DaveD
Well-Known Member
I've got some questions about factors of safety, and how these are dealt with when designing in composites. But first some “thinking out loud”:
Taking at FAR23 as a starting point there are limit loads and ultimate loads. Limit loads being the maximum loads the structure can expect to see in service and ultimate loads being the maximum load the structure must withstand. There is a requirement that there be no permanent distortion to the structure at the limit load which for metals corresponds to design using the material yield stress as the maximum acceptable. For ultimate loads the requirement is that the structure must withstand the load without collapsing, but permanent deformation is acceptable. i.e. material yield is acceptable providing the structure doesn’t collapse. Ultimate loads for metal structures are typically 1.5× the Limit load, but with higher values for castings (to account for material uncertainty), fittings and joints (to account for loading uncertainty, among other things).
The ×1.5 factor to derive the ultimate load in metals is apparently about loading uncertainty and not material properties. Tightly controlled production means that metal properties are highly consistent, and their long history of use means that design data is readily available. For example DOT/FAA/AR-MMPDS is available for download, running to 1728 pages of metal material data, including A- and B-basis properties [A-basis = 99% of material will achieve the given property with a 95% confidence, B-basis = 90% of material will achieve the given property with a 95% confidence].
So where does this leave the would-be composite designer? Composites don’t have a convenient yield point to use, so another approach is required. I’ve heard a factor of safety of 2 being mentioned a lot (I think I may have even quoted this number myself!), but what is this value actually trying to capture?
To answer my own question; It looks a like a fudge to cover a multitude of material uncertainties, but which ones?
1. Variability in the basic materials – fibres and matrix
2. Quality of the laminate – i.e. bubbles, fibre/matrix ratio etc.
3. Hygroscopic (water absorption) effects on strength
4. Thermal expansion/contraction loads
5. Thermal effects on strength
6. In service damage / aging / UV exposure
If it’s all of the above, then a factor of two doesn’t feel like a big enough number! But that’s just the beginning; we also have another problem – the material properties data.
If we make our own test coupons then we can… Make multiple batches; test them at different temps; soak them in water etc. But this all takes time and $$$ and doesn’t really help for preliminary design when you haven’t yet decided what layups you are going to use.
If, on the other hand, you find some published data, then where do you go? (NIAR-AGATE has some data… but only for prepreg materials, and mostly carbon, but at least it is clearly defined what the test conditions were.) You’re left to interpret the values you find and incorporate some “knockdown factors”, (on the assumption that you are dealing with values for a room temp/dry/undamaged laminate):
×0.8 – for hot/wet conditions
×0.8 – for material scatter
×0.65 – for the worst case of “Barely Visible Impact Damage”
Combining the above .8×.8×.65 = 0.416 (i.e. 42% of the original value)
So by now we’re doubled the loads and more than halved the strength data!
Next, what do we define as “failure”? We could treat “first ply failure” as the requirement for ultimate load, but some laminates can retain a lot of their strength well beyond failure of the first ply. At the expense of more calculation we could make “last ply failure” the ultimate condition and check that first ply failure does not occur before the limit load is reached.
It seems sensible to be conservative, but taking the most conservative approach to all of the above points will end up being prohibitively heavy.
For what it’s worth, I think I’ll keep the usual ×1.5 factor for determining the ultimate load and use knockdown factors on the material properties (I’ll make some layups and test them myself) to account for material scatter and damage and thus get a “composite appropriate” factor of safety. At least that way it’ll be clear whether I’m fudging a loading uncertainty or a materials uncertainty!
I‘m interested to know how do others have dealt with this?
…And if anyone has some material properties for hand layup epoxy-glass (to get me into the ballpark!) I’d love to see them!
Taking at FAR23 as a starting point there are limit loads and ultimate loads. Limit loads being the maximum loads the structure can expect to see in service and ultimate loads being the maximum load the structure must withstand. There is a requirement that there be no permanent distortion to the structure at the limit load which for metals corresponds to design using the material yield stress as the maximum acceptable. For ultimate loads the requirement is that the structure must withstand the load without collapsing, but permanent deformation is acceptable. i.e. material yield is acceptable providing the structure doesn’t collapse. Ultimate loads for metal structures are typically 1.5× the Limit load, but with higher values for castings (to account for material uncertainty), fittings and joints (to account for loading uncertainty, among other things).
The ×1.5 factor to derive the ultimate load in metals is apparently about loading uncertainty and not material properties. Tightly controlled production means that metal properties are highly consistent, and their long history of use means that design data is readily available. For example DOT/FAA/AR-MMPDS is available for download, running to 1728 pages of metal material data, including A- and B-basis properties [A-basis = 99% of material will achieve the given property with a 95% confidence, B-basis = 90% of material will achieve the given property with a 95% confidence].
So where does this leave the would-be composite designer? Composites don’t have a convenient yield point to use, so another approach is required. I’ve heard a factor of safety of 2 being mentioned a lot (I think I may have even quoted this number myself!), but what is this value actually trying to capture?
To answer my own question; It looks a like a fudge to cover a multitude of material uncertainties, but which ones?
1. Variability in the basic materials – fibres and matrix
2. Quality of the laminate – i.e. bubbles, fibre/matrix ratio etc.
3. Hygroscopic (water absorption) effects on strength
4. Thermal expansion/contraction loads
5. Thermal effects on strength
6. In service damage / aging / UV exposure
If it’s all of the above, then a factor of two doesn’t feel like a big enough number! But that’s just the beginning; we also have another problem – the material properties data.
If we make our own test coupons then we can… Make multiple batches; test them at different temps; soak them in water etc. But this all takes time and $$$ and doesn’t really help for preliminary design when you haven’t yet decided what layups you are going to use.
If, on the other hand, you find some published data, then where do you go? (NIAR-AGATE has some data… but only for prepreg materials, and mostly carbon, but at least it is clearly defined what the test conditions were.) You’re left to interpret the values you find and incorporate some “knockdown factors”, (on the assumption that you are dealing with values for a room temp/dry/undamaged laminate):
×0.8 – for hot/wet conditions
×0.8 – for material scatter
×0.65 – for the worst case of “Barely Visible Impact Damage”
Combining the above .8×.8×.65 = 0.416 (i.e. 42% of the original value)
So by now we’re doubled the loads and more than halved the strength data!
Next, what do we define as “failure”? We could treat “first ply failure” as the requirement for ultimate load, but some laminates can retain a lot of their strength well beyond failure of the first ply. At the expense of more calculation we could make “last ply failure” the ultimate condition and check that first ply failure does not occur before the limit load is reached.
It seems sensible to be conservative, but taking the most conservative approach to all of the above points will end up being prohibitively heavy.
For what it’s worth, I think I’ll keep the usual ×1.5 factor for determining the ultimate load and use knockdown factors on the material properties (I’ll make some layups and test them myself) to account for material scatter and damage and thus get a “composite appropriate” factor of safety. At least that way it’ll be clear whether I’m fudging a loading uncertainty or a materials uncertainty!
I‘m interested to know how do others have dealt with this?
…And if anyone has some material properties for hand layup epoxy-glass (to get me into the ballpark!) I’d love to see them!
