Ask what failure looks like and people picture units washed downstream. That happens, and it is not the common case.
The common case is an installation where everything is still there and the protection is gone.
The five ways it actually goes
1. Settlement, uniform and invisible
Fines are drawn out from beneath the armor by flow moving through it, voids form, and the layer settles into them.
What you see: nothing. The surface looks exactly as installed.
How you find it: level survey against the as-built. A layer that has dropped is a layer that has lost that much protection, and only a comparison against a datum finds it.
This is winnowing, and it is a filter problem rather than a material problem. See why scour protection fails.
2. Edge retreat
Erosion just outside the protection lowers the bed there. The outermost units lose support beside them and slump into the developing hole. The next row is exposed and does the same.
What you see: at the perimeter, units sitting lower than the rest, a step down at the boundary, or a ragged edge where there was a clean line.
Where it starts: the toe, most often. Then the flanks. See toe protection.
3. Displacement
Individual units moved by flow, ice, debris, or impact.
What you see: gaps in the coverage, units downstream of where they were placed, a thinner layer at high points.
Note the direction. Displaced units downstream tell you it was flow. Units pushed shoreward or lifted onto the bank suggest ice. See ice damage to bank protection.
4. Mesh damage
Cuts, tears, and abrasion wear on individual units.
What you see: open mesh, exposed fill, in severe cases a unit that has lost fill and slumped.
What limits it: the weave keeps damage local, so one damaged unit is one damaged unit. See raschel weave explained.
Where to look: the exposed face, the waterline band where ice works, and anywhere the units bear on sharp rock or a structure.
5. The installation is fine and the problem moved
Worth naming because it gets misread as a failure of the armor.
The protection is intact and performing, and erosion has developed just beyond it, or the channel has continued to degrade, or a headcut has arrived from downstream. The armor did its job over the area it covered; the area it covered was not the whole problem.
See contraction, local, and degradation scour and stormwater outfall erosion.
An inspection sequence
In the order that finds problems earliest:
- Level survey against the as-built. Catches uniform settlement, which nothing else catches
- The perimeter, all of it. Look for lowering just outside the armor, not inside it
- The toe, at low water. Exposure, undercutting, a step down to the bed
- The flanks, where the protection meets bank or adjacent treatment
- Behind and above the crest, for evidence flow got round or over
- The face, for displaced units, gaps, and mesh damage
- Beyond the protection, upstream and downstream, for the problem having moved
- Probe the bed around the edges. Loose material where there was consolidated bed means a hole formed and refilled. See post-flood scour inspection
When to do it
- After any significant event, not on an annual cycle. The bed condition at the peak is not what you can see afterwards
- After the first winter, which shows whether the design suits the site
- After breakup, in ice environments
- Periodically against a datum, so settlement is detectable
What repair looks like
The practical difference between a unit-placed system and a continuous one.
Damaged or displaced units are replaced individually. Coverage is extended by adding units at an edge. Neither requires taking apart a section of structure, and neither usually requires dewatering.
That is a genuine advantage over rigid systems, where the repair is a section, and over cabled systems, where the repair is a panel. It is covered in rock bags vs concrete revetment and rock bags vs articulated concrete blocks.
It is not an argument that maintenance is unnecessary. The realistic model is inspect, and replace units as needed, and a maintenance expectation stated in those terms is more honest than one that implies none.
The thing worth internalising
Most armor failures are not material failures.
Three of the five modes above are about the interfaces — what is beneath the armor, what is at its edges, whether the bed it sits on is stable. A better mesh does not fix any of them.
Which is why, when we are asked what the product’s failure mode is, the accurate answer starts with the filter and the toe rather than with the polymer.