Advanced numerical modelling in geotechnical engineering is often described as the sophisticated end of analysis. That description is incomplete. A model can be technically complex and still add very little to a project. The real value of advanced modelling is not the software, the mesh or the number of output contours. It is whether the analysis changes the quality of the engineering decision.
For many projects, conventional calculations, limit-equilibrium methods, empirical correlations and established design procedures are entirely appropriate. They are efficient, transparent and often more than sufficient. Advanced analysis becomes valuable when the governing behaviour depends on mechanisms that simplified methods cannot represent adequately: stress redistribution, staged construction, nonlinear soil response, seepage and consolidation, soil–structure interaction, seismic loading, three-dimensional effects or the interaction of several of these at once.
The practical question is therefore not “Can this be modelled numerically?” Almost anything can. The better question is “Will the model materially improve our understanding of the governing mechanism, reduce uncertainty or help the project team make a better technical or construction decision?”
When Is Advanced Numerical Modelling Justified?
A decision framework for complex geotechnical problems
Define the engineering decision
What question must the analysis answer?
Identify the governing mechanism
Deformation · staging · SSI · seepage / consolidation · seismic response · 3D effects
Do conventional methods capture the controlling behaviour adequately?
Use conventional analysis
- Efficient
- Transparent
- Often sufficient
Advanced modelling may add little value.
Use advanced numerical modelling
- Represent the governing mechanism
- Test alternatives and staging
- Reduce uncertainty in the decision
Build a defensible model
- Ground model + parameters
- Constitutive formulation
- Boundary + drainage
- Appropriate dimensionality
Check credibility
- Calibration
- Sensitivity
- Verification
- Independent checks
Support the project decision
- Design
- Construction
- Risk
- Performance
Key principle: Use the simplest model that captures the mechanisms controlling the engineering decision.
Figure 1. A decision-focused framework for selecting the appropriate level of geotechnical analysis.
The decision should drive the model
A numerical model should begin with a clearly defined engineering question. Are we trying to estimate settlement beneath a staged embankment? Understand movements around a deep excavation? Evaluate load transfer between a foundation and the surrounding structure? Compare two support sequences? Test the seismic response of a soil–foundation–structure system? Reconcile monitoring data with expected behaviour? Or determine whether a particular failure mechanism is credible?
Those are different questions and they do not necessarily require the same model, constitutive formulation or level of detail. The model should be built around the mechanism that controls the decision—not around the capabilities of the software being used [1,4].
A useful model is not the most complicated model. It is the simplest model that captures the mechanisms controlling the engineering decision.
When advanced modelling becomes particularly valuable
Conventional Methods vs. Advanced Modelling
Choosing the right level of geotechnical analysis
Conventional Methods
Often sufficient when
Advanced Modelling
More valuable when
Geometry is simple and the governing mechanism is well understood.
Geometry, loading, staging, or interacting mechanisms are complex.
The primary question is overall stability or routine design capacity.
Deformation, movement distribution, or construction performance governs.
Construction sequence has limited influence on the outcome.
Staged excavation, fill placement, dewatering, or temporary conditions change the response.
Soil–structure interaction effects are modest or can be represented simply.
Interaction between the ground and structure materially affects forces or deformations.
Groundwater and time-dependent effects are secondary.
Seepage, consolidation, pore-pressure change, or cyclic / seismic response is important.
The project decision is insensitive to refined modelling.
The decision depends on understanding uncertainty, sensitivity, and consequence.
Best practice: Use advanced modelling when it materially improves the quality of the engineering decision—not simply because the software can model it.
Figure 2. Conventional methods remain appropriate for many problems; advanced modelling is most valuable when the governing behaviour or project decision requires additional resolution.
1. Deformation matters as much as stability
Traditional stability calculations are powerful when the principal question is whether a system has adequate resistance against a defined failure mechanism. They are less suited to projects where the magnitude, distribution and sequence of movement are themselves critical.
This is common for deep excavations beside sensitive structures, tunnels below existing infrastructure, foundations supporting deformation-sensitive equipment, marine structures, embankments on compressible soils and projects governed by strict track, utility or building movement criteria. Finite-element or finite-difference analysis can help evaluate how stresses and strains develop through the ground, where yielding may concentrate and how movement is transferred to adjacent assets.
2. Construction sequence changes the answer
Geotechnical systems rarely appear instantaneously in their final configuration. Excavations advance in stages. Struts and anchors are installed progressively. Embankments are raised over time. Ground improvement may occur between loading stages. Foundations may be loaded before the surrounding permanent works are complete. Pore pressures may build and dissipate while construction is still underway.
When the stress path depends strongly on how the project is built, analysing only the final geometry can miss the condition that actually governs performance. Staged analysis can be especially important where temporary works, preload stages, partial excavation, construction access or intermediate pore-pressure conditions are more critical than the permanent state.
This is also where modelling can provide direct construction value: comparing sequences, identifying stages that drive movement or instability, and testing whether a proposed change in means and methods is likely to improve or worsen ground response.
3. Soil–structure interaction controls performance
Foundations, retaining systems, shafts, tunnels, buried structures and marine infrastructure do not respond independently of the ground. Structural stiffness influences soil deformation, while soil stiffness, yielding and interface behaviour influence structural force distribution.
Where that interaction materially affects the design, uncoupled assumptions can become overly conservative in some locations and unconservative in others. Coupled analysis can provide a more coherent representation of load transfer, deformation compatibility, interface response and structural demand. For complex foundations and below-grade structures, this can be more informative than applying soil springs or earth pressures that were derived independently of the structural response they are intended to represent.
4. Soil behaviour is nonlinear, stress-dependent or time-dependent
Soil stiffness is not constant. It varies with strain level, stress history, confinement, drainage condition and loading path. Strength may also evolve with effective stress, consolidation or cyclic degradation. For routine problems, simplified parameters may be sufficient. For movement-sensitive or highly stressed systems, however, the selected constitutive model and parameter set can materially change the predicted response [1,2].
Advanced modelling becomes more useful when the analysis must distinguish between small-strain and larger-strain stiffness, represent yielding or hardening, simulate drained and undrained response appropriately, capture creep or consolidation, or assess the effect of stress history and anisotropy on behaviour.
5. Seepage, consolidation and loading interact
Some problems cannot be understood adequately by treating groundwater as a fixed line on a cross-section. Excavation, dewatering, staged fill placement, low-permeability soils and ground improvement can produce transient pore-pressure conditions that directly affect effective stress, stability and settlement.
Coupled seepage–deformation or consolidation analysis can be valuable where the rate of construction matters, where excess pore pressures control short-term stability, or where long-term settlement and pore-pressure dissipation influence sequencing and handover criteria. In these cases, time is part of the geotechnical problem rather than simply a reporting variable.
6. Seismic or dynamic response is important
Dynamic problems introduce additional questions that static calculations cannot answer directly: how seismic waves propagate through the soil profile, how stiffness and damping evolve with strain, how foundations and underground structures interact with moving ground, and whether cyclic loading changes pore pressures or deformation patterns.
Numerical analysis can be valuable for seismic soil–structure interaction, site response, deformation-based assessment, dynamic earth pressures, cyclic loading and other cases where the project decision depends on system response rather than a single static capacity check. The challenge is not simply applying an earthquake record; it is developing a model whose constitutive behaviour, damping, boundaries, input motion and calibration are appropriate for the physical problem [3].
7. Geometry or interacting mechanisms are genuinely three-dimensional
Irregular excavation layouts, adjacent foundations, complex retaining geometries, tunnel interfaces, marine structures, localized loads and transitions between ground conditions can make two-dimensional idealizations difficult to justify.
This does not automatically mean that a full 3D model is required. Often a carefully selected 2D section remains the best engineering tool. Three-dimensional analysis should be used when the physical mechanism is genuinely three-dimensional and when representing that mechanism is important to the decision. Model dimension should follow behaviour—not software availability.
A defensible model starts with the ground model—not the finite-element mesh
Advanced software does not remove uncertainty from geotechnical engineering. It can, in fact, make uncertainty harder to see by presenting highly detailed output from poorly constrained inputs.
A defensible analysis begins with the geological and geotechnical interpretation: stratigraphy, spatial variability, groundwater conditions, stress history, drainage behaviour and the mechanisms that the available investigation can actually support. Only then should model parameters be selected.
The parameter set must also be consistent with the constitutive model. A modulus taken from one test or empirical correlation is not automatically appropriate for every strain range or stress path. Similarly, undrained strength, effective-stress strength, interface parameters, permeability, small-strain stiffness and cyclic properties should be selected for the behaviour being represented—not simply because values are available in a report.
Calibration, sensitivity and verification are part of the analysis
A numerical model should not be accepted simply because it converges. Where monitoring data, field performance, laboratory testing, historical observations or comparable case histories are available, they should be used to test whether the model produces credible behaviour.
Sensitivity analysis is equally important. The objective is not to vary every parameter mechanically, but to identify which uncertain assumptions can actually change the project decision. In many cases, a focused sensitivity study around stiffness, groundwater, strength, interface behaviour or construction sequence tells the project team more than a single highly refined deterministic run.
Separate verification checks should remain part of the workflow. Simplified calculations, hand checks, limit-equilibrium analyses, closed-form solutions and observed behaviour can all provide essential reference points. Advanced analysis is strongest when it is consistent with physical reasoning and can explain why it differs from simpler methods where differences occur.
Numerical verification also matters. Mesh refinement, boundary location, initial stress generation, drainage assumptions, interface formulation, construction-stage activation and convergence behaviour should be checked to confirm that the numerical solution is not being controlled by modelling artefacts [1,4]. Where structural actions are important, force equilibrium and load paths should also be reviewed rather than relying only on displacement contours.
More output does not mean more certainty
Numerical models can generate displacement fields, stress contours and force distributions to several decimal places. That precision should not be confused with accuracy. Geotechnical predictions remain dependent on the ground model, parameter uncertainty, construction assumptions and the engineer's interpretation of the problem.
The strongest analyses therefore focus less on one predicted number and more on the range of plausible behaviour, the sensitivity of the result, the mechanisms controlling performance and the consequences if those assumptions prove wrong.
Where advanced modelling creates project value
The benefit of advanced analysis is not limited to design calculations. Used at the right stage, it can support several high-value project decisions.
For owners and developers, it can test whether critical assumptions are sufficiently robust, help quantify the consequences of uncertainty, and support design assurance where ground behaviour materially affects cost, programme or asset performance.
For contractors, it can support pursuit and preconstruction decisions, compare construction sequences, evaluate temporary conditions, investigate unexpected movement and assess whether proposed means-and-methods changes are likely to alter geotechnical risk.
For engineering teams, it can provide specialist capacity for complex soil–structure interaction, seismic response, difficult constitutive behaviour or technically sensitive peer review without replacing the broader design team.
In all three cases, the purpose is the same: turn a complex ground problem into a clearer technical decision.
When advanced modelling may not be warranted
Advanced modelling may add little value where the geometry and ground conditions are straightforward, the governing mechanism is well understood, established analytical methods address the design question directly, and the project decision is insensitive to deformation or interaction effects.
It may also be premature when the site investigation is too limited to support the level of sophistication being proposed. A detailed model built on poorly constrained inputs can create an appearance of certainty without actually reducing risk. In some cases, the best next step is not a more sophisticated model but better ground characterization, additional testing or a clearer definition of the decision that needs to be made.
What should a project team ask before commissioning advanced analysis?
Before investing in advanced numerical work, the project team should be able to answer five questions:
1. What decision will the model support?
2. Which physical mechanisms must be represented for that decision to be credible?
3. Are the available investigation and parameters sufficient for the proposed level of modelling?
4. Which uncertainties are most likely to change the result?
5. How will the model be checked against simpler calculations, monitoring data, observed behaviour or separate technical checks?
If those questions cannot be answered clearly, the modelling scope is probably not mature enough.
The objective is better engineering judgement
Advanced numerical modelling is most valuable when it allows a project team to examine behaviour that cannot be understood adequately through simpler methods. It can test assumptions, identify controlling mechanisms, compare construction sequences, evaluate soil–structure interaction and quantify the consequences of uncertainty.
But the model should remain a tool for engineering judgement, not a substitute for it. The quality of the decision still depends on understanding the ground, selecting defensible inputs, choosing a modelling approach that reflects the physical problem, testing sensitivity and interpreting the results in the context of how the project will actually be designed, constructed and monitored.
References
[1] Potts, D.M. & Zdravković, L. (1999). Finite Element Analysis in Geotechnical Engineering: Theory. Thomas Telford.
[2] Atkinson, J.H. (2000). “Non-linear soil stiffness in routine design.” Géotechnique, 50(5), 487–508. https://doi.org/10.1680/geot.2000.50.5.487
[3] Kramer, S.L. & Stewart, J.P. (2025). Geotechnical Earthquake Engineering, 2nd Edition. CRC Press.
[4] Muir Wood, D. (2004). Geotechnical Modelling. Spon Press.
Discuss a complex ground or modelling challenge
Geotechnics Plus supports owners, contractors and engineering teams with advanced numerical modelling, soil–structure interaction, seismic and dynamic geotechnics, staged construction analysis and specialist technical review where ground behaviour materially affects project decisions.