FOD-Detection.com

Changi rebuilt iFerret itself, and cut false alarms more than tenfold

Here is the answer to the question this site left open in 2020, and it is not the answer we expected.

Changi Airport Group did not replace iFerret when its supplier collapsed. It took the system over and made it better. Working with its technology partner NCS, CAG’s engineering team spent more than fifteen months reworking the detection algorithm. The result, iFerret 2.0, cut false alarms by more than ten-fold. Changi has not restated a detection accuracy figure for 2.0; the above-95-per-cent number in circulation was published for the original system. It went in first on Changi’s new Runway 3, with deployment to Runways 1 and 2 to follow.

The failure mode they attacked is the specific one that has always plagued electro-optical detection at this site’s favourite airport: standing water. Reflections off puddles on a wet runway generate detections that are not objects, and Singapore has 170 or so rain days a year.

Liu Yanling, senior vice president of airside operations at CAG:

iFerret is an important system that helps safeguard every flight’s take-off and landing at Changi Airport.

Three things follow from this, and they matter more than the announcement.

The first is that the vendor was not the system. Stratech built iFerret, sold it, went under, and the product got materially better afterwards, in the hands of the customer. Every airport evaluating a detection system worries about supplier longevity, and rightly so. Changi has just demonstrated that the risk is survivable if the airport has — or can buy — the engineering capability to own the software. Most airports cannot. That, and not the sensor, is the real dependency.

The second is what they chose to fix. Not detection range. Not scan time. Not resolution. False alarms. Because a detection system that cries wolf on every wet morning is switched off, ignored, or its alerts are cleared without an inspection, and a system in that state has a probability of detection of zero no matter what the datasheet says. The whole argument of this site is that probability of detection dominates detection time. Changi has just spent fifteen months and a partnership with a systems integrator acting on exactly that, and quantified the result.

The third is that they published a number at all. More than ten-fold, on false alarms. That is a real figure attached to a real installation, and in fifteen years of covering this industry it is one of the very few. Every vendor release in the last decade leads with speed, coverage or artificial intelligence, and none of them will tell you what fraction of objects the thing actually finds.

Half of the 2020 question is now answered: Changi kept it and improved it. Hong Kong commissioned its system in April 2018 and has said nothing since. Dubai and Miami remain blank. If you know, get in touch.

Sources (International Airport Review, NCS)