5. Wear Mechanisms and Wear Modelling

Introduction and context

Wear is the progressive loss, displacement, or transformation of material from a solid surface as a consequence of relative motion, contact, or impact. It is one of the principal ways in which tribological systems lose accuracy, efficiency, reliability, and service life. Unlike friction, which primarily represents resistance to motion, wear is a damage process. A component may therefore exhibit high friction with little material loss, or low friction while suffering severe subsurface fatigue or chemical degradation.

Wear is not a single mechanism. It is the outcome of interacting mechanical, thermal, chemical, and environmental processes. The same component may experience different mechanisms during its life: an initially clean metallic contact may develop adhesive transfer; transferred material may form abrasive particles; repeated loading may then produce fatigue cracks. Correct diagnosis consequently requires more than measuring mass loss. It requires examination of contact conditions, surface topography, debris, subsurface deformation, temperature, lubrication, and material chemistry.

Wear modelling provides a means of relating operating conditions and material properties to damage rate. Simple empirical relations, such as Archard’s law, are useful for engineering estimates, while mechanistic and numerical models are needed when wear depends strongly on particle geometry, stress fields, temperature, oxidation, lubrication, or evolving surface morphology.

Learning objectives

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

1. Distinguish adhesive, abrasive, fatigue, fretting, erosive, corrosive, oxidative, and impact wear.

2. Explain the physical origins and characteristic morphologies of major wear mechanisms.

3. Apply Archard-type equations and interpret their assumptions and limitations.

4. Select suitable mechanistic, numerical, or empirical modelling approaches.

5. Construct and interpret wear maps and wear-regime transitions.

6. Use debris and damage morphology as evidence in failure diagnosis.

7. Evaluate practical control strategies for industrial tribosystems.

8. Recognise uncertainty, ethical responsibilities, and validation requirements in wear prediction.

Wear as a system response

A useful description of a wear problem begins with the tribosystem rather than with the material alone. The tribosystem includes the two bodies in contact, any intermediate material such as lubricant or debris, the surrounding environment, and the loading and motion history. Important variables include normal load, contact pressure, sliding distance, speed, temperature, humidity, lubrication regime, surface roughness, hardness, toughness, microstructure, and chemical reactivity.

Wear rate is often expressed as volume loss per unit sliding distance, mass loss per unit time, or a dimensionless specific wear rate. A volume-based specific wear rate is commonly written as

k_s = V/(W L)

where V is wear volume, W is normal load, and L is sliding distance. The units of k_s are m³/(N·m), or equivalently m²/N. This quantity is useful for comparing tests, but it is not a universal material constant. It depends on the mechanism, counterface, environment, load, speed, and stage of running-in.

The observed wear may be divided into primary wear, caused directly by asperity interaction or impact, and secondary effects, such as oxidation, third-body abrasion, crack propagation, or debris compaction. The distinction is important because changing one variable can alter the dominant mechanism rather than simply changing the intensity of the same mechanism.

Principal wear mechanisms

Adhesive wear

Adhesive wear occurs when microscopic junctions form between contacting asperities and are subsequently sheared during sliding. Junction formation is promoted by high local pressure, clean and chemically compatible surfaces, inadequate lubrication, and removal of protective films. Material may be transferred from one surface to the other, or fragments may detach from either surface.

Typical evidence includes transfer films, smeared regions, torn surface features, ploughing associated with transferred fragments, and metallic debris. In severe cases, seizure occurs when the real area of contact grows rapidly and sliding becomes unstable. Adhesive wear is not determined simply by bulk hardness; chemical affinity, oxide-film stability, solubility, and surface energy are also important.

Adhesion can be reduced by separating surfaces with a lubricant film, using dissimilar material pairs, applying compatible coatings, controlling surface finish, or promoting stable protective tribofilms. However, excessive roughness may increase cutting and abrasion, so surface modification must be designed for the complete tribosystem.

Abrasive wear

Abrasive wear results when a hard asperity or hard particle cuts, ploughs, or deforms a softer surface. In two-body abrasion, the abrasive feature is fixed to one of the contacting bodies. In three-body abrasion, loose particles roll or slide between the surfaces. Abrasive wear may be caused by contaminants, wear debris, hard inclusions, oxides, or coating fragments.

Cutting produces grooves and raised lips, whereas ploughing produces displaced material without complete removal. The severity depends on relative hardness, abrasive shape, particle size, concentration, contact pressure, and particle mobility. Sharp angular particles are generally more damaging than rounded particles. A hard surface can also suffer abrasion if it contains brittle phases that fracture into aggressive debris.

Control methods include filtration, sealing, improved cleanliness, increased surface hardness, tougher coatings, debris exclusion, and redesign of lubricant circulation. A coating that is hard but poorly bonded may worsen wear when it delaminates and becomes a third-body abrasive.

Fatigue wear

Fatigue wear develops under repeated or cyclic contact stresses. Repeated rolling or sliding can initiate subsurface or near-surface cracks, particularly when stress concentrations arise from roughness, inclusions, residual stresses, or geometrical defects. Crack growth eventually produces pits, spalls, flakes, or delaminated sheets.

Rolling-contact fatigue is prominent in bearings, gears, cam systems, and rails. It is influenced by maximum Hertzian stress, contact conformity, lubrication film thickness, material cleanliness, hardness, residual stress, and load cycles. Unlike simple adhesive wear, fatigue wear may continue even when instantaneous sliding is small. Surface distress, micropitting, and pitting are often early indicators.

The distinction between fatigue and delamination is sometimes practical rather than absolute. Delamination models describe the progressive separation of thin surface layers after crack networks develop, while classical fatigue descriptions emphasise cyclic stress and crack propagation.

Fretting wear

Fretting is damage produced by small-amplitude oscillatory motion between nominally contacting surfaces. The displacement may be only a few micrometres to hundreds of micrometres, but high local stresses and repeated cycles can cause severe damage. Fretting commonly occurs in bolted joints, spline connections, press fits, electrical contacts, and aerospace structures subject to vibration.

Fretting regimes are often described using displacement amplitude, normal load, and tangential force. In partial slip, much of the contact sticks while an outer annulus slips. In gross slip, the entire contact undergoes relative motion. A third regime, often called mixed or transitional, involves changing stick-slip behaviour. Fretting may produce oxidised powder, known as reddish-brown debris in many ferrous systems, but debris colour alone is not diagnostic.

Fretting wear may coexist with fretting fatigue. The former is dominated by material loss and surface alteration; the latter involves crack initiation and propagation under cyclic stresses. Increasing contact pressure, changing joint stiffness, applying suitable coatings, eliminating vibration, and controlling assembly preload are possible remedies.

Erosive and impact wear

Erosive wear is caused by repeated impingement of solid particles, liquid droplets, or gas-borne particles on a surface. Particle velocity, impact angle, size, hardness, shape, concentration, and target properties control the response. Ductile materials commonly show cutting and ploughing at shallow impact angles, while brittle materials may suffer cracking and chipping at larger angles. This distinction is a tendency, not a universal rule.

Impact wear involves repeated collision or high-energy contact, with substantial deformation, fragmentation, rebound, and sometimes plastic shakedown. It occurs in crushers, hammers, railway crossings, valves, chutes, and forming equipment. Impact and erosion may occur together, but impact wear is generally defined by discrete high-energy contact events, whereas erosion often involves many particles or droplets transported by a fluid.

Design responses include reducing particle velocity, changing flow direction, using sacrificial liners, selecting tough rather than merely hard materials, and avoiding abrupt geometry that concentrates impacts.

Corrosive and oxidative wear

Corrosive wear occurs when chemical or electrochemical reactions weaken a surface and mechanical action removes the reaction products. The environment may contain water, acids, alkalis, salts, dissolved oxygen, or reactive process chemicals. Wear can expose fresh material, accelerate corrosion, and create a positive feedback known as tribocorrosion.

Oxidative wear is a specific interaction between mechanical removal and oxidation. A compact, adherent oxide may reduce metal-to-metal contact and lower wear. Conversely, a brittle or poorly adherent oxide may fracture into abrasive debris and increase wear. Oxidative wear is therefore not synonymous with high oxidation rate; the mechanical stability of the oxide is decisive.

Diagnosis requires chemical analysis as well as microscopy. Surface colour, oxide composition, electrochemical potential, humidity, and lubricant additives may be relevant. Control includes environmental isolation, corrosion-resistant materials, inhibitors, cathodic or barrier protection where appropriate, and lubricant formulations that form stable boundary films.

Modelling wear: from empirical laws to mechanistic descriptions

Archard’s model

The classical Archard relation is

V = K W L/H

where V is wear volume, W is normal load, L is sliding distance, H is the hardness of the softer material, and K is a dimensionless wear coefficient. In rate form,

dV/dL = K W/H.

The model is based on the idea that wear results from the removal of asperity volumes and that the characteristic contact pressure is related to hardness. It predicts proportionality to load and sliding distance and inverse proportionality to hardness. It is valuable for preliminary sizing, comparison of materials, and estimating service life when the operating regime is stable.

The coefficient K incorporates junction formation, asperity geometry, fracture probability, lubrication, environment, and debris behaviour. It can vary by several orders of magnitude between mild and severe wear. Consequently, it should be obtained from representative experiments or validated literature rather than treated as an intrinsic constant.

For a pin or component with nominal contact area A, the wear depth h can be estimated from

h = V/A = K W L/(H A).

If the nominal pressure p = W/A is approximately constant, then h = K p L/H. This result is convenient but can be misleading when contact area changes, wear alters alignment, or pressure is strongly non-uniform.

Mechanistic models

Mechanistic models attempt to connect wear to identifiable physical events. In an asperity-fracture model, the volume removed depends on asperity size, junction strength, and fracture or shear probability. In abrasive models, wear may be related to abrasive geometry, penetration depth, hardness ratio, and particle concentration. In fatigue models, crack initiation and propagation are linked to cyclic stress, stress intensity, material toughness, and residual stress.

Energy-based models use frictional work or dissipated energy as a predictor. A general form is

V = K_E E_d,

where E_d is dissipated energy and K_E is an experimentally calibrated wear-energy coefficient. Such models can be useful when load and friction vary simultaneously, but frictional energy is not always converted into wear. Much of it becomes heat, elastic energy, plastic deformation, or chemical reaction.

Contact-mechanics models calculate pressure and shear distributions using Hertzian, elastic-plastic, rough-surface, or finite-element formulations. Wear is then updated incrementally, for example by applying Archard’s law locally:

Δh(x,y) = K p(x,y) Δs/H.

The geometry is updated after each increment, allowing changes in contact area, pressure, and alignment to be captured. This approach is often called wear-debris or geometry-evolution simulation when coupled to finite-element or multibody analysis.

Probabilistic and data-driven approaches

Wear measurements usually contain scatter because surfaces, debris populations, and microstructures vary. Reliability-based models may represent K or crack-growth parameters as random variables. Bayesian updating can combine prior information with inspection data, while survival models can estimate the probability of reaching a wear limit.

Machine-learning methods can classify wear mechanisms from surface images, vibration signals, acoustic emissions, lubricant particles, or sensor histories. Their usefulness depends on representative training data and physically meaningful labels. A high classification accuracy on laboratory data does not guarantee reliable prediction in a different machine, lubricant, temperature range, or material system. Hybrid models that combine physical equations with data-driven correction terms are often more defensible than unconstrained black-box prediction.

Wear maps and regime transitions

A wear map displays the dominant mechanism or wear rate as a function of two or more controlling variables. Common axes include load or nominal pressure, sliding speed, temperature, displacement amplitude, lubrication parameter, impact angle, or hardness ratio. Regime maps may be constructed from experiments, analytical criteria, or numerical simulations.

A useful example is a load–speed map for a sliding pair. At low load and adequate lubrication, mild oxidative or film-controlled wear may occur. Increasing load can break the film and produce adhesive wear. At still higher load, temperature rise, plastic deformation, and seizure may create severe wear. If hard debris is generated, the system can transition from adhesion to three-body abrasion. The boundaries are not exact lines; they are probabilistic transitions influenced by surface history and environment.

Wear maps must therefore be interpreted as system-specific. A map developed for one counterface roughness, lubricant, and temperature should not be transferred uncritically to another. Dimensionless groups can improve generalisation. Examples include a lubrication parameter based on film thickness divided by composite roughness, a hardness ratio, a contact-stress-to-hardness ratio, and a fretting amplitude normalised by contact size.

Reading debris and damage morphology

Wear debris is often a time-resolved record of the mechanism. Large metallic flakes suggest delamination or severe adhesive transfer; long curled chips suggest cutting abrasion; fine angular particles suggest brittle fracture or severe oxidation; spherical particles may indicate molten, re-solidified, or highly plastic material, although morphology must be interpreted cautiously. Oxide particles, polymer transfer films, and composite fragments provide additional clues.

Surface examination should proceed at multiple scales. Macroscopic observations reveal scoring, discolouration, pitting, and directional patterns. Optical microscopy identifies grooves, pits, smeared layers, and spalls. Scanning electron microscopy can reveal microcracks, lips, fatigue facets, particle impacts, and transfer layers. Profilometry quantifies roughness, wear volume, groove depth, and waviness. Energy-dispersive spectroscopy, X-ray diffraction, Raman spectroscopy, and ferrography can identify chemical phases and debris composition.

Diagnosis should compare the worn surface with the unworn counterpart and with the counterface. A groove on one surface may correspond to a transferred ridge on the other. The presence of oxide does not prove oxidative wear as the primary mechanism; oxide may form after an initially mechanical event. Likewise, a polished surface does not necessarily indicate low wear, because material may have been removed by subsurface fatigue.

An illustrative industrial case: premature wear in a lubricated gearbox

Background

A hypothetical industrial gearbox drives a conveyor under variable load. Its hardened steel gear pair was designed using a conventional contact-stress calculation and a wear coefficient obtained from a laboratory pin-on-disc test. After several months, the operator observed increasing backlash, metallic particles in the oil, and a rise in vibration.

Situation and analysis

Initial inspection found polishing near the pitch line, isolated micropits, and patches of dark transfer film. Oil analysis showed an increasing concentration of iron particles and a reduction in the lubricant’s cleanliness. Microscopy revealed both fatigue-like pits and angular particles capable of three-body abrasion. The evidence indicated a transition rather than a single mechanism: marginal lubrication permitted asperity interaction and oxidative film formation; repeated contact initiated micropitting; detached particles accelerated abrasive damage.

The original calculation had used Archard’s equation with a constant coefficient and nominal load. It did not account for load variation, transient lubrication starvation, evolving roughness, or debris circulation. A revised analysis calculated the local pressure distribution, estimated the film-thickness ratio over the duty cycle, and introduced a regime-dependent wear coefficient. Oil-flow measurements also showed intermittent starvation near one gear position.

Approach and outcome

The corrective strategy combined improved oil filtration, correction of the lubricant delivery path, alignment adjustment, and a surface-finish review. Inspection intervals were shortened temporarily, and vibration and oil-particle trends were monitored. The model predicted that removing abrasive debris and restoring film separation would reduce the transition into severe wear more effectively than simply increasing material hardness.

Learning points

The case illustrates four principles. First, Archard’s law is a useful baseline, not a complete diagnosis. Second, wear mechanisms may occur sequentially or simultaneously. Third, debris analysis should be combined with contact and lubrication analysis. Fourth, prevention is often more effective when it targets the system variable causing regime transition rather than only strengthening the component.

Practical decision-making and professional responsibilities

Wear control begins with a clearly defined failure criterion: allowable dimensional loss, leakage, roughness, vibration, efficiency loss, electrical resistance, or probability of crack initiation. Material selection should consider hardness, toughness, corrosion resistance, thermal stability, compatibility, coating adhesion, and manufacturability together. Test conditions must reproduce the important features of the service duty cycle, including transients and contamination.

Common mistakes include using a wear coefficient from an unrelated test, confusing friction reduction with wear reduction, ignoring the counterface, measuring only mass loss, overlooking running-in, and assuming that harder always means better. Another frequent error is extrapolating a short laboratory test linearly over a long service life after the mechanism has changed.

Wear prediction also has an ethical and professional dimension. An engineer should state model assumptions, uncertainty, calibration range, and inspection requirements clearly. Safety-critical components require conservative validation, traceable data, and appropriate standards or regulatory approval. Automated condition-monitoring systems should not conceal low confidence or generate maintenance recommendations without human review when failure consequences are severe.

Emerging directions

Current developments include multifunctional coatings, solid-lubricant and self-lubricating composites, surface texturing, environmentally compatible lubricants, and additive-manufactured materials designed with controlled microstructures. Digital twins increasingly combine contact simulation, sensor data, oil-debris monitoring, and maintenance history. Physics-informed machine learning offers a way to use data while preserving constraints from contact mechanics and conservation laws.

The main research challenge is transferability. Wear is strongly path-dependent: the history of temperature, debris, roughness, chemistry, and loading changes future behaviour. Better models will therefore combine mechanism identification, uncertainty quantification, in situ measurement, and adaptive updating rather than relying on a single universal wear coefficient.

Best-practice workflow for wear diagnosis and modelling

A defensible engineering workflow is to:

1. Define the functional failure limit and the relevant wear metric.

2. Describe the complete tribosystem and duty cycle.

3. Inspect both bodies, debris, lubricant, and environmental conditions.

4. Form competing mechanism hypotheses rather than assuming one cause.

5. Use morphology and chemical evidence to test those hypotheses.

6. Select a model whose assumptions match the dominant mechanism.

7. Calibrate with representative tests and quantify uncertainty.

8. Validate predictions against independent measurements.

9. Monitor regime indicators such as debris, temperature, vibration, and friction.

10. Reassess the model when operating conditions or surface states change.

Chapter Summary

Wear is a system-level damage process involving mechanical, thermal, chemical, and environmental interactions. Adhesive wear arises from shearing of microscopic junctions; abrasive wear from cutting or ploughing by asperities or particles; fatigue wear from cyclic stresses and crack growth; fretting from small-amplitude oscillation; erosive and impact wear from particle or body collisions; and corrosive or oxidative wear from chemical reactions coupled with mechanical removal. Mechanisms commonly coexist and may change during service.

Archard’s law relates wear volume to load, sliding distance, hardness, and a calibrated wear coefficient. It is useful for first estimates but cannot represent every mechanism or evolving contact condition. Mechanistic, energy-based, contact-mechanics, probabilistic, and hybrid data-driven models provide progressively richer descriptions. Wear maps identify regime transitions, while debris analysis, microscopy, profilometry, and chemical examination help diagnose the active processes. Reliable practice requires representative testing, explicit uncertainty, validation, and continuous attention to changes in lubrication, contamination, temperature, geometry, and surface state.

Key Terms

Adhesive wear: Material loss caused by formation and shearing of microscopic interfacial junctions.

Abrasive wear: Material removal or displacement by hard asperities or particles.

Archard’s law: An empirical relation connecting wear volume with load, sliding distance, hardness, and a wear coefficient.

Corrosive wear: Wear accelerated by chemical or electrochemical reactions at the contact.

Delamination: Separation of surface layers following crack initiation and subsurface damage.

Erosive wear: Material loss caused by repeated impingement of particles or droplets carried by a fluid.

Fatigue wear: Wear produced by cyclic stresses, crack growth, and eventual material detachment.

Fretting: Damage caused by small-amplitude oscillatory relative motion.

Impact wear: Damage associated with repeated high-energy collisions or contact events.

Oxidative wear: Wear involving formation and mechanical removal of oxide films or particles.

Running-in: Initial period during which roughness, contact area, and wear rate adjust toward a new state.

Severe wear: A high-rate wear regime often associated with film failure, transfer, plastic deformation, or abrasion.

Specific wear rate: Wear volume normalised by load and sliding distance.

Tribocorrosion: Interaction in which mechanical wear and corrosion mutually accelerate one another.

Tribofilm: A protective or modifying surface film formed during tribological operation.

Three-body abrasion: Abrasion involving loose particles between two contacting surfaces.

Transfer film: Material deposited from one contacting surface onto the other.

Wear coefficient: A calibrated parameter representing the effectiveness of a particular wear process under specified conditions.

Wear map: A graphical representation of wear regimes or rates as functions of controlling variables.

Review Questions

Short Answer Questions

1. Why is a wear coefficient not generally an intrinsic constant of a material?

2. Distinguish two-body from three-body abrasive wear.

3. What physical process is represented by the hardness term in Archard’s law?

4. How does fretting differ from ordinary sliding wear?

5. Why should both the worn surface and the counterface be examined?

Long Answer Questions

1. Explain the mechanisms, characteristic evidence, and control strategies for adhesive, abrasive, and fatigue wear.

2. Derive the wear-depth form of Archard’s equation and discuss its assumptions.

3. Describe how corrosive and oxidative processes interact with mechanical wear.

4. Explain the construction and interpretation of a load–speed wear map.

5. Compare empirical, mechanistic, numerical, probabilistic, and data-driven wear models.

Analytical / Application Questions

1. A sliding contact operates at constant nominal pressure, but its wear rate increases sharply after contamination enters the lubricant. Construct a mechanism-based explanation and identify the evidence you would collect.

2. A hardened gear develops micropits and angular iron particles despite apparently adequate hardness. Use the tribosystem concept to diagnose likely interacting mechanisms and propose a validation plan.

3. Two coatings have equal hardness, but one delaminates and becomes abrasive debris while the other remains adherent. Explain why Archard’s law alone cannot predict their performance and identify the additional variables required.

22. Multiple Choice Questions with Answers

Question 1

Which observation most strongly supports three-body abrasive wear?

A. A continuous transfer film with smeared metallic regions

B. Loose angular particles accompanied by parallel grooves on the surface

C. Subsurface cracks and isolated fatigue spalls without debris

D. A chemically uniform passive film with no surface grooves

Correct Answer: B. Loose angular particles accompanied by parallel grooves on the surface

Question 2

In Archard’s equation V = KWL/H, what does K primarily represent?

A. Only the elastic modulus of the softer body

B. The nominal contact area alone

C. The combined effectiveness of the prevailing wear process under specified conditions

D. The lubricant viscosity at operating temperature

Correct Answer: C. The combined effectiveness of the prevailing wear process under specified conditions

Question 3

A component experiences small-amplitude oscillations, oxidised powder, and cracks at a bolted interface. Which diagnosis is most appropriate?

A. Fretting with possible fretting fatigue

B. Pure rolling-contact fatigue

C. Liquid droplet erosion only

D. Thermal fatigue unrelated to contact motion

Correct Answer: A. Fretting with possible fretting fatigue

Question 4

Why can increasing hardness fail to solve a severe wear problem?

A. Hardness has no relation to wear under any conditions

B. The dominant mechanism may involve chemistry, debris, coating adhesion, or fatigue rather than simple indentation resistance

C. Hard materials always have higher friction

D. Hardness can only be used for polymers

Correct Answer: B. The dominant mechanism may involve chemistry, debris, coating adhesion, or fatigue rather than simple indentation resistance

Question 5

Which evidence most directly indicates fatigue wear?

A. Long, continuous cutting grooves

B. Transferred metallic smears without pits

C. Subsurface cracks followed by pits, spalls, or flakes

D. A uniform oxide film with no material loss

Correct Answer: C. Subsurface cracks followed by pits, spalls, or flakes

Question 6

A local Archard-type finite-element wear update is most useful when the engineer needs to account for:

A. A constant, uniform contact pressure for all time

B. Changes in geometry, pressure distribution, and contact area during wear

C. Only the chemical composition of the lubricant

D. The colour of debris without measuring volume loss

Correct Answer: B. Changes in geometry, pressure distribution, and contact area during wear

Question 7

Which statement about a wear map is most accurate?

A. Its regime boundaries are universal material constants

B. It is valid only when the surface and environmental conditions represented in its construction are relevant

C. It eliminates the need for experimental validation

D. It describes friction but cannot describe wear mechanisms

Correct Answer: B. It is valid only when the surface and environmental conditions represented in its construction are relevant

Question 8

A ductile target struck by particles at a shallow angle most commonly exhibits which dominant response?

A. Cutting and ploughing

B. Only subsurface rolling fatigue

C. Electrochemical dissolution without deformation

D. Pure elastic rebound with no surface change

Correct Answer: A. Cutting and ploughing

Question 9

Why is debris colour alone insufficient to diagnose oxidative or fretting wear?

A. Debris never contains chemical information

B. Colour can arise from multiple materials and mechanisms and must be correlated with morphology and composition

C. Oxides are always colourless

D. Fretting produces no debris

Correct Answer: B. Colour can arise from multiple materials and mechanisms and must be correlated with morphology and composition

Question 10

Which modelling practice is most defensible for a safety-critical wear prediction?

A. Use a convenient literature coefficient without checking test conditions

B. Extrapolate a short test linearly regardless of mechanism changes

C. State assumptions, calibrate with representative data, quantify uncertainty, and validate independently

D. Replace physical inspection entirely with a black-box algorithm

Correct Answer: C. State assumptions, calibrate with representative data, quantify uncertainty, and validate independently