Roof Crush & Roof Crush Testing--Technical Considerations
TASA ID: 189
The PROBLEM and a SOLUTION:
The problem of evaluating greenhouse structural integrity for rollover accidents, i.e., the roof crush issue, is still unresolved after several decades of consideration. The current federal regulation and the long anticipated upgrade are both inadequate, and it is not clear that NHTSA either can or will come up with a reasonable standard. While a reliable testing protocol with adequate performance standards has been achieved in this area for other types of vehicles, especially for school busses, there is no real reason to believe that a regulation adequate to the task of eliminating the harm caused by weak roofs in light vehicles will ever be developed. The technical problems associated with the regulatory task are analyzed below. They are formidable and cast doubt on the whole regulatory approach even before political considerations are introduced.
A test program like the very successful series produced by the IIHS for offset frontal impacts would be more effective, in the short run at least, for improving roof strength. Televise the results of drop tests, or Jordan tests, or CRIS tests, and let the public see the differences between those which only pass the federal 216 test and the best from Europe. These presentations need not even involve grading standards. Just put the best against the worst and let the public decide what they want to ride in. No arcane, tedious discussion of the Malibu tests, the "drop" theory of injury, or the "problem" of non -repeatability, just bent steel and unbent steel. Vehicle makers would come around quickly and without regulation just as they have for the offset frontal test.)
TESTING
Engineering testing is different from scientific testing. Product or component testing is done to establish performance, not to discover or validate new scientific theories. Engineering testing does not require invariant test conditions or reproducible results. If the results are adequate in terms of system performance, it makes no difference what variations of input parameters or test results develop, as long as the inputs are representative and include the "worst case" and the results, even if the "worst case" is acceptable. Since the "worst case" may be hard to identify, guaranteeing performance may require a number of tests in which the input parameters are varied along the full range of probable values.
Reproducibility is a matter of convenience and economy for the testers; it has no fundamental engineering significance. Reproducible results with tests that result in catastrophic failures are an unreasonable and unnecessary goal. Vehicle makers could do hundreds of rollover tests on each new platform, (basic vehicle family) covering the full range of parameters of concern, just as they are currently doing for other certification testing. The idea that roof strength must be evaluated with only one test or even one type of test is an unnecessary concession. Thus, the new Jordan and CRIS rollover test protocols, while they may be of great illustrative or forensic value, do not represent a necessary or even necessarily a desirable approach to greenhouse testing, and their advancement may represent an unnecessary compromise with respect to vehicle producers' true obligations.
STRUCTURAL FAILURE
Let us consider the possibility of developing a formal definition for structural adequacy for something like a vehicle greenhouse under conditions like those encountered in rollover accidents. Let us define a structure, most generally, as a set of mechanically connected components which is capable of transmitting significant loads to components remote from the point(s) of loading. A component is an identifiable piece of the structure, e.g., a beam element, or a joint.
Now, what we need in a vehicle greenhouse is something like what is commonly thought of as rigidity. But what is rigidity, and how is it to be defined for evaluative or regulatory purposes? Consideration of the mechanisms of structural failure might suggest a definition which includes a concept like this: A rigid structure is one that does not exhibit plastic hinges when loaded (to some significant level). A plastic hinge results when a component or an area or element in a component is subjected to a bending moment (force acting through a distance) beyond its elastic limit causing it to exhibit plastic failure. Note that a "rigid structure," so defined, can still fail in other ways, essentially with pure compressive or tensile loads, although many failures with these loading modes may still involve localized plastic hinges, e.g. slender column buckling.
Plastic hinges are designed in elements of certain structures, notably buildings expected to be subject to seismic loads. The hinge elements at the base can absorb energy and prevent the complete collapse of the structure by "giving" a little as the structure is loaded. Plastic hinges are largely unknown, however, in mechanical engineering design. The problem is that they occur with roof collapse in rollover accidents.
(A caveat: Strength is a property of materials, not of structures. A given material, say a specific, well-defined steel alloy, will have various strengths as determined by various types of tests -tensile, shear- and will yield consistent strength values under well-controlled testing conditions. Structures, on the other hand, have no simple "strength"; rather they have the ability to resist a specific load with a given method of application of that load. The "method" includes the point at which the load is applied, the direction of loading, and duration and the rate of the loading process. Change any of these, and the "strength" of the structure may change.)
These diagrams may make the issues a little clearer. As we load a structure, it will deform up to someelastic limit, beyond which the loading is in the plastic range and results in non-recoverable damage, i.e., permanent deformation. Prior to this point, if we release the load, the object returns to its original shape. The point or load where we go from elastic to plastic is called the "yield point"; "failure" occurs beyond the yield point- somewhere. "Somewhere" because "failure" can involve a subjective judgment. Note also that in the left diagram, the applied load can exceed the yield load. The structure can resist more even while it is failing, even though the damage is permanent, and just specifying a peak load alone is meaningless. (Or at least we don't know otherwise, a priori. Note further: The measured magnitude of the applied load is determined by the ability to resist it; you can't put a 100 lb. load on a structure that breaks at 50 lb. If you put a 100 lb. weight on an object that breaks with a 50 lb. load, the object will fail before experiencing the full weight internally.)
A simpler case, which may be relevant for structures like vehicle roofs, is shown in the right diagram above. Here, column buckling, a severe geometric-dimensional change, occurs at the point of maximum sustainable load. Again, "failure" is in the eyes of the evaluator. Roof failures probably involve both of the situations illustrated above- and that's a real problem. What's worse, roof failure probably involves a mechanism something like this:
Here a component element has experienced plastic failure in a limited area while the rest of the component has not been stressed beyond its elastic limit. When the load is released, the plastic area stays deformed while the under- stressed area returns to its original shape. The result is a deformed structure without complete failure of any one element. Will this deformation be significant? Can we have a bending moment "failure," without full plastic hinges? We suggest that there is no theoretical answer to this question. It depends entirely on the overall design.
Thus, the "no plastic hinge" approach or some other deformation theory- based approach to evaluating roof strength might look interesting from a technical definition point of view, but it is not at all obvious that the standard can be practically applied or that anything less than an absolute prohibition against elastic failure -nothing breaks or bends permanently- will be of value. Such an absolute prohibition would work; it just isn't doable. You have to tolerate some damage. The question is: How much? Consider this illustration from Volvo:
www.autosteel.org
Note the small difference between the inadequate structure on the left and the reinforced good roof on the right. Is there a simple, easily identifiable parameter that we could use to condemn the left one and pass the reinforced structure? The improved structure is still bent, but without "structural collapse," whatever that is.
This load/crush curve from a Ford Explorer test looks similar to the stress/strain curves above, but the similarityis deceiving. The truly elastic portion of the strain or crush curve is probably too small to show up on this scale. What looks to be a linear stress/strain curve out to about 50mm ( 2") is actually a line of plastic failure. (You crush a Ford roof 2," and it isn't bouncing back; a Volvo XC90 may be different.) The resistance is actually decreasing as the point of maximum allowable crush is reached -5" or 127 mm. Somewhere out around 15' of crush, additional resistance is felt because something has "bottomed out"- probably against the door or firewall.
We can now see the real problems with the current FMVSS 216 test protocol and anything similar. They break the roof, and they keep on pushing. Now, while a "failed" structure might conceivably be serviceable, field experience with the Explorer and almost all domestic and Asian makes suggests that this level of failure -5" at the associated load- is not a serviceable failure; it is a catastrophic failure. In order for a static load test to be of value here, it must be shown first that the load is somehow realistic and secondly that the load does not result in a catastrophic failure. Failing these two determinations, field experience must show, as it does in the case of school busses, that the static test does happen to result in structures of adequate strength.
Establishing equivalence between a static load and a dynamic load, the sort that is actually encountered in highway rollovers, is notoriously difficult. Indeed, it is not at all clear what it would mean to say that a static and a dynamic load are equivalent, except perhaps in the level of damage done, much less how such equivalence is to be established, again, except by looking at the results. And if we have the results of a dynamic test, it's not clear why we need the static test at all. The assumption seems to be that the results of various dynamic "tests," e.g., real world rollovers, can somehow be anticipated by a single static load test. This assumption, to say the least, requires some proof, like that available for school busses where several decades of field experience have proven the adequacy of the test protocol. (Contact the author for the relevant documentation) Similarly, it is not really clear how the issue of "catastrophic failure" is to be decided, a priori. All this suggests that what is really required is actual rollover testing- of some sort- to evaluate the performance of production vehicles.
There is another regulatory advantage to a dynamic loading protocol that is sometimes overlooked. Dynamic loads are transient. Static loads, even quasi static loads, are not, except by fiat. Thus, we can drop, hammer, roll or bang a structure and stipulate that strain or crush must stop increasing before theloading event, e.g., ground contact, terminates. This is the guarantee against complete structural failure and what it means to say that it didn't break, at least for impact events defined or constrained by kinematics. This is possible with dynamic loading but not with static loading except by arbitrary stipulation. With a static loading protocol, if the structure stops deforming, it must necessarily become infinitely strong and rigid as the static load increases and approaches infinity unless the loading operation is stopped at an arbitrary point. With a dynamic load, the structure can bounce, roll or rebound through the point of maximum impact loading with limited damage. Indeed, the maximum load in an unrestrained dynamic loading situation is defined by the structure and not by constraints imposed by the mounting of the structure, a very fundamental difference. The difference eliminates the requirement for a force or load (energy) parameter in the test standard itself. Just roll it or drop it, with realistic kinematics, stipulate the maximum damage, and let the forces be what they will. It really doesn't make any difference. Impact events of the type of concern here are really defined by kinematics anyway.
Those who are technically sophisticated will realize that design validation admits to analytical techniques and does not necessarily require any sort of physical testing, either static or dynamic, today. While it is true that simulation using multi-physics software is probably feasible (see the ALGOR website for example), there is an underlying potential problem associated with using simulation for design validation: Cheating! Ford "misused " ADAMS when "evaluating" the Bronco II's stability, according to the guy who wrote the program, and there is no reason to think that somebody else won't cook the analytical books if allowed to rely on computer validation for determining compliance with a legal standard. The input data sets for these sorts of problems are large, complex, and hard to evaluate, and even if were possible to confirm the validity of the data input - to validate the model- if cheaters got their hands on the source code and modified it in house or played with the data sets, the results could be virtually impossible to debunk. What you would then have are two analyses, assuming you could get and run their input file, and two different result sets. Proving which was correct to non-technical audiences would be next to impossible.
But this does not mean that it is not feasible to determine the standard itself by analytical means, even if computer analysis is not allowed for determining compliance. (The latter, incidentally is now accepted by the EEC for highway busses.) "Impact on surface" contact analysis might yield a quasi -static force value of interest, i.e., tell us how strong the roof has to be to survive contact with a concrete roadway. But since the force that develops along the line of contact can be no greater than the resistance offered by the weaker surface or object (Newton's Third Law), it is not clear, at least to the current author, how this analysis would be done. Perhaps we increase the strength requirements until some maximum level of crush or penetration is reached. But how do we determine what this maximum value should be?
But let us consider further the problems associated with static testing before recommending something as "radical" as actual rollover testing. First of all, engineers don't really know why things break, although they may be very good at predicting when things will break. Theories of failure that have their devotees include:
- Maximum principal stress theory
- Maximum shear stress theory
- Shear strain energy theory
- Constant distortion energy theory
(And these are only theories for ductile material failure in the elastic range. Theories of why things break would necessarily be generally more complicated.)
Secondly, while the notion of a Newtonian Force- the essential element in a static load test -is an invaluable idea, it is a complete fiction, and nothing happens because of the application of a Newtonian Force, contrary to the assumptions made by the advocates of static load tests. The ontology of modern physics includes the following: the Strong Force, the Weak Force, Electromagnetism and Gravity. None of these field influences is really a Newtonian force although the latter two can be described as such if you're far enough away and operating at an appropriate scale. Newtonian forces are macroscopic summations of microscopic events, and failure starts at the micro level.
Things happen in the real world, at the scale and with the processes we are concerned with here, because of momentum and energy exchanges. In order to change a physical system, energy must be supplied, perhaps with concomitant momentum changes. Static loads impart deformation energy to a system stored in the form of elastic strain until the thing breaks; deforming the object takes energy. Dynamic loading just does it more quickly in the form of a bang. While a force or system of forces might be associated with either loading method, the real story is the imparted energy. A work function, a measure of energy absorption without failure, ala SAE J2422, might then be a more appropriate method of determining structural integrity then, but it still leaves us with the question: What is failure? Can this be decided a priori without looking at injuries and deaths in the field?
We could specify a test protocol involving imparting a certain amount of energy to the greenhouse, say that associated with a drop height consistent with typical vehicle rollover kinematics. But how do we know that it has passed the test? One standard might be the "no new holes criteria" necessary to minimize ejection harm. Another standard, apparently used by Volvo, is no more than 2" permanent deformation. The current 5" standard might even work if it turned out that rolled vehicles never crushed beyond this point and that occupants were protected when crush was limited to this value. Other possibilities include no, or limited, loss of interior volume, say 5% (European heavy trucks), except that associated perhaps with "matchboxing" (parallelogram displacement). But a standard involving matchboxing would probably have to include a prohibition against plastic hinge failure to prevent the greenhouse from collapsing completely. A standard involving the loss of interior volume could allow a lot of external deformation if the roof panel and supporting structure were sufficiently isolated from the interior strike surfaces and the vehicle could still make FMVSS 201 standards in rollover testing. Space and height considerations, however, might come into play.
What we really want, of course, is an occupant harm-based standard, not a vehicle-based standard, and for vehicle producers to prove that they have minimized the mechanisms of harm production in rollover accidents. We want them to prove it before the vehicles go into production and before field experience proves that they haven't succeeded in harm reduction. It is difficult to see how they can do this without rollover testing with instrumented dummies, and a lot of it.
This article discusses issues of general interest and does not give any specific legal or business advice pertaining to any specific circumstances. Before acting upon any of its information, you should obtain appropriate advice from a lawyer or other qualified professional.
This article may not be duplicated, altered, distributed, saved, incorporated into another document or website, or otherwise modified without the permission of TASA.