As a composite designer how do you deal with Matrix Microcracking

Homebuilt Aircraft & Kit Plane Forum

Help Support Homebuilt Aircraft & Kit Plane Forum:

Voidhawk9

Well-Known Member
HBA Supporter
Joined
Mar 26, 2012
Messages
662
Location
Timaru, NZ
There are 'toughened' resins out there that are less brittle, if that is still your concern. You don't want to sand the stuff, though, and it'll give your cutting tools a rougher time as well.
But remember, the reinforcements carry the load, the matrix just holds it rigid, and there seems to be no evidence that micro-cracking is of any consequence, at least not at the level that a homebuilder ever needs to be concerned about.
Use a quality epoxy. Don't roll the dice with it and start experimenting with fillers.
 

wsimpso1

Super Moderator
Staff member
Log Member
Joined
Oct 18, 2003
Messages
8,489
Location
Saline Michigan
The original article described the issue as one seen with composites of glass and polyester and described the solution as being to use generous FOS in that material.

I went on to point out that when you use epoxy or vinylester resin, you already get very large FOS in the resin. We already use FOS of 2 to first fiber failure and get resin failure at 3-5 times the strain of first fiber failure with epoxy and vinylester, so we already use FOS of 6-10 for epoxy and vinylester resins in our typical structures. Using min FOS of 8 in bolted or riveted joints is also the norm, so our typical practice in composites using epoxy or vinylester are already covered for microcracking.

So do not use polyester in serious parts... QED.

Billski
 

AeroER

Well-Known Member
Joined
Oct 6, 2021
Messages
53
The original article described the issue as one seen with composites of glass and polyester and described the solution as being to use generous FOS in that material.

I went on to point out that when you use epoxy or vinylester resin, you already get very large FOS in the resin. We already use FOS of 2 to first fiber failure and get resin failure at 3-5 times the strain of first fiber failure with epoxy and vinylester, so we already use FOS of 6-10 for epoxy and vinylester resins in our typical structures. Using min FOS of 8 in bolted or riveted joints is also the norm, so our typical practice in composites using epoxy or vinylester are already covered for microcracking.

So do not use polyester in serious parts... QED.

Billski
What ae the failure modes you analyze with factor of safety = 8?

The norm at the major aerospace companies is FS = 1.50 for all modes in tested primary structure and later repairs of fighter aircraft. FS=2.0 is common in untested structure such as a modification for flight test. But, we had well characterized allowables, including for bearing. I can just about guarantee the matrix is cracked at every structural fastener due to bearing loads.

On the one hand, FS=8 is awful conservative, on the other it might not be in amateur designed and built aircraft.
 

wanttobuild

Well-Known Member
Joined
Jun 13, 2015
Messages
816
Location
kuttawa, ky
AeroER

May I ask how how the big guys address micro cracking?
Drilling holed in a composite structure freaks me out, cutting the primary load bearing fiber just where it is needed the most.
 

AeroER

Well-Known Member
Joined
Oct 6, 2021
Messages
53
It's ignored, but also Incorporated into the bearing allowables. Allowables are written as the bypass stress v. bearing stress. The bypass load at end fasteners loaded normal to the edge of the part is zero, and this is the location of the basic bearing allowable.

Bearing allowables are very small compared to metals. This drives in added thickness to handle the stress, and larger diameter fasteners to handle the fastener bending. That drives added edge distance, and the combined effect is more weight than anyone wants.
 

wanttobuild

Well-Known Member
Joined
Jun 13, 2015
Messages
816
Location
kuttawa, ky
aeroER

Yep, gotcha. Thank you for your answer, and it seems to me you have a Very Firm Grasp of design!!
I can only assume, the above design criteria applies to all composite materials.

The industry needs a faster that is applied like rivet, has a slightly larger bearing surfaces on both sides of the matrix like a bolt but without expanding in the hole
 

AeroER

Well-Known Member
Joined
Oct 6, 2021
Messages
53
The matrix cracking is caused by load transfer at the fastener. Generally interference fits are avoided, but not the extremely close clearance fits required for efficient joints.

The selection of fasteners is plentiful. Some are too costly for homebuilders, and more so when special installation tools are required.
 

wsimpso1

Super Moderator
Staff member
Log Member
Joined
Oct 18, 2003
Messages
8,489
Location
Saline Michigan
What ae the failure modes you analyze with factor of safety = 8?

The norm at the major aerospace companies is FS = 1.50 for all modes in tested primary structure and later repairs of fighter aircraft. FS=2.0 is common in untested structure such as a modification for flight test. But, we had well characterized allowables, including for bearing. I can just about guarantee the matrix is cracked at every structural fastener due to bearing loads.

On the one hand, FS=8 is awful conservative, on the other it might not be in amateur designed and built aircraft.
A lot depends upon your approach. You might get away with FOS of 4 in non-life safety situations, low cycle counts, or in iterative build-test validated designs. Let's follow the thought through.

If you simply apply a load across a joint with a field of rivets or bolts, and call the load evenly distributed you had better go FOS of at least 4 and maybe 8 for standard failure modes of joined parts. Folks still design joints this way for axial loading, but work out distributed loading when faced with moments. Basic FOS of 2 is used for composite members, then stress concentration of 1.4 to 2.0, which already gets you to overall FOS of 2.8-4. This with the caveat that composites under good load distribution do not need stress concentration factors while composites under less than good load distributions need a lot of stress concentration factor. Shear splitting and resin failures in the bearing area await those who use small FOS in the members.

Then there is the entire issue of uneven load distribution across the joint due to strain being picked up differentially in the mated elements. Even in double shear joints in tension, strain starts at zero (at the free ends) and increases to beyond nominal value as you pass by the last fastener, with the next layer picking up strain in the opposite direction. This adds nominally 2.0 to the scheme (equal stiffnesses from both ends). So now you are up to 5.6-8. When joined parts are of differing stiffness the correction goes higher.

Go to a detailed FEA of the joint with preloaded fasteners in the analysis and you might go back down to just material based FOS. This area has been notoriously unreliable. When used they are dependent both upon analysts skilled in this particular art and upon uniformly applied methods validated to produce reliable results. I have seen failures of these analyses to predict behavior and produced joint fatigue failures at short cycle counts.

FEA approaches are commonly used that do not include the fastener or preload, simply tying the the two members together at the holes. This approach includes differential strain in the joined elements, but exaggerates shear load in the fasteners and requiring raised FOS for reliability. Depending upon how critical the joint is, FOS drift up into the range of 4-8. Get fancy and make the spider elastic, and again validation is usually required before the approach can be considered reliable or drive fatigue tests of the joints before release. In a typical design cycle the modest upsizing upfront is preferable to program delays after failures on test...

These issues drive folks to reduce bearing allowables, which amounts to raising the FOS in these materials. This effect is generally larger in composites than in metals.

There are more and the topic can get complicated.

One of the most interesting points in designing and analyzing joints is how either adding another fastener (or two) or bumping to the next size up runs up the FOS. The weight differences for going with enough FOS to know they will be OK is usually pretty small unless one has a massive number of fasteners. Designing to reduce fasteners is a wise strategic approach.

Billski
 

David L. Downey

Well-Known Member
Joined
Aug 7, 2019
Messages
158
Location
Harleysville, PA
aeroER

Yep, gotcha. Thank you for your answer, and it seems to me you have a Very Firm Grasp of design!!
I can only assume, the above design criteria applies to all composite materials.

The industry needs a faster that is applied like rivet, has a slightly larger bearing surfaces on both sides of the matrix like a bolt but without expanding in the hole
!!! kind of like a tension Huckbolt or Hylock!!!
 

David L. Downey

Well-Known Member
Joined
Aug 7, 2019
Messages
158
Location
Harleysville, PA
The matrix cracking is caused by load transfer at the fastener. Generally interference fits are avoided, but not the extremely close clearance fits required for efficient joints.

The selection of fasteners is plentiful. Some are too costly for homebuilders, and more so when special installation tools are required.
I spent years at Boing looking at virgin, unfilled fastener holes. The very best holes were generated rather than drilled or bored. Even then there was some degree of cantilevered, protruding filaments and damaged matrix at the surface of the bores. I retired before the serious "on-the-floor" utility of picosecond and femptosecond pulsed laser for this task in a productive and cost effective way. But early experiments showed nearly zero protruding filaments and virtually zero heat affected zone or microcracking in the matrix. Those holes had to be generated though so the the spot only touched the material to be ablated. We were never able to measure any heat increase in the laminate after the power, pulse, and feeds were worked out.
 

AeroER

Well-Known Member
Joined
Oct 6, 2021
Messages
53
A lot depends upon your approach. You might get away with FOS of 4 in non-life safety situations, low cycle counts, or in iterative build-test validated designs. Let's follow the thought through.

If you simply apply a load across a joint with a field of rivets or bolts, and call the load evenly distributed you had better go FOS of at least 4 and maybe 8 for standard failure modes of joined parts. Folks still design joints this way for axial loading, but work out distributed loading when faced with moments. Basic FOS of 2 is used for composite members, then stress concentration of 1.4 to 2.0, which already gets you to overall FOS of 2.8-4. This with the caveat that composites under good load distribution do not need stress concentration factors while composites under less than good load distributions need a lot of stress concentration factor. Shear splitting and resin failures in the bearing area await those who use small FOS in the members.

Then there is the entire issue of uneven load distribution across the joint due to strain being picked up differentially in the mated elements. Even in double shear joints in tension, strain starts at zero (at the free ends) and increases to beyond nominal value as you pass by the last fastener, with the next layer picking up strain in the opposite direction. This adds nominally 2.0 to the scheme (equal stiffnesses from both ends). So now you are up to 5.6-8. When joined parts are of differing stiffness the correction goes higher.

Go to a detailed FEA of the joint with preloaded fasteners in the analysis and you might go back down to just material based FOS. This area has been notoriously unreliable. When used they are dependent both upon analysts skilled in this particular art and upon uniformly applied methods validated to produce reliable results. I have seen failures of these analyses to predict behavior and produced joint fatigue failures at short cycle counts.

FEA approaches are commonly used that do not include the fastener or preload, simply tying the the two members together at the holes. This approach includes differential strain in the joined elements, but exaggerates shear load in the fasteners and requiring raised FOS for reliability. Depending upon how critical the joint is, FOS drift up into the range of 4-8. Get fancy and make the spider elastic, and again validation is usually required before the approach can be considered reliable or drive fatigue tests of the joints before release. In a typical design cycle the modest upsizing upfront is preferable to program delays after failures on test...

These issues drive folks to reduce bearing allowables, which amounts to raising the FOS in these materials. This effect is generally larger in composites than in metals.

There are more and the topic can get complicated.

One of the most interesting points in designing and analyzing joints is how either adding another fastener (or two) or bumping to the next size up runs up the FOS. The weight differences for going with enough FOS to know they will be OK is usually pretty small unless one has a massive number of fasteners. Designing to reduce fasteners is a wise strategic approach.

Billski
That doesn't answer the question.

If you are applying FS=8, then what is that written against? The fasteners (shear, bending, tension, or interaction), tearout of the laminate, bearing in the laminate, net section strains in the laminate?

If FS=8 is applied to the fasteners due to approximations that really don't rise to first order analysis, then your freebody game should be improved to calculate reasonably accurate fastener loads, especially in light of peaking at the ends of joints.

That requires good understanding of the laminate modulus, at minimum enough to bracket the value in a shear critical joint. An argument can be made for using identical loads at each fastener in a thin bearing critical joint under ultimate loads, but that does not dismiss the need to check the loads at limit levels.

Another consideration is whether using a giant factor of safety to cover the fastener bending in thick laminates is acceptable. That might be okay for the amateur that barely understands the analysis, but it's not rigorous and carries a fair bit of risk that the resulting margins of safety are not correct.
 

Bille Floyd

Well-Known Member
Joined
Sep 26, 2019
Messages
534
A few here , may not know this ; so i will mention
it now , since it is topic related :

The addition of more than 5% of a reactive diluent
to Epoxy ; it will terminate molecular weight chains.
So use heat to change the viscosity ;NOT a reducer.

Bille
 

AeroER

Well-Known Member
Joined
Oct 6, 2021
Messages
53
Billski, I wish I had seen your question about 787 on page 45 when it was fresh.
 

Rob de Bie

Member
Joined
Feb 7, 2021
Messages
16
The addition of more than 5% of a reactive diluent
to Epoxy ; it will terminate molecular weight chains.
So use heat to change the viscosity ;NOT a reducer.
Bille, do you know whether this also applies to epoxies that are formulated to have a low viscosity? For example the types developed for vacuum injection? I worked with one that had a pretty low Tg (like 80 deg C after postcure) - I always wondered about that.

Rob
 
Top