Virtual vs physical testing

LinkedIn +

In the great rush to reduce development time, computer modeling and CAE is clearly the way forward. Computer models are quicker to generate and easier to modify.

Costs are also a consideration, forget expensive test chambers, and proving grounds, soon you will be able to do this all in the confines of a small box under your desk, in your air conditioned office.

The guys with the dirty finger nails (development engineers) will be consigned to history together with the dodo and the Corvair (all credit to Ralph Nader).

Moving directly from design to production componentry will eliminate the wasteful iterative processes of development. ‘Right first time’ will be the watch words. The pilot production run will provide the test and certification fleet necessary for legislative compliance, but can be considerably reduced in size, from 80+ vehicles to perhaps a dozen or so. Full scale production can then roll forward almost immediately to build launch stocks, while legal testing is completed, again speeding the process from concept to launch. A brave new world indeed.

On the contrary, trusting the companies’ new product purely to the realms of cyberspace is folly. The gradual reduction of test vehicle fleets seen over the last 10 years has highlighted this.

Reduced durability testing and general development validation programs have lead to errors being missed and resulted in major high profile recalls for a number of manufacturers.

Consider also that computers are basically dumb. This is not the statement of a luddite, it is merely the fact that any computer simulation is only as good as the man who programmed it. Production vehicles are subjected to a wider variety of operating conditions (owners) than any simulation can possibly cover.

By way of illustration let us consider the A300 airbus. This was the first drive by wire commercial airliner. Great care was taken in the programming of the flight control computer to cover all operating conditions. However, when an A300 was carrying out a low fly past at the Paris Air Show  – at low speed, with flaps down, wheels down, engines throttled back and the nose raised – the flight computer mistakenly assumed the pilot intended to land. The plane did land but unfortunately not on the runway.

Traditional development/validation test programs with large fleets of test vehicles are more likely to uncover design limitations simply by sample size –it is and always has been common knowledge (at least among developers) that you should never work with a sample size of one. A sample size of none needs no further comment.

Mark Berry: Itinerant development engineer, apprenticed at Rolls Royce, prior to five year-long spells at both Jaguar Land Rover and Lotus before 25 years as a “jobbing engineer” specialist in EMS calibration and emissions, while being widely conversant in most aspects of type approval. Particular skills in managing climatic and altitude field test programs.

August 18, 2016

Share this story:

Comments are closed.