2. Process Physics and Surface Integrity

Introduction and Context

Advanced manufacturing processes are governed by the interaction of energy, material, tools, workholding systems, fluids, and the surrounding environment. A process may remove material by mechanical fracture, plastic deformation, melting, vaporization, chemical dissolution, electrochemical action, or a combination of these mechanisms. The visible result—a machined surface, a drilled hole, an eroded cavity, or an additively built layer—is therefore the outcome of a complex physical history rather than merely a geometric operation.

This history determines surface integrity, a broad term describing the condition of a manufactured surface and its near-surface region. Surface integrity includes surface texture, dimensional accuracy, residual stress, microstructure, hardness, plastic deformation, cracks, inclusions, contamination, and metallurgical transformations. Two components may have the same nominal dimensions and similar roughness values but radically different fatigue lives if one contains tensile residual stress, white layers, thermal cracks, or severe work hardening.

For postgraduate engineers, process selection cannot be separated from process physics. Cutting speed, discharge energy, abrasive concentration, laser power, electrolyte chemistry, feed rate, and cooling conditions are not simply operating settings; they control heat generation, force transmission, material removal, tool wear, defect formation, and the final functional performance of the component. This chapter develops a framework for relating these mechanisms to accuracy and surface integrity across advanced manufacturing processes.

Learning Objectives

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

1. Explain mechanical, thermal, electrical, chemical, and fluid-energy mechanisms involved in advanced manufacturing.

2. Relate process variables to material removal rate, energy consumption, force, temperature, and tool wear.

3. Distinguish geometric accuracy, surface texture, and metallurgical surface integrity.

4. Analyze the formation of residual stresses, heat-affected layers, recast material, microcracks, and work-hardened regions.

5. Select suitable models and measurements for evaluating process performance.

6. Diagnose common defects and propose technically justified corrective actions.

7. Interpret the influence of process physics on fatigue, wear, corrosion, and service reliability.

8. Apply process-monitoring, digital, and data-analytic methods to surface-integrity control.

The Physical Architecture of Material Removal

Energy pathways

Every manufacturing process supplies energy to a workpiece. In conventional machining, mechanical energy is transmitted through a cutting edge. In electrical-discharge machining (EDM), electrical energy is converted into localized thermal energy through plasma channels. In laser processing, electromagnetic radiation produces heating, melting, and sometimes vaporization. In electrochemical machining (ECM), electrical energy drives anodic dissolution. Abrasive waterjet machining combines high-velocity fluid momentum with abrasive impact, while chemical machining uses controlled chemical reactions to remove material.

The useful fraction of supplied energy produces material removal. The remainder is distributed among the tool, workpiece, chips, dielectric or coolant, fixtures, and environment. A general energy balance may be written as

\[

E_{in}=E_{removed}+E_{tool}+E_{workpiece}+E_{fluid}+E_{loss}

\]

The terms are not always measured independently, but the balance is useful for understanding why a process can remove material efficiently while still causing unacceptable thermal damage. A small removed volume may receive a large amount of energy per unit volume, producing melting, phase transformation, or residual stress.

Mechanical removal and deformation

In cutting, the tool compresses material ahead of the cutting edge. A primary shear zone forms where the work material undergoes intense plastic deformation and separates into a chip. Additional deformation occurs at the tool–chip interface and along the machined surface. Cutting force is often resolved into tangential, feed, and radial components. These forces influence power, tool deflection, vibration, dimensional error, and residual stress.

The approximate cutting power is

\[

P=F_c v_c

\]

where \(F_c\) is the principal cutting force and \(v_c\) is cutting speed. Specific cutting energy, defined as power divided by material removal rate, is a useful indicator of process efficiency. It generally increases when the undeformed chip thickness becomes very small because ploughing, rubbing, size effects, and edge-radius effects become significant.

Plastic deformation is not confined to chip formation. The tool edge may elastically and plastically deform the surface, particularly during finishing, burnishing, grinding, and abrasive processes. This can improve bearing properties in some cases, but it may also create a hardened layer, directional texture, or tensile stress if excessive heat accompanies the deformation.

Thermal mechanisms

Heat is generated by plastic deformation in the shear zone, friction at the tool–chip interface, rubbing at the flank face, abrasive action, electrical discharges, laser absorption, and chemical reactions. The temperature field depends on heat-generation rate, contact time, thermal conductivity, specific heat, density, heat-transfer coefficient, and cooling conditions.

A simplified transient heat-conduction relation is

\[

\frac{\partial T}{\partial t}=\alpha \nabla^2T+\frac{\dot q}{\rho c}

\]

where \(T\) is temperature, \(\alpha\) is thermal diffusivity, \(\dot q\) is volumetric heat generation, \(\rho\) is density, and \(c\) is specific heat. The equation illustrates why two materials exposed to the same nominal process settings may develop different surface conditions: thermal properties and phase-transformation behavior differ.

Thermal gradients produce expansion and contraction. If the heated surface is constrained by cooler underlying material, compressive stress may develop during heating and tensile stress during cooling. The final residual stress depends on the thermal cycle, plastic deformation, phase transformation, and material constraint. High temperatures may also reduce hardness, dissolve precipitates, temper martensite, generate a recast layer, or create a heat-affected zone.

Electrical and discharge mechanisms

In EDM, a voltage difference across a small dielectric gap initiates a discharge when the electric field exceeds the breakdown strength of the dielectric. A plasma channel forms, and localized temperatures are sufficiently high to melt and vaporize material. The discharge crater is subsequently flushed by dielectric circulation. The machined surface consists of overlapping craters and may contain recast material, microcracks, and altered microstructure.

Pulse duration, peak current, duty factor, gap voltage, polarity, dielectric condition, and flushing govern crater size and thermal damage. High discharge energy generally increases material removal rate but also increases roughness, recast thickness, and crack risk. Finishing discharges use lower energy to reduce these effects.

Chemical and electrochemical mechanisms

Chemical machining removes material through controlled dissolution. Maskants protect regions that must remain unchanged, while exposed material reacts with the etchant. Undercutting beneath the mask is a central accuracy limitation. Etchant temperature, concentration, agitation, exposure time, and workpiece composition affect removal rate and dimensional control.

In ECM, the workpiece is the anode and the shaped tool is the cathode. Material dissolves anodically according to electrochemical reactions. The interelectrode gap and electrolyte flow determine current distribution and therefore the shape produced. Because the tool does not physically contact the workpiece, mechanical forces and tool wear are low; however, stray corrosion, passivation, gas formation, electrolyte contamination, and poor insulation may reduce accuracy.

Fluid-energy mechanisms

Abrasive waterjet machining converts pump pressure into jet velocity. The water entrains abrasive particles, which impact and cut the workpiece through erosion, micro-ploughing, and brittle fracture. Jet coherence, stand-off distance, abrasive size, traverse speed, pressure, and material properties determine kerf geometry and striation. Excessive stand-off allows jet dispersion and reduces accuracy; excessive traverse speed leaves uncut or rough regions.

In abrasive flow machining, a viscoelastic medium containing abrasive particles is forced through passages. Material removal is concentrated at restrictions and edges, making the process suitable for polishing complex internal passages. Removal is controlled by abrasive concentration, extrusion pressure, medium rheology, and number of cycles.

Material Removal, Accuracy, and Process Models

A useful model must connect controllable variables to measurable outcomes. For conventional machining, material removal rate is commonly approximated by

\[

MRR=v_c f a_p

\]

where \(f\) is feed per revolution or equivalent feed rate and \(a_p\) is depth of cut, subject to the selected kinematic convention. In EDM, MRR is related to pulse energy, pulse frequency, discharge efficiency, and material properties. In ECM, Faraday’s law links removed mass to charge passed:

\[

m=\frac{MIt\eta}{zF}

\]

where \(M\) is molar mass, \(I\) current, \(t\) time, \(\eta\) current efficiency, \(z\) ionic valence, and \(F\) Faraday’s constant.

No model is universally sufficient. Analytical models provide insight into force, temperature, stress, and removal mechanisms, but often simplify geometry and material behavior. Numerical methods, including finite-element, finite-difference, computational-fluid-dynamics, and multiphysics models, permit more realistic treatment of moving heat sources, phase change, fluid flow, and coupled deformation. Empirical and statistical models remain essential in industrial settings because they can incorporate machine-specific effects and are often faster to use for parameter optimization.

Accuracy should be separated into at least four related quantities:

• Dimensional accuracy: agreement between produced and specified dimensions.
• Form accuracy: deviation from the intended geometric form.
• 定位 or positional accuracy: deviation in location, orientation, or feature relationship.
• Surface texture: small-scale roughness, waviness, lay, and surface defects.

A process may achieve low roughness but poor form accuracy, or excellent dimensions but severe metallurgical damage. Consequently, surface-integrity qualification must not rely on roughness alone.

Tool Wear and Its Physical Consequences

Tool wear changes the process while it is running. Common mechanisms include abrasion, adhesion, diffusion, oxidation, fracture, plastic deformation, and electrochemical attack. In cutting tools, flank wear changes the effective tool geometry and increases rubbing. Crater wear alters the rake face and chip flow. In grinding, wheel loading, glazing, and grain fracture change the active cutting behavior. In EDM, electrode wear modifies the tool shape and can distort cavities.

Wear is influenced by hardness, temperature, chemical affinity, contact pressure, cutting speed, lubrication, and the presence of hard inclusions. A worn tool generally produces higher forces and temperature, which accelerate further wear—a process sometimes described as a positive feedback loop. The resulting surface may show dimensional drift, increased roughness, burns, chatter marks, or residual stress changes.

Tool-life models such as Taylor’s relationship,

\[

V T^n=C

\]

are useful for preliminary planning, but advanced processes require caution because wear may be nonuniform and strongly dependent on coating, interruption, thermal cycling, and workpiece microstructure. Tool replacement should therefore be based not only on elapsed time but also on functional indicators such as force, vibration, acoustic emission, temperature, dimensional trend, or surface condition.

Surface Integrity: From Texture to Subsurface Condition

Surface integrity has two principal dimensions. The first is the surface topography, including roughness, waviness, lay, burrs, pits, scratches, and form. The second is the metallurgical and mechanical condition of the surface and subsurface. The latter may include:

• plastic deformation and work hardening;
• residual stress and stress gradients;
• phase transformation and altered grain structure;
• heat-affected or recast layers;
• microcracks, laps, tears, and pores;
• chemical contamination and corrosion susceptibility;
• changes in hardness, toughness, and residual ductility.

A roughness parameter such as arithmetic mean roughness, \(R_a\), is informative but incomplete. Two surfaces with identical \(R_a\) may differ in peak sharpness, spacing, directionality, bearing area, and subsurface damage. Functional parameters, three-dimensional areal measurements, and process-specific inspections are often more meaningful.

Residual stress

Residual stress is self-equilibrating stress remaining after external loads and thermal gradients are removed. Tensile residual stress near a surface can accelerate fatigue-crack initiation and stress-corrosion cracking. Compressive residual stress can improve fatigue performance by opposing crack opening, although excessive compression accompanied by distortion or surface damage is not necessarily beneficial.

Residual stress is generated by mechanical deformation, nonuniform heating and cooling, phase transformation, and chemical or electrochemical effects. Grinding burn, for example, may produce a thermally softened layer and tensile stress if cooling is inadequate. Controlled shot peening produces near-surface plastic deformation and compressive stress, but overpeening can increase roughness or cause surface damage.

Residual stress is measured using methods such as X-ray diffraction, hole drilling, neutron diffraction, contour methods, and incremental sectioning. Each method has a characteristic penetration depth, spatial resolution, and sensitivity to texture and phase composition.

Microstructure and altered layers

Rapid heating and cooling can produce non-equilibrium structures. In EDM, a typical surface may contain a recast layer, a heat-affected zone, and the unaffected base material. The recast layer can be harder or more brittle than the substrate and may contain microcracks. Laser processing can create a remelted layer and a heat-affected zone whose properties depend on energy density and scan strategy.

Mechanical processes may create a plastically deformed layer with elongated grains and increased dislocation density. In machining age-hardened alloys, thermal exposure may overage or dissolve strengthening precipitates. In titanium alloys, excessive temperature can produce undesirable phases or chemical reactions with the environment. Surface integrity must therefore be evaluated in relation to the specific alloy and service requirement.

Defect Formation and Diagnosis

Defects are best diagnosed by connecting their morphology with the underlying physics. Chatter marks indicate dynamic instability involving machine stiffness, tool geometry, cutting parameters, and regenerative feedback. Burrs form when material plastically bends or fractures at an exit edge; their size depends on tool sharpness, workpiece ductility, feed, and edge geometry. Taper in a drilled or wire-cut feature may arise from tool deflection, nonuniform energy distribution, jet dispersion, or electrode wear.

Thermal damage commonly appears as discoloration, softening, rehardening, tensile stress, cracks, or a recast layer. In abrasive processes, glazing reduces cutting ability and increases friction, while loading embeds work material in the wheel. In ECM, stray machining and overcutting result from unintended current paths. In waterjet machining, striations and taper are associated with jet wear, deflection, and insufficient energy at the lower part of the cut.

A systematic diagnosis follows four stages: define the functional defect; measure its spatial distribution; identify the likely energy or deformation mechanism; and verify the proposed correction through controlled trials. Corrective action should address the cause rather than merely reducing the visible symptom.

Measurement and Process Monitoring

Surface-integrity assessment normally combines dimensional metrology, surface-texture measurement, and subsurface characterization. Coordinate-measuring machines and optical systems evaluate form and position. Stylus and areal optical instruments measure texture. Metallographic sectioning, microscopy, microhardness mapping, scanning electron microscopy, energy-dispersive spectroscopy, and X-ray diffraction reveal subsurface and chemical changes.

In-process monitoring is increasingly important. Cutting force signals can reveal tool wear, chatter, and abnormal engagement. Acoustic emission is sensitive to fracture, grinding events, and discharge activity. Infrared cameras and pyrometers estimate temperature, although emissivity and line-of-sight limitations must be considered. Electrical waveform analysis is valuable in EDM and ECM. Vibration and spindle-current signals can support indirect estimation of tool condition.

Digital manufacturing systems can combine sensor data with process models. Statistical process control identifies drift, while machine-learning models can classify defects or predict tool life. Such systems require representative training data, traceable labels, sensor calibration, and attention to false alarms. A prediction model should not replace physical reasoning; it should complement it.

Applied Case: Surface Integrity in Wire EDM of a Hardened Die Insert

Background

The following is an illustrative industrial case. A hardened tool-steel die insert must be produced by wire EDM because conventional milling would impose excessive tool forces and tool wear. The insert requires accurate cavity geometry, low roughness, and adequate fatigue resistance during repeated forming cycles.

Situation and problem

An initial rough-cutting schedule achieves the required geometry within a reasonable time, but inspection identifies a relatively rough surface, a visible recast layer, and scattered microcracks near sharp corners. The component meets dimensional requirements, yet early trials show premature crack initiation during cyclic service.

Analysis

The roughing schedule uses high pulse energy to maximize removal rate. Each discharge creates a molten pool; rapid solidification leaves overlapping craters and a recast layer. At corners, flushing is less effective and local discharge conditions become unstable. Thermal contraction and repeated heating generate tensile residual stress. The dimensional result is acceptable because the geometry is controlled by the wire path, but functional surface integrity is inadequate.

Approach

The manufacturing team separates the operation into roughing, skim cutting, and finishing passes. Roughing parameters are retained for efficient bulk removal. Subsequent passes use lower peak current and shorter pulse duration, improved wire tension and dielectric filtration, and a controlled offset strategy. Corner speed is reduced to stabilize energy delivery. The insert is then inspected using roughness measurement, metallographic cross-sections, microhardness mapping, and X-ray diffraction.

Outcome

The finishing sequence reduces crater size and recast thickness and greatly lowers the frequency of visible microcracks. Dimensional accuracy is maintained through the final skim pass. The additional processing time is justified because the insert now satisfies both geometric and functional requirements.

Learning points

The case demonstrates that dimensional accuracy is not equivalent to surface integrity. It also shows the value of separating roughing and finishing objectives, controlling local thermal conditions, and selecting inspection methods that reveal subsurface damage. A process decision based solely on material removal rate would have been inadequate.

Challenges, Limitations, and Professional Considerations

The main challenge in surface-integrity control is the coupling of variables. Increasing productivity may increase heat, force, vibration, wear, or defect probability. Cooling may reduce temperature but introduce thermal shock, contamination, corrosion, or disposal problems. A parameter window that works for one alloy, machine, tool coating, or geometry may fail for another.

Common mistakes include treating roughness as the only quality indicator, using nominal rather than actual tool condition, ignoring machine dynamics, applying a model outside its calibration range, and changing several parameters simultaneously without a designed experiment. Another error is to inspect only the final dimension while neglecting residual stress or microstructure that controls service life.

Professional practice requires traceability of material certificates, tool condition, coolant or dielectric state, calibration records, process parameters, and inspection results. Chemical etchants, electrolytes, dielectric fluids, and metallic dusts require appropriate environmental and occupational controls. Waste handling, fluid recycling, ventilation, electrical isolation, laser safety, and guarding are not separate from process engineering; they are part of responsible process design.

Current Developments and Best Practices

Current research and industrial development emphasize hybrid and adaptive processes. Examples include laser-assisted machining, ultrasonic-assisted cutting, electrochemical-assisted machining, minimum-quantity lubrication, cryogenic cooling, and additive–subtractive manufacturing. These approaches seek to modify local material behavior, reduce force, improve tool life, or control heat, but they introduce additional interfaces and process variables.

Best practice is to establish a process–structure–property chain: identify the service requirement, determine the critical integrity attributes, select mechanisms and parameters that can produce them, and verify the result with appropriate measurements. Design of experiments is more reliable than one-factor-at-a-time trials for studying interaction effects. Statistical control should be based on functional quality characteristics, not only easily measured geometric quantities.

For digitally enabled production, sensor fusion and model-based control offer opportunities for adaptive feed, energy regulation, tool replacement, and defect prevention. Nevertheless, physical validation remains essential, particularly when failure consequences are high. Explainable models, secure data handling, calibration, and human oversight are important professional requirements.

Chapter Summary

Advanced manufacturing outcomes arise from coupled mechanical, thermal, electrical, chemical, and fluid-energy mechanisms. These mechanisms determine material removal, force, power, temperature, tool wear, dimensional accuracy, and defect formation. Surface integrity extends beyond roughness to include residual stress, microstructure, hardness, altered layers, cracks, contamination, and subsurface deformation. High removal rate may conflict with functional quality because concentrated energy can generate thermal damage or excessive deformation.

Analytical, numerical, empirical, and data-driven models each provide different kinds of insight. Effective engineering practice combines them with dimensional metrology, texture measurement, metallography, residual-stress analysis, and in-process monitoring. The illustrative wire-EDM case shows why roughing and finishing must be designed for different objectives. Reliable process design therefore begins with the service function of the component and ends with evidence that both geometry and surface integrity meet that function.

Key Terms

Surface integrity: The combined topographical, mechanical, metallurgical, and chemical condition of a manufactured surface and its subsurface region.

Specific cutting energy: Energy required per unit volume of material removed in a cutting operation.

Residual stress: Self-equilibrating stress remaining in a component after external loads and major process forces are removed.

Recast layer: Rapidly resolidified material deposited on a surface after local melting.

Heat-affected zone: Region whose microstructure or properties change due to process heating without complete melting.

Material removal rate: Volume of material removed per unit time.

Plastic deformation: Permanent deformation caused when stress exceeds the material’s elastic limit.

Tool wear: Progressive loss or alteration of tool material and geometry during processing.

Flank wear: Wear occurring on the tool face that contacts the newly machined surface.

Crater wear: Depression formed on the tool rake face by chip contact and thermochemical action.

Chatter: Self-excited or forced vibration that produces periodic surface marks and process instability.

Areal roughness: Three-dimensional statistical description of surface height variations over an area.

Work hardening: Increase in hardness and strength caused by plastic deformation.

Thermal gradient: Spatial variation of temperature within a material or process zone.

Electrochemical machining: Anodic dissolution process controlled by electrical current and electrolyte flow.

Electrical-discharge machining: Material-removal process using localized thermal discharges across a dielectric gap.

Specific energy density: Energy supplied per unit volume or area of processed material.

Tool-life model: Mathematical relationship used to relate tool life to process variables such as cutting speed.

Process monitoring: Measurement and interpretation of process signals to detect state changes or defects.

Review Questions

Short Answer Questions

1. Define surface integrity and distinguish it from surface roughness.

2. What physical mechanisms generate heat during conventional machining?

3. State two causes of residual stress in a manufactured surface.

4. Why does tool wear influence dimensional accuracy?

5. What is the function of a dielectric in EDM?

Long Answer Questions

1. Explain how mechanical deformation and thermal effects jointly determine surface integrity in machining.

2. Discuss the formation, effects, and measurement of residual stresses in manufactured components.

3. Compare EDM and ECM with respect to energy mechanism, tool contact, accuracy, and surface condition.

4. Explain the role of tool wear in process drift and describe suitable monitoring signals.

5. Develop a process-control strategy for producing a fatigue-critical component with controlled surface integrity.

Analytical / Application Questions

1. A grinding operation produces acceptable roughness but causes subsurface tensile stress and reduced fatigue life. Diagnose the likely thermal and mechanical causes and recommend process changes.

2. An EDM cavity has good dimensions but excessive recast material and microcracks at corners. Analyze the discharge and flushing conditions and propose a revised parameter strategy.

3. A waterjet-cut component shows increasing taper and striation toward the bottom of the kerf. Relate the defect to jet physics and recommend changes in pressure, abrasive delivery, stand-off distance, and traverse speed.

22. Multiple Choice Questions with Answers

Question 1

Which statement best explains why surface roughness alone is insufficient for evaluating surface integrity?

A. Roughness can be measured only on polished surfaces.

B. Two surfaces with similar roughness may differ in residual stress and subsurface microstructure.

C. Roughness is unrelated to manufacturing parameters.

D. Surface integrity concerns only dimensional accuracy.

Correct Answer: B. Two surfaces with similar roughness may differ in residual stress and subsurface microstructure.

Question 2

In conventional machining, approximate cutting power is most directly calculated as:

A. Cutting force multiplied by cutting speed

B. Feed multiplied by depth of cut

C. Temperature multiplied by thermal conductivity

D. Tool wear divided by machining time

Correct Answer: A. Cutting force multiplied by cutting speed

Question 3

High-energy EDM discharges generally produce:

A. Smaller craters and no thermal effects

B. Higher removal rate with increased risk of recast material and microcracks

C. Purely mechanical deformation without melting

D. Anodic dissolution controlled by electrolyte flow

Correct Answer: B. Higher removal rate with increased risk of recast material and microcracks

Question 4

Which condition most directly promotes tensile residual stress after a thermally severe process?

A. Uniform cooling without constraint

B. Rapid heating and cooling with thermal expansion constrained by cooler material

C. Low temperature elastic loading only

D. Removal of all surface material before cooling

Correct Answer: B. Rapid heating and cooling with thermal expansion constrained by cooler material

Question 5

Faraday’s law is particularly relevant to estimating material removal in:

A. Electrochemical machining

B. Abrasive waterjet machining

C. Conventional turning

D. Laser cutting without chemical reaction

Correct Answer: A. Electrochemical machining

Question 6

A worn cutting tool commonly causes dimensional drift because it:

A. Eliminates friction at the tool–chip interface

B. Changes effective tool geometry and increases rubbing and heat generation

C. Always decreases cutting forces

D. Prevents plastic deformation in the workpiece

Correct Answer: B. Changes effective tool geometry and increases rubbing and heat generation

Question 7

In wire EDM, a finishing pass using lower pulse energy is primarily intended to:

A. Increase crater size

B. Increase electrode wear

C. Reduce roughness and thermal damage after roughing

D. Eliminate the need for dielectric circulation

Correct Answer: C. Reduce roughness and thermal damage after roughing

Question 8

Which measurement combination is most appropriate for detecting both geometric accuracy and subsurface thermal damage?

A. Only visual inspection

B. Coordinate measurement together with metallography and microhardness mapping

C. Only arithmetic roughness measurement

D. Only measurement of machining time

Correct Answer: B. Coordinate measurement together with metallography and microhardness mapping

Question 9

Increasing abrasive waterjet stand-off distance excessively is likely to:

A. Improve jet coherence indefinitely

B. Increase jet dispersion and reduce cutting accuracy

C. Eliminate kerf taper

D. Convert erosion into electrochemical dissolution

Correct Answer: B. Increase jet dispersion and reduce cutting accuracy

Question 10

Why should a machine-learning monitoring model be validated with physical process knowledge?

A. Sensor data never contain useful information.

B. Physical reasoning helps identify calibration errors, false alarms, and conditions outside the training data.

C. Machine-learning models automatically measure residual stress directly.

D. Physical models make all measurements unnecessary.

Correct Answer: B. Physical reasoning helps identify calibration errors, false alarms, and conditions outside the training data.