Arc Flash Analysis Step-by-Step Instructions for Vietnam - Carelabs Vietnam

Most arc flash studies that produce unreliable labels do not fail in the calculation. They fail in the preparation. The IEEE 1584 model is well defined and the software is mature; what varies between a study you can rely on and one you cannot is the quality of what goes into it.

This is a practical guide to running the work in a Vietnamese industrial facility, organised around the stages where it actually goes wrong.

Before the engineer arrives

Find out what your single line diagram really describes. In most plants the drawing set records the installation at handover, not the installation today. Extensions, temporary supplies made permanent, and transformer changes accumulate without the drawings following. The study will reconcile this, but knowing in advance that the drawings are ten years old changes the survey time you should budget for.

Request your fault level from EVN early. This is the single most common cause of delay. The figure needed is the fault level at your point of connection, as both a maximum and a minimum. Requests can take weeks, and the analysis cannot be completed without them, so this should be raised at the point of enquiry rather than after mobilisation.

Identify which panels are genuinely opened live. Not every enclosure needs assessment. What needs assessment is anywhere a person works energised: testing, fault-finding, racking, thermographic survey. Sites frequently under-report this, because work that has always been done a certain way stops being noticed.

Locate the protection settings records, and expect them to be wrong. They will be checked against the devices regardless. Knowing whether a record exists at all determines how much time the protection audit takes.

During the site survey

The survey collects what the model is sensitive to, and the sensitivities are not intuitive.

• Enclosure dimensions and electrode configuration. IEEE 1584-2018 treats a vertical electrode arrangement inside a box differently from horizontal electrodes or an open-air configuration. Choosing the wrong one moves the result substantially.
• Working distance. Incident energy falls off sharply with distance, so the working distance recorded must be the distance the technician's torso actually occupies, not a standard figure.
• Gap between conductors. Determined by equipment class and taken from the installation rather than assumed.
• Nameplate data. Transformer rating and impedance, breaker interrupting capacity, cable sizes and lengths.

A survey that records only nameplate data and skips enclosure geometry will produce a number. It will not produce a trustworthy one.

Where the analysis goes wrong

Using assumed fault current. Arc flash energy derives from bolted fault current, which comes from a short circuit study to IEC 60909. Where a study is run on estimated fault levels, the labels carry the authority of a calculation without the basis for one.

Ignoring the minimum fault case. This is counter-intuitive and it matters. Lower fault current can push the arcing current below a protective device's instantaneous threshold, so the device that would have cleared in a few cycles now takes seconds. Lower current, far higher energy. The worst case is frequently the minimum source condition, not the maximum.

Taking clearing time from the settings schedule. The schedule and the relay disagree more often than not. Clearing time has more influence over the result than almost any other input, so it is read from the device.

Applying one working distance across the whole site. LV switchgear, MCCs and HV panels are worked at different distances, and applying a single figure produces results that are wrong in both directions.

After the report

A study becomes useful at the point the label is applied and understood, which is where Vietnamese facilities with mixed local and expatriate engineering teams often need more attention than the calculation received. A label the technician cannot read is not a control. Bilingual labelling and a briefing in the working language of the floor are part of the deliverable, not an optional extra.

The obligation behind this sits in the Law on Occupational Safety and Hygiene 2015, which requires the employer to identify and control workplace hazards, alongside TCVN 7447 for the installation itself and QCVN 01:2008/BCT for electrical safety. DOLISA assesses whether the hazard was identified. A study on a shelf does not demonstrate that; a labelled installation with trained staff does.

Keeping it valid

The result is a snapshot of a specific network with specific protection settings. Change either and the labels stop describing reality. An EVN reinforcement, a transformer upgrade, added generation, a switchgear extension or a protection setting change should each trigger a re-run of the affected positions. Where the calibrated model has been retained, that re-run is short. Where it has not, the whole survey is repeated. This is the main reason the model is worth insisting on as a deliverable.

Related reading: short circuit analysis establishes the fault current this study depends on, and relay coordination governs the clearing times that drive the result.