A machine guard fails in one of three distinguishable ways, and a photograph of the aftermath rarely tells them apart. The guard was never adequate for the hazard it faced. It was adequate and had been removed. Or it was in place and its interlock had been made to report a state the machine was not in. Those findings point at three different parties — the builder, the employer, whoever performed the last service — and the difference is written in hardware, wiring and control logic well before it appears in testimony.
Three failures that photograph identically
Inadequacy is established by comparing the guard as built against the hazard it controls: opening size, reach distance, the stopping performance behind it. Removal is established at the mounting — fastener condition, thread wear, witness marks. Defeat is established at the interlock, and it is the most consequential, because it means the machine ran while the guard was open.
The three also coexist. A guard awkward by design invites removal, and a guard removed often enough gets a bypass fitted so removal stops interrupting the line.
What the safety function was designed to guarantee
Safeguarding is a designed function with a specified reliability, not simply a barrier. ISO 12100 sets the risk assessment framework; ISO 13849 expresses how far the safety-related parts of a control system are meant to be trusted, in performance levels. The ANSI B11 series carries the equivalent obligations in the United States — B11.0 for risk assessment, B11.19 for safeguarding performance criteria.
So the design record states what the machine should have done when the guard opened, and with what tolerance for a single fault. Where the installed arrangement was single-channel, or its monitoring disabled, it no longer met the level the assessment called for.
How a bypass leaves a signature
Defeat is usually cheap and improvised, and improvisation leaves traces. A spare coded actuator taped inside the enclosure. A magnet resting against a reed switch. A jumper across interlock terminals in a wire gauge matching nothing else in the cabinet.
Photograph the arrangement before anything is disconnected; a jumper is more persuasive in place, terminal numbers visible, than in an evidence bag. Wiring that departs from the machine's own schematic is a clean finding: the schematic states what the builder intended, the deviation what somebody chose.
ISO 14119, which governs interlocking devices associated with guards, treats defeat as foreseeable behaviour to be designed against. The question is not only who fitted the jumper, but whether the safeguard made fitting one likely.
The controller knows what the guard reported
An interlock is an input, and inputs are visible to the control system. Safety controllers and safety relays often retain reported guard state, discrepancy faults between redundant channels, and reset events as diagnostics. A door logged as closed through a continuous production run, while its actuator sat taped to a bracket, is not an ambiguous record.
The same data establishes duration, and a bypass fitted minutes before an injury implies something different about supervision than one in place for months.
Motivation belongs in the design record
Bypasses are rarely acts of indifference. They cluster around tasks the safeguarding made difficult: clearing jams, threading material, adjusting a fixture, observing a process that must run to be observed. When a guard is opened many times a shift and every opening costs a restart sequence, defeat becomes the path of least resistance.
That pattern is evidence about design and work organisation, not only about an individual, and it shows up in downtime and jam records.
Adequacy is a reach-and-distance question
A guard that is present can still be inadequate. Fixed guards are assessed against opening size versus distance to the hazard, and against reach over, under, around and through. The safe-distance tables in ISO 13857 and the criteria in ANSI B11.19 exist because intuition performs badly here. OSHA's requirements at 29 CFR 1910.212 and 1910.219 set the regulatory floor.
The measurement is straightforward and frequently never taken. Documenting the actual opening, distance and reach path converts an argument into arithmetic against a published table.
Presence-sensing devices fail in their own ways
Light curtains, mats, scanners and two-hand controls depend on the machine's stopping performance. Required safety distance is a function of measured stopping time, and stopping time degrades as brakes, clutches and valves wear. A curtain mounted at a correct distance on commissioning day can be too close years later without anything visibly changing.
Muting and blanking are configurable and easily widened to stop nuisance trips; whether the configuration still matched the risk assessment is recoverable from the device parameters. Robot cells raise the perimeter and teach-mode questions ANSI/RIA R15.06 covers.
The paper record around the hardware
Guard complaints, work orders mentioning a jammed or removed guard, prior near misses, and internal audit findings establish knowledge. A defeat documented beforehand and one nobody could have known about produce very different outcomes on identical facts. The builder's manual and residual-risk information are the other half: what the manufacturer warned about, and what the employer did with the warning.
Where these opinions are challenged
Predictably: that the bypass was fitted after the incident, that inadequacy is asserted rather than measured, that the control-system record was collected only after the machine had been cycled, and that the reconstruction assumes a body position no evidence supports.
Work that photographs the arrangement in place, measures rather than estimates, and separates what the controller proves from what it merely permits survives that. A conclusion resting on one artefact does not.
This article is general technical orientation, not a failure analysis, an engineering opinion, or advice on any specific matter. Determining the cause of a particular incident requires hands-on examination by a credentialed expert.