Safety interlock switches are guard-mounted devices that prevent machine start-up or hazardous motion while a movable guard is open, and — in guard-locking variants — keep the guard itself locked until the machine has run to a safe state. They sit under ISO 14119 (interlocking devices associated with guards) and are typically designed into safety chains that meet ISO 13849-1 up to PLe or IEC 62061 up to SIL 3 [S4][S7].
On the bench, the device is deceptively simple: an actuator key, a cam or coded target, and a block of positively-driven contacts. The honest engineering work lives in the integration — risk assessment, mounting geometry, and the wiring that converts a mechanical event into a stop signal. Misuse is the failure mode, not the switch itself.
Where the switch adds value — concrete benefits
Guard-locking interlock switches are specified when the hazard has a run-down time: the door must stay shut until the spindle, blade, or robot arm has stopped, even if an operator pulls the handle. Tongue-operated designs such as the Schmersal AZ16 series or the IDEC HS series use a positive-mode cam that mechanically forces the NC contacts open — a failure mode the standard recognises as "positively driven" rather than spring-dependent [S9][S5].
Mechanically, the published numbers are tight and verifiable. IDEC's interlock line carries a direct-opening travel of 8 mm minimum, direct-opening force of 60 N minimum, thermal current (Ith) of 2.5 A, rated insulation voltage (Ui) of 300 V, and enclosure rating of IP67 per IEC 60529, with an operating window of –25 to +70 °C [S5]. Eaton's safety interlock line quotes a B10d of 20,000,000 cycles — the duty-life figure that feeds directly into an ISO 13849-1 PFH calculation [S7]. For the mounting itself, the IDEM KLP datasheet specifies 4.0 Nm on M5 fixing bolts, 1.5 Nm on the lid and head bolts, and a 3 mm actuator-to-stop gap when the guard is closed [S1].
Non-contact RFID-coded versions (to ISO 14119 type 4) sidestep the wear problem of mechanical tongues, resist defeat with low-level tools, and tolerate up to several millimetres of misalignment. Mechanical tongue switches remain the lower-cost, easier-to-replace option for Category 1, 2, and 3 circuits that do not need coding [S4][S7].
Where the switch gets it wrong — honest limitations
Every mechanical interlock has a weak point: the actuator key. The Schmersal literature is explicit that actuator keys are tamper-resistant, not tamper-proof, and a high-bypass guard can still be defeated with a second key, a bent pin, or — on poorly-installed units — by prying the door against the switch body [S9]. The mitigation is mechanical: a separate door-mounted stop, not the switch face, must take the closing impact, and alignment guides must keep the actuator from side-loading the aperture [S1].
Electrical contacts have a finite life. Even at 20 million B10d, an application that cycles the guard every 10 seconds runs the mechanism through ~3 million operations per year, which is the territory where preventive replacement schedules, not warranties, decide uptime [S7].
Environment also bites. IP67 keeps out splash and brief immersion but is not a substitute for stainless heads and food-grade housings in washdown pharmaceutical lines. Operating temperature is bounded — IDEC quotes –25 to +70 °C operating and –40 to +80 °C storage, with relative humidity 45–85 % non-condensing [S5]. Outside that envelope, condensation freezes inside the head and the positively-driven contacts stick. The Banner and Rockwell summaries both flag that interlock switches do not, on their own, satisfy every safeguarding need: presence-sensing safety mats, light curtains, and two-hand controls cover hazards that a door switch cannot [S6][S8].
Options lined up against the criteria that actually matter
The selection question is rarely "which brand" — it is which sub-type fits the risk graph. Across the four dominant families, the engineering trade-offs look like this:
1. Tongue / mechanical (ISO 14119 type 1 or 2). Lowest cost, simplest wiring, B10d typically 1–20 million, force-guided NC contacts, positive mode operation. Best for hinged or sliding guards on Category 1–3 circuits; not suitable where low-level defeat is foreseeable. Schmersal AZ16, Rockwell MT-GD2, IDEC HS1L, Eaton LSM [S3][S7][S9].
2. Hinge-pinned. The switch replaces the hinge pin; the door cannot be opened without breaking the contact. Compact, but every door needs its own switch and adjustment is limited. Common on small panels and electrical enclosures [S3].
3. Guard-locking (solenoid or spring). Used when run-down time is non-zero — presses, robot cells, mixers with rotating agitators. Holding force typically 1,000–2,500 N; either energise-to-release (fail-safe) or energise-to-lock (fail-secure) variants. The IDEM KLP and Schmersal AZM161 sit in this class [S1][S9].
4. Non-contact coded (RFID or magnetic, ISO 14119 type 3 or 4). Highest defeat resistance, highest misalignment tolerance, no mechanical wear, but the highest unit cost and a separate coded actuator that must be specified as a matched pair. Mandatory under ISO 14119 type 4 when a fault-masking analysis shows a single mechanical switch can be defeated with common tools [S4][S7].
Standards the designer actually invokes
ISO 14119 is the master document for "interlocking devices associated with guards" — it defines the four types, the coding levels (low to high), and the defeat-resistance logic that decides whether a mechanical or coded device is required [S4][S8]. The functional safety side is covered by ISO 13849-1 (Performance Level a–e) and IEC 62061 (SIL 1–3), with IEC 60947-5-1 governing the electromechanical contact block itself [S5][S7]. On the North American side, ANSI/NFPA 79 (industrial machinery electrical), ANSI B11.19 (safeguarding performance criteria), and ANSI/RIA R15.06 (industrial robots) are referenced in parallel [S8].
Installation discipline is not optional. The IDEM KLP datasheet spells it out: a risk assessment for the application, competent personnel, 4.0 Nm on M5 fixing bolts, 1.5 Nm on lid and head bolts, 0.7 Nm on terminal screws, a separate mechanical door stop, a 3 mm actuator gap, weekly functional check, and replacement of any unit with a bent actuator or damaged head [S1]. Skipping the stop is the single most common field failure I have seen — the guard is closed hard, the switch face takes the impact, and the positive-mode cam goes out of alignment within a year.
Where it fits, and where it should be left out
Interlock switches are the right call for movable guards around defined mechanical hazards — presses, mixers, conveyors with accessible nip points, robot work envelopes with restricted access, and electrical cabinet doors that must not be opened under load. They are not the right tool for open-process hazards, perimeter access control, or anywhere a person can reach the hazard before a door is involved; those cases want a light curtain, safety mat, or area scanner instead [S8].
For engineers wiring the downstream safety relay, the safety relay installation guide covers the contact expansion and EDM feedback that the switch's NC/NO pair feeds into. A broader map of safety interlock switch types and ISO 14119 classifications is useful when the decision is between type 1, 2, 3, or 4 — the type drives the coding and the defeat-resistance argument on the risk-assessment form. For background on the device family as a whole, the safety interlock switch encyclopedia entry consolidates the standards and contact terminology.
Trackable signals to watch next
Two engineering signals are worth monitoring over the next planning cycle: (1) wider adoption of ISO 14119 type 4 RFID-coded switches in mid-range Category 3 / PL d applications, driven by the cost gap to mechanical type 2 narrowing; (2) tighter harmonisation between ISO 13849-1 PFH calculations and the B10d figures manufacturers publish — at present, the 20,000,000-cycle B10d rating has to be converted to a PFH value via the formulas in ISO 13849-1:2023 before the safety integrity claim is auditable [S7].
Component reference pages worth checking: safety barrier, and safety fence.