By now you certainly know that Volkswagen, AG of Germany has been caught fudging emissions test results for its cars with diesel engines. Engineers with the California Air Resources Board found during compliance testing that while VW cars with diesel engines passed the standard emissions tests, they failed when subjected to real-world driving tests.
What actually happened, according to the California Air Resources Board (CARB), is that VW’s emissions computer failed to spray enough of a substance called diesel emissions fluid (DEF) into the exhaust stream to neutralize the oxides of nitrogen that are produced in all diesel engines.
The DEF, which is a mixture of the organic chemical urea and water, reacts in the catalytic converter to produce nitrogen and water, both of which are harmless and hugely abundant in the atmosphere.
DEF is used in a wide variety of diesel car and truck engines, and it’s widely available and relatively cheap. However, Volkswagen, along with fellow German car makers Mercedes Benz and BMW, uses a proprietary version of DEF called AdBlue, which is somewhat more expensive.
The amount of AdBlue used by a car in normal use depends greatly on the type of car, the type of diesel engine and how the car is driven. On a Volkswagen, the usage appears to be about five gallons every 10,000 miles. A parts department representative in a Washington, D.C., area Volkswagen dealership told me that the price of a 2.5-gallon bottle is $10.70.
While neither the U.S. Environmental Protection Agency nor VW has released any details on what the EPA is calling a “Defeat Device,” the engineers in California who did the actual testing were more forthcoming. The engineers developed a special test to see whether VW was reducing NOx emissions in real-world use and determined that emissions were only partially reduced at the beginning of the test.
Then CARB found that the DEF dosing of the exhaust stream became insufficient and eventually stopped. While it’s referred to as a device, the software code that delivered the bogus results was actually introduced into the VW emission control software during an effort by CARB and VW to clean up diesel emissions beginning in early 2014. The software was able to determine when a car engine was undergoing emissions testing, at which point it used enough DEF. Otherwise it didn’t.
It turns out that Volkswagen engineers were aware of this problem for at least a year, but whether they informed the company leadership remains a question.
The statement of Volkswagen CEO Martin Winterkorn announcing an internal investigation suggests he was not aware. His position is not surprising, and the fact that a group of engineers might take it upon themselves to create a way to fudge test results is certainly not unique to VW.
In fact, during the years when I’ve tested digital systems including performance-enhancing software code became so routine that I developed tests that were designed specifically not to look like tests.
Volkswagen Emission Test Scheme Probably Not Industry’s Only Case
I first noticed this during the 1980s when I was reviewing products for Byte Magazine and we discovered one brand of Ethernet cards that performed much better than others during tests, but didn’t seem to be much faster otherwise.
In those days, the most common test of network interface cards, one developed by Novell, had a predictable pattern and when the Ethernet card saw that pattern, it simply passed packets through without processing.
A few years later while I was testing Gigabit Ethernet switches at the University of Hawaii, I noticed something similar and as a result worked with Spirent Communications to develop a network test that mirrored actual traffic at UH.
Once faced with what appeared to be actual traffic, the switch being tested failed spectacularly. Fortunately, the VW cars didn’t have that problem and are considered safe to drive even if they add more pollutants to the atmosphere. Owners of affected VW and Audi diesel models are also fortunate the company is committed to fixing the problem.
But the problem doesn’t end there. Just because Volkswagen got caught doesn’t mean it’s the only company that’s delivering false results. My experience shows that creating software that can detect tests to produce positive results is a long-standing practice.
The only way we were able to routinely get past such efforts was to find a way to simulate real-world conditions during the tests. The CARB did the same thing in testing cars.
But what’s to keep other automobile companies from doing the same thing? Nothing. While I haven’t done this testing myself, now that CARB has discovered what is going on, it’s clear that real-world testing should be conducted on every vehicle that the Board tests, not just Volkswagens and not just cars with diesel engines.
If this is done, I’d be willing to wager something significant, say a case of fine craft beer, that other instances of auto test result fudging show up.
Of course, there’s the question of why VW engineers went to the trouble to create the ability to fudge test results. I can think of two reasons. The first being that they weren’t given the time required to produce the software they needed to make the engine run efficiently enough to pass the test legally. The second is that the engineers were under pressure to keep costs under control and produced an engineering answer to a cost problem.
It’s possible the engineers developed the test-fudging software to keep maintenance costs from going through the roof. DEF is fairly expensive as an ongoing maintenance requirement, and Volkswagen was bearing that cost as part of its warranty service. Just stretching out the time before more AdBlue was needed would save millions.
The downside to this kind of solution is that someone always finds out that it was done, and now Volkswagen will pay dearly in fines, in repairs to cars, in loss of customer confidence and in overall damage to the reputation of its brand. Volkswagen’s stock price has already taken a huge hit.
Welcome to the real world, Volkswagen.