder and lighter at the extremities, with transitions that are gradual rather than abrupt. Abrupt changes in thickness create hot spots at the junction, because the heavier side holds heat while the lighter side does not.
Where two sections meet, the combined mass is greater than either one alone. A T junction, an L junction or the intersection of a rib with a wall all produce a local volume that freezes later than the surrounding material. These junctions are the places where shrinkage porosity concentrates, and they are also the places where a designer is most likely to have stopped thinking about thickness because the geometry looks simple.
Fillets help by spreading the thermal mass over a larger volume, but a fillet does not eliminate the junction. It moves the problem slightly and reduces its severity. What actually solves it is a feeding path that reaches the junction before the junction freezes, which is a riser question as much as a section question.
This is the point at which section thickness and riser design stop being separate topics. A drawing that specifies wall thickness without saying anything about how the heavy junctions will be fed is only half a specification.
The drawing needs to state the nominal wall thickness for each region, identify any section that is deliberately heavier and explain why, and show the transitions between regions. It should also state which surfaces are machined, because a machined surface removes the sound skin and exposes whatever lies beneath it.
Where these are absent, the foundry is left to infer them, and inference on a thin wall casting is a poor substitute for a decision. The cost of a section review at drawing stage is trivial against the cost of a first article that has to be scrapped and redesigned.
A sound section design also makes the casting easier to inspect. When the thicknesses are deliberate, the places where defects can occur are known in advance, and the inspection plan can be targeted rather than uniform. That is worth more than it sounds, because uniform inspection on a part with one known risk area wastes effort everywhere and still misses the risk.
For components that will hold fluid, such as the tanks shown in the aluminium fuel tank range, the section question is inseparable from the leak test. A shrinkage cavity at a junction does not always show on a radiograph and does not always leak on the first test. It leaks after a period in service, when the pressure cycles have opened it. That is the worst possible time to discover a section that was never fed.
Section thickness, in the end, is the map of where the metal will be liquid last. Everything else in the casting process is arranged around that map. Drawing it deliberately is the difference between a casting that is designed and one that is merely shaped.
Thickness is measured on the first casting rather than assumed from the pattern. A wall thinner than drawn has lost strength and feeding capacity, and the loss stays invisible until the part is cut. The TIG aluminium welding route sidesteps the question by starting from plate of known gauge.
A wall with stock on one face is thinner after the cut, and what remains has to satisfy the drawing. aluminium water tank range work starts from plate instead, which keeps the welded route attractive for simple shells.
If a section cannot be fed, it changes. Local thickening to create a feeding path costs less at drawing stage than a pattern change later, and the custom parts, TIG stainless steel welding and titanium fabricated components routes are the comparison case.
