The Fraud of Benchmarking

by Jonathan Tennenbaum

This week's pedagogical discussion illuminates the issue of "non-linearity in the small" from a somewhat different angle than the preceeding ones. This week we shall take a look at the celebrated case of Mercedes' famous "A-Class" automobile. The following documentation was researched by Rudiger Rumpf and Ralf Schauerhammer, and will be featured in a forthcoming article in the German magazine FUSION. I have translated the core of their manuscript and added some comments. --

Jonathan Tennenbaum

To remind readers who are not "car freaks" and may not have followed this affair from the beginning: Two years ago the world-famous automobile manufacturer Daimler-Benz -- manufacturer of the legendary Mercedes-Benz automobile and now partner of Chrysler in the greatest mega-merger in industrial history -- triumphantly introduced into the world market for low-priced compact cars its own, specially-designed model called "A-Klasse." Not only would Class A offer people with small pocket-books the prestige and "feeling" of driving a "Mercedes," but the car itself boasted extraordinary features. Instead of the engine being located in front of (or behind) the passenger cabin as in other cars, the A-class has its engine placed {underneath}, with the passenger cabin built on top. This so-called "sandwich" construction, never before utilized in this sort of car, offers greater flexibilities in the use of space, while at the same time the passengers sit significantly higher above the ground than in other cars.

On Sept. 23, 1997 test drivers in Denmark found that the Class A tilted onto two wheels during a swerving manouver (i.e. sharp turn of the sort needed to steer around an object) at 55 km/hour (34 mph). A month later, on Oct. 30, a Class A Mercedes flipped upside down during the so-called "Elk Test" at 60 km/hour (37 mph), slightly injuring three test drivers in the car.

The serious issue raised by these events is not so much the obvious design weaknesses of Mercedes Class A; what is significant is the kind of faulty {thinking}, embedded in the process of design and development of the car, which ultimately caused the embarassing and commercially disastrous result. In fact, the main weakness in the design and development process of Class A lay in a dependence on so-called "benchmarking" methods.

Up until recently, Mercedes has traditionally devoted much more time and investment to develop new models, than Japanese firms for example. In order to improve its "competiveness," in the case of the Class A, Mercedes set the goal of reducing the development time from the traditional 7 years (84 months) or so, to a mere 32 months! Yet, the comparison with the development time of Japanese makes is unrealistic and misleading, because Japanese producers typically concentrate on improving already established and proved designs. In that case only about 20-40% of the components must be newly constructed, even to make a new model. But in the case of the Mercedes Class A, for example, 100% of the components had to be newly developed.

Although Mercedes is one of the leading automobile manufacturers in the world, Mercedes engineers had never before built a compact car with front wheel drive. Furthermore, the planned "sandwich" construction had never before been used in a small car. Setting these goals while at the same time cutting the development time from 84 to 32 months, placed huge pressure on the development department of Mercedes. That pressure greatly strengthened the already-existing tendency to think that computer simulations could replace actual, real-life driving tests.

In order to save development time and costs, in April 1993 Daimler-Benz engineers input the then-available, projected basic design parameters of the Class A into a computer simulation system designed to simulate the dynamic behavior of the car. This was done before even the components and parts of the automobile had been constructed. These imaginary driving tests, according to Mercedes, were supposed to be sufficient to provide "all the answers" concerning certain important design decisions as, in particular, which of three alternative construction types ("Mehrlenker," "Verbundlenker" or "Laengslenker") should be chosen for the rear axle assembly. On the basis of those simulations, the cheapest of the three alternatives was chosen as fully acceptable.

As a further test for the projected driving characteristics of the Class-A (which had not yet been built, even in prototype), "engineers put themselves behind the wheel of a Mercedes S 280 (a completely different model-JT) which had been programmed to simulate the dynamic behavior of the Class-A," a Mercedes report boasts. "During a double lane change (the similar type of manouver to the now-infamous "Elk Test"- JT) which reveals the handling characteristics of an automobile as well as its safety reserves in marginal situations, the Laengslenker axle (the cheapest of the three alternatives - JT) performed convincingly.... In the context of the total concept of the Class A, the Laengslenker demonstrated itself to be the best compromise." The "compromise" referred to, was a compromise between handling characteristics and cost. And here the fact that the Laengslenker was much less costly, clinched the decision in its favor. All other Mercedes-Benz models are equiped with the "Mehrlenker" axle, which is much costlier, including in comparison with the systems used in competing models.

In an information brochure on the Class A Mercedes also wrote: "This time there was not enough time to carry out the extensive basic investigations with different axle types, which are normal procedure for the development of a completely new automobile type."

Other German producers calculate up to 12 months for the costly process of harmonizing and adjusting the chassis and related assemblies in connection with electronic systems such as ABS (anti-lock brake system) and ESP (electronic stability program). This is normally done only after 3 years of testing with prototypes or rebuilt older models in order to determine the proper design of the axle. This is what Mercedes itself had always done before.

For example: In 1989, driving tests in extremely demanding, mountainous areas revealed -- contrary to the results of computer simulations -- that the braking system of the then-newly-designed Class S auto (V-12 motor with up to 400 horsepower) was far from meeting the full stress performance requirements. The entire brake system had to be completely redesigned. But at that stage the projected beginning of mass production of the auto was still two years in the future.

In the case of the Class A, Mercedes not only did not have the necessary time, but also lacked sufficient capacities in its development department: just in the period from June 1993 to October 1997, 9 new models had their premieres. Juergen Stockmar, Director of Development at Opel, was quoted as saying that many employees in the development departments of automobile producers today are working beyond the limits of their endurance, and that the overworked condition of the employees ran like a red thread through a series of technical breakdowns and other problems.

After the disaster in October 1997 Mercedes was finally able to bring the problems of the Class A under control -- although only after three separate attempts and after company head Schrempp had intervened to halt deliveries until further design and development had been carried out. But the methods used to "solve" the problem were rather dubious.

The electronic stability system (ESP), which was originally planned to be sold as an option for 1700 DM, was now included as standard equipment on all Class A autos and delivered to the buyer without additional cost. The ESP had originally been conceived as a supplementary program for safe and stable cars, to assist control of the vehicle under extreme conditions such as wet or slippery roads. But in the Class A, the ESP became indispensable even to carry out a simple avoidance maneuver on a dry surface -- something which competing models had never had problems with.

The fact that Mercedes now claims it has solved the problems of Class A by supplementary installation of the ESP system, demonstrates that the fundamental problem behind the Class A disaster has not penetrated to the conciousness of the company's board members. They are still holding to their belief in benchmarking and computer simulations.

This becomes obvious from the fact that the board of Daimler-Chrysler still refuses to withdraw its even bigger disaster, the regrettably-misnamed "Smart," from the market. "Flop" would be a more appropriate name for this totally misconceived and technically defective product. Besides the problems shared with the Class A -- shorter wheelbase and elevated center of gravity -- the developers thought they could simply ignore problems that had been known for decades. These are the problems which arise when one attempts, as with the "Smart," to realize rear-motor, rear-wheel drive in a vehicle with short wheelbase, leading to a situation where nearly two-thirds of the car's weight falls on the rear axle.

Ignoring the long and problematic history of constructions of this type, the Mercedes engineers even installed an over-powerful engine (after all, the car was supposed to be "smart"!), with the philosophy that "electronics will fix everything." Since the car was known to be unstable, the maximum speed was set at 130 km/h. But even below that speed the electronics cannot compensate for the fundamental fallacies in the design. Physics prevents this! The cheap electronic stabilizing program called "Trust" has revealed itself, in all tests which were not designed in advance to give a positive result, as a failure.

Readers will not have failed to recognize a recurring syndrom of today's larger world in our story of the "A-Klasse": Rather than correct fundamental, axiomatic fallacies in the design of policy, the reponse to each ensuing disaster is: "we'll fix it!" The "successful" result is to carry the axiomatic fallacies forward into the next, even worse phase of disaster, whose onset has been rendered inevitable by the follies of such linear "crisis management."

Note, also, a second point: in a multiply-connected manifold, "dimensionalities" can never be treated as Cartesian independent variables. In substituting or modifying even an apparently minor technical component within a complete functional system such as an automobile or a space vehicle, the potential nonlinear impact of that change upon the characteristic functioning of the whole system is an issue of physics, not mathematics. In a unique experiment, the components of the experimental apparatus and their characteristics, taken in and of themselves, seem to be fully "known." But the composition of the experiment generates an irreducible anomaly, refuting exactly the sort of linear "curve fitting," which turned Mercedes' proud creation into a total flop.