A functional check that was not a functional check
At 06:15 a gas tester performs a functional (bump) check on a four-gas personal monitor before a shift of confined space standby work. He uses a cylinder labelled 25 ppm H2S, 100 ppm CO, 2.5% CH4 and 18% O2; the certificate is nine months old against a stated twelve-month life, and the cylinder pressure gauge sits in the red zone. He does not fit the demand-flow regulator's cup over the sensor; he holds the open regulator nozzle near the instrument in still air for about five seconds. The display flickers: H2S 11 ppm, CO 40 ppm, CH4 1.1%, O2 19.2%. He marks the check as pass because all the channels moved, and signs the register with the cylinder number. During the shift the instrument reads 20.4% oxygen in fresh air outside the vessel. The vessel entry limit is 19.5% oxygen and the entrant is ready to go in.
1. Assess the validity of this functional check. Identify each defect and state what it does to the reliability of the readings.
Read the worked answer
The check is invalid, and the pass was wrong on the tester's own numbers.
- Cylinder pressure in the red. Below the regulator's minimum operating pressure the delivered concentration is unreliable, and the composition can shift as the cylinder empties. H2S in particular decays and adsorbs onto regulators and lines, which is why reactive mixtures have the shortest lives.
- An ageing standard. Nine months into a twelve-month life is still in certificate, but a near-end-of-life, low-pressure reactive cylinder is doubly suspect. Had the certificate expired, the check would be void outright.
- No demand-flow cup. Holding an open regulator near the sensor depends on gas reaching the sensor undiluted. In still air most of it does not, so the instrument is sniffing a mixture of test gas and room air. The displayed values are not traceable to the cylinder label and cannot verify span.
- Five seconds is nowhere near long enough to see a sensor's full response, still less to confirm the alarm function.
- The recorded numbers would have failed a 10% tolerance on every channel: H2S 11 against 25 ppm, CO 40 against 100 ppm, CH4 1.1 against 2.5%, O2 19.2 against 18.0%. Marking it pass because the numbers moved turns a functional check into decoration.
A valid check exposes each sensor through the correct regulator and cup, allows full response, compares the display with the label within the manufacturer's tolerance, and confirms the alarms.
2. The instrument reads 20.4% oxygen in fresh air during the shift. Is it fit to support an entry decision against a 19.5% limit?
Read the worked answer
No. Fresh air is 20.9% oxygen and the instrument is reading 0.5% low, which is at or beyond the usual ±0.5% accuracy tolerance for an electrochemical cell. Possible causes all matter: a partly consumed or poisoned cell, a temperature shock from moving between a warm vessel and cooler outside air, a blocked or wet filter, or a sensor at end of life.
The entry limit is 19.5%. With a display that is half a percent low, a reading of 20.0% could mean 20.5% or 19.5% - the safe side of the line cannot be distinguished. Worse, electrochemical cells commonly fail by drifting low and then collapsing quickly, so a cell already 0.5% out may not survive the shift.
Actions: remove it from service; perform a fresh-air calibration in clean outdoor air (not inside the vessel, not in a vehicle); then a full span calibration with in-date gas. If it will not calibrate to 20.9 ±0.1%, replace the cell. Issue a calibrated instrument. Treat readings taken since the start of shift as suspect, and re-test with the replacement before anyone stays in or goes back into the vessel.
Do not mentally apply a correction. A meter whose reading you are adjusting in your head is a meter you cannot rely on during an emergency.
3. What should the register have recorded, and what should have happened at the bench?
Read the worked answer
Record for every check: date, time, instrument make, model and serial, firmware and sensor set, cylinder number and certified concentration of each component, cylinder expiry, regulator and cup used, the displayed value on every channel, pass or fail against tolerance, the alarm test result, technician name and signature, and the action taken after any failure. The shift entry record should cross-reference the instrument serial and the most recent successful bump test.
What should have happened: replace the cylinder with an in-date one of adequate pressure; fit the demand-flow regulator and cup; allow the manufacturer's warm-up time; apply gas until each reading is stable (typically 30–60 seconds per channel, per the manual); compare with the label within tolerance; test the audible, visual and vibrating alarms; on any failure, carry out a full calibration, and if calibration fails, quarantine the instrument and label it clearly.
Two further points. A bump test confirms response, not accuracy; it does not replace the periodic full calibration (commonly every six months for catalytic and electrochemical sensors, following the manufacturer). And a record showing pass when the numbers were outside tolerance is worse than no record at all, because in an investigation it becomes false assurance that the atmosphere was ever verified.