VW's In A Fix With Their Fix

Back in September, the U. S. Environmental Protection Agency (EPA) accused Volkswagen of cheating with regard to emissions controls of many of its cars that use diesel engines.  VW admitted as much, its CEO resigned, and now the firm faces the problem of fixing all the cars that violate emissions standards.  One way or another, some 11 million cars worldwide are implicated, with about half a million in the U. S. alone.  How did VW get into this fix, and how are they going to dig themselves out?

As new information has emerged on exactly how the cheating was done, it's pretty easy to tell that this was no single-line software tweak by a lone rogue engineer.  According to a Nov. 4 BBC report, someone (probably several someones) designed software to detect when the car was on a test stand designed for EPA checks.  This typically involves running the car while it is on a dynamometer, which uses rollers underneath the wheels to load the engine to simulate actual road conditions.  But in order for the stationary test equipment to be connected to the vehicle, the car is usually sitting still in a laboratory somewhere during the test.  I'm not saying that I know how the software guys did it, but if I were faced with the problem of how to figure out if a test-stand situation like this was going on, I'd look at the built-in accelerometers that every airbag-equipped car has.  If nobody's at the steering wheel and the car isn't going anyplace even if it's in "drive" and the engine's running, chances are it's on a test stand. 

However they did it, when an emission-test situation was detected the car switched into a mode that made it pass the emissions test.  But the price was severely crippled power and lowered engine performance, which however would not typically show up on an emissions test—after all, nobody's actually driving it to tell.  Once the test was over, the software readjusted the engine settings to produce normal power and performance—and as much as forty times more nitrous oxides (NOx) than the EPA allows.  But hey—it passed the test.  That's all that counts, right?

This mode of cheating is why fixing the problem with many diesel models, especially older ones, is not going to be some simple reload-new-software exercise.  If you've gone on a road trip recently and looked around in a truck-stop convenience store, you may have noticed piles of plastic bottles full of something called "diesel exhaust fluid."  Turns out that this stuff is now needed for many tractor-trailer diesel engines in order to meet the EPA's requirements for NOx emissions.  There's machinery on board the truck that squirts the fluid—which contains urea—into the exhaust, and the urea solution vaporizes to form ammonia and carbon dioxide.  The ammonia, in the presence of a catalyst in a thing called a selective catalytic reduction system (SCR), combines with the nasty NOx molecules to form nitrogen and water, which finally leave the exhaust pipe and rejoin Mother Nature, leaving her nearly as pristine as she was before the truck came by. 

It's one thing for truck engineers to see the regulations coming down the pike, and take time to redesign the power plant so as to accommodate another anti-pollution system which requires valves, heaters to keep the urea solution from freezing, pipes, level-monitoring systems, and all the other stuff needed to do the NOx-killing job.  It's quite another thing for VW to be under the gun to retrofit small diesel passenger cars that are maybe four or five years old, with a kit of SCR stuff they were not designed to have.  You'll need someplace to stick the SCR unit in the exhaust line, a way to get a pipe from the SCR to the urea tank, a place to put the urea tank, control lines, etc.  Engineers estimate the cost per vehicle could range up to $1000 or more.  With some cars, it may be cheaper for VW simply to buy them back from the owners and send them to the scrapyard.  Software-only fixes may be possible for some diesel models, but it looks like millions of cars worldwide will need expensive hardware installations to meet current emissions requirements.

VW says its internal investigation into how all this happened is still continuing.  For their sake, I hope they wind it up pretty soon, at least well enough to publish a timeline with names and actions.  But even without such information, it's obvious by now that deception with regard to emissions controls was an established policy.  Maybe the conspiracy—that's not too strong a term at this point—was concealed from upper management, and that's one of the things we need to know.  But even if it was, it's clear that there was a group of engineers inside VW who deliberately set out to cheat the system of pollution controls.  And they got away with it for several years.

It's not often that such a clear-cut case of wrongdoing by engineers makes the headlines.  Far more often, engineers will face a dilemma in which either choice has advantages and disadvantages, both morally and otherwise.  And sometimes engineers make the wrong choice, basing their decisions on incomplete information.  But in most engineering situations, information is always incomplete.  There's always more you'd like to know, but at some point the project must go on, choices must be made, and sometimes they turn out to be wrong ones.

But the VW emissions case is different.  Deception was intended from the start.  I don't know what internal company dynamics brought pressure to bear on engineers to the extent that developing a software evasion of emissions controls seemed like a good idea, but clearly something was wrong with the way ethical principles were stated and handed down. 

Sometimes, companies who do bad things are unrepentant and fight tooth and nail despite being in the wrong.  In such cases, large government fines are sometimes the only thing that will make an impression.  But in VW's case, its CEO resigned, sales are dropping, and there are news stories with graphics that show the famed chrome VW emblem breaking apart.  It's starting to look like the market and news media will do more punishing than the EPA is likely to do.  Whether that's fair or not is almost beside the point.  To survive, VW will have to own up fully, fix the mess it made to the best of its ability, and be a different company from the inside out—from now on.

Sources:  An Associated Press article on the types of fixes needed by VW was published in numerous outlets, including the U. S. News and World Report website on Nov. 19 at http://www.usnews.com/news/business/articles/2015/11/19/vw-has-only-a-few-costly-options-to-fix-polluting-diesels.  Information on the details of how the cheating software worked was carried by the BBC on Nov. 4 at http://www.bbc.com/news/business-34324772.  I also referred to the Wikipedia article on diesel exhaust fluid.  I last blogged on the VW emissions scandal on Sept. 21, 2015.

EPA Accuses VW of Software Cheat in Diesel Autos

Last Friday, Sept. 18, the U. S. Environmental Protection Agency (EPA) announced that it had discovered a "defeat device" installed in nearly half a million diesel vehicles made by Volkswagen (VW) and sold in the U. S. from 2009 to 2015.  Specifically, EPA claims that VW engineers have admitted to designing and installing software that implements full emissions controls on their diesel engines only when the software detects that the car is undergoing emissions testing.  The rest of the time, some of the emissions controls are disabled, allowing the vehicle to produce as much as forty times the maximum allowed levels of NOx, a type of pollutant that can lead to respiratory problems and smog.  When queried about the accusations, VW spokespersons declined comment, citing the ongoing investigation.

Until VW has their day in court, or wherever this case ends up, fairness dictates that we give them the benefit of the doubt.  But when both the EPA and the California Air Resources Board (CARB) issue notices that VW is in violation of clean-air ordinances, citing admissions made by VW personnel, it's a fairly safe bet that something is amiss.

In 2014, some researchers at West Virginia University who were working for the International Council on Clean Transportation discovered that certain VW diesels emitted far more pollutants when operating under actual road conditions than one would expect from the fact that they are certified by the EPA for sale in the U. S.  When the researchers notified the EPA about this, EPA asked VW about it, and VW said they would issue a recall to recalibrate the systems involved, which they did in December of 2014.  However, the California Air Resources Board checked some of the supposedly fixed VWs in May of 2015, and found that some of them were still out of compliance—hence, more meetings with VW.  According to a letter from the CARB, its staff and EPA staff held a technical meeting with VW personnel on Sept. 3, 2015.  Reading between the lines, we can surmise that the question they asked was along the lines of, "Okay, guys, what's really going on here?"  Faced with the inevitable, VW admitted that they had deliberately designed the vehicle's software to detect an official emissions test, and to turn on all the pollution controls only during testing.  The rest of the time, some of the controls were inactive. 

Faced with this smoking gun (so to speak), EPA and CARB had no choice but to declare the affected vehicles in violation and to order VW to issue a recall to remove the defeat-device software. 

As it turns out, if the allegations prove true this isn't the first time that regulators have found diesel-engine defeat devices deployed on a massive scale.  Back in 1998, diesels in trucks and construction machinery made by Caterpillar, Renault, and Volvo were found to have two different sets of software.  One set was used when the EPA was running emissions tests on the engines, and adjusted the injection timing for low NOx emissions.  The second set of software used a different injection timing that delivered better fuel economy, but also caused more NOx emissions.  The manufacturers ended up paying about a billion-dollar fine for that infraction. 

There seems to be something about software that tempts engineers to bend the rules.  With hardware, it's relatively easy to dig into the machinery and find the gizmo that's doing its nefarious work—that's the kind of thing that the term "defeat device" brings to mind.  It reminds me of a scene from the autobiography of Vannevar Bush, who was in charge of the U. S. Office of Scientific Research and Development during World War II.  In the 1920s, he was a professor at MIT and got involved with a startup company named Raytheon.  At the time, Raytheon's hot product was a type of rectifier tube that was useful in the rapidly growing production of radios that operated from power-line current (earlier radios used messy and expensive batteries).  In a dispute with rival radio manufacturer Westinghouse, Bush claimed that Westinghouse was using Raytheon's patented tube structure.  The patent attorney for the rival firm rival denied it.  In response, Bush told Westinghouse's patent attorney to pick up a Westinghouse tube (which had an opaque coating on the glass) and crack it over a trash can.  He did so, and there was Raytheon's patented tube structure.  As Bush put it, the patent attorney agreed to advise his client Westinghouse to "keep off the grass."

You can't do that sort of dramatic stunt with software so easily.  If the accessible form of the software involved is in the form of machine code (which it usually is in production systems), often nobody other than the people who wrote it can really tell what it does.  So sneaky evasions such as the one VW engineers are accused of doing with the defeat-device software are hard to pin down, which means that indirect evidence such as performance measurements have to be used instead.  And it's not often that regulatory agencies go to such trouble to track down violations.  Further investigation may reveal exactly who at VW was responsible for the defeat-device software, and how high in the firm the decision was made.  And then, if the charges are proven, VW will have to pay—at least with a recall fixing the problem, and perhaps with fines or other penalties. 

The contrast between the way cars used to pollute before environmental regulations and what comes out the tailpipe today was brought home to me recently when we started working on a 1955 Oldsmobile owned by my late father-in-law.  It now starts up pretty reliably without help, but whenever it does, a blue cloud appears behind it and the sharp tang of volatile organic compounds (VOCs) fills the air.  Exhaust just doesn't smell like that any more, by and large, and that's thanks to catalytic converters, selective catalytic reduction for diesels that uses urea to reduce NOx emissions, and many other measures that make the air cleaner than it would otherwise be.

If the charges against VW prove to be true, that firm will have the opportunity to make the air behind its cars even cleaner.  And we will all be thankful for that.

Sources:  Numerous news outlets carried reports of the EPA's press release of Sept. 18, which can be found on the EPA website at yosemite.epa.gov/opa/admpress.nsf/21b8983ffa5d0e4685257dd4006b85e2/dfc8e33b5ab162b985257ec40057813b!OpenDocument.  I referred to reports on the issue by the Washington Post at http://www.washingtonpost.com/news/energy-environment/wp/2015/09/18/epa-volkswagen-used-defeat-device-to-circumvent-air-pollution-controls/ and a letter from the CARB at http://www.arb.ca.gov/newsrel/in_use_compliance_letter.htm.  I also referred to an article on the 1998 defeat-device actions in the Los Angeles Times for Oct. 23, 1998 at http://articles.latimes.com/1998/oct/23/news/mn-35220.  The patent dispute between Raytheon and Westinghouse is described on p. 198 of Vannevar Bush, Pieces of the Action (William Morrow, 1970).