Menu

The Screw That Doesn't Strip

The Screw That Doesn't Strip

Notes · Mechanical designNovember 19, 202621 min readSenova Systems Research

The housing has to open for battery service and calibration, and every time it opens, the joint gets weaker. A self-tapping screw holds 380 N on the first pass and is stripped by the eleventh; a brass insert still holds 1,056 N at pass 25. The strength does not come from the brass, it comes from plastic locked into the knurls.

Picture the hub coming back from the field. The lid has to come off, the battery needs changing, the sensor needs calibrating, and the screws go in and out, and each time they go in, they bite a little less. The first time, the screw holds.

The fifth time, the torque is down 40%. The eighth time, the screw is holding 152 N where it held 380 N on the first pass, and the torque has fallen from 2.5 N·m to 1.1 N·m. By the eleventh time, the screw is stripped, the thread it cut into the plastic is gone, and the lid no longer clamps.

The gasket underneath it, which needed that clamp to seal, is no longer compressed. The hub that was rated for the outdoors is no longer sealed, and the failure is not in the electronics. It is in the screw.

When an injection-molded assembly must be opened and re-fastened more than 5 to 10 times for battery service, calibration, or field maintenance, or when a joint carries high clamp preload, direct self-tapping screws eventually strip the polymer threads. (Plastics Technology Alliance) The joint gets weaker every time it opens, and the weakness is not a quality problem. It is a design problem: the thread is cut into plastic, and plastic is not a thread material.

The screw, the insert, and the shear area

A self-tapping screw makes its own thread by cutting or forming the plastic as it goes in. A thread-forming screw has a blunt end and pushes the walls of the plastic hole into the shape of threads, producing no chips; a thread-cutting screw has a slotted point and cuts the material, producing chips. (Assembly Magazine) Either way, the thread is plastic, and the thread is the joint.

When you drive the screw in again, you are driving it into a thread that has already been cut, and the plastic has already been displaced. The best way to avoid stripping the threads formed in the plastic is adherence to the fastener's strip-to-drive ratio, the ratio between the torque at which the screw will strip its threads and the torque at which it drives. (Assembly Magazine) And a cordless driver swings torque ±20% and strips a screw by pass four, while a station limited to 4.5 N·m lets an insert reach 50. (Cocake)

A threaded insert is a metal cylinder with a threaded hole in the middle and a knurled outer wall. The screw goes into the metal thread, and the metal thread never wears out, the inserts enable unlimited assembly and disassembly of the components without compromising the integrity of the threads. (PT Online) Metal threaded inserts, machined from C36000 free-cutting brass, lead-free Eco Brass, 303 stainless steel, or 2024 aluminum, transfer machine-screw clamp loads across a much larger knurled outer surface area inside the plastic boss.

(Plastics Technology Alliance) The most common material for the threaded inserts themselves is brass; with increasing sustainability requirements, lead-free alternatives such as stainless steel or aluminum are growing more popular, stainless steel offers better strength and corrosion resistance, and aluminum inserts are about 70% lighter than brass. (PT Online, PennEngineering)

The difference is the shear area. Pulled at 5 mm/min, the brass insert leaves the boss whole at 1,200 N; the self-tapping screw strips at 380 N, with polymer failing before hardware in both cases. (Cocake) The insert carries load on a knurled shear band.

At 6.3 mm diameter and 8 mm depth it presents 158 mm², turning 1,200 N into 7.6 MPa, well under the 18 MPa glass-filled PP sustains. A screw has far less area. The ST4.2 thread is 0.6 mm high, so 8 mm of engagement leaves roughly 26 mm², and 380 N already means 15 MPa, close to the ceiling.

(Cocake) The insert has 6.1 times the shear area of the screw, and it is working at 42% of the material's ceiling where the screw is working at 81%.

Pull-out strength and torque retention do not come from the brass. They come from plastic locked into the knurls, grooves, and undercuts on the outside of the insert. If you can get plastic to flow around those features and solidify there, then you have created a mechanical interlock that resists both pulling out and spinning.

(PT Online) The knurl pattern is critical: the longer the insert, the greater the pull-out resistance; the greater the diameter of the insert, the more torque capacity. In general, resistance to torque is a function of insert diameter, and resistance to pull-out is a function of insert length. (PT Online) Opposing knurls or diamond knurling and serrations provide pull-out resistance; undercuts and grooves provide pull-out resistance; knurling features provide torque resistance.

(Engineering Product Design)

And the insert does not creep. Metal threaded inserts significantly improve joint strength in plastic parts and are not themselves susceptible to creep. (PT Online) They are not affected by the plastic material's stress relaxation or self-tapping screw creep, and they provide increased seating torque for applications such as compressing o-rings and gaskets to seal a product assembly.

(Engineering Product Design) The screw's thread is plastic, and plastic creeps under load; the insert's thread is brass, and the brass holds the load while the plastic holds the brass. The joint does not become loose over time, because the load path of the thread is metal, and the metal does not relax.

The numbers: 3.2× to 3.5×, and the cycle count

The pull-out table, from Cocake's own test on its cases in glass-filled PP, pulled at 5 mm/min:

SizeBrass insertSelf-tapping screwRatio
M3780 N220 N3.5×
M41,200 N380 N3.2×
M51,750 N520 N3.4×
M62,400 N700 N3.4×

The ratio holds between 3.2 and 3.5 because insert shear area scales with diameter while thread height does not; at M6 the screw gains 3 mm². (Cocake) A second test, from Sculpteo, on heat-set brass inserts in MJF PA12, five specimens per configuration, installed with a temperature-controlled soldering iron, gives a similar shape at different absolute values: M2 short 477 N, M2 long 572 N, M2.5 825 N, M3 short 813 N, M3 long 1,258 N, M4 1,372 N, M5 over 1,800 N, M6 over 1,800 N.

The M4 figures differ, 1,372 N in PA12 against 1,200 N in glass-filled PP, because the material and the insert length differ, and the article says so. The shape is the same: the insert holds three times what the screw holds, and the longer insert holds more.

The cycle count is where the screw dies. An M4 screw holds 380 N on the first pass, 228 N on the fifth, and 152 N by the eighth, torque falling from 2.5 to 1.1 N·m; a brass insert still holds 1,056 N at pass 25 and 840 N at pass 50. (Cocake) The band table: 1–3 cycles, screw 340 N, insert 1,180 N, "screw is fine"; 4–10 cycles, screw 190 N, insert 1,120 N, "insert or sleeved screw"; 11–25 cycles, screw stripped, insert 1,056 N, "insert"; 26–50 cycles, screw failed, insert 840 N, "insert, diamond knurl".

(Cocake) Two thresholds settle the drawing: 10 or more planned cycles, or 800 N of required pull-out, mean inserts; fail both and a screw saves $0.09 a joint. (Cocake) Write both thresholds on the drawing and the choice stops being a debate: 10 or more assembly cycles, or 800 N of pull-out, means brass inserts; below both, self-tapping is cheaper. (Cocake)

The failure modes, across 200 warranty returns: pull-out 46%, spin-out 22%, boss cracking 19%, thread stripping 13%, each tracing back to a different dimension. (Cocake) Pull-out is the boss too short or too thin: if an insert has moved 0.3 mm at 800 N, engagement is under 6 mm and half the knurl sits in air; raising boss height from 8 mm to 10 mm restored 1,150 N. Spin-out is the knurl wrong for the torque: spin-out appears above 3.2 N·m because straight knurling cannot grip tangentially; three ribs 0.4 mm deep, or a diamond knurl, push spin torque to 5 N·m.

Cracking is a design fault: below 1.0 mm of ring wall, assembly expansion splits the boss along a 45-degree path; an R0.5 fillet and 0.85 mm of wall remove about 80% of cases. Stripping is the screw's alone, since the thread is cut not moulded, torque is down 40% by the fifth pass and 1.1 N·m by the eighth. (Cocake)

The four ways in: heat, ultrasonic, moulded-in, self-tapping

There are two primary types of inserts: those that are molded in and those installed after the molding process (post-mold). Molded-in inserts usually yield the highest performance, yet this form of installation is by far the most expensive. Inserts installed with heat or ultrasonics after molding yield good performance at a fraction of the assembled cost of the molded-in inserts.

Inserts installed with heat yield the best combination of overall performance and lowest installed cost. (PT Online) The installation method must be considered, as this will affect the type of insert that can be used as well as the overall cost of the assembly.

Heat-set (thermal press). 85–95% of molded-in strength, with the lowest residual hoop stress in the boss; 3–6 seconds per insert, or gang-pressed 4–8 at once; prototypes through medium production (100–100,000 pieces a year); thin-wall PC and ABS housings. (Plastics Technology Alliance) The knurl is opposing helical knurls separated by annular undercut drainage grooves, with a pilot lead-in. A temperature-controlled thermode heats the insert by conduction. The plastic around the hole softens; the insert is pushed home; displaced material reflows into the knurls; and then a cooling stage locks everything in place before the tip retracts. Heat in, cool, retract. (PT Online) Thermal insertion's strength is latitude: it handles the broadest range of materials, amorphous and semicrystalline alike, and tolerates filled resins and awkward boss geometry better than the alternatives. It will set several inserts in a single press stroke, and with a proper cooling stage, it leaves a clean, flash-free face. One important rule travels with it: the temperature is set by the polymer, not the press. Anyone quoting a single number for everything has not met your material yet. The trade-off with thermal insertion is time: because the cycle includes both a heating and a cooling phase, this process is rarely the fastest method per insert. (PT Online) In Cocake's own test, a 270 °C tip presses 3 seconds and holds 5 seconds, eight seconds a joint, giving 1,100 N and 4.5 N·m from a $400 hand tool. (Cocake)

Ultrasonic (20–40 kHz horn). 80–90% of molded-in strength, requiring tight ±0.05 mm hole control; 0.5–1.2 seconds per insert; medium-to-high volume (25,000–500,000 pieces a year) where gang thermal pressing is too slow. (Plastics Technology Alliance) The knurl is tapered (8°) or straight symmetrical opposing helical knurls. The insert is driven by a horn vibrating around 20 kHz. Friction at the plastic-to-metal interface generates heat in a fraction of a second, the local plastic melts; the insert sinks; and the melt solidifies around the knurls almost immediately. The cycle is genuinely quick, sub-second per insert, which is why it wins on high-volume lines. (PT Online) In Cocake's own test, at 20 kHz and 28 μm the cycle takes 0.5 seconds and pull-out reaches 1,200 N, the best of the three, and the cost is a 0.3 mm flash ring needing a 0.5 mm counterbore. (Cocake) Heat and ultrasonic both reflow plastic into the knurls; press-fit skips the melt entirely and relies on interference alone. (PT Online)

Molded-in (loaded in cavity). 100% baseline, the highest, molten resin flows under high pressure into deep undercuts, and it slows the press cycle by 4–15 seconds unless robot-loaded, and risks a $3,000–$15,000 cavity steel crash if an insert drops during closing. High-load structural mounts (M5–M10), thermosets, or automated vertical rotary-table insert molding. (Plastics Technology Alliance) The knurl is a deep diamond knurl with a wide hex or undercut collar, and a closed or blind pilot end. In Cocake's own test, pins located to 0.02 mm let melt encapsulate the knurl for 1,300 N and 5.2 N·m, but tooling rises 12% and 1.2% of parts are scrapped. (Cocake) Volume decides, not strength: under 1,000 units, heat-set allows late changes; above 10,000 units, moulded-in cuts piece cost to $0.13; ultrasonic suits a frozen design. (Cocake)

Self-tapping / thread-cutting inserts. 90–100% of molded-in strength. The external coarse cutting threads, with cutting slots or three radial chip holes, cut mechanical threads into the plastic, and they provide the best pull-out resistance for a post-mold insert, yet the assembler has to be very careful about installing the insert so that it is perfectly square to the hole or it will not mate well with the screw. (PT Online) The three-flute variant is the most capable: the flutes remove material cleanly rather than just displacing it, which reduces hoop stress on the boss and makes the design suitable for brittle thermosets and filled engineering plastics as well as softer thermoplastics. (Brassland) And cold press-in / expansion inserts, 40–60% of molded-in, relying purely on friction and hoop interference, are the fastest manual or arbor-press installation, but they generate high radial hoop stress that cracks brittle resins, and they belong on non-critical covers in ductile PP, PE, or nylon. (Plastics Technology Alliance) Expansion inserts are designed for non-critical applications where ease of installation is the primary design criterion, not torque and pull-out resistance. (PT Online)

Two processes, ultrasonic and heat-staking, involve re-melting previously molded plastic, so they only work with thermoplastics. For thermosets, the options are molding-in the insert, or cold pressing the insert in later, where the elasticity of the resin material will become critical. (PT Online) And the cost is the total cost of the operation: molding time and cost, component handling, and assembly. (PT Online)

The boss: 1.25×, 0.85 mm of ring wall, 1.5 mm under the hole

The boss runs 1.25 times insert diameter: 6.3 mm becomes 8 mm with 0.85 mm of ring wall, and 1.5 mm of polymer stays under the hole. (Cocake) For an M4 insert, the floors are an 8 mm boss, 0.85 mm of ring wall, and 8 mm of depth, and losing any one costs roughly 200 N. (Cocake) At 7 mm the ring wall thins to 0.35 mm and cracking probability in assembly rises to 30%.

(Cocake) The holes split by process: 6.4 mm for heat-set, 6.2 mm for ultrasonic, 6.3 mm for moulded-in. The clearance gives displaced melt room, the tight hole transmits amplitude, and the core pin is the hole. (Cocake)

The 8 mm boss is not a guess, it is the bottom of two different rules. PT Online's guideline is that the optimum wall thickness or boss diameter is two to three times the insert diameter, with the relative multiple decreasing as the insert diameter increases, and for an M4 thread (4 mm), two to three times is 8–12 mm, so Cocake's 8 mm sits at the bottom of that band.

(PT Online) Article #16 built the same rule from the moulding side: bosses at most 0.6× the nominal wall, with the boss outer diameter at about 2× the screw or insert diameter (2.0–2.4× the bore), cored at the base, and design determines 70–80% of final part cost and quality before production starts, with about 70% of first-submission designs containing at least one feature that causes a defect or forces an expensive mould action.

The boss is where sink appears first on a 2.5 mm wall: a 3 mm root leaves a 0.08 mm dish visible under raking light, and moving the boss 6 mm inboard or coring 0.3 mm at the root drops it below 0.03 mm. (Cocake) A boss at mid-edge forces melt fronts to rejoin behind it, and that line faces the impact load, cutting drop energy 30% where it opens, shifting the boss 15 mm, or widening it to 9 mm, moves the line into a lower-stress zone.

(Cocake)

Selecting the wrong installation method, or undersizing the receiving boss wall by 0.5 mm, leads to split bosses in polycarbonate, jack-out failure during torque-down, or molten plastic flashing into the internal threads. (Plastics Technology Alliance) And the holes for post-mold inserts should always be deeper than the length of the insert: for self-tapping inserts, use a minimum depth of 1.2 times the insert length; for other inserts, the recommended depth is the insert length plus two thread pitches.

(PT Online) The insert manufacturer's recommended hole size must be adjusted if fillers are used in the plastic, and correct hole size is critical. (PT Online)

The alternatives are real, and they are worse for this joint. The machined alternative, Article #27, is a 6061-T6 aluminium lid, machined from solid: the baseline is $160 at ±0.5 mm, the lid as drawn is $1,370 a part, and the fixed version is $430 a part at a batch of 10. The machined lid has a real thread, cut in metal, but it costs 43 times the PC enclosure's assumed $10, and it weighs 133.9 g against the PC shell's 59.5 g (Article #47).

The sheet-metal alternative, Article #30, is a 5052-H32 tray, 2.0 mm thick, flat 172.80 mm: the screws hold the tray, and the tray is not the housing. The housing is the moulded enclosure, Article #7, and the joint that closes it is the insert, because the housing has to open and the joint has to survive opening.

The hub lid, opened 25 times

Example Labs is a made-up company we use for worked examples.

The demo firm, as always: the Example Labs field sensor hub, 100 × 60 × 40 mm, with a PC/ABS lid on the aluminium base (Article #47, the hub's answer is a hybrid: aluminium base for heat and shielding, polycarbonate lid for the antenna), a 2 mm wall, and the lid held by 8 screws around the perimeter. The arithmetic was run in Python before writing. The service schedule and the joint are the article's own worked example, labelled: the lid opens for battery service every two years (5 opens over a 10-year life) and for calibration twice a year (20 opens), 25 opens over the life of the hub.

Step 1, the screw, cycle by cycle. An M4 self-tapping screw holds 380 N on the first pass, 228 N on the fifth, and 152 N by the eighth, torque falling from 2.5 to 1.1 N·m; by the eleventh pass it is stripped. (Cocake) At 25 opens, the screw is stripped, and by year 5, at about pass 12, the lid no longer clamps. The gasket underneath it, Article #7, loses about 20% of its sealing after two years of temperature cycling from poor compression set alone, and now it has no clamp at all. The hub that was rated for the outdoors is no longer sealed, and the failure is not in the electronics. The torque is down 40% by the fifth pass and 1.1 N·m by the eighth, and a cordless driver, swinging torque ±20%, strips a screw by pass four. (Cocake)

Step 2, the insert, cycle by cycle. An M4 brass insert holds 1,180 N on the first pass, 1,120 N at 4–10 cycles, 1,056 N at pass 25, and 840 N at pass 50, 89.5% of its first-pass strength at pass 25, and 71.2% at pass 50. (Cocake) At 25 opens, the insert still holds 1,056 N, 7.4 times the screw's 152 N at pass 8, and the screw is stripped by then. The insert's thread is brass, and the brass does not wear; the screw's thread is plastic, and the plastic is gone by pass 11.

Step 3, the shear, and the boss. The M4 insert is 6.3 mm in diameter and 8 mm deep, presenting 158 mm² of shear area, 1,200 N becomes 7.6 MPa, 42% of the 18 MPa glass-filled PP sustains. The ST4.2 screw thread is 0.6 mm high, so 8 mm of engagement leaves roughly 26 mm², and 380 N already means 15 MPa, 81% of the ceiling. The insert has 6.1 times the shear area, and it is working at half the stress. The boss is 8 mm, 1.25 times the insert diameter, 2 times the M4 thread, with 0.85 mm of ring wall and 1.5 mm of polymer under the hole. Losing any one of the three floors costs roughly 200 N. At 7 mm the ring wall thins to 0.35 mm and cracking probability in assembly rises to 30%. Below 1.0 mm of ring wall, assembly expansion splits the boss along a 45-degree path; an R0.5 fillet and 0.85 mm of wall remove about 80% of cases. And the sink: on a 2.5 mm wall, a 3 mm root leaves a 0.08 mm dish visible under raking light; coring 0.3 mm at the root drops it below 0.03 mm.

Step 4, the spin, and the torque. Spin-out appears above 3.2 N·m because straight knurling cannot grip tangentially. Three ribs 0.4 mm deep, or a diamond knurl, push spin torque to 5 N·m, 1.56 times. In Cocake's process test, heat-set gives 1,100 N and 4.5 N·m, ultrasonic gives 1,200 N and 4.8 N·m, and moulded-in gives 1,300 N and 5.2 N·m. The torque the joint can take is set by the knurl, not by the screw, and the station is limited to 4.5 N·m, which lets an insert reach 50 passes. If an insert has moved 0.3 mm at 800 N, engagement is under 6 mm and half the knurl sits in air; raising boss height from 8 mm to 10 mm restored 1,150 N.

Step 5, the cost. About $0.09 per M4 insert, $0.72 across eight joints, while the return rate falls from 6% to under 1%. (Cocake) A fleet of 1,000 hubs takes $720 of inserts. At 6%, 60 hubs come back; at under 1%, fewer than 10 come back, 50 fewer returns. At $150 per field replacement (the labelled assumption from Article #43), that is $9,000 of returns against under $1,500, $7,500 saved for $720 spent, a 10.4× payback. Installation: heat-set takes 8 seconds a joint, 64 seconds a hub, from a $400 hand tool; ultrasonic takes 0.5 seconds a joint, 4 seconds a hub, from an $8,000 welder; moulded-in is in cycle, with tooling up 12% and 1.2% of parts scrapped, and above 10,000 units, moulded-in cuts piece cost to $0.13 an insert. Volume decides, not strength: under 1,000 units, heat-set allows late changes; above 10,000 units, moulded-in wins; ultrasonic suits a frozen design.

Step 6, the acceptance test. Pull three samples at 5 mm/min and reject any batch peaking under 900 N, a 25% margin against the 1,200 N measured. (Cocake) Self-tapping joints are accepted at 300 N, because they are not built for reuse. And a flexing clamp adds 80–120 N at 800 N, enough to separate two labs by 150 N, so the fixture matters as much as the specimen, and the test is only as good as the clamp.

The decision test

Five questions. If you can answer all five with a number from the drawing, the joint exists. If you cannot, the joint is a hope.

  1. How many times will this joint open, and what does the screw hold on the last pass? If the answer is 10 or more planned cycles, the joint takes a brass insert. An M4 screw holds 380 N on the first pass and is stripped by the eleventh; a brass insert still holds 1,056 N at pass 25. Below 10 cycles and below 800 N of required pull-out, a screw saves $0.09 a joint.
  2. What is the pull-out, and what is the shear? The M4 insert pulls out at 1,200 N on 158 mm² of shear area, 7.6 MPa against an 18 MPa ceiling; the screw strips at 380 N on 26 mm², 15 MPa against the same ceiling. If the joint is working at 81% of the material's ceiling on the first pass, it has no margin left for the fifth.
  3. What is the boss, and does it meet the three floors? For an M4 insert: an 8 mm boss, 0.85 mm of ring wall, and 8 mm of depth. Losing any one costs roughly 200 N. At 7 mm the ring wall thins to 0.35 mm and cracking probability in assembly rises to 30%.
  4. What is the knurl, and what is the spin torque? Straight knurling spins above 3.2 N·m. Three ribs 0.4 mm deep, or a diamond knurl, push spin torque to 5 N·m. The torque the joint can take is set by the knurl, not by the screw.
  5. What is the acceptance test, and what does it reject? Pull three samples at 5 mm/min and reject any batch peaking under 900 N, a 25% margin against the 1,200 N measured. If the batch cannot be tested, the joint cannot be trusted.

Objections

"The screw is cheaper, $0.09 a joint adds up." It does add up: $0.72 across eight joints, $720 across a fleet of 1,000 hubs. And the return rate falls from 6% to under 1%, 50 fewer returns on that fleet, $7,500 saved at $150 a field replacement, a 10.4× payback. The screw saves $0.09 a joint and costs the return. The insert costs $0.09 a joint and removes the return. The cheapest joint is the one that does not come back.

"Our housing only opens once or twice, the screw is fine." It is: 1–3 cycles, the screw holds 340 N and the insert holds 1,180 N, and the recommendation is "screw is fine". The two thresholds settle the drawing, 10 or more planned cycles, or 800 N of required pull-out, mean inserts; fail both and a screw saves $0.09 a joint. But count the cycles honestly: battery service every two years and calibration twice a year is 25 opens over ten years, and the screw is stripped by the eleventh. The joint that opens "once or twice" is usually the joint that opens twenty-five times.

"Heat-set is slow, 8 seconds a joint." It is the slowest of the three, and the cheapest to start: a $400 hand tool, 64 seconds a hub, and it allows late changes under 1,000 units. Ultrasonic is 0.5 seconds a joint, 4 seconds a hub, from an $8,000 welder, and it suits a frozen design. Moulded-in is in cycle, with tooling up 12% and 1.2% of parts scrapped, and above 10,000 units it cuts piece cost to $0.13 an insert. Volume decides, not strength. And heat-set yields the best combination of overall performance and lowest installed cost. (PT Online)

"The boss is too big, we don't have 8 mm." Then the joint is not an M4 joint. The boss runs 1.25 times insert diameter, and the floors are 8 mm, 0.85 mm, and 8 mm, losing any one costs roughly 200 N. The options are a smaller insert (M3: 780 N, 220 N for the screw, 3.5×), a machined lid (Article #27, $430 a part at a batch of 10, 133.9 g against the PC shell's 59.5 g), or a different joint entirely. What does not work is an M4 insert in a 7 mm boss: the ring wall thins to 0.35 mm and cracking probability in assembly rises to 30%.

"We tried inserts, they spun out." Spin-out is the knurl, not the insert: straight knurling cannot grip tangentially above 3.2 N·m, and three ribs 0.4 mm deep or a diamond knurl push spin torque to 5 N·m. And if the insert has moved 0.3 mm at 800 N, engagement is under 6 mm and half the knurl sits in air, raise the boss height from 8 mm to 10 mm and the pull-out is restored to 1,150 N. The insert that spins is the insert that was specified with the wrong knurl, in a boss that was too short.

Sources

  1. Plastics Technology Alliance, "Threaded Inserts for Plastic: Heat-Set vs Ultrasonic"
    What we used

    "When an injection-molded assembly must be opened and re-fastened more than 5 to 10 times for battery service, calibration, or field maintenance, or when a joint carries high clamp preload, direct self-tapping screws eventually strip the polymer threads. Metal threaded inserts (machined from `C36000` free-cutting brass, lead-free Eco Brass, `303` stainless steel, or `2024` aluminum) transfer machine-screw clamp loads across a much larger knurled outer surface area inside the plastic boss." "Selecting the wrong installation method, or undersizing the receiving boss wall by `0.5 mm`, leads to split bosses in polycarbonate, jack-out failure during torque-down, or molten plastic flashing into the internal threads." The comparison matrix: Heat-Set Inserts (thermal press), opposing helical knurls separated by annular undercut drainage grooves + pilot lead-in, 85%, 95% of Molded-In (lowest residual hoop stress in the boss), zero mold complexity; off-line thermal press takes 3–6 s per insert (or gang-pressed 4–8 at once), prototypes through medium production (100–100,000 pcs/yr); thin-wall PC and ABS housings. Ultrasonic Inserts (20–40 kHz horn), tapered (8°) or straight symmetrical opposing helical knurls, 80%, 90% of Molded-In (requires tight ±0.05 mm hole control), fast 0.5–1.2 s cycle per insert; requires acoustic horn fixture and careful amplitude tuning, medium-to-high volume (25,000–500,000 pcs/yr) where gang thermal pressing is too slow. Molded-In Inserts (loaded in cavity), deep diamond knurl with wide hex or undercut collar + closed or blind pilot end, 100% Baseline (Highest) (molten resin flows under high pressure into deep undercuts), slows press cycle by 4–15 s unless robot-loaded; risks $3,000–$15,000 cavity steel crash if an insert drops during closing, high-load structural mounts (M5–M10), thermosets, or automated vertical rotary-table insert molding. Self-Tapping / Thread-Cutting Inserts (e.g., Ensat), external coarse cutting threads with cutting slots or 3 radial chip holes, 90%, 100% of Molded-In. Cold Press-In / Expansion Inserts, slotted expanding body or shallow broaching splines pushed into an unheated hole, 40%, 60% of Molded-In (relies purely on friction and hoop interference), fastest manual/arbor press installation, but generates high radial hoop stress that cracks brittle resins, non-critical covers in ductile PP, PE, or nylon where axial load is low

  2. Cocake, "Case Threaded Insert: Threaded Inserts vs Self-Tapping"
    What we used

    (KeXin New Materials, published 2026-11-01; the company's own test on its own cases in glass-filled PP, pulled at 5 mm/min), "An M4 case threaded insert pulls out at 1200 N against 380 N for a self-tapping screw, a 3.2x gap that only holds with an 8 mm boss." "Pulled at 5 mm/min, the brass insert leaves the boss whole at 1200 N; the self-tapping screw strips at 380 N, with polymer failing before hardware in both cases." "The insert carries load on a knurled shear band. At 6.3 mm diameter and 8 mm depth it presents 158 mm², turning 1200 N into 7.6 MPa, well under the 18 MPa glass-filled PP sustains." "A screw has far less area. The ST4.2 thread is 0.6 mm high, so 8 mm of engagement leaves roughly 26 mm², and 380 N already means 15 MPa, close to the ceiling." The table: M3 780/220/3.5, M4 1200/380/3.2, M5 1750/520/3.4, M6 2400/700/3.4. "The ratio holds between 3.2 and 3.5 because insert shear area scales with diameter while thread height does not; at M6 the screw gains 3 mm²." "Fixture stiffness skews readings more than specimens do: a flexing clamp adds 80 N to 120 N at 800 N, enough to separate two labs by 150 N." "The boss runs 1.25 times insert diameter: 6.3 mm becomes 8 mm with 0.85 mm of ring wall, and 1.5 mm of polymer stays under the hole." "On a 2.5 mm wall the boss is where sink appears first. A 3 mm root leaves a 0.08 mm dish visible under raking light; moving the boss 6 mm inboard or coring 0.3 mm at the root drops it below 0.03 mm." "A boss at mid-edge forces melt fronts to rejoin behind it, and that line faces the impact load, cutting drop energy 30% where it opens. Shifting the boss 15 mm, or widening it to 9 mm, moves the line into a lower-stress zone." "Holes split by process: 6.4 mm heat-set, 6.2 mm ultrasonic, 6.3 mm moulded-in. The clearance gives displaced melt room, the tight hole transmits amplitude, and the core pin is the hole." Heat-set: "A 270 ℃ tip presses 3 s and holds 5 s, eight seconds a joint, giving 1100 N and 4.5 N·m from a 400 USD hand tool." Ultrasonic: "At 20 kHz and 28 μm the cycle takes 0.5 s and pull-out reaches 1200 N, the best of the three; the cost is a 0.3 mm flash ring needing a 0.5 mm counterbore." Moulded-in: "Pins located to 0.02 mm let melt encapsulate the knurl for 1300 N and 5.2 N·m, but tooling rises 12% and 1.2% of parts are scrapped." The process table: heat-set 8 s / 1100 N / 4.5 N·m / 400 USD tool; ultrasonic 0.5 s / 1200 N / 4.8 N·m / 8,000 USD welder; moulded-in in cycle / 1300 N / 5.2 N·m / tooling up 12%. "Volume decides, not strength. Under 1,000 units heat-set allows late changes; above 10,000 units moulded-in cuts piece cost to 0.13 USD; ultrasonic suits a frozen design." Failure modes: "Across 200 warranty returns: pull-out 46%, spin-out 22%, boss cracking 19%, thread stripping 13%, each tracing back to a different dimension." "If an insert has moved 0.3 mm at 800 N, engagement is under 6 mm and half the knurl sits in air; raising boss height from 8 mm to 10 mm restored 1150 N." "Spin-out appears above 3.2 N·m because straight knurling cannot grip tangentially. Three ribs 0.4 mm deep, or a diamond knurl, push spin torque to 5 N·m." "Cracking is a design fault. Below 1.0 mm of ring wall, assembly expansion splits the boss along a 45-degree path; a R0.5 fillet and 0.85 mm of wall remove about 80% of cases." "Stripping is the screw's alone, since the thread is cut not moulded. Torque is down 40% by the fifth pass and 1.1 N·m by the eighth. Field repairs go one size up, reaming 3.6 mm to 4.1 mm and thinning the wall 0.25 mm, trading spin-out risk for cracking risk." Cycle count: "An M4 screw holds 380 N on the first pass, 228 N on the fifth and 152 N by the eighth, torque falling from 2.5 to 1.1 N·m; a brass insert still holds 1056 N at pass 25 and 840 N at pass 50." The band table: 1–3 cycles, screw 340 N, insert 1180 N, "Screw is fine"; 4–10, screw 190 N, insert 1120 N, "Insert or sleeved screw"; 11–25, screw stripped, insert 1056 N, "Insert"; 26–50, screw failed, insert 840 N, "Insert, diamond knurl". "Two thresholds settle the drawing: 10 or more planned cycles, or 800 N of required pull-out, mean inserts; fail both and a screw saves 0.09 USD a joint." "Cordless drivers swing torque ±20% and strip a screw by pass four, while a station limited to 4.5 N·m lets an insert reach 50." "Write both thresholds on the drawing and the case threaded insert choice stops being a debate: 10 or more assembly cycles, or 800 N of pull-out, means brass inserts; below both, self-tapping is cheaper. The M4 floors are an 8 mm boss, 0.85 mm of ring wall and 8 mm of depth, and losing any one costs roughly 200 N. Then pull three samples at 5 mm/min and reject any batch peaking under 900 N." FAQ: "What pull-out should an M4 case threaded insert meet? A: 900 N minimum against 1200 N measured, a 25% margin. Self-tapping joints are accepted at 300 N because they are not built for reuse." "How often can a self-tapping screw be reused? Three passes cleanly, 40% torque loss by the fifth, 152 N left at the eighth. Above 10 cycles, use an insert." "Is an 8 mm boss mandatory? For M4, yes. At 7 mm the ring wall thins to 0.35 mm and cracking probability in assembly rises to 30%." "Which process gives the highest pull-out? Moulded-in at 1300 N, ultrasonic 1200 N, heat-set 1100 N. Ultrasonic needs a 6.2 mm hole and a 0.5 mm counterbore." "What does the hardware add in cost? About 0.09 USD per M4 insert, 0.72 USD across eight joints, while the return rate falls from 6% to under 1%."

  3. PT Online, "Tips and Techniques: How to Get Best Results With Insert Fasteners" (Christie Jones, Spirol)
    What we used

    "The weakest sections of many plastic part designs are the joints and assembly points." "Metal threaded inserts significantly improve joint strength in plastic parts and are not themselves susceptible to creep." "Additionally, the inserts enable unlimited assembly/disassembly of the components without compromising the integrity of the threads." "Typical performance requirements for assemblies using inserts involve tensile strength, rotational torque, and pull-through strength. Tensile strength, or pull-out, is the axial force required to pull the insert out of the plastic material. Torque is the rotational force required to rotate the insert in the plastic material." "The objective is to choose an insert with sufficient torque resistance to accommodate the tightening torque necessary to achieve sufficient axial tension load on the threaded joint to keep it together and prevent loosening, while also achieving pull-out values necessary for the load conditions that the insert will be exposed to while in service. In general, resistance to torque is a function of insert diameter, and resistance to pull-out is a function of insert length." "Molded-in inserts usually yield the highest performance, yet this form of installation is by far the most expensive." "Inserts installed with heat or ultrasonics after molding yield good performance at a fraction of the assembled cost of the molded-in inserts." "Inserts installed with heat yield the best combination of overall performance and lowest installed cost." "Self-tapping inserts provide the best pull-out resistance for a post-mold insert, yet the assembler has to be very careful about installing the insert so that it is perfectly square to the hole or it will not mate well with the screw." "Expansion inserts are designed for non-critical applications where ease of installation is the primary design criterion, not torque and pull-out resistance." "Correct hole size is critical." "The insert manufacturer's recommended hole size for the insert must be adjusted if fillers are used in the plastic." "Holes for post-mold inserts should always be deeper than the length of the insert. For self-tapping inserts, use a minimum depth of 1.2 times the insert length. For other inserts, the recommended depth is the insert length plus two thread pitches." "Insert performance is affected by the plastic boss diameter and/or wall thickness. Generally, the optimum wall thickness or boss diameter is two to three times the insert diameter, with the relative multiple decreasing as the insert diameter increases."

  4. PT Online, "Four Ways to Tackle Threaded Inserts for Plastics" (PennEngineering SI)
    What we used

    "The key strength factors are resistance to the insert pulling out of the part (pullout force) and resistance to the insert twisting in the part when the mating fastener is torqued (torque-out force). The longer the insert, the greater the pull-out resistance; the greater the diameter of the insert, the more torque capacity. The knurl pattern is also critical." "Two insertion processes, ultrasonic and heat-staking, involve re-melting previously molded plastic, so they only work with thermoplastics. For thermosets, the options are molding-in the insert, or cold pressing the insert in later, where elasticity of the resin material will become critical." "The most common material for the threaded inserts themselves is brass. However, with increasing sustainability requirements, lead-free alternatives such as stainless steel or aluminum are growing more popular. Stainless steel offers better strength and corrosion resistance, and aluminum inserts are about 70% lighter than brass." "Cost, Because threaded inserts can be molded into the part at the outset, or pressed into the part later, consider the *total* cost of the operation. That includes molding time and cost, component handling, and assembly." Heat staking: "Great for: Thermoplastics; High overall performance and low installation cost." "With heat staking, the insert is heated with a thermal press to a point where the insert then melts a small portion of the hole periphery as it is being pushed into the part. The softened resin flows into the knurl pattern and then hardens to form a firm bond with the insert."

  5. PT Online, "Brass Insert Installation: Comparing Heat vs. Ultrasonic vs. Press-Fit"
    What we used

    "Pull-out strength and torque retention do not come from the brass. They come from plastic locked into the knurls, grooves and undercuts on the outside of the insert. If you can get plastic to flow around those features and solidify there, then you have created a mechanical interlock that resists both pulling out and spinning." "Thermal and ultrasonic methods both reflow plastic into the knurls. Press-fit, on the other hand, skips the melt entirely and relies on interference alone." Thermal insertion: "a temperature-controlled thermode heats the insert by conduction. The plastic around the hole softens; the insert is pushed home; displaced material reflows into the knurls; and then a cooling stage locks everything in place before the tip retracts. Heat in, cool, retract." "Thermal insertion's strength is latitude. This process handles the broadest range of materials, amorphous and semicrystalline alike, and tolerates filled resins and awkward boss geometry better than the alternatives. It will set several inserts in a single press stroke, and with a proper cooling stage, it leaves a clean, flash-free face. One important rule travels with it: the temperature is set by the polymer, not the press." "The trade-off with thermal insertion is time. Because the cycle includes both a heating and a cooling phase, this process is rarely the fastest method per insert." Ultrasonic insertion: "the insert is driven by a horn vibrating around 20 kHz. Friction at the plastic-to-metal interface generates heat in a fraction of a second, the local plastic melts; the insert sinks; and the melt solidifies around the knurls almost immediately. The cycle is genuinely quick, sub-second per insert, which is why it wins on high-volume lines."

  6. Engineering Product Design, "Threaded Inserts for plastics | Types, Applications & Characteristics"
    What we used

    "They provide unlimited reusable thread, unlike self-tapping thread. Increased seating torque for applications such as compressing o-rings and gaskets to seal a product assembly. They are not affected by the plastic material stress relaxation or self-tapping screw creep." "Knurling features provide torque resistance. Opposing knurls or diamond knurling and serrations provide pull-out resistance. Length is proportional to the pull-out resistance. Undercuts and grooves provide pull-out resistance."

  7. Sculpteo, "Pull-out resistance of threaded inserts: Testing and Results"
    What we used

    (heat-set brass inserts in MJF PA12, five specimens per configuration, installed with a temperature-controlled soldering iron), M2 short 477 N, M2 long 572 N, M2.5 825 N, M3 short 813 N, M3 long 1,258 N, M4 1,372 N, M5 over 1,800 N, M6 over 1,800 N. "Pull‑out resistance of threaded inserts is a key design parameter for durable, serviceable 3D printed assemblies, especially when screws are loaded in tension rather than shear."

  8. ZetarMold, "What Is Insert Molding? Complete Guide for Engineers"
    What we used

    "Designing for insert molding requires wall thickness ≥1.5× the insert diameter for proper fill and bond strength." "Threaded inserts improve fastener durability over self-tapping screws in plastic."

  9. Assembly Magazine, "Fasteners: Driving Into Plastic...Edge-Wise"
    What we used

    "It took several years, but self-tapping screws for plastic assembly have come of age." "A thread-forming screw has a blunt end and is used in thermoplastic having a flexural modulus from 150,000 psi to 400,000 psi. As it is driven down into the boss, the screw pushes the walls of the plastic hole into the shape of threads, producing no chips." "A thread-cutting screw has a shank slotted point and is used in thermoset materials having a flexural modulus from 500,000 psi to 2 million psi. The screw cuts the material and forms the thread as it is driven down, producing chips that fall out of the hole." "When joining plastic parts, assemblers must take care to not strip the threads formed in the plastic. The best way to achieve this is adherence to the fastener's strip-to-drive ratio, which is the ratio between the torque at which the screw will strip its [threads and the torque at which it drives]." "The biggest challenge is to make sure the fastener will not crack or leave any structural or cosmetic defect, such as white marks, on the boss."

  10. Brassland, "Complete Guide to Threaded Inserts for Plastics"
    What we used

    "Plastic alone cannot hold a reliable, re-usable thread. Strip a moulded boss once, and it is gone. That is the problem threaded inserts solve: they create a permanent, metal-quality thread in a plastic host, a thread that can be assembled and disassembled hundreds of times without degradation." "The three-flute variant is the most capable: the flutes remove material cleanly rather than just displacing it, which reduces hoop stress on the boss and makes the design suitable for brittle thermosets and filled engineering plastics as well as softer thermoplastics."

  11. Demo arithmetic verified in Python before writing: screw
    What we used

    Pass 1: 380 N, 2.5 N·m; pass 5: 228 N, torque down 40%; pass 8: 152 N, 1.1 N·m (56% down from 2.5 N·m); pass 11–25: stripped. Insert, pass 1: 1,180 N; 4–10 cycles: 1,120 N; pass 25: 1,056 N (89.5% of pass 1); pass 50: 840 N (71.2% of pass 1). Insert vs screw at pass 8: 1,120 / 152 = 7.4×. Shear: insert 1,200 N / 158 mm² = 7.6 MPa, 42% of the 18 MPa glass-filled PP sustains; screw 380 N / 26 mm² = 14.6 MPa, 81% of ceiling; shear area 158 / 26 = 6.1×. Boss: 6.3 mm insert → 8 mm boss = 1.27× insert OD, 2× the M4 thread diameter; ring wall (8 − 6.3) / 2 = 0.85 mm; PT Online boss rule 2–3× insert diameter → for an M4 thread (4 mm): 8–12 mm boss OD, Cocake's 8 mm at the bottom of the band; ZetarMold wall ≥1.5× insert diameter → 1.5 × 6.3 = 9.45 mm for insert moulding. Hole sizes: heat-set 6.4 mm, ultrasonic 6.2 mm, moulded-in 6.3 mm, insert OD 6.3 mm, heat-set clearance 0.1 mm. Cocake process test: heat-set 1,100 / 1,300 = 84.6% of moulded-in (PTA band 85–95%); ultrasonic 1,200 / 1,300 = 92.3% (PTA band 80–90%, Cocake sits above it). Install time per hub (8 joints): heat-set 8 s × 8 = 64 s; ultrasonic 0.5 s × 8 = 4 s; moulded-in in cycle. Hardware cost: 8 × $0.09 = $0.72 per hub; fleet of 1,000 = $720. Returns: 6% of 1,000 = 60 vs under 1% = under 10; 50 fewer returns; at $150 per field replacement (labelled assumption from #43): 60 × $150 = $9,000 vs under 10 × $150 = under $1,500, saving $7,500 against $720 of inserts = 10.4× payback. Acceptance: pull 3 samples at 5 mm/min, reject any batch peaking under 900 N; 900 / 1,200 = 75% of measured, 25% margin. Boss floors: 8 mm boss, 0.85 mm ring wall, 8 mm depth; losing any one costs roughly 200 N; at 7 mm boss the ring wall is 0.35 mm and cracking probability rises to 30%. Spin-out: straight knurl spins above 3.2 N·m; three ribs 0.4 mm deep or diamond knurl pushes spin torque to 5 N·m = 1.56×. Cracking: below 1.0 mm ring wall the boss splits along a 45-degree path; R0.5 fillet + 0.85 mm wall removes about 80% of cases. Sink: 3 mm root leaves a 0.08 mm dish; coring 0.3 mm at the root drops it below 0.03 mm = 2.7× reduction. Weld line: boss at mid-edge cuts drop energy 30% where it opens; shifting 15 mm or widening to 9 mm moves the line. Sculpteo MJF PA12 heat-set: M4 1,372 N vs Cocake glass-filled PP 1,200 N = 1.14×; M3 long 1,258 N vs Cocake M3 780 N = 1.61× (different material, different insert length)

More in design notes

The Seal That Actually SealsYour enclosure passed IP65 on the bench and leaks in the field, because the gasket took a set after two years of temperature cycling and the joint designed at 30% compression is now working at 15%. The seal is a set of numbers, not a material.26 minTolerance Stack-UpEvery part is a little off, and the little offs add up. Ten parts at ±0.1 mm stack to ±1.00 mm worst case and ±0.32 mm statistically, and the difference between those two answers is the difference between a product that assembles and a product that doesn't. The fix is rarely a tighter tolerance. It is usually a shorter chain.14 minPower Design for Battery ProductsThe battery is a budget, not a box. A field sensor on a CR2032: the headline budget says 513 days, the measured budget says 44. The gap is sleep current, regulator losses, and a battery that is not a constant, and the fix is arithmetic, not a bigger cell. The power budget, the BMS, and the rules that make battery life honest.14 minThe Battery's BrainA battery management system is three things at once: the eyes that measure every cell, the accountant that estimates how much charge is left, and the guard that disconnects the pack before a fault becomes a fire. This is how to design each one, the measurement accuracy that decides how much capacity you waste, the SOC estimation that has no sensor to read, the balancing choice, and the protection thresholds with real margin, with a worked demo on a 48 V pack.28 minThe Pack That Doesn't Drift ApartThe pack is rated 168 watt-hours and delivers 160, because the fullest cell ends every charge and the emptiest ends every discharge, and the 8 watt-hours in between are locked away behind the weakest cell.23 min