What is 3D printing? Starting from a drawing that can't be printed

A drawing that "can't be printed": what problem does 3D printing actually solve?

Last week we reviewed a bracket drawing: 1.2 mm wall thickness, an enclosed internal channel, and two mating holes marked ±0.05. The customer wrote in the notes, "3D print it." Behind that sentence is actually an unthought-through question—not all shapes are suitable for layer-by-layer deposition, and not all tolerances can be guaranteed by printing.

The essence of 3D printing is layer-by-layer deposition: slicing a 3D model into 2D contours and stacking them layer by layer. This principle determines what it is good at and what it is not. What it is good at is things traditional subtractive manufacturing cannot make—enclosed internal cavities, conformal cooling channels, irregular structures after topology optimization, and rapid iteration from drawing to physical part. What it is not good at is mass replication, isotropic strength, and mirror-grade mating surfaces. The bond strength between layers is usually lower than the bulk material; Z-direction tensile strength may be only 60–70% of XY, and the specific ratio varies with process and material, so it must be judged from measured data.

So when is 3D printing the right choice? Here are three criteria: First, the part shape is so complex that CNC fixturing is difficult or the tool cannot reach inside; Second, the quantity is small and revisions are frequent, making tooling uneconomical; Third, internal channels or lattice structures are needed, features that traditional processes cannot form. Conversely, if the part is a simple body of revolution, requires mirror-grade mating, or the batch size is already in the hundreds, don't rush to print—CNC or injection molding may be cheaper and more reliable. The correct response to that line on the drawing, "3D print it," is: first look at shape and quantity, then decide the process.

Comparison of mainstream 3D printing processes: how to choose among FDM, SLA, SLS, MJF, and DMLS

The selection order is often reversed. The correct order is: first determine the material, then the process, and finally discuss precision. If the drawing specifies PA12, don't agonize over layer thickness on FDM; if it specifies 316L, only DMLS will do. Conversely, if you pick the process first and then look for a material, you often end up reworking because of strength or heat resistance.

Three quick criteria

• For form and assembly interference verification only: FDM or SLA. SLA has a better surface, but don't use resin parts for load-bearing or long-term outdoor use.
• For assembly strength, batches of tens to hundreds of parts: SLS or MJF. Both suit nylon-type functional parts; MJF batch consistency is usually more stable, while SLS has a more mature material system.
• For metal load-bearing, heat resistance, conformal cooling: DMLS. But accept the post-processing costs upfront: support removal, heat treatment, and machining of mating surfaces often account for more than half of the quote.

A common misconception: using SLA resin parts as structural parts—they crack after two assemblies. Another: using FDM for sealing surfaces—layer lines leak directly. There's only one criterion—material determines the process, not the process determines the material. When the drawing doesn't specify a material, first ask about the application: load-bearing, sealing, heat resistance, or appearance only, then work back to the process.

Six things that must be clearly stated on the drawing: wall thickness, tolerances, mating surfaces, threads, supports, post-processing

3D printing quotes are low and lead times are fast, but many projects fail because of incomplete drawing information. The model is a solid, but process information isn't in the model. Miss any of the six items below and you may have to rework.

Wall thickness and tolerances: give ranges by process, don't specify by machining habits

Minimum wall thickness varies by process. FDM is commonly 0.8–1.2 mm, SLA can be thinner, SLS/MJF commonly around 1.0 mm; these are common industry references, and actual values should be confirmed with the equipment, material, and drawing. Writing only "thin wall" on a drawing is as good as writing nothing. Tolerances must be directional: XY is usually better than Z, and layer thickness determines Z resolution. If a mating surface is specified as ±0.05 while the model is a curved surface, first ask whether the process can reliably achieve it, then decide whether to change the specification or the process.

Mating surfaces, threads, and supports: assembly datums determine success or failure

Place mating surfaces in the XY plane as much as possible; Z-direction layer lines make sliding fits gritty. Don't expect to print threads directly; the common approach is to leave a pilot hole and tap it, or design an insert hole; directly printed threads have low strength and incomplete thread profiles. Support contact surfaces leave rough marks, so the drawing must state whether that surface is acceptable; if not, change the orientation or add allowance for machining. Post-processing (support removal, sanding, sandblasting, dyeing) changes dimensions and appearance and also affects lead time, so it must be stated on the drawing or RFQ.

Orientation determines success or failure: interlayer strength, warping, and support failure modes

The same model, with a different orientation, can differ in tensile strength by more than 30%, and a mating hole can go from passing a go gauge to not fitting at all. FDM, SLA, SLS, and MJF all build layer by layer, and the bond between layers is always weaker than within a layer. Don't design printed parts as if they were isotropic materials.

Load-bearing parts: keep principal stress in the XY plane

If a part bears load—snap fits, fixture clamp plates, load-bearing brackets—first look at the principal stress direction. Lay the part "flat" so tensile stress runs along the XY plane, with the Z direction bearing only compressive stress or no stress. A vertical printed column splits between layers with one bend; the same beam printed flat can withstand significantly greater bending moment. This criterion is more effective than adjusting infill density.

Seals and mating surfaces: don't let layer lines cross the sealing line

If a part needs to seal or mate—O-ring grooves, face seals, sliding fits—the orientation should make the sealing or mating surface an in-layer plane rather than an interlayer step. Layer lines are themselves microscopic grooves, and crossing the sealing line means a leak point. For mating holes, keep the axis horizontal as much as possible (holes formed along the XY direction) to avoid ovality and interlayer steps when the hole axis runs along Z, which cause pin shafts to seize.

Warping and supports: the cost of large flat surfaces and overhangs

With a large flat surface facing down, uneven cooling shrinkage causes corners to lift, a common failure; adding a brim or changing orientation usually controls it. Overhang angles exceeding about 45° (a common industry reference, to be confirmed by specific process and material) generally require supports; support contact surfaces leave rough marks, so don't place assembly surfaces or appearance surfaces facing down. Supports aren't free: one more support means one more support-removal operation and one more dimensional uncertainty.

Order of judgment: first determine the assembly datum and load direction, then determine orientation; if orientation can't be settled, tolerances and post-processing are empty talk. When the drawing doesn't specify orientation requirements, confirm with engineering before quoting—don't wait until it's printed to find it won't fit.

A real case: from "printed but won't fit" to revising the drawing and printing again

Last year, a batch of inspection fixtures, small parts: an 80×60×12 mm locating block with two φ6H7 pin holes, with hole axes originally perpendicular to the large flat surface. The customer was in a hurry to verify assembly and sent the model straight to FDM printing. When the parts arrived, the pins wouldn't go in—the holes measured about 0.15 mm small, and after being forced in they couldn't be removed; the Z-direction layer lines on the hole walls were like fine threads and seized as soon as they were turned.

The review found three problems stacked together: orientation put the hole axis in the Z direction, with weak interlayer bonding and hole diameter shrinkage accumulating along the layer height; tolerance copied machining H7 without leaving allowance for printing; process chose FDM for mating parts. The fix was straightforward: rotate the hole axis to the XY plane, leave 0.1 mm per side on the hole diameter and then ream; change the mating surface to SLA to ensure the surface; change the pin holes to inserts rather than printing them directly. The second version fit in one go.

The lesson from this case isn't that "FDM doesn't work," but that the order was wrong: first determine the assembly datum, then determine the printing direction, and only then choose the process. Once the datum is set, there are criteria for how to orient holes, how much allowance to leave, and whether post-machining is needed. Choosing the process first leaves you passively patching things up afterward.

Inspection and acceptance: how to measure 3D printed parts, what to measure, and which dimensions not to fuss over

The most argument-prone item on a 3D printed part acceptance sheet is measuring a layer-line surface with calipers and then asking "why isn't it Ra1.6." Layer lines, support marks, and step marks are process characteristics of layer-by-layer buildup, not machining defects; holding them to machining surface roughness standards leaves the supplier able only to rework or refuse the order. What should really be held tight are mating dimensions and functional surfaces; appearance surfaces can be accepted as "no obvious visual defects."

Set acceptance criteria by dimension type, not one-size-fits-all

Measuring tools are not better the more expensive they are. For mating holes, plug gauges go/no-go are fastest; save the CMM for first articles and disputed parts; calipers measuring a layer-line surface inherently have several hundredths of uncertainty, and the numbers measured cannot be used as arbitration evidence. If the part has sealing surfaces or sliding mating surfaces, the first article must go on the CMM, and batch parts are sampled with plug gauges plus visual inspection.

The most easily misaligned statement on the procurement side: "quote by model, inspect by machining"

When requesting a quote, only throwing over an STP file means the supplier estimates labor hours by volume and orientation; after arrival, QC takes the machining tolerance table and measures item by item, and as a result the layer-line surfaces and Z-direction hole spacing all "exceed tolerance." The way to avoid misalignment is to write three things clearly at the quoting stage: orientation requirements (which surfaces must face down or up), post-processing requirements (support removal, sanding, sandblasting, dyeing, tapping), acceptance standards (which dimensions are held tight, which only get functional checks). Write these three into the RFQ, and the quote and acceptance will match up. When you need to quickly verify pricing against drawings, you can go through the 3D printing quote entry and submit the requirements together.

One last criterion: dimensions that can assemble, rotate, and seal are held tight; dimensions that merely do not look glossy enough are loosened. Spending inspection resources on functional surfaces is far more worthwhile than spending them on layer lines.

When not to use 3D printing: boundary conditions and small-batch economics

3D printing is not a master key. Batch size is the first hard boundary: when demand for the same part exceeds dozens to hundreds of pieces (the specific threshold fluctuates with part size, material, and post-processing amount, and must be calculated per part), the unit cost of injection molding or hand-made duplicate molds will overtake it. Injection molding requires amortizing mold costs upfront, but once amortized, the unit cost may be only a fraction of printing; the lifespan of vacuum-cast silicone molds is usually dozens of times, suitable for low-to-medium batches.

The second boundary is performance requirements. Layer-by-layer deposition determines that Z-direction bonding is weaker than XY direction. If the part requires isotropic strength, mirror-fit surfaces, long-term weather resistance, or food-contact certification, 3D printing may not be applicable—such needs should turn to CNC, injection molding, or sheet metal. Large solid parts are also not cost-effective: printing time rises steeply with volume and height, and material and machine-time costs rise in tandem.

Simple algorithm: total cost = unit cost × quantity + one-time costs (mold/programming/fixtures). Calculate each of the four paths—3D printing, CNC, duplicate molding, injection molding—and take the one with the lowest total cost. If the quantity falls near the boundary, first print a small batch to verify assembly, then decide whether to open a mold. When requesting quotes, provide the quantity tiers, material, post-processing, and acceptance standards together to avoid the misalignment of "quote by model, inspect by machining." For related process capabilities, refer to the injection molding capabilities and rapid prototyping pages.

Common questions FAQ

The following questions are the ones most often asked during quoting and drawing review. The answers all carry boundary conditions; for your specific part, you still need to look at the drawing and actual working conditions.