Моделирование CFD снижает затраты на защиту мостов от размыва для Network Rail

Computational fluid dynamics (CFD) modelling is trimming bridge scour prevention costs across Network Rail’s asset base by bringing greater accuracy to assessments. Bridge scour – the erosion and removal of sand, gravel and soil from around a bridge’s foundations caused by fast-moving water and multi-directional currents – is the leading cause of structural collapse for bridges over water. In Europe, 500 scour related incidents were recorded over the past century, according to the Journal of Bridge Engineering, (Vol. 28, No. 4); and given the rising incidence of flooding due to climate change, it’s a problem that’s likely to worsen. Engineers have traditionally prevented scour by building deep underwater pillars, putting down heavy rocks or laying concrete “mattresses” to protect the riverbed. And in instances where granular accuracy isn’t achievable, the tendency has been to opt for a “belt-and-braces” approach that covers possible unknown risks as well as known ones. A key reason for this is the use of conventional methods to assess bridge scour, including 1D/2D hydraulic models. 1D models average flow across cross-sections using the energy equation (typically used to represent the river channel), 2D models compute depth-averaged velocities across a spatial grid (typically used to represent the floodplain) and this hybrid 1D/2D approach blends both methods to balance speed and localised detail. The downside with these otherwise effective approaches – particularly 1D models – is their reliance on parametric formulas and averaging out complex localised flow dynamics, resulting in a tendency to overestimate scour depth. This leads to inflated priority ratings for moderate-risk structures, prompting expensive interventions. In the current era of limited maintenance budgets and growing carbon requirements – together with an ageing asset base – any method that reduces that tendency is a welcome one. Accuracy for complexity Computational fluid dynamics modelling, or CFD modelling – which couples hydrodynamics and sediment transport predictions – is proving its worth as such a method. CFD is essentially 3D modelling for hydraulics, but it can also simulate the complex 3D turbulence and water behaviour around irregular structures relevant to local scour around engineered bridge elements. That’s the message that Network Rail senior design engineer Iain Fox and Andy Collier, technical director of environmental, engineering and risk management specialist JBA Consulting, are driving. It was explained in a paper – Physically Realistic Scour Modelling using CFD to Enhance Infrastructure Resilience – presented at the recent International Conference on Natural Hazards and Infrastructure (ICONHIC2026) in Crete. Collier explains: “When we think about an individual bridge or engineering problem, we naturally lean towards being conservative, a little cautious in what we tell people. If we think the answer is 9.5, we’ll often call it 10. “What we started to realise when we were writing the paper was that if everybody’s rounding up to ‘just be safe’ and then if everything becomes high risk, the idea of high risk has lost its meaning.” CFD modelling can play a role in building a more nuanced picture of risk, Fox says. He explains that Network Rail uses a two-stage process to assess scour risk, the first of which is a high-level look at averages. “For example, to work out how fast a river flows, at stage 1 we’ll look at the contour levels at different points of the river and work out what the gradient of the river is, which will give us roughly what the velocity of the water is. “If we get a high risk at that point, we go on to a stage 2 assessment, which then looks at doing a hydraulic model, drilling through the structure to find out exactly where the foundation depth is in relation to the riverbed. That will then tell us how much of that riverbed needs to erode before the structure becomes at risk. It involves lots more measurements and understanding of flow rates and velocities within that channel – the way we currently do it is with the one- dimensional or 1D hydraulic model. “The CFD modelling solves more complex equations and shows you in detail what is happening with the water at different heights.” CFD’s inputs rely on the Navier-Stokes equation – a set of mathematical formulas that describe how fluids move. CFD modelling can capture velocities from the bed and predict outcomes from a range of impacts, for example shear stresses – the forces of moving water against a riverbed, which cause scour by pushing and lifting sediment particles away from bridge piers and abutments. The complexity captured by CFD is a key point of difference says Collier. For one thing, the 1D model is limited in how it captures water movement. “1D only captures one direction that water is moving, either backwards or forwards, but it can’t go left and right, and it can’t calculate the water moving up and down. It doesn’t allow for any variation in the velocity through the vertical axis and it doesn’t allow for any variation across the width axis. “Then we’ve got another option, which is 2D, meaning the water can go in two directions. That means the water now can vary within the channel or whatever we’re modelling, and it can vary by the water moving forwards and backwards as it did before, but it can also now vary that the water can move left and right.” By contrast, with 3D CFD, “the water is now able to move as it does in reality,” he says. “Turbulence is now fully represented along with other physical phenomena like superelevation, where the water is higher on one side than the other, which might be when it’s changing direction or momentum. “This means the water wants to keep going in one direction, but the physical structure forces it to change direction. That’s when we see it slow down and then kinetic energy gets transferred into potential energy and it rises basically on that side. CFD represents all this using the Navier-Stokes equation.” A key benefit is that it supports modelling of the riverbed itself, says Collier. “What that means is we can put in a mixture of materials, or we can say that we’ve got layers that are also either single-property or a mix.” The variability of materials is a critical factor in terms of assessing scour risk, Fox notes. “If you start digging down in that riverbed, that material changes very quickly. It’s not silt all the way down to the Earth’s core, it changes in layers. Scour is self-limiting, because a growing scour hole enlarges the flow area, which reduces water velocity and vortex strength until the flow can no longer lift and carry away local sediment. Silt has really fine particles, so it doesn’t take much to move it away – as the scour hole gets deeper, the material changes.” CFD  uses  The Shields parameter – a dimensionless number used in fluid mechanics and sediment transport to determine when sediment particles will begin to move along the bed of a waterway – to work out the energy required to lift particles of varying sizes up and out of the scour hole. And the model can show the scour hole at its deepest point at which it can’t lift any more material out, Fox explains. “We can then utilise that information to determine where that appears to be in relation to the foundation of the structure and is it going to cause a risk?” Comparative case study Collier and Fox shared an animated presentation with  NCE  about a recent project at Shrewsbury railway station bridge over the River Severn, demonstrating the capabilities of CFD modelling as against conventional approaches. Collier notes: “You’ve got the main channel of the river (left hand side) that expands during flooding and runs through three spans of the bridge (right hand side) that are usually dry until a flood event. When the river floods, the arches allow more water to go under the railway.” The CFD shows the riverbed and areas in front of the bridge piers where the riverbed has dropped, with yellow representing a 3m-deep scour hole, while red is a 4m-deep scour hole. The model also uses streamlines – lines drawn to illustrate how the water varies. “At first, it’s very uniform,” says Collier. “The lines are parallel and side by side, with the spacing between them the same. And then when we reach the object, we can see how the flow pattern varies and how it impacts the velocities.” Critically, the Stage 2 scour assessment using a 1D model had predicted that the scour would occur in one area towards the middle of the bridge with a depth of around 12m. However, Collier notes that the CFD model produced a “drum roll” revelation. “In fact, the scour is in a different location and at a vastly reduced depth,” he says. “There’s still four-and-a-bit metres of scour in that hole, imagining an awful lot of energy that’s driven that scour process – that’s the reality of the 200-year flood event, which is a large event – but that’s considerably less than 12m.” Aside from the discrepancy between the predicted and actual depth, “the key for us is that if we had just used the stage 2 results, we would have defended the wrong bit of the structure,” he says. “The river comes along and turns as it goes under the structure, so during the stage 2 assessment it is not unreasonable to conclude that water would be hitting the central section of the bridge (main channel) and that would be the bit that’s going to be most at risk. “However, when you think about what’s happening during a flood event, the water is now out of channel and it’s flowing through the whole width of the bridge, not just the channel itself. So the central part of the river is in fact in what we call the right-hand abutment before you get to the flood relief arches. “It’s the CFD model that says ‘this is where it is now’. That wouldn’t have been evident looking at the Stage 2 1D model.” Embedding adaptation Fox confirms that in 2023, CFD was integrated into the standards as an advanced three-dimensional (3D) modelling tier under Issue 3 of the Network Rail standard NR/L2/CIV/295 (Scour Assessment of Bridges, Culverts and Retaining Walls). He says this development aligns with Network Rail’s efforts to CFD model around 30 to 40 structures to date, adding that “there are more in the pipeline to do”. “We are using this software to make engineering decisions and to decide where money is going to be channelled to the highest risk structures,” he says. Considering the total potential cost of intervention, those decisions could be game-changing. “There are many things that we can do to protect a structure from scour, but the cost of doing that is a lot per structure. One of the costliest factors is getting to them; most Network Rail bridges are not conveniently at the end of a motorway, so access is terrible and when we’re trying to get machinery and plant and so on there, we’ve got to build a new trackway and numerous other temporary solutions. There are also obviously added complexities with working in a river.” Aside from the potential cost savings of avoiding such complicated works when they’re not strictly necessary, an additional benefit of CFD modelling is that it can envision the wider impact of planned interventions, he says. “Have we passed the risk downstream? Have we increased flood risk? Have we interrupted a flow or ecology path? Bodies such as the Environment Agency, Natural Resources Wales and the Scottish Environment Protection Agency are increasingly asking for proof that any engineering work that we do is not going to impact the wider catchment area. So this is of benefit in that respect too.” The overarching theme that supports wider uptake of CFD modelling is ultimately one of resilience – and not just for railways, Collier concludes. “This is ultimately about how we enhance the resilience of not just the rail network, but the road network and properties and so on in adapting to climate change,” he says. “The potential for us to remove or reduce the number of at-risk structures is making it more manageable for asset owners. And most importantly, the money saved can be diverted to assets that really do need protection.”