Introduction to Composite Materials

The basic idea behind a composite structure is that by using two materials together the builder can get better performance than either one is capable of on its own.  This is not a new idea.  Rebar is a classic example, the concrete is reinforced by steel rods that can better support tensile loads.  Concrete itself is a composite, gravel and ash reinforces the cement that holds the material together.  The oldest man-made composite material is called cob, a building material of mud or clay brick that uses sticks and straw for support.

Each of these composite materials is composed of two parts: matrix and reinforcement.  Matrix gives the structure its shape and provides rigidity.  Reinforcement provides strength to the final structure.  Reinforcing materials are often much stronger than the matrix, but would make a poor building material because they bend easily like straw, or don't form a structure like the gravel and ash used in concrete.
Long before humans were making cob nature had produced composite structures.  The best example is also the oldest composite building material used by man: wood.  Wood is composed of long fibers made from the cell walls of the original tree bound together by a glue-like substance called lignin.  In this case lignin is the matrix and the fibers are the reinforcement.  The strength of wood comes from the fibers which are supported and held together by lignin.  In the direction the fibers run, called the grain, the wood is stronger than in any other direction.  However, because the lignin soaked fibers are more stiff than the lignin itself, against the grain the wood is more flexible.

It's rare that the fibers themselves break.  When a piece of wood breaks it is usually in the lignin first, cracks form where it begins to fail or where it separates from the fibers.  That's because the fibers are so much stronger than the lignin is.  When you glue two pieces of wood together the glue becomes the weakest part.  Most glue isn't as strong as lignin.  Even if it was there's no fiber in the glue bond for reinforcement, only the glue is carrying any loads.
Materials like wood are very different from those like metal or stone.  Metal and stone have nearly the same properties in all directions, given the term isotropic.  Wood is an anisotropic material, meaning that its properties depend on what direction it is assembled and loaded in.  Composites like wood, which act differently in different directions, can often act in ways that seem counter-intuitive to people who aren't accustomed to dealing with them.

Modern composites are very much like wood, they are composed of fibers supported by a resin matrix.  Composites are formed of thin layers of combined fiber/resin bound together by resin.  They are strongest in their fiber direction (called the principle direction), weaker in the other direction within a ply (called the transverse direction), and weakest between plies (the normal direction).  In a manufactured composite the fibers can be either long, like in wood, or short, like in concrete.

Short-fiber composites are made of a matrix supported by chopped fibers or pellets of random orientation.  These materials are generally isotropic, because the fiber reinforcement is in all directions evenly.  They are also very easy to produce.  Adding chopped fibers of a much stronger material significantly increases the strength of the final product, just as adding gravel increases strength over the base cement.  However the matrix is still carrying much of the load, since it has to transfer forces between the various pieces of reinforcement. 

In long, or continuous, fiber composites the fibers extend the entire length of the part.  In this case it is the fiber carrying the load and the matrix serves to bind fibers together and ensure they don't bend or buckle.  Fibers are strong but they are also flexible and will bend rather than support a load in compression.  Matrix material gives the fiber their rigidity and holds them in shape.  Typically long fiber composites are nonisotropic, since they are strongest in their fiber direction.  In the other direction they act more like a short fiber composite, where the fibers provide reinforcement but the matrix has to carry the load between them.
Fibers can be produced in many ways.  Terms for composite fibers are borrowed from textiles, since that's where many of the processes that produce them came from:

• filament - a single fiber, much longer than it is wide.
• yarn - a bundle filaments that are lightly wound together, may or may not be twisted
• tow - a large (3K - 12K) bundle of loosely bound filaments that are not twisted together
• fabric - a large flat weave of filaments, yarn, or tows interlaced
• mat - a random assortment of short fibers laid flat

             

When fiber is added to produce a part it can be either unidirectional or bidirectional.  As these terms imply unidirectional material has all (or most) fibers pointed in one direction.  Fabric can be produced this way with relatively small reinforcing fibers woven around the main fibers, to hold the fabric together.  These light fibers don't end up supporting the final part.  Bidirectional fabric has two primary directions, ninety degrees apart.  While bidirectional parts are strong in two directions, rather than one, because half their fibers are pointed in a different direction they aren't as strong as unidirectional material is in the primary direction.  Uni, as it can be called for short, is only strong in one direction but in that direction it is twice as strong as bidirectional material.  Both types tend to be weak in shear without some other support.

Once produced fiber must have matrix added.  This can either be done by the manufacturer of the fiber or when the composite part is being produced.  If the material is pre-impregnated with resin by the manufacturer then it is called pre-preg, for short.  Although pre-preg material is often more convenient it also has a shelf life, should be kept refrigerated, and requires heat to cure properly.  Dry material can sometimes be easier to shape, but care must be taken when adding wet resin that the part is wetted fully.

Pre-preg is often sold as tape.  A unidirectional or bidirectional fabric is impregnated with resin, then wound around a spool with backing material.  Just like an ordinary roll of duct-tape, itself a composite material of fabric saturated with adhesive.  Prepreg tape should be stored in a freezer until it is ready to be made into a part and should be heated in an oven to cure it.

Lay-up is the process of shaping a composite part.  Typically parts are laid up around a mold, or tool.  Most tools are either male, meaning the part is laid up over the tool, or female, meaning the part is laid up in the tool.  Two-part tools squeeze the lay-up between male and female halves.  When laying up the part in or around a mold there are several options for getting the fiber and matrix where you want them.

The simplest process is wet lay-up, or hand lay-up.  In this case technicians carefully drape wetted fabric on the tool, one layer at a time.  Each layer is either wetted by hand or prepreg is used.  Technicians must be careful to squeeze each layer down to ensure that the layers are closely bonded, the denser the fibers are the stronger the final part will be.  Hand lay-up is easiest with fabric materials or tapes, since it isn't easy to handle individual filaments or tows.  The limitation is naturally that parts must be designed so that fabric can be easily draped over the tool and the natural tendency for human error.  Fabric can also fold at corners, or bridge up when it should fit tightly.   

Another method is to have a machine lay wetted fiber tows over a mold a few at a time.  A filament winding machine tracks over a male mold laying down fibers in the desired direction.  The mandrel with the mold rotates as a carriage moves back and forth along its axis.  On the carriage a head dispenses impregnated tows over the mandrel.  The simplest machines have a one-dimensional head, it moves back and forth along the axis alone, as pictured.  For this simple winder there is a limitation on fiber angle, as you can see from looking at the picture it would be impossible for the carriage above to lay fibers that ran completely along the axis of the mandrel.  Simpler winders have a minimum and maximum angle that they can lay. 
More complex machines have two axis heads that also move up and down and three axis heads that move around the mandrel.  Mandrels can also be designed to rotate in two or three directions, rather than the simple lathe model pictured.  Six axis filament winders can produce parts with virtually any fiber orientation.  But the more complicated the machine the more it costs.  Even simple filament winding machines can become very expensive.  Filament winders are also limited in the size of part they are capable of producing, the larger the winder the more its price goes up. 

To get around that we can use a system called pultrusion, a combination of the terms pull and extrude.  In this kind of production fibers are pulled off a reel, through a resin bath and cure process, and then chopped off at the desired length.  Since the cure process is a part of the machine parts are ready to be trimmed and used as soon as they come out.  A pultrusion machine can produce linear parts of a simple cross-section at any desired length, simply by cutting off the desired length as it comes off the machine.  After the machine itself is purchased this process is cheap, simple, and fast.  Obviously, however, the machine is restricted to only those simple shapes.

Typically after laying up a part it is necessary to cure the resin, when the adhesive sets and hardens in place.  Many wet lay-up parts are made using a form of two-part epoxy and can be allowed to air cure.  Air cured parts are simply those that cure on their own at room temperature given time.  Most adhesives will eventually start to cure some time after they are mixed, but many won't do so at a rate that is reasonable for production.
Many resins cure faster, or only, at elevated temperature, so an oven is often used. In principle no different than an ordinary convection oven, most composite ovens are much larger but much more limited in temperature.  Ovens are limited in what size part they can handle.  Larger parts require larger ovens, which become increasingly expensive.  The size of an oven also limits how many parts can be cured at once.
Oven cures do introduce another problem, though.  The mold a part is built around, the matrix, and the fibers all expand when they are heated.  Each expands differently, which creates a new source of stress in the part when it cures.  Each layer of the final part also wants to expand in a different direction, because their fibers point in different directions.  This can be dealt with somewhat by carefully laying up the part and choosing the mold material.

It's desirable that as the part cures it also be compacted, to bring plies closer together.  Once each layer is fully wetted extra matrix only weakens the final part, and extra matrix between plies adds more area that can fail.  After laying up the part we want to see the extra resin squeezed out and removed.  This can be accomplished by vacuum bagging the part.  In a vacuum bagging process the part is laid up, covered by fabric to absorb the excess resin and distribute air-flow evenly.  Then it is covered by an airtight bag.  Air is removed from the bag by a vacuum pump, pulling the bag tight against the part and squeezing excess resin out of the part.  The part is left under the bag until the remaining adhesive completely cures and is then removed and trimmed.  Bagged parts can be placed in an oven to cure, a vacuum pump outside is connected through ports in the oven to parts inside.  Heat can also be applied using electric heat blankets which, much like their consumer counterparts, are insulating blankets with heating elements built in.  Placing the blankets over a bagged part provides heat and compaction.

Instead of removing the air and allowing atmospheric pressure to compress the part another system is to apply pressure to the outside of the bag, which accomplishes the same thing.  An autoclave is a large pressure vessel designed to accommodate bagged parts and molds.  The autoclave ramps up pressure then holds it at some desired amount while the part cures.  Autoclaves are often combined with vacuum bagging.  Inside the autoclave are ports connected to a vacuum pump outside.  As pressure is increased in the autoclave it is also removed from the bag, pulling the vaccuum down tight.  Vaccuum can be maintained for the entire cure cycle or released when the autoclave reaches pressure. 
Heating elements are incorporated into autoclaves, so in addition to adding pressure the autoclave adds heat.  So autoclave bagged parts are also cured at raised temperature as if an oven.  Autoclaves are limited to a certain size of part and can only accommodate so many parts at once, which leads to limitations on the size of part that can be cured and limits the production rate of a facility.

Compression molding ensures that all molds that are in use are providing compaction.  It does this quite simply by making the tool itself the source of compaction.  A piece is laid up in a two part mold and the two parts press together, providing compaction and squeeze out.  Pressure can be applied to two hard tools by a press.  Instead of a hard tool one of the two parts can also be an inflatable bladder, as the bladder inflates it provides compaction to the part.  This process produces parts with a smooth finish and allows for complex part shapes.

If we extend the idea of compression molding slightly we can actually lay up the part with dry fabric and then add wet adhesive after clamping the two-part mold together.  Resin transfer molding can greatly reduce material costs by limiting the amount of trim needed and allows for complex parts.  Since resin transfer uses dry fabric it avoids the shelf life issue that comes with prepreg. 

The issue with both is the dedicated, expensive tooling required.  Each mold is designed specifically for one part, so high part count assemblies require a very high tooling investment.  Molds can be built with heating elements included in their design, which allows the part to be heated to cure without placing it in an autoclave.

During lay up parts are formed from multiple layers, or plies, of material held together by resin.  Each ply is called a single lamina, which are laid up together into a laminate.  The whole laminate together has properties that depend on the individual lamina that make it up.  Because each ply has different properties depending on what direction its fibers are directed the properties of the whole laminate depend on the order those plies are laid up. 

Though there are an infinite number of choices a composite designer could make when deciding how to design laminates, many of these will react in ways that are difficult to predict.  There are some special lay-ups, however, that are worth considering here:

• balanced - for each ply facing in one direction there is a ply facing perpendicular.  That is, for each ply facing one direction there is another facing at ninety degrees to it.  This is always the case with bidirectional fabric.
• symmetric - plies are arranged so that those on opposite sides of the laminate are at the same angle.  So if the top ply points straight ahead so does the bottom ply.  If the second ply is at ninety degrees then so is the second to last.  A symmetric laminate can help reduce the stress a part suffers when it is heated during cure.
• isotropic - laminates that are both balanced and symmetric behave as if they were an isotropic material.  So they have the same properties no matter what direction they are loaded in and their reactions become easy to predict.

Anisotropic lamina can be used in specific situations but they must be very carefully analyzed to anticipate their reactions.  For example the X-29 experimental plane used the fact that its composite wings would tend to bend under load to limit the angle of attack.  As lift increased the wings bent forward, which reduced the lift.  This ensured that the plane would remain under control.  For most purposes, though, designers still prefer isotropic laminates.

There are cases where fiber and matrix plies are not the only components that must be included in a structural composite.  Parts that need to resist bending or buckling work best when their strongest components are removed from the center, the reason why I beams have their shape.  In some cases parts need to have a specific cross-section for aero- or hydro-dynamic reasons.  In both cases we want to increase the area of the part without adding too much weight.  The easiest way to do that is to separate the faces using some sort of filler core material, creating a sandwich structure with the two face sheets as the bread and the filler in the center.
One of the most common fillers is honeycomb core, composed of thin walled, mostly empty cells.  The core is light but surprisingly strong and good at resisting compression in the cell direction.  Face sheets are bonded to both sides and cured in place, with the core present.  Honeycomb supports the face sheets and provides the shape desired without much additional weight. 
An everyday example of a honeycomb sandwich material is corrugated cardboard, the faces are flat cardboard but between them is a thin walled paper material.  The combined structure is surprisingly strong for its light weight and cheap construction.  Core can be made from a number of materials, including Nomex, aluminium, and even paper.  Most core only bends in one direction, and resists bending in the other.  If the face is punctured core cells can also fill with water, dramatically increasing weight.

Foam can also be used as a filler.  Porous foam tends to be more flexible than honeycomb material but softer and easier to compress.  Since it's more flexible it can be used to create more complex shapes than honeycomb, but is typically used as a filler between honeycomb core sections where honeycome cannot be used.
Potting compound is a porous, putty like bonding agent that can be used on its own or between sections of other core material.  Potting compound is easy to shape and fairly strong but tends to be heavier in bulk sections that either foam or honeycomb.  Typically potting is used as a way of filling in between other core sections or building up a non structural surface.
Laminate cross section can be varied through a part by adding or removing plies and core material.  This can be done to reinforce high stress areas or to build up a shape for other reasons.  It's rare that we want core material to be exposed to the environment; it should be completely covered.  Stepping down and tapering the cross-section is one way of doing this.  Another is to wrap material around the edge of a part to cover the exposed core at the edge.

After we make the composite parts we need to assemble them into the final product.  Early on when designers didn't have much experience with composite parts they would simply replace existing aluminum pieces with the new, lighter, carbon fiber parts.  These black aluminum designs didn't use many of the biggest advantages of composite design and often encountered problems that the aluminum parts didn't have, because carbon fiber isn't aluminum and doesn't act like it.
When attaching two composite pieces together, or when attaching a composite part to a metal one, there are two methods that can be used.  One is to use fasteners, like bolts and rivets, as is frequently done with metal parts.  Installing a fastener means drilling a hole through the part, disrupting its fibers and creating a location of increased stress.  Fasteners often pull against the laminate, which pushes some plies against their fiber direction.  Over tightening a fastener causes additional damage as it crushes the laminate and any core material, creating unbonds that grow out away from the fastener.  Under certain circumstances fasteners can be torn completely through the laminate, causing significant damage where they pull out, even if the part is held in place by other fasteners.

Another method is to bond the laminate in place, the same way it is held together.  Adhesive can be used to affix laminates in place.  There are several different kinds of bonded joints that can be made, each with its own advantages and limitations.  In general the more surface area of connection there is between the two parts in question the stronger the joint.  It's also important that the two fit as snugly as possible, with only as much adhesive as necessary.  Areas that are just adhesive will be much weaker than the parts themselves.  Anyone familiar with wood working will be familiar with the issue of when and how to use glue and nails, adhesive joints and fasteners follow nearly the same guidelines.

In general to avoid black aluminum construction it is smart to avoid joints and fasteners as much as possible.  Low part counts are good for a number of reasons.  First, remember that the fibers are what carries most of your load, long continuous fibers can do that job better than ones that are broken up or interrupted.  Second, joints and fasteners are weak points, fasteners weaken the parts and joints bond using weaker adhesive.  Third, adding connections increases weight.  Fasteners and joints require reinforcement, they add new parts, they need overlap, all this increases the weight of the final product.
Of course there are also reasons why you might need to fasten things.  We already said there's a limit to how big parts can be made, so breaking a structure down into smaller parts has to be done so you can build it at all.  Smaller parts are easier to replace during maintenance.  Sometimes it just isn't possible to have a composite part do something, so metal parts have to be attached to the structure.  There are always reasons why assembly is needed, but in general it's something that should be reduced as much as possible.

Once a part is built it can be damaged by use or carelessness.  Since matrix is the weakest component it is typically where failure occurs.  Adhesive failure occurs when the adhesive itself breaks, but remains attached to the fibers, the adhesive itself it weaker that its bond to the fiber.  Bond failure occurs when the adhesive fails by releasing the fiber or ply it is attached to, but remains intact.  Usually we see some combination of the two, but talk about whichever it was that started the break as the source of failure.
Fiber failure is less common, but can occur since the matrix can often by more flexible than the fibers are when the part is under tension.  When the part begins to fail it stretches, when it stretches the fiber breaks first, then without the fiber support the matrix fails.  Fiber pullout occurs, with broken fibers sticking out of broken matrix.
When a crack grows through a ply we call it in-plane fracture.  This means that it is the adhesive in a single play that failed, either by releasing the fiber or by breaking itself.  These cracks look very much like the cracks that grow through a piece of wood, rapidly expanding along the matrix between fibers.

               

It's more common to see the resin holding two plies together fail, called interlaminar fracture.  Remember that although the whole laminate together wants to behave a certain way, each ply wants to act differently.  The plies are forced to act together by matrix between them, stressing the matrix even if nothing else is.  Eventually this can allow two plies in a laminate to come unbonded from each other.

Even if the face sheets are undamaged it is still possible to see core damage occur, especially when the core is weaker than the face material.  It is also possible to see both undamaged but the adhesive between them allow skin to core unbond.  This frequently happens when the core is poorly shaped or placed, leaving a gap between the face and core, and when impact damaged crushes the core slightly, without damaging the face sheet.  It's important to remember that every part of a composite material has different properties so each part wants to do something the others don't, which creates stress when they are all held together.

It's often difficult to spot damage to a composite structure, because the damage is often below the surface while surface plies look completely undamaged.  Spotting damage early can become very important since composite materials are often more brittle than metal components.  Athough ripping a piece apart will definitely spot any damage, we want to keep parts in service if we can.  So we need some form of non-destructive testing (NDT) for composites that will let us find sub-surface damage. 
The first step in scanning a part for damage is a simple visual inspection.  Trained technicians can sometimes spot signs of damage below the surface of the part, even though the surface is undamaged.  Damage can appear in the form of light spots, sudden color changes, or changes in light transmission through a part. 
Another simple inspection technique is the coin tap test.  Disbonds below the surface of a laminate often sound hollow when tapped.  So by tapping a suspect area with a coin, key, small hammer, or similar item an inspector can quickly check for near surface failures.  Both of these methods are simple, easy, and cheap.  They are also imprecise and prone to human error.
Since voids in the laminate have different sound transmission than the undamaged laminate sound is an obvious choice for inspection.  Ultrasonic testing (UT) uses high frequency sound transmitted through a part for a check.  A receiver listens to the sound emitted by a transmitter and reports changes.  The probe listens for either the echo of the initial pulse, reflection, or is placed on the opposite side and listens for the through transmission, attenuation. 

UT systems come in many varieties.  Small, portable kits are available that can be readily moved to a part under inspection and used to inspect suspect areas spotted by visual inspection or known problem areas.  These kits don't record any information, so notes have to be taken by the operator.  Larger UT machines can automatically scan a part by moving the two probes in unison across both faces of a part.  A computer controls the movement of the probes and records the results of the scan.  UT machines can even be designed to record the depth of voids, creating a 3 dimensional map of the part.  However increasingly complex UT systems are increasingly expensive, and there is a limit on the size of parts that can be fit into an automated system.
Fibers are thicker and more dense than matrix.  So the fibers block more electromagnetic radiation than the matrix.  Just like bone is more dense than muscle, and for the same reason X-rays can spot damage.  Many fabrics and tapes also include by design additional, nonstructural, fibers that are made specifically to be visible to X-ray, called tracers.  X-ray inspection involves taking X-ray photographs of a part, looking for regions of less dense material or flaws in the tracer direction.  The downside of X-rays is that the image is two dimensional.  Multiple layers stacked above each other can create a confused image, because the fibers of multiple layers are superimposed on the same image.

  

The issue of superimposed plies can be resolved with a computed tomography (CT) scan.  This kind of system uses a digital X-ray detectors to generate slices of a part.  A collection of X-ray detectors and emitters are positioned on opposite sides of a rotating head, as the head moves it generates images of thin slices of the part.  The computer can then combine all these slices to create a three dimensional image of the interior of the part.  This allows inspectors to detect internal voids that are too deep for UT and would be disguised by X-ray superposition.  It suffers the same drawback as X-ray scans, in that it is limited by the available space in the CT machine, just as X-rays can only be taken when the part fits in an X-ray booth.

Once damage is identified we'll want to fix it.  Most repairs that we conduct on a part are either cosmetic, temporary, or permanent.  Cosmetic repairs are surface deep, repairing damage that doesn't affect the load bearing ability of a piece.  This can be done not just for looks, but also to ensure a surface will be the right shape to fit when it is assembled or to produce a smooth, low drag part or to seal the core to prevent water from getting in.  Temporary repairs are intended only to last until the next regular maintenance cycle a part goes through.  Though the repaired part might not have the same profile it once did, it should be as close as possible to its original strength.  Permanent repairs attempt to restore a part as close as possible to its original properties.  Once a permanent repair is done the part should be like new, though the repair will usually be slightly weaker than the original part was.
When a scratch or depression on the part doesn't get through the topmost layer a very easy surface repair can be done.  In this repair the defect is filled with adhesive, cured, and then sanded smooth.  The repair doesn't add any strength, it just smoothes out the part.  It is a purely cosmetic repair.  In principle is isn't much different from patching a hole in a wall with stucco, it seals the hole but is much weaker than the original material.

Another easy repair when two plies have separated is to simply drill down to the void and fill it with adhesive.  The operator drills two holes and injects adhesive through one of them.  Air is forced out the other hole until the void is filled with adhesive.  To improve flow a vaccuum pump can pull air through one hole as resin is injected into the other.  Once the matrix cures the two layers are bonded together again.  Unfortunately it does mean damaging the top ply by adding drill holes, but the top ply shouldn't be weakened too much since the holes will be fairly small.
When simple repairs are not enough a patch over the damaged area is needed.  Patch repairs work just like the patches on clothes.  By covering the tear with undamaged fabric you can keep the hole from getting any bigger.  Patches bridge the load gap the hole creates, carrying the stress around the tear instead of to it.
A simple patch is just layers of material bonded over the top of a damaged area.  Though relatively easy and quick such a patch also adds thickness and weight to the final part.  This changes the profile of the part, which can hurt its fit in place or increase drag.  The added weight is also usually undesirable.  Simple patches are often done as temporary repairs, to hold until other facilities are available.

To reduce the added weight and thickness of a patch we can sand away the damaged area first.  A taper sand repair removes the original damaged material, then builds it back up with undamaged plies.  The operator sands away material at an angle, taking off a little less material as he goes down.  This requires a skilled technician and a wider area available for repair.  But the result is a stronger, thinner patch that comes closer to restoring the original part strength.  Because there is more connection between the patch and the laminate it is able to transfer the load more effectively, which means the new patch acts more like undamaged skin.

Similar to the taper sand an operator can sand away part of each ply one at a time.  So as he moves down he removes less and less of the area around the damaged area.  Known as a step sand repair, the goal is to remove the plies individually, instead of removing material at an angle.  This kind of repair requires a very skilled operator, and is more difficult than a taper sand.  The resulting repair has more contact area between the repair plies, without the bending that occurs at the edge of plies in a taper sand, potentially making a stronger repair.  Step sand repairs don't substantially affect the total thickness of the patch compared to a taper sand repair, and the potential increase in strength isn't often seen since the difficulty of the patch means the two come out very similar.
If the core itself is damaged then the only repair is to remove it.  A core unbond can be repaired by a drill and fill operation, but damaged core requires a core replacement.  In this operation the technician sands through the face of a part down to the core, then drills or cuts out the damaged core section.  Another section of core material is cut that fits in the hole left by removing the damaged section and fixed in place with adhesive.  Potting compound can be added to fill in any gaps.  Then a step or taper sand repair replaces the removed face plies.  Sometimes the damage penetrates the whole laminate.  In this case on face is removed to do the core replacement.  Then the other face is removed in a step or taper sand.