Casting defects: what causes them, how to measure yield, and how to trace a bad casting back

A defect investigation goes wrong when the name already contains a guess. Call a cavity a 'slag defect' and the team will chase slag practice even if the real cause is turbulent filling. A Casting Source article on defect investigation puts it plainly: the International Atlas of Casting Defects gives each defect a code based on appearance, and recommends that unknown defects be classified by appearance rather than cause. Identify what you see first, then test the possible causes.

You do not need the full Atlas to benefit from the idea. A short, controlled list of defect codes that everyone on the floor uses the same way is worth more than a long list nobody agrees on.

Record two things for every rejected casting: the defect code and the stage where it was found. A shrinkage cavity found at gate cut-off and one found by the customer are the same defect with very different costs.

In investment casting the mold is a ceramic shell built in layers on a wax pattern, then dewaxed and fired before pouring. That adds a whole stage where things can go wrong before any metal is melted, and many defects found after knockout started in the shell room.

The Investment Casting Institute's Atlas of Shell Defects traces many of them to a few controllable conditions:

• Humidity and temperature during drying. Evaporating slurry cools the wax, which shrinks and then expands again as it warms. Because wax moves faster than the shell, that cycle can crack prime coats. The Atlas advises running dip and prime drying areas at relative humidity above 50 percent, and keeping ambient temperature controlled around the clock.
• Slurry control. The Atlas recommends logging viscosity, temperature, pH and specific gravity daily, with weekly checks on binder silica, total solids and refractory content. If specific gravity falls while viscosity holds, the slurry is becoming unstable.
• Dewax speed. Wax must start to melt at the surface before the body of the pattern expands. The Atlas gives a rule of thumb of reaching 100 PSIG in 10 seconds or less in the autoclave, and releasing pressure slowly, over about two minutes, to avoid delamination.
• Shell thickness. The Atlas notes that if shell thickness doubles, burst strength roughly quadruples. Too thin and the shell cracks in dewax; too strong and it can restrain the casting and cause hot tears. Some shops weigh shells after dewax against a band per pattern as a quick check on coat build.

Good castings

Castings that passed every inspection stage, counted at the end of the route.

Castings poured

Every casting position actually filled with metal: trees poured multiplied by castings per tree, or molds poured.

Weight of good, degated castings

Finished casting weight after gates and risers are removed.

Weight of metal poured

Everything that went into the mold, including the gating system, risers and the tree itself.

Casting Source defines casting yield as the percentage of quality, degated castings produced relative to the total metal melted and poured, which is the weight-based measure. The count-based quality yield answers a different question: of the castings we tried to make, how many were good? Track both. Metal yield tells you about gating and feeding efficiency. Quality yield tells you about defects.

A Pareto chart sorts scrap by defect code from largest to smallest with a running cumulative share. Scrap is rarely spread evenly across codes, which is exactly why the chart is useful: it tells you where an hour of investigation pays back.

A Pareto across the whole foundry is a starting point, not an answer. Cut it by pattern, by alloy and by the stage found. Shrinkage concentrated on one pattern points at its gating. Gas porosity concentrated on one alloy points at melt practice. Shell cracking concentrated in one week points at the shell room or the dewax cycle in that week. Count matters, and so does cost: a defect found after machining costs far more than one found at knockout.

The worked example only works if each rejected casting can be tied to the records upstream of it. The Steel Founders' Society of America defines a heat as the total metal produced that can be represented by one analysis sample and one set of mechanical tests. That makes the heat the natural anchor for chemistry. The shell, the tree and the pour are the anchors for everything else.

Think of it as a tree of records. A casting belongs to a tree or mold. The tree was built in a shell batch on a known day, under logged slurry and humidity conditions, then dewaxed and fired in a known cycle. It was poured from a known heat, at a recorded temperature. The heat has its charge mix and its spectro result. If each link carries an ID, a scrapped casting points straight back to every condition it saw.

• Give every heat a number and record it on every pour.
• Give every tree or mold an ID that survives knockout and cut-off, and record which pattern it carries.
• Log shell room conditions and dewax and burnout cycles against dates and batches the tree IDs can be matched to.
• Record outcomes per casting or per tree with defect code and stage found, not as a single scrap total per day.
• Classify each heat's chemistry against the grade's element bands at the time, and keep that verdict with the heat.

Many of the causes in the defect table are equipment conditions. An autoclave that has lost its pressurization speed cracks shells. A drying room whose humidity control drifts produces rat-tailing. A worn furnace lining or a damp ladle adds inclusions and gas. A pyrometer that reads low turns a good pour temperature into a misrun.

That is why defect data and maintenance data belong side by side. When a defect cluster appears, the first question after 'which heat?' is often 'what was the state of the equipment that day?' Calibration records for pyrometers and spectrometers, PM records for autoclaves, burnout furnaces and humidity control, and lining campaign records for furnaces all answer it.

MaintenanceIQ's Foundry package records each stage as its own record: a pattern master, pour records tied to their heat, and outcome records with a defect code and the stage found, including NDT and FPI stages. Casting yield (good divided by poured, by pattern and alloy) and the defect Pareto are computed from those records rather than typed in. Each heat's spectro chemistry is classified against the alloy grade's element bands.

For investment shops it adds shell room control, dewax and burnout logs, and shell weight control, whose station works offline and syncs when the network returns. The production route module, licensed separately, stamps a lot at each stage and lets a dispatched casting drill back to the heat and chemistry it was poured from. The furnaces, autoclaves and instruments behind all of this sit on the same CMMS for PM and calibration. More on investment and metal casting.