The intuitive model of armor failure is that the design event exceeded the design capacity and the units washed away. That does happen. It is not the common case.
Far more often, a countermeasure is found to have failed while every component of it is still present and undamaged. The stone is on site. The mat is still cabled. The bags are whole. The protection is simply no longer between the flow and the foundation.
Four mechanisms account for most of it.
1. Winnowing
The armor layer stops the flow lifting the large material. It does not, on its own, stop the flow lifting the fine material underneath.
Water moves through the voids in any permeable armor layer. Where the bed beneath contains sand and silt, that flow entrains the fines and carries them up and out through the layer. The armor stays exactly where it was placed. The material supporting it leaves.
Eventually the layer has nothing beneath it and settles. From above, nothing has moved until the day the whole layer drops.
What it demands: a filter. Either a granular filter of intermediate gradation, or a geotextile, sized so that the bed material cannot pass but water can. Getting this wrong is probably the most common single defect in armor design, and it is invisible until it is not.
Note that this is a property of the interface, not of the armor product. Any permeable armor over an erodible bed has to answer the question. Where a system contains its own graded fill retained by a mesh, the fill is not being drawn out of the unit — but the question of what sits between the unit and the natural bed still has to be resolved on a site with fine material.
2. Edge failure
Every armor installation has a perimeter. At that perimeter there is a step from an armored surface to an unarmored one, and flow works at discontinuities.
Scour develops just outside the edge. The hole deepens. The armor at the boundary loses the support of the bed beside it and slumps into the hole. That exposes the next row, which does the same thing.
The failure travels inward from the perimeter, and the middle of the installation — the part everyone looks at — is the last to be affected.
What it demands: termination detail. The three that matter are the toe (keyed down below anticipated scour, or extended far enough that a developing hole cannot reach the structure), the flanks (tied into the bank or into adjacent protection with no gap), and the crest (carried high enough that flow cannot get behind it and attack from the landward side).
An armor layer is only as good as the weakest of its four boundaries. In practice, the toe is the one that fails.
3. Undermining and void formation
Related to winnowing but distinct in cause: a void forms beneath a rigid or semi-rigid armor system because the bed moves and the armor does not.
A concrete slab, a grouted mattress, or a cabled block mat holds its shape. If the bed beneath it degrades — through the long-term channel lowering described in what is scour, or through a scour hole developing at the edge and working underneath — the armor is left spanning a gap.
It can hold for a long time. Then it cracks, or it drops, and the void it was covering is exposed all at once.
What it demands: either enough rigidity and reinforcement to span whatever void may form, which is expensive and hard to guarantee, or a system flexible enough to follow the bed down and stay in contact with it. This is the trade at the heart of the choice between rigid and flexible armor, and it is covered in the comparison hub.
4. Displacement of individual units
The mechanism everyone designs for: hydraulic forces exceed what holds the unit in place, and it moves.
Worth noting what actually governs, because it is not simply weight:
- Weight resists lifting, but the relevant quantity is submerged weight, not dry weight.
- Interlock and packing matter enormously. A well-packed layer where units bear on each other behaves very differently from the same units placed loosely.
- Shape and surface affect drag and lift.
- The unit that moves first is at an edge or a high point. Once one moves, its neighbours are exposed on a new face.
Riprap loss is usually progressive in exactly this way: a few stones at the toe or the perimeter migrate, the layer thins there, more move, and the reach needs replenishment. That maintenance cycle is a normal and accepted model on many sites. It becomes a problem when the same reach is being replenished after every significant event, which is the point at which lifecycle cost rather than first cost should drive the decision.
What all four have in common
None of them is about the armor being insufficiently strong.
Three of the four are about the interfaces: what is under the armor, what is at its edges, and whether it stays in contact with a bed that moves. Those are design and detailing questions, not product-selection questions, and a better product specified with the same defective details fails the same way.
This is why a countermeasure design that spends all its attention on unit sizing and none on filter, toe, flank, and crest is incomplete regardless of what it selects.
The inspection implications
Because these mechanisms are mostly invisible from above, an inspection that looks at the armor surface is looking at the wrong thing. What is worth checking:
- The perimeter, all of it, for local lowering just outside the armor
- The toe, for exposure or slumping
- The flanks, where they meet the bank or adjacent protection
- Level, against a previous survey, to catch uniform settlement that is otherwise invisible
- Behind and above the crest, for evidence of flow getting round it
- After a significant event specifically, since the bed condition at the peak is not the bed condition you can see
A layer that has settled six inches uniformly looks completely normal and has lost six inches of protection. Only a comparison against a previous level finds it.
Where to go next
- What is scour? for the underlying process
- Bridge pier scour for why the protected footprint exceeds the structure’s
- The comparison hub for how systems differ on flexibility and interface behaviour
- When not to use rock bags for the cases our own data does not cover
- Installation and site operations for placement, keying, and post-installation inspection
If you have an installation that is not performing, level survey against the as-built and photographs of the perimeter are the two most useful things to send through the quote form.