12. Computational and Multiscale Tribology

Introduction and Context

Tribological behaviour emerges from interactions that span several orders of magnitude in length and time. A machine component may be designed at the millimetre or metre scale, while its friction and wear are governed by asperities measured in micrometres, molecular films measured in nanometres, and atomic interactions occurring in femtoseconds. Computational tribology provides the methods required to connect these scales and to predict contact pressure, film formation, temperature, friction, material removal and surface evolution.

Classical analytical solutions remain valuable because they reveal governing parameters and provide rapid estimates. However, real contacts commonly involve rough, elastic–plastic surfaces, non-Newtonian lubricants, thermal gradients, transient loading, chemical reactions and evolving wear geometry. Finite-element methods (FEM), computational fluid dynamics (CFD), molecular dynamics (MD), discrete-element methods and data-driven models extend analysis into these complex regimes. Their usefulness depends not merely on computational power, but on appropriate assumptions, numerical verification, experimental validation and explicit treatment of uncertainty.

This chapter develops a practical framework for selecting and combining computational methods in advanced tribology. It also emphasizes a central principle: a highly detailed simulation is not necessarily a reliable simulation. Reliability arises when the model represents the relevant physics, the numerical solution is verified, predictions are validated against independent evidence, and uncertainty is communicated honestly.

Learning Objectives

After studying this chapter, the reader should be able to:

1. Explain the role and scale of analytical, FEM, CFD, MD and multiscale models in tribology.

2. Formulate computational models for contact, lubrication, thermal behaviour and wear.

3. Distinguish verification, validation, calibration, sensitivity analysis and uncertainty quantification.

4. Select suitable numerical methods and boundary conditions for engineering problems.

5. Interpret computational results in relation to physical mechanisms and experiments.

6. Identify common modelling errors, ethical risks and limitations of predictive tribological simulations.

7. Construct a defensible multiscale workflow for an engineering application.

12.1 The Computational View of a Tribological System

A computational tribology problem consists of four linked elements: geometry and surface description, material and lubricant behaviour, operating conditions, and performance criteria. Geometry includes nominal shape, waviness, roughness and defects. Material description may require elastic, plastic, viscoelastic, temperature-dependent or rate-dependent properties. Operating conditions include load, speed, temperature, environment and lubricant supply. Performance may be expressed through friction coefficient, minimum film thickness, maximum temperature, wear volume, fatigue life or leakage.

The governing equations are usually coupled. Mechanical equilibrium determines pressure and deformation; pressure and velocity determine lubricant flow; viscous dissipation generates heat; temperature changes viscosity and material strength; and wear modifies the geometry on which all subsequent fields depend. This coupling can be represented schematically as

mechanics → flow → heat transfer → material response → geometry evolution.

The appropriate model is therefore determined by the dominant physical scale and mechanism, not by the availability of a particular software package.

Important terminology

A contact model calculates the pressure and deformation generated when bodies approach or slide against one another. A conformal contact has relatively matching surfaces, whereas a non-conformal contact involves localized contact, as in gears, rolling elements and cam followers. Direct numerical simulation resolves the relevant surface or flow features explicitly. A reduced-order model represents their effect through simplified equations or fitted parameters. Homogenization replaces small-scale heterogeneous behaviour by effective properties at a larger scale. Coarse-graining reduces atomistic degrees of freedom while preserving selected macroscopic behaviour.

The representative volume element (RVE) is a sufficiently large microscopic region whose averaged response represents a material or surface. The tribological state is the combined condition of surfaces, lubricant, temperature, load and history. A model is forward when outputs are predicted from specified inputs; it is inverse when unknown parameters are inferred from measurements.

12.2 Analytical Foundations and Reduced Models

Analytical models are the first level of computational reasoning. Hertzian contact theory estimates contact radius, pressure distribution and elastic deformation for idealized curved bodies. For a spherical contact, the contact radius increases with load and effective radius and decreases with effective modulus; the maximum pressure is obtained from the load divided by the contact area with the appropriate Hertzian distribution. These relations are useful for checking numerical results even when the actual contact is rough or layered.

For lubricated contacts, the Reynolds equation is obtained by simplifying the Navier–Stokes equations for thin-film flow. In its generalized transient form, it relates pressure gradients to film thickness, lubricant viscosity, entrainment velocity and squeeze motion. The equation is the basis of hydrodynamic and elastohydrodynamic lubrication calculations. Coupling Reynolds flow with elastic deformation produces elastohydrodynamic lubrication (EHL), while temperature-dependent viscosity and energy equations produce thermohydrodynamic lubrication (THD).

Dimensionless groups assist model selection. The Reynolds number indicates the importance of inertia relative to viscosity, the Peclet number compares convective and diffusive heat transport, and the Sommerfeld-type parameter combines viscosity, speed, load and geometry in bearing analysis. The film parameter, commonly expressed as the ratio of minimum film thickness to composite roughness, helps distinguish full-film from mixed or boundary lubrication. Such parameters are not substitutes for simulation; they identify dominant regimes and expose inconsistent assumptions.

Reduced models are particularly valuable in design optimization and digital twins. Examples include resistance-network models for bearing lubrication, asperity contact models for rough surfaces, thermal resistance networks, Archard-type wear laws and surrogate models created from high-fidelity simulations. Their limitation is that fitted coefficients may only be valid within the calibration range. Extrapolation across lubricant chemistry, surface finish or load regime can be unsafe.

12.3 Finite-Element Modelling of Contact and Wear

FEM divides a continuum into elements connected at nodes and approximates displacement, temperature or other fields using interpolation functions. Contact is imposed through a normal constraint preventing interpenetration and a tangential law describing friction. Common formulations include penalty methods, augmented Lagrangian methods and Lagrange multipliers. Penalty methods are robust and computationally convenient, but excessive penalty stiffness can cause ill-conditioning, while insufficient stiffness permits unrealistic penetration.

A credible contact model requires careful definition of master and slave surfaces, contact detection, friction regularization and mesh refinement. The mesh should be refined where pressure, stress or temperature gradients are high. Adaptive remeshing may be necessary when wear substantially changes the geometry. In elastoplastic contact, the mesh must resolve the plastic zone rather than merely the visible contact patch. For layered coatings, element dimensions should be compatible with coating thickness and expected stress gradients.

Thermomechanical contact models include frictional heat generation, heat partition between bodies, conduction through the solids and convection to the environment. The frictional heat flux is often approximated as the product of frictional traction and sliding velocity. The distribution of heat between bodies depends on thermal conductivity, diffusivity, contact duration and thermal effusivity. Assuming that all heat enters one body can significantly distort predicted temperature.

Wear can be represented through several levels of sophistication. An Archard formulation relates incremental wear volume to normal load, sliding distance, hardness and a wear coefficient. In FEM, this relation may be converted into nodal or surface recession. More mechanistic models incorporate contact stress, dissipated energy, fracture, oxidation or temperature. Wear simulation is path-dependent: the pressure field changes as material is removed, so one-step post-processing is often inadequate for long service periods. Cycle jumping, adaptive time stepping and geometry updates reduce computational expense while retaining the principal evolution.

12.4 CFD and Lubricant-Film Computation

CFD solves conservation equations for mass, momentum and energy in a fluid domain. In tribology, the domain may be a bearing clearance, seal gap, textured surface, lubricant feed passage or cavitation region. The choice between full Navier–Stokes equations and the Reynolds equation depends on film thickness, geometry, inertia, pressure variation and computational objectives.

Cavitation is a major modelling issue. A pressure solution that allows unbounded negative pressure is physically unacceptable for a liquid film. Mass-conserving cavitation algorithms, complementarity formulations and multiphase CFD are used to describe the rupture and reformation of the film. Boundary conditions must distinguish prescribed pressure, mass flow, lubricant supply and open cavitation boundaries.

Lubricants may be non-Newtonian, shear-thinning, viscoelastic or compressible. Temperature changes viscosity, density and sometimes phase state. In gas lubrication, rarefaction may require slip-flow or kinetic corrections. In textured contacts, the mesh must resolve texture depth and wavelength or use an effective-flow model. Surface roughness can be explicitly resolved, statistically represented or incorporated through flow factors.

The most useful CFD outputs are not isolated colour plots but integrated engineering quantities: load-carrying capacity, frictional power loss, leakage, film-thickness distribution, cavitation extent and thermal margin. A visually smooth pressure field can still be wrong if the mass flow or load balance is not satisfied.

12.5 Molecular Dynamics and Atomistic Tribology

MD computes the trajectories of atoms using an interatomic potential and numerical integration of Newton’s equations. It is suited to questions involving adhesion, nanoscale friction, asperity junctions, thin-film confinement, atomic wear, defect generation and lubricant–surface interactions. Typical outputs include friction force, normal force, temperature, coordination number, density profile and chemical or structural changes.

An MD model requires a simulation cell, boundary conditions, initial configuration, interatomic potential, thermostat strategy and loading protocol. Periodic boundaries approximate an extended surface, while fixed or driven layers apply macroscopic motion. The selected potential must reproduce the properties relevant to the question; a potential accurate for bulk elasticity may be inadequate for bond breaking or chemical reactions.

MD has severe scale limitations. Simulation times are commonly much shorter and dimensions much smaller than engineering tribological processes. High sliding velocities and idealized surfaces may be necessary computational compromises, but they can alter mechanisms. Thermostats may remove frictional heat unrealistically if applied directly to the shearing region. Results should therefore be interpreted mechanistically rather than automatically converted into engineering friction coefficients.

Atomistic simulations are most powerful when used to identify constitutive behaviour or mechanisms for larger-scale models. For example, MD can estimate interfacial shear strength, confined-fluid viscosity or temperature-dependent adhesion. These quantities can then inform continuum contact, lubrication or cohesive-zone models.

12.6 Linking Scales: Multiscale Strategies

Multiscale tribology is not a single algorithm. It is a modelling strategy for transferring information between scales. In a sequential, or hierarchical, approach, a fine-scale model supplies parameters to a larger-scale model. MD may provide interfacial friction or adhesion; an asperity model averages local contacts; a bearing model predicts system torque and temperature. This approach is relatively stable and efficient, but feedback from the large scale to the small scale is limited.

In a concurrent approach, two scales are solved simultaneously in different regions or domains. A quasicontinuum method, for example, uses atomistic resolution near defects and continuum resolution elsewhere. Domain-decomposition methods can couple molecular and continuum regions, but interface conditions and energy transfer are difficult to formulate.

Statistical multiscale methods describe rough surfaces through spectral density, height distributions and correlation lengths. A measured topography may be decomposed into wavelength bands. Long wavelengths influence nominal geometry and pressure distribution, while shorter wavelengths control asperity interactions and boundary film behaviour. Effective properties should be generated from measured or physically justified surface statistics rather than arbitrary roughness parameters.

A practical hierarchy is often:

• atomistic calculations for adhesion, chemical interaction and confined films;
• asperity or homogenized models for rough-contact response;
• FEM or EHL/CFD for component-level fields;
• system models for efficiency, temperature and life prediction.

The transfer variables must be defined clearly. They may include effective modulus, traction–separation law, friction law, permeability, viscosity, heat-transfer coefficient or wear coefficient. Each transfer introduces uncertainty and possible loss of physics. A multiscale model is credible only when these interfaces are documented and tested.

12.7 Verification, Validation and Uncertainty Quantification

Verification asks whether the equations have been solved correctly. It includes code verification against manufactured solutions, comparison with analytical solutions, mesh convergence, time-step convergence, residual monitoring and conservation checks. For contact and CFD problems, global force balance, mass conservation and energy balance are essential.

Validation asks whether the equations and assumptions represent reality for the intended application. Validation requires comparison with independent experimental data, not merely the data used to calibrate parameters. Relevant measurements may include friction torque, film thickness, contact temperature, wear volume, pressure distribution or surface morphology. Agreement in one output does not validate every predicted mechanism.

Calibration estimates uncertain model parameters from observations. Calibration should not be confused with validation. A wear coefficient fitted to a short-duration test may reproduce that test while failing under a different speed, temperature or material condition.

Sensitivity analysis determines how output changes when inputs vary. Local sensitivity uses derivatives around a nominal point; global methods, such as variance-based analysis or Latin hypercube sampling, explore interactions over a specified input range. Parameters with low sensitivity need not be measured with excessive precision, whereas highly influential parameters deserve careful characterization.

Uncertainty has at least two sources. Aleatory uncertainty describes natural variability, such as roughness variation or fluctuating load. Epistemic uncertainty arises from incomplete knowledge, such as an uncertain friction law or unknown heat partition. Monte Carlo simulation, polynomial chaos, Bayesian inference and interval methods can propagate these uncertainties. Results should be reported as distributions or credible intervals where appropriate, rather than as a single falsely precise value.

A useful engineering workflow is: define the decision and quantity of interest; identify mechanisms; select the lowest adequate model; specify inputs and distributions; verify numerical implementation; calibrate using suitable data; validate against independent observations; perform sensitivity analysis; propagate uncertainty; and document assumptions and applicability limits.

12.8 Applied Case: Multiscale Prediction of a Coated Journal Bearing

Background

The following is an illustrative case involving a high-speed journal bearing with a thin protective coating and a biodegradable lubricant. The design objective is to reduce friction while preventing coating failure during start–stop operation.

Situation and problem

A conventional isothermal EHL model predicts an adequate minimum film thickness during steady running. However, field experience indicates high torque and occasional coating damage during start-up. The likely causes are transient squeeze, mixed lubrication, frictional heating and adhesion at coating asperities.

Analysis and strategy

At the system level, a bearing model first calculates shaft speed, load, lubricant supply and time-dependent clearance. A transient Reynolds solver predicts pressure, film thickness and cavitation. The viscosity is coupled to a thermal model containing heat generation, conduction through the coating and heat transfer to the housing.

At the component level, FEM represents the coated journal and bearing shell. The pressure field is transferred to the structural model, which calculates coating stresses and deformation. A mixed-lubrication correction introduces asperity load when the film parameter becomes small. The asperity model uses measured roughness spectra rather than a single average roughness value.

At the smallest scale, MD simulations examine the lubricant molecules confined between representative coating surfaces. The simulations estimate an interfacial shear response and adhesion trend for selected temperatures. These results are not treated as direct engineering friction coefficients; instead, they inform a boundary-friction law used in the mixed-lubrication model.

The coupled model is verified through mesh refinement, time-step reduction, mass-balance checks and comparison with a known analytical bearing solution under simplified conditions. Validation uses measured start-up torque, steady friction and post-test coating wear from experiments not used for parameter fitting. Global sensitivity analysis identifies lubricant viscosity–temperature behaviour, coating hardness and heat partition as the most influential variables.

Outcome

The model shows that steady-state film thickness is not the controlling design metric. During start-up, transient squeeze reduces film thickness rapidly, increasing asperity load and localized coating temperature. A modest reduction in coating roughness and improved lubricant supply delay the transition to severe mixed lubrication more effectively than increasing steady-speed viscosity. The design team adopts a controlled start-up sequence, improves coating finish and selects a lubricant with a more stable viscosity–temperature response.

Learning points

The case demonstrates that multiscale modelling should be organized around a decision, not around the desire to use every available simulation method. Atomistic results are useful only when translated into a defensible larger-scale law. Experimental validation must include the transient regime that causes failure. Finally, uncertainty analysis can change the design recommendation by revealing which parameters actually control risk.

12.9 Practical Applications and Professional Use

Computational tribology is applied to gears, rolling bearings, seals, piston rings, magnetic storage devices, microelectromechanical systems, cutting tools, electric machines and biomedical implants. In industry, models support geometry optimization, lubricant selection, coating development, maintenance intervals, failure investigation and virtual testing.

For engineering organizations, the principal value is often comparative rather than absolute. A model may reliably rank two textures, coatings or lubricants even when its absolute wear prediction remains uncertain. This requires consistent assumptions and controlled validation. Computational results should be integrated with test planning, materials characterization and field monitoring rather than used as a replacement for all experiments.

Digital twins and reduced-order surrogates allow sensor data, operating histories and computational models to be combined for condition monitoring. Machine learning can accelerate surrogate construction, detect abnormal friction signatures or infer wear state. However, purely data-driven models may fail outside their training distribution. Physics-informed and hybrid models are preferable when extrapolation, interpretability and safety are important.

12.10 Challenges, Risks and Professional Considerations

Common mistakes include using smooth-surface theory for rough contacts without justification, prescribing unrealistic lubricant pressure boundaries, ignoring cavitation, applying excessive contact stiffness, using insufficient mesh resolution, fitting too many parameters to too little data and reporting only the most favourable simulation. Another error is confusing numerical convergence with physical validity: a converged solution can still represent the wrong problem.

Tribological data are often proprietary, variable and difficult to reproduce. Surface preparation, lubricant contamination, temperature history and measurement bandwidth can alter results. Models should therefore record geometry, material state, boundary conditions, solver settings, parameter sources and uncertainty. Reproducibility requires version-controlled input files and transparent post-processing.

Ethically, engineers must not present a deterministic prediction as a guarantee when uncertainty is material. Safety-critical decisions require conservative margins, independent review and traceable evidence. Regulatory requirements may apply to medical implants, environmental release of lubricants, machinery safety and software quality. Intellectual-property restrictions should not prevent the disclosure of assumptions essential to safe interpretation.

12.11 Emerging Directions and Best Practice

Current research is moving toward reactive and electrochemical tribology, data-assisted multiscale models, real-time reduced-order simulation, high-fidelity surface measurement, uncertainty-aware digital twins and coupled mechanical–chemical–thermal descriptions. Increasing attention is being given to sustainable lubricants, recyclable coatings and energy-efficient computation.

Best practice can be summarized as follows:

1. Begin with the physical decision and dominant mechanism.

2. Use analytical estimates to establish scales and sanity checks.

3. Select the simplest model that can answer the question.

4. Represent surfaces, materials and lubricant behaviour with traceable data.

5. Verify discretization, conservation and solver implementation.

6. Validate against independent measurements in the relevant regime.

7. Quantify sensitivity and uncertainty rather than hiding them.

8. Couple scales only through clearly defined, tested transfer laws.

9. Report limitations, extrapolation risks and numerical settings.

10. Combine simulation with targeted experiments and engineering judgement.

Chapter Summary

Computational and multiscale tribology connects analytical theory, continuum mechanics, fluid dynamics, atomistic simulation and system-level engineering. Analytical models provide scaling laws and checks; FEM resolves contact, deformation, thermal fields and geometry evolution; CFD predicts lubricant flow, cavitation and heat transport; MD investigates nanoscale adhesion, confined films and interfacial mechanisms. Multiscale strategies transfer information between these levels through hierarchical, concurrent or homogenized approaches.

Reliable prediction requires more than numerical sophistication. Verification establishes that equations are solved correctly, validation tests whether the model represents physical behaviour, calibration estimates parameters, and sensitivity and uncertainty analysis reveal the robustness of conclusions. Practical applications include bearing, gear, seal, coating, MEMS and biomedical design. The strongest engineering workflow begins with a decision, selects an adequate model, validates it in the relevant operating regime and communicates uncertainty transparently.

Key Terms

Analytical model: A model expressed through derived equations or closed-form approximations.

Finite-element method: A numerical method that discretizes a continuum into elements to approximate field variables.

Computational fluid dynamics: Numerical solution of fluid conservation equations for flow, pressure and thermal fields.

Molecular dynamics: Atomistic simulation based on Newton’s equations and an interatomic potential.

Multiscale modelling: Representation of a physical problem using linked models at different length or time scales.

Elastohydrodynamic lubrication: Lubrication in which fluid pressure significantly deforms the contacting solids.

Thermohydrodynamic lubrication: Lubrication analysis coupled with temperature-dependent fluid and solid thermal behaviour.

Cavitation: Formation of a vapour or gas-filled region when liquid-film pressure falls below the cavitation condition.

Homogenization: Replacement of fine-scale heterogeneous behaviour by effective large-scale properties.

Representative volume element: A microscopic region whose averaged response represents a larger heterogeneous medium.

Contact penalty method: A contact formulation that enforces non-penetration using a finite penalty stiffness.

Wear law: A mathematical relation describing material removal as a function of load, distance, energy or other variables.

Verification: Demonstration that a numerical implementation solves the intended equations correctly.

Validation: Demonstration that a model represents the physical system adequately for its intended use.

Calibration: Estimation of model parameters using measured observations.

Sensitivity analysis: Study of how changes in inputs influence model outputs.

Uncertainty quantification: Characterization and propagation of variability and incomplete knowledge in predictions.

Reduced-order model: A simplified model retaining the dominant behaviour of a higher-fidelity model.

Surrogate model: An inexpensive approximation used in place of a computationally expensive simulation.

Heat partition: Distribution of friction-generated heat between contacting bodies.

Review Questions

Short Answer Questions

1. Why are analytical contact solutions useful even when a detailed FEM model is available?

2. Distinguish verification from validation in computational tribology.

3. What is the function of a cavitation model in lubricant-film analysis?

4. State two limitations of molecular-dynamics simulations for engineering tribology.

5. Define a representative volume element.

Long Answer Questions

1. Explain the formulation and major numerical issues of a finite-element model for sliding contact with frictional heating and wear.

2. Compare Reynolds-based lubrication analysis with full CFD, including their assumptions and suitable applications.

3. Describe hierarchical and concurrent multiscale modelling and discuss the transfer of information between scales.

4. Explain how sensitivity analysis and uncertainty quantification improve engineering decisions based on tribological simulations.

5. Discuss the verification, calibration and validation requirements for a predictive EHL model.

Analytical / Application Questions

1. A bearing simulation predicts a converged pressure field but violates mass conservation near the cavitation boundary. Diagnose the likely problem and propose verification steps.

2. An MD simulation predicts lower friction at an unrealistically high sliding speed than an experiment conducted at low speed. Explain why direct comparison may be invalid and how the atomistic result could still be used.

3. Design a multiscale computational workflow for a coated journal bearing experiencing start–stop operation. Identify the models, transfer variables, validation measurements and major uncertainties.

22. Multiple Choice Questions with Answers

Question 1

Which statement best distinguishes verification from validation?

A. Verification compares predictions with experiments, while validation checks mesh quality.

B. Verification checks whether equations are solved correctly, while validation checks whether the model represents physical behaviour.

C. Verification estimates uncertain parameters, while validation calculates sensitivity.

D. Verification applies only to MD, while validation applies only to CFD.

Correct Answer: B. Verification checks whether equations are solved correctly, while validation checks whether the model represents physical behaviour.

Question 2

A penalty contact formulation with excessively large penalty stiffness is most likely to cause which problem?

A. Unbounded cavitation

B. Loss of lubricant viscosity

C. Numerical ill-conditioning

D. Zero contact pressure

Correct Answer: C. Numerical ill-conditioning

Question 3

Why is a mass-conserving cavitation treatment important in lubricant-film computation?

A. It prevents physically unrealistic negative-pressure predictions.

B. It eliminates the need for thermal analysis.

C. It ensures that all contacts are conformal.

D. It converts a liquid lubricant into a solid film.

Correct Answer: A. It prevents physically unrealistic negative-pressure predictions.

Question 4

Which is the most appropriate use of molecular dynamics in a multiscale tribology model?

A. Predicting the complete service life of a full-size bearing directly

B. Replacing all experimental measurements of wear

C. Estimating interfacial behaviour that can inform a continuum friction law

D. Calculating only the nominal shaft diameter

Correct Answer: C. Estimating interfacial behaviour that can inform a continuum friction law

Question 5

A model agrees with the experiment used to fit its wear coefficient but fails for a different speed. This most directly illustrates the difference between:

A. Calibration and validation

B. Mesh refinement and time stepping

C. CFD and FEM

D. Elasticity and plasticity

Correct Answer: A. Calibration and validation

Question 6

Which output is most useful for assessing whether a CFD lubrication solution is physically credible?

A. Colour smoothness of the pressure plot alone

B. Mass balance, load capacity and predicted film thickness

C. Number of graphical contours

D. File size of the simulation result

Correct Answer: B. Mass balance, load capacity and predicted film thickness

Question 7

In the illustrative coated-bearing case, why was steady-state film thickness insufficient as the sole design metric?

A. The bearing had no lubricant during steady running.

B. Start-up transients produced squeeze, asperity load and localized heating.

C. Film thickness is unrelated to friction.

D. Coating stress can never be calculated by FEM.

Correct Answer: B. Start-up transients produced squeeze, asperity load and localized heating.

Question 8

Which statement about uncertainty is correct?

A. Aleatory uncertainty represents only numerical round-off error.

B. Epistemic uncertainty results from natural random surface variation only.

C. Uncertainty may arise from both physical variability and incomplete knowledge.

D. A converged numerical solution contains no uncertainty.

Correct Answer: C. Uncertainty may arise from both physical variability and incomplete knowledge.

Question 9

What is the principal purpose of homogenization in multiscale tribology?

A. To replace fine-scale behaviour with effective properties at a larger scale

B. To increase the number of atomistic degrees of freedom

C. To impose zero temperature in a contact

D. To eliminate all surface roughness physically

Correct Answer: A. To replace fine-scale behaviour with effective properties at a larger scale

Question 10

Which modelling practice is most defensible for a safety-critical tribological application?

A. Report only the nominal prediction to avoid confusion.

B. Use the most complex available model without validation.

C. Combine verification, independent validation, uncertainty analysis and traceable assumptions.

D. Fit all parameters using one short-duration test.

Correct Answer: C. Combine verification, independent validation, uncertainty analysis and traceable assumptions.