A Short History of Real World Testing; What have we learnt? September 25, 2013 MIRA Ltd 2013
Real World Testing My view of durability and development testing attempts to replicate the real world. Introduce the history, and up to date methods for determining the targets and development tests used at MIRA. Presentation will cover: A brief history testing the real world under control? Usage targets & test equivalents Belgian Pave at MIRA Health & Safety less of an issue. Finding out about the real world Accurately representing how things are used Allowing for the linear and non-linear (i.e. Rubber) A simple example of developing to real usage September 25, 2013 We Deliver 2
The Early Years 1917 Aberdeen PG in Canada military Test Area earliest. 1921 Fiat Lingotto end of production roof top test track. Fiat Lingotto factory roof top test track. September 25, 2013 We Deliver 3
The Early Years 1917 Aberdeen PG in Canada military Test Area earliest. 1921 Fiat Lingotto end of production roof top test track. 1924 GM opened the Milford Car PG near Detroit A gravel loop plus a tarmac straight, 4 miles in all. "No other industry has gone forward so swiftly with so few basic facts if the industry is to continue it must know more facts about what happens as the car is being operated mile after mile in the hands of the user. Quote GM pamphlet at the time. Racing circuits do not count as they have a very limited part to play in testing and proving. Not to be outdone other manufacturers followed suit.. 1925 Ford Dearborn 1928 Packard September 25, 2013 We Deliver 4
1946 - MIRA first purpose built Proving Ground for automobiles having other than just roads Up to 70 s tracks allowed the car industry to check handling, high speed, breakages and other good inventive stuff. Standards and procedures were created for these tests. All different for each manufacturer. Nothing for accurately testing durability well except to do a lot of testing for a long time. Strength & performance was checked well on test tracks but durability only by 1000 s of miles on a representative road route. MIRA High Speed in the early days Belgian Pave at MIRA Health & Safety less of an issue. One of the first MIRA system test tracks Long Wave Pitch September 25, 2013 We Deliver 5
Empirical Experience Into the 80 s, durability tests laid out how to prepare the vehicle for a test, with stringent measurement processes; how to run the vehicle on the tracks, and how to measure & review the results. Little was made understood how realistic the test was, or what the overall goal was. Durability testing for a long time came down to two tests; Early failures or accelerated life using MIRA 1000 miles Pave. Through life testing on endurance road routes 100,000 miles. Fig 7: Does 1000 miles Belgian Pave = 100,000 road miles? Eventually, 1000 miles Pave, and 100,000 miles road route, became synonymous with a vehicle life, purely by repetition. In fact vehicle development engineers as a whole came to accept that these two tests actually represented the usage life. Fig 8: Only evidence vertical only, 100:1 Pave to Road correlation - the author 1988 September 25, 2013 We Deliver 6
Measured to Death The 90 s; durability endurance sign off road routes were measured, typically by strain gauging suspension or accelerometers. The established tests procedures were very accurately measured. Assumption was that loads represented what actually occurred. Even to applying fatigue damage analyses - hopefully producing the same damage as the representative road route The danger is to then apply same test to another vehicle type. It can be most inappropriate, even over test! Then test & usage must be equivalent! Myth - 1000 miles Pave equals 100,000 miles of the road started. Just because a test vehicle survives, an empirical test & is okay in production does not mean the test is the same as the users/usage. Fig 9: Applying assumed user conditions leads to some over test. Just because you get the same damage does not mean the mode of failure is the same! September 25, 2013 We Deliver 7
Assumed Failure It is easy to assume that just because we can reproduce a failure, that we have reproduced the process or mode of failure. Actual causes can be difficult to assess. Assumption Actual Mode September 25, 2013 We Deliver 8
Usage targets & test equivalents September 25, 2013 MIRA Ltd 2013
Durability at Design Stages???? A bold statement: Very few, if any, in the auto industry designs for durability. Criterion is performance & strength with a factor of safety. Durability & reliability comes from prototype empirical test. Important questions: Manufacturers are extending warranty periods, implying longer life. The goal of your next design could be to increase the warranty from 100,000 to 200,000 miles. Firstly Secondly What in your design process can you factor by two to account for this? Is it adequate to just double your existing durability or proving test? September 25, 2013 We Deliver 10
Performance v Robustness v Durability Major goals: Performance Vehicle does what it is supposed to. Get a customer interested. Robustness People will accidently (or on purpose) exceed performance criteria Customer expects safety factor. Durability Vehicle should last the period expected. Customer expects to sell vehicle on to another. Without over-engineering and/or costing too much. Just meeting performance or robustness does not guarantee durability. Repeated robustness test, or worst-casing, is not a substitute for durability testing. Even special cases such as racing, expectation is that it will last the practice and the race. September 25, 2013 We Deliver 11
Real World Testing Customer Usage Profiling & Usage Correlation are methods for creating tests that match in use criteria and reproduce multiple input conditions in the correct proportions in a shorter accelerated period. Fig 10: 1988 first publication of a multiple regression, real world correlation by the author at a 1988 E-I-S seminar. Fig 10 : Basic correlation- first published second E-I-S conference around 1986. Although still being advocated today, it has been vastly updated in recent years to statistically, software tools, reality & speed. September 25, 2013 We Deliver 12
Finding out about the real world Accurately representing how things are used September 25, 2013 MIRA Ltd 2013
The Target Most quoted European target typically 200,000 kms of road. Is this is a useful target to convert to design or durability test or load case targets. We have to ask too many extra questions:. Which roads, & for how long? What mix of loading? Is the longest distance user the worst? What terrain (hills, mountain) and at what speeds? What are the seasonal differences? What are the environmental conditions? and many more. How does a number of owners affect the usage pattern? How to take into account not only the first owner, but the second and even third ownership? It is interesting to note that warranty periods and design targets now span across many years and often more than one owner type. September 25, 2013 We Deliver 14
M MIRA has proving ground schedules used for design targets for a number of vehicle types. Termed CarDur (Car Durability) Schedules they accurately provide for the structural or power train usage equivalent inputs at the 90th percentile usage patterns of drivers. I will describe how we go about this.. September 25, 2013 We Deliver 15
MIRA 2012 European Durability Development Target & Test Schedule - CarDur September 25, 2013 We Deliver 16
Allowing for the linear and nonlinear (i.e. Rubber) September 25, 2013 MIRA Ltd 2013
Damage Intensity
Every Combination of Forces Force Z Each combination vector is a possible transfer function & a relative potential damage calculated Force X Force Y
Allowing for Frequency
Surface Damage Weighting Relative damage for every possible transfer function from surface to structure that can exist Normalise to Urban Surface to provide EUK weightings
Case Study: Physical data & test v Virtual data & test Some relevant info - extending the life of rubber components September 25, 2013 MIRA Ltd 2013
The Vehicle - Fun Part
CAE & Traditional
Non-Linear Modelling Dampers, bushes and flexible components modelled dynamically Components tested to produce basic response library for training a Neural Net Dynamic results interpreted by Neural Network providing fuzzy load case data for CAE work Results also provide NVH and other predictive results for ride and comfort estimations
Physical Portion
VPG - Random Surfaces
VPG - Tyre Models (etc) Tyre models : MF Tyre (Dynamics) TNO Swift Tyre (Basic durability) Cosin F Tire (Durability) CD Tyre (Durability)
Virtual Hybrid Development
Predicted Result Correlates
It Works
Contact Details David Ensor BSc (Hons) Ceng MIMechE Senior Consultant Engineer Direct T: +44 (0)24 7635 5295 E: david.ensor@mira.co.uk MIRA Ltd Watling Street, Nuneaton, Warwickshire, CV10 0TU, UK T: +44 (0)24 7635 5000 F: +44 (0)24 7635 8000 www.mira.co.uk September 25, 2013 We Deliver 32