Engineering design is the process of turning a need into a product, system, part, or process that can be built, checked, used, maintained, and changed in a controlled way.

A design is not just a CAD model. A complete engineering design explains:

1. What the thing must do.

2. What is being built.

3. Why the chosen solution should work.

4. How much variation is allowed.

5. How it will be manufactured, inspected, and maintained.

6. How failures and future changes will be controlled.

The subject flows best as a sequence:

Each topic in engineering design fits somewhere in that chain.

1. The design loop

Engineering design is iterative. Engineers rarely move straight from problem to finished product. They define the need, propose a solution, test the solution against constraints, revise it, and repeat until the design is good enough to release.

A simple design loop is:

1. Define the problem.

2. Convert the problem into measurable requirements.

3. Generate possible solutions.

4. Select the most promising concept.

5. Define the geometry, materials, interfaces, and parts.

6. Analyze whether the design works.

7. Check whether it can be manufactured and inspected.

8. Build and test prototypes or first articles.

9. Release controlled documentation.

10. Manage changes through revision control.

The loop matters because every design decision affects several others. A tighter tolerance may improve fit, but it may also increase cost. A stronger material may reduce failure risk, but it may be harder to machine or more prone to corrosion. A lighter part may improve performance, but it may vibrate more.

Main categories of design work

Running example

Use a simple example throughout this guide: an aluminum wall bracket that holds a small sensor box.

The bracket must:

• Support the sensor weight.

• Bolt to a wall.

• Keep the sensor aligned.

• Survive vibration and temperature changes.

• Be easy to manufacture, inspect, and replace.

Even this simple part involves requirements, loads, material selection, CAD, drawings, tolerances, GD&T, fasteners, inspection, failure modes, and revision control.

2. From need to requirements

Big idea: requirements define what success means.

A requirement is a clear statement of what the design must do or satisfy. Without requirements, there is no objective way to decide whether the design is correct.

A weak requirement is vague:

The bracket should be strong.

A better requirement is measurable:

The bracket shall support a static vertical load of 500\ \mathrm{N} with a minimum factor of safety of 2.0 against yield.

The second version can be analyzed, tested, and accepted or rejected.

Need, requirement, and acceptance criterion

These three ideas are related but not the same.

A requirement says what must be true. An acceptance criterion says how the requirement will be judged.

Requirement types

Good requirements are specific, measurable, realistic, traceable, and verifiable.

Verification and validation

Verification checks whether the design meets the written requirements.

Validation checks whether the requirements actually solve the user’s problem.

A bracket can pass every written load test and still be a bad design if the real sensor cable cannot be installed. That is a validation problem.

Traceability

Traceability connects the original need to design decisions and evidence.

Example:

Traceability prevents random design features. Every critical feature should exist for a reason.

Common mistakes with requirements

• Using words like strong, light, cheap, durable, or easy without numbers.

• Mixing a design solution into a requirement too early.

• Forgetting environment, service, or inspection requirements.

• Writing requirements that cannot be tested.

• Failing to identify which requirements are critical.

3. Concept design and architecture

Big idea: before defining details, engineers decide what kind of solution they are building.

A concept is a possible way to satisfy the requirements. At this stage, the design may be sketches, rough CAD, block diagrams, prototypes, or simple calculations.

The purpose is not to finish the design. The purpose is to choose a direction before spending time on detailed drawings and analysis.

Functional decomposition

Functional decomposition breaks a problem into smaller functions.

For the bracket example:

This helps avoid jumping directly to geometry before understanding the job of each feature.

Concept generation

Common ways to generate concepts:

• Sketch multiple layouts.

• Change material or manufacturing process.

• Change load path.

• Reduce part count.

• Use standard parts.

• Move interfaces.

• Separate functions into modules.

• Combine functions into one part.

For the bracket, possible concepts might be:

Trade studies

A trade study compares options against criteria. A simple weighted score can organize the decision, but it should not replace judgment.

Example scoring form:

Weighted score for one concept:

where w i is the criterion weight and s i is the concept score.

The score is only useful if the criteria are chosen honestly and the major risks are discussed.

Design intent

Design intent is the reason behind the geometry and specifications.

It answers questions like:

• Which surfaces locate the part?

• Which features carry the load?

• Which dimensions are critical?

• Which tolerances protect assembly?

• Which features are noncritical and can remain loose?

• What should not change during a future revision?

Example:

The large rear face of the bracket is datum A because it contacts the wall. The two lower holes are datum features B and C because they locate the bracket pattern. The top surface is not a datum because it does not control function.

This kind of reasoning keeps drawings, CAD constraints, GD&T, and inspection aligned.

4. Interfaces and system boundaries

Big idea: many design failures happen where parts or teams meet.

An interface is a boundary where one part, subsystem, supplier, or team connects to another. Interfaces must be controlled because a change on one side can break the other side.

Types of interfaces

The bracket has mechanical interfaces with the wall and sensor box. It may also have cable clearance and service interfaces.

Interface control document

An interface control document, or ICD, records what each side must obey.

Typical ICD content:

Why interface control matters

Suppose the wall bracket team moves a mounting hole by 2\ \mathrm{mm} to make the bracket easier to machine. If the sensor box team is not aware of the change, the parts may no longer assemble. Interface control prevents this kind of mismatch.

Interfaces should be identified early, controlled clearly, and changed formally.

5. Product definition: CAD, drawings, and BOMs

Big idea: product definition is the controlled description of what will be built.

Once a concept is selected, the design must become specific enough for manufacturing, inspection, purchasing, assembly, and service.

CAD

Computer-aided design, or CAD, is the digital model of a part or assembly. CAD helps engineers create geometry, check fit, generate drawings, support simulation, and communicate design intent.

Common CAD file types by role:

CAD is not automatically a complete design. A model may show shape, but it may not show tolerances, materials, finishes, inspection rules, or revision status.

Parametric modeling and intent

Parametric CAD uses dimensions, constraints, and relationships. Good models are built so important design intent survives changes.

Poor modeling intent:

• A hole is located from a random edge that may later move.

• A fillet is added before the feature it depends on is stable.

• Symmetric parts are modeled with unrelated dimensions.

Better modeling intent:

• Holes are located from functional datums.

• Symmetry is controlled with a center plane.

• Critical dimensions are named and easy to edit.

• Nonfunctional cosmetic features are added late.

Engineering drawings

An engineering drawing is a controlled 2D document that defines a part or assembly for manufacture and inspection.

A drawing usually includes:

Model-controlled and drawing-controlled design

Different organizations use different authority structures.

The important rule is consistency. Do not define the same requirement in two places if they can disagree.

Bill of materials

A bill of materials, or BOM, lists the parts required to build a product.

Common BOM fields:

The BOM connects engineering to purchasing, manufacturing, assembly, cost, inventory, and service.

Assembly documentation

Assemblies often need more than part drawings. They may require:

• Exploded views.

• Item balloons tied to the BOM.

• Fastener callouts.

• Torque values.

• Adhesive or lubricant notes.

• Assembly order.

• Inspection checks.

• Functional tests.

• Service instructions.

For the bracket, an assembly drawing might specify bolt type, washer use, installation torque, cable routing, and sensor orientation.

6. Dimensions, tolerances, and variation

Big idea: real parts are imperfect, so design must control acceptable imperfection.

No manufacturing process produces exact geometry. Tolerances define how much variation is allowed while still preserving function.

A dimension has a nominal value and limits.

Example:

This means:

For a general bilateral tolerance:

For a symmetric tolerance \pm T :

Why tolerances matter

Tolerances affect:

• Assembly fit.

• Interchangeability.

• Manufacturing cost.

• Inspection cost.

• Scrap rate.

• Supplier capability.

• Product reliability.

A loose tolerance may make assembly unreliable. A tight tolerance may make the part expensive or impossible to produce consistently.

The design question is not “how tight can this be?” It is:

How loose can this be while the design still works?

Tolerance types

Fits and clearances

A clearance is the space between mating parts. A fit describes the relationship between two mating features, usually a hole and shaft.

For a hole and shaft:

Minimum clearance:

Maximum clearance:

Interpretation:

Common fit types:

Tolerance stack-up

Tolerance stack-up predicts how multiple part variations combine in an assembly.

Suppose a bracket must leave a clearance gap G between the sensor and a cover. If several dimensions affect that gap, each tolerance can increase or decrease the final clearance.

Worst-case stack-up assumes every dimension is at its worst possible limit:

This is conservative. If it passes, the assembly should work even in the worst allowed combination.

A statistical root-sum-square estimate is:

RSS is less conservative and depends on assumptions about independent, centered process variation.

Stack-up workflow

1. Define the functional gap, clearance, or alignment requirement.

2. Draw the dimension chain from one side of the requirement to the other.

3. Decide which dimensions increase and decrease the gap.

4. Insert nominal dimensions and tolerances.

5. Calculate minimum and maximum possible gap.

6. Compare to the functional requirement.

7. Tighten only the dimensions that actually control the result.

Practical tolerance rule

Do not tolerance every feature tightly. Tighten features that control function:

• Bearing seats.

• Seal surfaces.

• Locating holes.

• Critical gaps.

• Alignment surfaces.

• Interfaces to other parts.

Leave nonfunctional cosmetic or clearance features with general tolerances when possible.

7. GD&T and datums

Big idea: GD&T controls geometry using functional references instead of only plus/minus dimensions.

Geometric Dimensioning and Tolerancing, or GD&T, is a symbolic language for controlling form, orientation, location, profile, and runout.

GD&T is useful when ordinary coordinate dimensions do not clearly describe how the part must function or how it should be inspected.

Datums

A datum is a theoretically exact reference used for measurement and assembly. A datum feature is the real part feature used to establish that reference.

For the bracket:

• The wall-contact face might define datum A.

• One mounting hole might define datum B.

• A second mounting hole or slot might define datum C.

This creates a repeatable coordinate frame for inspection.

Datum priority

Datums should usually be functional features: surfaces, holes, or axes that actually locate the part in use.

Common GD&T controls

Feature control frame

A feature control frame states the geometric rule.

Conceptually:

Example meaning:

The axis of this hole must lie inside a cylindrical tolerance zone relative to datums A, B, and C.

Position tolerance

Position tolerance is often used for holes because it controls the allowable location of a feature relative to datums.

For a circular position tolerance zone of diameter T , the allowed radial error from true position is:

This is different from separate x and y coordinate tolerances. A circular tolerance zone better matches how a round hole functions.

Material condition modifiers

For a hole, MMC is the smallest allowed hole. For a shaft or pin, MMC is the largest allowed shaft or pin.

GD&T workflow

1. Identify how the part is located in the assembly.

2. Choose datum features from functional interfaces.

3. Identify which features need form, orientation, or position control.

4. Apply GD&T only where it improves function or inspection clarity.

5. Check that the tolerance can be measured.

6. Check that the tolerance is compatible with manufacturing capability.

Common GD&T mistakes

• Choosing datums from nonfunctional surfaces.

• Overusing GD&T where simple dimensions are enough.

• Creating datum schemes that inspection cannot reproduce.

• Tightening position tolerances without stack-up justification.

• Forgetting that datum order matters.

8. Materials, surfaces, and manufacturing process selection

Big idea: geometry only works if the material, surface, and process are compatible with the job.

Material selection and manufacturing process selection should happen together. A material that is strong on paper may be expensive, hard to machine, difficult to weld, prone to corrosion, or incompatible with the required surface finish.

Material selection

Material selection is the choice of what a part is made from.

Important material properties:

Strength versus stiffness

Strength and stiffness are different.

Strength asks whether the material fails:

Stiffness asks whether the part moves too much:

A part can be strong enough but too flexible. A rubber part may survive the load but deflect too much. A steel part may be stiff but too heavy.

Specific properties

For weight-sensitive design, compare properties per unit density.

Specific strength:

Specific stiffness:

These help compare materials such as aluminum, steel, titanium, composites, and plastics when mass matters.

Thermal expansion

Temperature changes alter dimensions:

where \alpha is the coefficient of thermal expansion, L 0 is original length, and \Delta T is temperature change.

Thermal expansion matters for:

• Precision fits.

• Mixed-material assemblies.

• Electronics housings.

• Bearings and seals.

• Long rails or frames.

• Hot or cold environments.

Surface finish

Surface finish describes surface texture, often using roughness R a .

Surface finish affects:

• Friction.

• Wear.

• Sealing.

• Adhesion.

• Coating quality.

• Fatigue strength.

• Appearance.

• Cleanability.

Examples:

Coatings and treatments

Common treatments include:

• Anodizing.

• Plating.

• Painting.

• Passivation.

• Heat treatment.

• Case hardening.

• Shot peening.

• Polishing.

• Dry-film lubrication.

Coatings can change dimensions. If a coating adds thickness, it can change a fit, close a clearance, or affect thread engagement.

Manufacturing process awareness

Different processes have different design rules.

A design is not finished until the chosen geometry can actually be produced by the chosen process.

9. Loads, free body diagrams, and load paths

Big idea: before calculating stress, understand where the forces go.

Mechanical analysis starts by identifying loads and supports. The most important first tool is usually a free body diagram.

Free body diagram

A free body diagram, or FBD, isolates one body and shows all external forces and moments acting on it.

For static equilibrium:

In vector form:

How to make an FBD

1. Choose the body or part to isolate.

2. Remove surrounding parts.

3. Replace contacts with reaction forces or moments.

4. Add known applied loads.

5. Add weight if relevant.

6. Choose coordinate axes.

7. Write equilibrium equations.

8. Solve for unknown reactions.

For the wall bracket, the sensor creates a downward load and a moment at the wall. The bolts and wall contact must react those loads.

Load path

The load path is the route force takes through the structure.

For the bracket:

Load path questions:

• Where does the load enter?

• Where does it leave?

• Which features carry tension, compression, shear, bending, or torque?

• Are there sharp corners or thin sections in the path?

• Are fasteners loaded in a good direction?

• Is there an alternate load path if one feature fails?

A strong-looking part can fail if the load path passes through a weak feature.

Load cases

A load case is one defined loading condition.

Common load cases:

Analysis should cover the load cases that matter for requirements and credible use.

10. Stress, strain, deflection, and margin

Big idea: mechanical analysis checks both failure and motion.

A design can fail by breaking, yielding, bending too much, wearing out, overheating, buckling, or vibrating. The basic mechanical checks start with stress, strain, deflection, factor of safety, and margin.

Stress

Stress is internal force per area.

Axial normal stress:

Average shear stress:

Bending stress:

where M is bending moment, c is distance from the neutral axis to the outer fiber, and I is the second moment of area.

Stress predicts whether a material will yield, crack, or fail.

Strain

Strain is deformation compared to original length:

For linear elastic behavior:

where E is elastic modulus.

Stress describes internal loading. Strain describes deformation.

Deflection

Deflection is how much a part moves under load.

For an axially loaded member:

For a cantilever beam with an end load:

Deflection matters for:

• Alignment.

• Gaps.

• Sealing pressure.

• Gear or bearing contact.

• Optical paths.

• Vibration behavior.

• User feel.

A bracket may have low stress but still allow the sensor to move too much.

Factor of safety

Factor of safety compares capacity to demand:

For yielding:

A factor of safety greater than 1 means estimated capacity exceeds estimated demand.

Required factor of safety depends on:

• Uncertainty in load.

• Material variability.

• Manufacturing quality.

• Inspection quality.

• Consequence of failure.

• Whether people can be injured.

• Whether standards specify a value.

Design margin

Design margin expresses remaining capacity:

Since:

then:

Interpretation:

Basic mechanical analysis workflow

1. Identify the requirement.

2. Identify the load case.

3. Draw the FBD.

4. Find reactions and internal loads.

5. Calculate stress and deflection.

6. Compare to allowable limits.

7. Calculate factor of safety and margin.

8. Check whether assumptions are conservative and realistic.

11. Failure modes: fatigue, buckling, wear, and details

Big idea: a design can fail even when the first static calculation looks acceptable.

A failure mode is a way the design stops satisfying its requirements.

Common failure modes:

Fatigue

Fatigue is failure caused by repeated loading. A part can fail below yield strength if the load repeats enough times.

Fatigue risk increases with:

• High cyclic stress.

• Stress concentrations.

• Rough surfaces.

• Welds.

• Corrosion.

• Tensile mean stress.

• Poor material quality.

• Vibration.

For the bracket, fatigue may matter if the wall is attached to vibrating equipment.

Stress concentration

A stress concentration occurs where geometry creates locally high stress, such as at holes, sharp corners, grooves, notches, threads, and weld toes.

A simple relation is:

where K t is the theoretical stress concentration factor.

Design responses include:

• Larger fillets.

• Smooth transitions.

• Better surface finish.

• Lower local load.

• Moving holes away from high-stress areas.

• Avoiding abrupt section changes.

Buckling

Buckling is instability under compression. A slender member can collapse sideways before the material reaches yield.

For an ideal column:

where:

Buckling matters for struts, columns, thin panels, lightweight frames, sheet metal walls, and compression members.

Fasteners and joints

Fasteners are common failure points because they involve preload, friction, vibration, fatigue, and assembly variation.

Important joint questions:

• Is the joint loaded in tension, shear, bending, or torque?

• Is preload required?

• Can vibration loosen the joint?

• Is thread engagement sufficient?

• Is the material under the head or nut strong enough?

• Is torque controlled during assembly?

• Is there tool access?

Bolt torque is often used as an approximate way to create preload:

where T is tightening torque, K is a nut factor, F is preload, and d is nominal bolt diameter.

This estimate is rough because friction dominates the result.

12. Thermal, vibration, and electrical analysis

Big idea: real products heat up, cool down, vibrate, and consume power.

Engineering design is not only static strength. A product can fail because it overheats, resonates, loses voltage, expands too much, or draws more power than expected.

Heat transfer

Heat transfer is the movement of thermal energy.

The three basic modes are conduction, convection, and radiation.

Conduction through a simple wall:

Convection:

Radiation:

where q is heat transfer rate.

Thermal resistance

Thermal systems can often be modeled like electrical resistance:

For conduction through a flat layer:

This is useful for heat sinks, electronics enclosures, motors, batteries, insulation, and thermal interfaces.

Vibration

Vibration is repeated motion caused by dynamic forces. It can cause noise, fatigue, loosened fasteners, sensor error, discomfort, or failure.

For a simple mass-spring system:

Frequency in hertz is:

where k is stiffness and m is mass.

Modal analysis

Modal analysis finds natural frequencies and mode shapes.

A design is at risk when excitation frequencies are near natural frequencies. This can cause resonance, where vibration amplitude becomes large.

Common excitation sources:

• Motors.

• Fans.

• Engines.

• Pumps.

• Road input.

• Rotating imbalance.

• Gear mesh.

• Human input.

Design responses include increasing stiffness, reducing mass, adding damping, isolating the source, or shifting excitation frequency.

Electrical load analysis

Electrical load analysis checks whether the system can safely provide required voltage, current, and power.

Ohm’s law:

Power:

For a resistive load:

Voltage drop in a conductor:

Too much voltage drop can cause resets, overheating, poor performance, or failure.

Power budget

A power budget lists electrical loads and compares demand to available supply.

Total power:

For a battery-powered system, a rough runtime estimate is:

Power budgets should include:

• Continuous loads.

• Startup loads.

• Peak loads.

• Duty cycles.

• Conversion efficiency.

• Thermal limits.

• Reserve margin.

13. CAE, FEA, and engineering judgment

Big idea: simulation is useful only when the model represents the real problem well enough.

Computer-aided engineering, or CAE, uses software to analyze engineering behavior. Finite element analysis, or FEA, divides geometry into small elements to estimate stress, deflection, temperature, vibration, or other fields.

FEA is powerful, but it is not automatically correct. The output depends on assumptions.

What FEA can help with

FEA workflow

1. Define the engineering question.

2. Identify the requirement and pass/fail rule.

3. Simplify geometry without removing important behavior.

4. Assign material properties.

5. Apply loads and constraints.

6. Define contacts, joints, and connections.

7. Mesh the model.

8. Solve.

9. Check reactions and units.

10. Refine the mesh and compare results.

11. Compare to hand calculations or test data.

12. Report assumptions, limitations, results, and margin.

Boundary conditions

Boundary conditions often dominate simulation error.

An over-constrained model may look too stiff. An under-constrained model may move unrealistically. A load applied over a tiny artificial area may create meaningless peak stress.

Good boundary conditions should represent how the real part is supported, loaded, and connected.

Mesh convergence

Mesh convergence checks whether a result stabilizes as the mesh is refined.

A simple percent-change check is:

where R is the result being tracked.

Peak stress at a sharp corner may not converge because ideal sharp corners create singularities. In that case, evaluate a realistic fillet, averaged stress, or a more appropriate failure metric.

Hand calculations still matter

Hand calculations are used to:

• Check units.

• Estimate magnitude.

• Reveal bad assumptions.

• Verify load paths.

• Provide independent comparison.

• Interpret simulation results.

A simulation result should be questioned if it is not within the same order of magnitude as a simple estimate.

14. DFMA and manufacturability

Big idea: a design is not complete unless it can be made and assembled repeatedly.

Design for Manufacturing and Assembly, or DFMA, means designing parts so they are easy to make, assemble, inspect, and service.

DFMA goals

DFMA tries to:

• Reduce part count.

• Simplify geometry.

• Use standard materials and fasteners.

• Avoid unnecessary tight tolerances.

• Reduce setup changes.

• Improve tool access.

• Prevent incorrect assembly.

• Lower cost and cycle time.

• Improve inspection clarity.

Manufacturing constraints by process

Assembly design principles

Good assembly design usually tries to:

• Make parts self-locating.

• Avoid ambiguous orientation.

• Use common fasteners.

• Provide tool clearance.

• Minimize hidden operations.

• Avoid long tolerance chains.

• Use keying to prevent incorrect assembly.

• Make critical features visible or inspectable.

• Keep service parts accessible.

Cost drivers

Manufacturing cost is affected by:

• Material choice.

• Part size and weight.

• Tolerance tightness.

• Surface finish.

• Number of setups.

• Tooling cost.

• Inspection burden.

• Scrap rate.

• Assembly labor.

• Supplier availability.

• Production volume.

A common design mistake is to specify precision that function does not require.

15. Quality, inspection, and metrology

Big idea: inspection connects the written design to real manufactured parts.

Quality is not created by inspection alone. Quality is designed into requirements, tolerances, materials, processes, assembly methods, supplier controls, and measurement systems.

Metrology

Metrology is the science of measurement.

Measurement quality depends on:

• Instrument resolution.

• Calibration.

• Fixturing.

• Operator method.

• Temperature.

• Datum setup.

• Surface condition.

• Part flexibility.

• Measurement uncertainty.

A measured value should not be treated as exact:

where U is measurement uncertainty.

Inspection plan

An inspection plan defines what will be measured, how it will be measured, how often it will be measured, and what happens if it fails.

Typical fields:

For the bracket, an inspection plan may include material certification, hole diameter, hole position, bracket flatness, coating thickness, and visual inspection.

Gauge R&R

Gauge Repeatability and Reproducibility, or Gauge R&R, checks whether the measurement system is reliable.

• Repeatability: variation when the same operator measures the same part with the same gauge.

• Reproducibility: variation between operators, setups, or gauges.

A poor measurement system can reject good parts or accept bad parts.

Process capability

Process capability describes whether a stable manufacturing process can consistently meet tolerance.

Let USL be the upper specification limit, LSL the lower specification limit, \mu the process mean, and \sigma the process standard deviation.

Potential capability:

Centered capability:

C p measures spread compared to tolerance width. C {pk} also accounts for whether the process is centered.

SPC

Statistical Process Control, or SPC, tracks process data over time to detect drift, instability, and abnormal variation.

SPC can reveal problems caused by:

• Tool wear.

• Temperature changes.

• Material variation.

• Machine changes.

• Operator changes.

• Fixture wear.

Common SPC tools:

• Control charts.

• Run charts.

• Histograms.

• Process capability studies.

• Pareto charts.

• Out-of-control rules.

Quality control versus quality assurance

Inspection catches problems. Good design and process control prevent them.

16. Reliability and risk

Big idea: a design must keep working in real service, not just pass once.

Reliability is the ability of a product to perform its required function for a specified time under specified conditions.

Reliability model

For a constant failure rate \lambda , a simple reliability model is:

Mean time between failures for a repairable system is often approximated as:

This model is useful as a basic reference, but it should not be applied blindly to all failure types.

MTBF, MTTR, and availability

MTBF is mean time between failures.

MTTR is mean time to repair.

Simple availability:

A system can improve availability by failing less often, being faster to repair, or both.

Derating

Derating means using a component below its maximum rating.

Examples:

• Use a capacitor below its voltage rating.

• Use a power supply below maximum load.

• Keep electronics below maximum junction temperature.

• Use a bearing below rated load.

• Avoid running a motor continuously at its limit.

Derating reduces stress and usually improves reliability.

Redundancy

Redundancy means adding backup components, paths, or functions.

Redundancy can improve reliability, but it also adds cost, weight, complexity, and possible new failure modes.

FMEA

Failure Mode and Effects Analysis, or FMEA, is a structured way to predict failures and reduce risk before they happen.

Common FMEA fields:

Traditional risk priority number:

where S is severity, O is occurrence, and D is detection ranking.

RPN helps sort risks, but engineering judgment is still required. A high-severity failure may deserve action even if the RPN is not the largest.

Fault tree analysis

Fault tree analysis starts with a top-level failure and maps combinations of lower-level failures that could cause it.

Common logic:

Fault trees are useful for safety-critical systems and complex failure logic.

17. Failure investigation and root cause analysis

Big idea: solving the visible problem is not enough; the cause must be removed.

Root cause analysis finds why a failure happened so the fix prevents recurrence.

Root cause workflow

1. Define the problem clearly.

2. Contain the issue if needed.

3. Collect failed parts, measurements, photos, and process data.

4. Reconstruct what happened.

5. Identify possible causes.

6. Test the most likely causes.

7. Identify the root cause.

8. Choose corrective action.

9. Verify the corrective action works.

10. Update drawings, processes, training, or controls.

5 Whys

5 Whys is a method of asking why repeatedly until the cause chain is clear.

Example:

1. Why did the bracket fail? Because it cracked at the mounting hole.

2. Why did it crack there? Because cyclic stress was high at a sharp corner.

3. Why was there a sharp corner? Because the drawing did not specify a fillet.

4. Why did the drawing omit the fillet? Because fatigue loading was not reviewed.

5. Why was fatigue not reviewed? Because the requirement only listed static load.

A shallow corrective action would replace the broken bracket. A better corrective action would update the requirement, add fatigue review, revise the drawing, and inspect the critical corner.

Fishbone diagram

A fishbone diagram, or Ishikawa diagram, organizes possible causes into categories.

Common categories:

• Machine.

• Method.

• Material.

• Measurement.

• Manpower.

• Environment.

It is useful because teams often jump to one explanation too early.

Corrective and preventive action

Corrective action fixes the existing cause.

Preventive action reduces the chance of similar failures elsewhere.

Examples:

18. Maintainability and serviceability

Big idea: a product should be designed for the people who must inspect, repair, and replace it.

Maintainability is how easily a system can be kept working.

Serviceability is how easily parts can be accessed, removed, repaired, or replaced.

Maintainability design features

Good service design includes:

• Clear access panels.

• Standard tools.

• Replaceable modules.

• Visible labels.

• Safe isolation points.

• Drainage or cleanup access.

• Reasonable fastener access.

• Diagnostic information.

• Replacement procedures.

Serviceability questions

Ask:

• Can the part be replaced without removing unrelated parts?

• Are fasteners visible and reachable?

• Can the technician tell which revision is installed?

• Are sharp edges, stored energy, hot surfaces, or electrical hazards controlled?

• Can the repaired system be tested afterward?

• Does the design prevent incorrect reassembly?

For the bracket, serviceability might mean the sensor can be removed without removing the wall bracket, and the bracket can be replaced with common tools.

19. Revision control, ECOs, and configuration management

Big idea: released designs must be controlled so everyone builds the same thing.

Once a design is released, uncontrolled changes can cause major problems. Manufacturing may build the wrong version, suppliers may use outdated drawings, tests may be performed on unknown hardware, or field units may not match documentation.

Revision control

Revision control records released changes.

A revision history should explain:

• What changed.

• Why it changed.

• Who approved it.

• When it became effective.

• Which parts, drawings, CAD files, BOMs, tools, tests, or field units are affected.

Engineering Change Order

An Engineering Change Order, or ECO, is a formal method for changing a released design.

Typical ECO workflow:

1. Identify the issue or improvement.

2. Define affected parts, drawings, CAD, BOMs, tools, tests, inventory, and field units.

3. Evaluate technical impact.

4. Evaluate cost, schedule, supplier, manufacturing, and service impact.

5. Review and approve.

6. Release revised documentation.

7. Communicate effective date and implementation plan.

8. Keep records for traceability.

Configuration management

Configuration management ensures the correct version of every part, document, software file, tool, and process is used together.

Configuration problems include:

• Drawing and CAD mismatch.

• Wrong revision ordered.

• Supplier using outdated files.

• Prototype configuration not recorded.

• Test result tied to unknown hardware.

• Field unit different from released design.

Part numbers and revisions

A part number identifies a controlled item. A revision identifies the version of that item.

A change to form, fit, or function usually requires formal revision control. Some organizations require a new part number for changes that are not backward compatible.

Release package

A release package may include:

• Requirement specification.

• CAD models.

• Drawings.

• BOM.

• Material specifications.

• Interface documents.

• Analysis reports.

• Test procedures and results.

• Inspection plans.

• Manufacturing instructions.

• ECO and approval records.

The goal is that manufacturing, inspection, suppliers, and service all use the same controlled definition.

20. Integrated design workflow

Big idea: engineering design works when all the pieces are connected.

A good design process does not treat requirements, CAD, tolerances, analysis, manufacturing, and reliability as separate topics. They feed each other.

Full workflow

1. Define the need.

2. Write measurable requirements.

3. Identify interfaces and constraints.

4. Generate concepts.

5. Compare concepts against requirements, risk, cost, and schedule.

6. Select a design direction.

7. Build preliminary CAD.

8. Identify load paths and failure modes.

9. Select materials and manufacturing processes.

10. Analyze stress, deflection, thermal behavior, vibration, and power as needed.

11. Define datums, tolerances, GD&T, fits, and surface finishes.

12. Perform tolerance stack-ups for critical interfaces.

13. Review manufacturability and assembly.

14. Create drawings, BOM, and inspection plans.

15. Build prototypes or first articles.

16. Test and inspect against requirements.

17. Revise the design based on evidence.

18. Release controlled documentation.

19. Manage changes through ECOs.

20. Monitor field performance and feed lessons back into future designs.

Choosing the right tool

Design review questions

Use these questions during design reviews:

• What requirement does this feature satisfy?

• What is the load path?

• What are the credible failure modes?

• What happens at tolerance limits?

• Which dimensions actually control assembly?

• Are datums chosen from functional interfaces?

• Is this tolerance tighter than necessary?

• Can the process make this repeatedly?

• Can inspection measure this reliably?

• Is the material compatible with the environment?

• Can the design be assembled incorrectly?

• Can it be serviced without damaging other parts?

• Does the analysis match real boundary conditions?

• Do CAD, drawings, BOM, test reports, and suppliers use the same revision?

Common design pitfalls

• Writing vague requirements.

• Treating CAD geometry as a complete product definition.

• Adding tight tolerances everywhere.

• Dimensioning from nonfunctional references.

• Choosing datums that do not match assembly function.

• Ignoring tolerance stack-up until parts fail to fit.

• Selecting material based only on strength.

• Forgetting stiffness, corrosion, wear, temperature, or fatigue.

• Using FEA without checking loads, constraints, mesh, and units.

• Reporting factor of safety without identifying the failure mode.

• Designing parts that can be made but not inspected.

• Ignoring maintainability until field repair is difficult.

• Letting drawings, CAD, BOMs, and suppliers drift out of revision sync.

• Fixing symptoms instead of root causes.

Pre-release sanity checks

Before release, check:

• Requirements are measurable.

• Acceptance criteria are defined.

• Critical interfaces are controlled.

• CAD and drawings agree.

• BOM matches the assembly.

• Materials and finishes are specified.

• Critical dimensions have tolerances.

• Datums match functional interfaces.

• Stack-ups pass critical fit requirements.

• Load paths are understood.

• Safety factors and margins are positive.

• Thermal, vibration, and electrical limits are addressed where relevant.

• Manufacturing process can hold required tolerances.

• Inspection method is capable.

• Failure modes have been reviewed.

• Service access is acceptable.

• Revision and configuration records are controlled.

21. Formula summary

Requirements and scoring

Tolerances and fits

Tolerance stack-up

Mechanics

Safety and margin

Materials and thermal expansion

Heat transfer

Vibration

Electrical

Quality and reliability

Compact intuition

Engineering design is controlled problem-solving.

The design starts with a need, becomes requirements, turns into geometry and interfaces, gets checked by analysis and testing, becomes manufacturable through tolerances and process choices, becomes trustworthy through inspection and reliability work, and stays controlled through revision management.

A complete engineering design answers:

1. What must it do?

2. What exactly is being built?

3. Will it work under real loads and environments?

4. Can it be manufactured and inspected repeatedly?

5. What can fail, and how is that risk controlled?

6. Can it be maintained?

7. How are future changes controlled?

If those questions are answered clearly, the design is much more likely to survive manufacturing, testing, service, and revision.

Sources

• Engineering LibreTexts

• Hibbeler, Engineering Mechanics

• Shigley et al., Mechanical Engineering Design

• Leake and Borgerson, Engineering Design Graphics

• Groover, Fundamentals of Modern Manufacturing

• Callister and Rethwisch, Materials Science and Engineering

• Montgomery, Introduction to Statistical Quality Control

• Oppenheim and Willsky, Signals and Systems

• Nise, Control Systems Engineering

• Incropera et al., Fundamentals of Heat and Mass Transfer

• Nilsson and Riedel, Electric Circuits

• Kerzner, Project Management: A Systems Approach to Planning, Scheduling, and Controlling

• Law, Simulation Modeling and Analysis

• Fraden, Handbook of Modern Sensors

• Parell GitHub repository