Self-Tapping Screws: Setting a Driving Torque That Actually Holds

A self-tapping screw can clear incoming inspection and still split heads on the assembly line. Nothing is defective. The part met its minimum. The process was simply set from the wrong number.

The confusion comes from treating one figure as two things. A standard breaking torque is a floor for the part — the level below which a screw is not acceptable. A driving torque is a setting for the process — the level at which the tool stops. They are related, but they are not the same value, and the gap between them is where assembly problems live.

This article is about closing that gap. If you want the derivation of the standard numbers instead, that is a separate question with a short answer: they were measured, statistically floored and thBonnieen frozen — not calculated.

Three torques, one screw

Every self-tapping joint involves three distinct thresholds, and only one of them is written into the product standard.

The standard constrains the third. Your line is governed by the first two.

A screw with a high breaking torque driving into 0.6 mm mild steel will strip the sheet long before the screw is stressed — the failure is in the panel, not the fastener. Conversely, a screw with a marginal core hardness will snap in a thick, hard joint where the sheet would have held. Same screw, two different failures, decided by the joint rather than the part.

What the standard value is — and is not
GB/T 3098.5 and ISO 2702 both specify a minimum breaking torque per thread size, and both define it by test rather than by formula. The procedure is short:

Clamp the shank in a split die or blind-hole die matched to the thread, with at least two full threads engaged.
Apply torque with a calibrated measuring device until the screw fractures.
Accept the lot if the maximum recorded torque meets or exceeds the tabulated value.
Reference values from GB/T 3098.5-2025:

Two things follow from how that table was built, and both matter commercially.

It is a floor, not a target. Passing means "not below the minimum". It says nothing about the upper end, nothing about consistency between lots, and nothing about whether the screw suits your joint.

It barely moves between editions. ST9.5 was added in the 2016 revision; a handful of small sizes were trimmed slightly in 2000 to align with ISO 2702. Otherwise the numbers have been stable since 1985. That stability is the signature of a threshold set by consensus on what industry can reliably achieve — not a value derived from first principles.

Why you cannot calculate it

Bolts have a closed-form calculation: breaking torque as torsional shear strength times torsional section modulus, with the strength ratio depending on property class (GB/T 3098.13 / ISO 898-7). It is tempting to run the same arithmetic on a self-tapping screw.

It does not transfer, for one structural reason: a bolt is through-hardened and fails in clean shear, while a self-tapping screw is case-carburised — hard outside, tough inside — and fails by thread deformation, brittle fracture or torsional rupture depending on which zone gives first.

That difference is not a correction factor. Case depth and breaking torque are not monotonically related: push the carburised layer too deep and the screw becomes brittle, so strength can fall while surface hardness rises. No two-variable formula reproduces a non-monotonic relationship.

For buying decisions, this has a practical consequence: any torque figure you derive yourself is an estimate, not a specification. Use it to sanity-check a supplier's report; never write it into an acceptance criterion.

A sanity check you can actually run

There is one legitimate use for the calculation: testing whether a reported value is plausible.

Take the bolt arithmetic, invert it, and solve for the tensile strength that would produce the torque your supplier reports. If the answer sits far outside the strength band implied by the specified core hardness, the report deserves a second look.

The spread is the finding. The standard's own values behave as though each size carries a different effective strength — roughly 820 to 1100 MPa across the range, consistent with a core hardness window of 270–390 HV but not with a single nominal figure applied uniformly.

Two uses follow. First, a supplier who reports the same percentage over the minimum on every size is reporting a calculation, not a test — ask for the raw data. Second, blanket torque limits copied across a product family are wrong at both ends: generous enough to pass a weak ST2.2, strict enough to reject a sound ST6.3.

From standard value to line setting

Four steps, in order.

1. Fix the joint before the screw. Material grade, sheet thickness, hole diameter and engagement length determine stripping torque far more than the fastener does. A torque setting validated on 1.5 mm cold-rolled steel means nothing on 0.8 mm aluminium.

2. Measure, do not compute. Run 30 pieces to failure in the actual joint with the actual tool. Log two curves: torque at seating, and torque at stripping. The distance between them is your usable window.

3. Set below the window, not inside it. Allow for driver repeatability, bit wear and material variation between lots. A shut-off tool's accuracy is quoted as a percentage of setting — read your tool's spec, then subtract that plus a margin for the joint. If you are setting torque within a few percent of the standard minimum, you are not controlling the process; you are gambling on the tail of a distribution.

4. Control the tool, not just the number. Calibration interval, driver speed, bit condition and operator technique all shift delivered torque. A calibrated driver with a worn bit is an uncalibrated driver.

Three lines to put in your purchase order

Most torque problems are traceable to a specification that was never written down. These three close the gap:

• Breaking torque ≥ the value for that thread size, tested to GB/T 3098.5 or ISO 2702, with the test method and instrument named in the report.
• Core hardness 270–390 HV, with case depth specified as a range — not left to the heat treater's discretion.
• Lot-level test data, not a single certificate reused across shipments.

One caution before you write any of this into an inspection document: confirm the figures against the current edition of the standard you are citing. The values above are from GB/T 3098.5-2025; earlier editions differ slightly on small sizes.