Plasticity, Uncertainty and Structural Model Verification

Conceptual overview

A linear elastic model answers a narrower question than a performance or collapse assessment. Plastic hinges can redistribute bending after local yielding, but the mechanism is highly dependent on cross-section class, lateral restraint and ductility. Stiffness degradation and connection slippage also change predicted deformation. Quantifying uncertainty adds a second layer: rather than treating a single E or load as perfectly known, one can explore response sensitivity to plausible parameter variation. Yet Monte Carlo samples drawn from invented distributions do not constitute a measured reliability model. A good computational report therefore includes assumptions, governing equations, units, verification benchmarks, reproducible inputs, mesh checks and a clear limitation statement.

Learning outcomes

After working through this unit you should be able to: (1) construct a physically consistent idealization, (2) express the relevant force, stiffness and displacement variables using compatible units, (3) solve a demonstrative problem by hand or a transparent numerical method, (4) interpret response patterns and their sensitivity to input parameters, and (5) identify what the simplified model cannot prove. In this unit the most important observation is not merely a numerical answer but the connection between structural behavior and the assumed support or constitutive conditions.

Unit engineering case — Model-audit exercise

A contractor submits a continuous-beam analysis with a smooth deflected-shape plot and a single stated maximum moment. Demand the stiffness matrix, support assumptions, input load cases and a reaction summary. Compare an interior deflection against the exact uniform-load beam formula, and run mesh refinement at a fixed nonnodal coordinate. Add a deterministic parameter-uncertainty plot while warning that arbitrary probability distributions cannot establish failure risk. Finally document which results have passed verification and which assumptions remain to be validated physically.

Investigation sequence

Begin by preparing a diagram that names every member, support, load, imposed movement and unknown. Choose an independent coordinate system and assign a sign to every generalized force. Derive a minimal reference calculation before using a computational model. Then run a sensitivity study in which only one physical assumption changes at a time. Record both the predicted movement and the reaction or internal-force counterpart. Check conservation of force and moment where the physical model supports those checks. Finally compare the calculation with a plausible measurement or with a more complete independent model. Label the evidence as a verification benchmark, an assumption or a genuinely observed fact.

Suggested written report

A useful report has a one-paragraph engineering question, a table of parameter values with units, an explicit statement of boundary conditions, the governing equations, a readable figure of the system, a table of numerical results, at least one independent verification, and a conclusion that identifies unverified assumptions. State what happens when the governing input doubles, when stiffness increases, and when a support is relaxed or stiffened. Explain whether the observed changes follow from algebraic scaling or from a change in load path. Keep the original units in the file even if a graph displays a converted unit.

8.1 Plastic hinge and mechanism concepts

Learning target and physical interpretation

A fully plastic section may carry a plastic moment capacity Mp if sufficient ductility and stability are available. A collapse mechanism forms when enough hinges develop to turn the ideal structure into a kinematic mechanism.

Governing relation

For an ideal fixed-fixed beam with central point load, a symmetric three-hinge mechanism gives Pcollapse=8Mp/L by virtual work.

Worked engineering interpretation

For Mp=200 kN·m and L=8 m, the ideal fixed-fixed central-load mechanism reaches 200 kN, whereas a simple-span central hinge mechanism reaches 100 kN.

What the idealization omits

The rigid-perfectly-plastic result does not contain a load–deflection curve: a plotted straight line to collapse is only illustrative.

Verification and interpretation of results

Write the virtual displacement and hinge rotation pattern, then confirm external work equals internal plastic dissipation.

Applied case and decision boundary

Use Lab 13 to compare ideal collapse mechanisms under different boundary conditions. Before drawing a design conclusion, identify which operating condition governs and which variation in span, load, support restraint or stiffness most affects the answer. If the calculation is to be used beyond teaching, a competent engineer must apply the relevant local code, material specification, site conditions, partial factors and independent review. This book's interactive demonstrations do not implement those obligations. Use the result to formulate better questions about the structure, not to replace a complete engineering assessment.

Practice and reflection

1. Derivation: Starting from the stated model, identify the equilibrium equation and the additional compatibility or constitutive equation required. Rewrite the derivation with explicit coordinate signs.

2. Numerical perturbation: Repeat the worked interpretation after increasing the principal load parameter by 10%, keeping the other inputs fixed. Predict the response before calculating it and state whether proportionality is theoretically expected.

3. Boundary-condition test: Replace one ideal support or restraint with a more flexible or more restrictive condition. Explain qualitatively which actions and movements are likely to change and what additional data are required.

4. Error diagnosis: Construct a plausible incorrect solution based on the warning in this lesson: The rigid-perfectly-plastic result does not contain a load–deflection curve: a plotted straight line to collapse is only illustrative. Explain which physical check exposes the error.

5. Professional judgment: Using the case Use Lab 13 to compare ideal collapse mechanisms under different boundary conditions., identify one observable site measurement that would be needed to validate the idealization and one design-code question the classroom model cannot answer.

Model-check prompt: State in one sentence what would falsify the predicted response. State separately what numerical residual could be tested without any field data.

8.2 Sensitivity and parameter uncertainty

Learning target and physical interpretation

Calculated responses are conditional on selected loads, support stiffness and material properties. Parameter sensitivity identifies which inputs most affect deformation and which are relatively benign.

Governing relation

For a simple-span beam with central force, δ=PL³/(48EI); logarithmic sensitivities are +1 to P, +3 to L, −1 to E and −1 to I.

Worked engineering interpretation

A 1% span increase changes the leading-order elastic deflection by approximately 3% when all other variables are held fixed.

What the idealization omits

Sampling an elastic formula with arbitrary normal distributions does not provide a calibrated failure probability, confidence interval or design reliability index.

Verification and interpretation of results

Compare the first-order sensitivity estimate against an explicit +5% finite difference for span and stiffness.

Applied case and decision boundary

Use Lab 14 to distinguish numerical spread from real measured population variability. Before drawing a design conclusion, identify which operating condition governs and which variation in span, load, support restraint or stiffness most affects the answer. If the calculation is to be used beyond teaching, a competent engineer must apply the relevant local code, material specification, site conditions, partial factors and independent review. This book's interactive demonstrations do not implement those obligations. Use the result to formulate better questions about the structure, not to replace a complete engineering assessment.

Practice and reflection

1. Derivation: Starting from the stated model, identify the equilibrium equation and the additional compatibility or constitutive equation required. Rewrite the derivation with explicit coordinate signs.

2. Numerical perturbation: Repeat the worked interpretation after increasing the principal load parameter by 10%, keeping the other inputs fixed. Predict the response before calculating it and state whether proportionality is theoretically expected.

3. Boundary-condition test: Replace one ideal support or restraint with a more flexible or more restrictive condition. Explain qualitatively which actions and movements are likely to change and what additional data are required.

4. Error diagnosis: Construct a plausible incorrect solution based on the warning in this lesson: Sampling an elastic formula with arbitrary normal distributions does not provide a calibrated failure probability, confidence interval or design reliability index. Explain which physical check exposes the error.

5. Professional judgment: Using the case Use Lab 14 to distinguish numerical spread from real measured population variability., identify one observable site measurement that would be needed to validate the idealization and one design-code question the classroom model cannot answer.

Model-check prompt: State in one sentence what would falsify the predicted response. State separately what numerical residual could be tested without any field data.

8.3 Beam element interpolation and convergence

Learning target and physical interpretation

Finite-element convergence is a property of a sequence of models and quantities of interest. Beam elements with cubic displacement interpolation may reproduce particular nodal values accurately but still miss interior curvature or displacement.

Governing relation

Hermite beam field v(x)=N1v_i+N2θ_i+N3v_j+N4θ_j; exact uniform-load deflection for simple span is v(x)=−w x(L³−2Lx²+x³)/(24EI).

Worked engineering interpretation

At x=0.37L, an interior coordinate that does not always coincide with a node, varying the mesh reveals interpolation error that would be hidden by examining only a nodal midspan value.

What the idealization omits

A converged displacement is not proof of converged extreme shear or stress, which may exhibit stronger sensitivity to loading representation or discontinuities.

Verification and interpretation of results

Plot absolute error against element count and compare the numerical field with a closed-form reference at the same physical coordinate.

Applied case and decision boundary

Use Lab 15 as a regression test when changing a beam-element calculation engine. Before drawing a design conclusion, identify which operating condition governs and which variation in span, load, support restraint or stiffness most affects the answer. If the calculation is to be used beyond teaching, a competent engineer must apply the relevant local code, material specification, site conditions, partial factors and independent review. This book's interactive demonstrations do not implement those obligations. Use the result to formulate better questions about the structure, not to replace a complete engineering assessment.

Practice and reflection

1. Derivation: Starting from the stated model, identify the equilibrium equation and the additional compatibility or constitutive equation required. Rewrite the derivation with explicit coordinate signs.

2. Numerical perturbation: Repeat the worked interpretation after increasing the principal load parameter by 10%, keeping the other inputs fixed. Predict the response before calculating it and state whether proportionality is theoretically expected.

3. Boundary-condition test: Replace one ideal support or restraint with a more flexible or more restrictive condition. Explain qualitatively which actions and movements are likely to change and what additional data are required.

4. Error diagnosis: Construct a plausible incorrect solution based on the warning in this lesson: A converged displacement is not proof of converged extreme shear or stress, which may exhibit stronger sensitivity to loading representation or discontinuities. Explain which physical check exposes the error.

5. Professional judgment: Using the case Use Lab 15 as a regression test when changing a beam-element calculation engine., identify one observable site measurement that would be needed to validate the idealization and one design-code question the classroom model cannot answer.

Model-check prompt: State in one sentence what would falsify the predicted response. State separately what numerical residual could be tested without any field data.

8.4 Verification versus validation

Learning target and physical interpretation

Verification asks whether the equations are solved correctly. Validation asks whether the selected equations and assumptions describe the physical system sufficiently for the intended decision. Both are needed.

Governing relation

A useful error metric is ε_rel=|Q_num−Q_ref|/max(|Q_ref|,Q_floor), with a declared floor to avoid division by near-zero reference values.

Worked engineering interpretation

A numerical model can pass a reaction-balance check perfectly while still be invalid for a deep beam if shear deformation is neglected.

What the idealization omits

Reporting a visually smooth graph as evidence of physical correctness confuses graphical polish with conservation laws and governing mechanics.

Verification and interpretation of results

Use manufactured solutions, independent benchmarks, limiting cases, unit checks and browser interaction tests as separate items in the audit trail.

Applied case and decision boundary

Create a model-review sheet comparing a hand calculation, an independent solver and experimental observations. Before drawing a design conclusion, identify which operating condition governs and which variation in span, load, support restraint or stiffness most affects the answer. If the calculation is to be used beyond teaching, a competent engineer must apply the relevant local code, material specification, site conditions, partial factors and independent review. This book's interactive demonstrations do not implement those obligations. Use the result to formulate better questions about the structure, not to replace a complete engineering assessment.

Practice and reflection

1. Derivation: Starting from the stated model, identify the equilibrium equation and the additional compatibility or constitutive equation required. Rewrite the derivation with explicit coordinate signs.

2. Numerical perturbation: Repeat the worked interpretation after increasing the principal load parameter by 10%, keeping the other inputs fixed. Predict the response before calculating it and state whether proportionality is theoretically expected.

3. Boundary-condition test: Replace one ideal support or restraint with a more flexible or more restrictive condition. Explain qualitatively which actions and movements are likely to change and what additional data are required.

4. Error diagnosis: Construct a plausible incorrect solution based on the warning in this lesson: Reporting a visually smooth graph as evidence of physical correctness confuses graphical polish with conservation laws and governing mechanics. Explain which physical check exposes the error.

5. Professional judgment: Using the case Create a model-review sheet comparing a hand calculation, an independent solver and experimental observations., identify one observable site measurement that would be needed to validate the idealization and one design-code question the classroom model cannot answer.

Model-check prompt: State in one sentence what would falsify the predicted response. State separately what numerical residual could be tested without any field data.

8.5 Ethical reporting and engineering decisions

Learning target and physical interpretation

Engineering reports should separate educational demonstrations from design deliverables. Inputs, output units, coordinate conventions, load factors, code versions and unresolved limitations belong alongside figures.

Governing relation

Factor-of-safety or utilization ratios are meaningful only when the load and resistance bases are compatible and the applicable design methodology is explicit.

Worked engineering interpretation

A 10 mm educational deflection model may teach mechanics, but does not establish compliance with serviceability limits unless span limits, support conditions, load combinations and code clauses are checked.

What the idealization omits

Presenting an unvalidated lab as a certified design engine can expose users to unsafe decisions despite correct algebra in selected demonstrations.

Verification and interpretation of results

List what must be independently checked before a model could inform real structural assessment, including site measurements and licensed engineering review.

Applied case and decision boundary

Compose a concise peer-review request that includes model files, assumptions, test evidence and open issues. Before drawing a design conclusion, identify which operating condition governs and which variation in span, load, support restraint or stiffness most affects the answer. If the calculation is to be used beyond teaching, a competent engineer must apply the relevant local code, material specification, site conditions, partial factors and independent review. This book's interactive demonstrations do not implement those obligations. Use the result to formulate better questions about the structure, not to replace a complete engineering assessment.

Practice and reflection

1. Derivation: Starting from the stated model, identify the equilibrium equation and the additional compatibility or constitutive equation required. Rewrite the derivation with explicit coordinate signs.

2. Numerical perturbation: Repeat the worked interpretation after increasing the principal load parameter by 10%, keeping the other inputs fixed. Predict the response before calculating it and state whether proportionality is theoretically expected.

3. Boundary-condition test: Replace one ideal support or restraint with a more flexible or more restrictive condition. Explain qualitatively which actions and movements are likely to change and what additional data are required.

4. Error diagnosis: Construct a plausible incorrect solution based on the warning in this lesson: Presenting an unvalidated lab as a certified design engine can expose users to unsafe decisions despite correct algebra in selected demonstrations. Explain which physical check exposes the error.

5. Professional judgment: Using the case Compose a concise peer-review request that includes model files, assumptions, test evidence and open issues., identify one observable site measurement that would be needed to validate the idealization and one design-code question the classroom model cannot answer.

Model-check prompt: State in one sentence what would falsify the predicted response. State separately what numerical residual could be tested without any field data.

End-of-unit self-check

1. Which unknown in the five lesson models is governed principally by equilibrium, and which requires a deformation condition? Explain for two different examples.
2. Identify a boundary-condition assumption in this unit that could change if a real bearing loses contact or a joint begins to slip.
3. Form a dimensionless or unit-consistent ratio from a model response and a characteristic load or span, then explain how that ratio changes with structural stiffness.
4. Explain how a model could satisfy overall force balance and nevertheless predict the wrong deformation field.
5. Design a two-point numerical verification plan that does not reuse the same solver for both checks.
6. Identify the limitation of treating code design or material yielding as if it were included in these ideal elastic analyses.

Short answer guidance: A complete answer should use a named force path or deformation coordinate rather than generic phrases. A verification answer should specify the numerical quantity being compared and the expected reference; a validation answer should identify a relevant observation or tested physical condition.

Unit 8 — Plasticity, uncertainty and structural model verification

Conceptual overview

A linear elastic model answers a narrower question than a performance or collapse assessment. Plastic hinges can redistribute bending after local yielding, but the mechanism is highly dependent on cross-section class, lateral restraint and ductility. Stiffness degradation and connection slippage also change predicted deformation. Quantifying uncertainty adds a second layer: rather than treating a single E or load as perfectly known, one can explore response sensitivity to plausible parameter variation. Yet Monte Carlo samples drawn from invented distributions do not constitute a measured reliability model. A good computational report therefore includes assumptions, governing equations, units, verification benchmarks, reproducible inputs, mesh checks and a clear limitation statement.

Learning outcomes

After working through this unit you should be able to: (1) construct a physically consistent idealization, (2) express the relevant force, stiffness and displacement variables using compatible units, (3) solve a demonstrative problem by hand or a transparent numerical method, (4) interpret response patterns and their sensitivity to input parameters, and (5) identify what the simplified model cannot prove. In this unit the most important observation is not merely a numerical answer but the connection between structural behavior and the assumed support or constitutive conditions.

Unit engineering case — Model-audit exercise

A contractor submits a continuous-beam analysis with a smooth deflected-shape plot and a single stated maximum moment. Demand the stiffness matrix, support assumptions, input load cases and a reaction summary. Compare an interior deflection against the exact uniform-load beam formula, and run mesh refinement at a fixed nonnodal coordinate. Add a deterministic parameter-uncertainty plot while warning that arbitrary probability distributions cannot establish failure risk. Finally document which results have passed verification and which assumptions remain to be validated physically.

Investigation sequence

Begin by preparing a diagram that names every member, support, load, imposed movement and unknown. Choose an independent coordinate system and assign a sign to every generalized force. Derive a minimal reference calculation before using a computational model. Then run a sensitivity study in which only one physical assumption changes at a time. Record both the predicted movement and the reaction or internal-force counterpart. Check conservation of force and moment where the physical model supports those checks. Finally compare the calculation with a plausible measurement or with a more complete independent model. Label the evidence as a verification benchmark, an assumption or a genuinely observed fact.

Suggested written report

A useful report has a one-paragraph engineering question, a table of parameter values with units, an explicit statement of boundary conditions, the governing equations, a readable figure of the system, a table of numerical results, at least one independent verification, and a conclusion that identifies unverified assumptions. State what happens when the governing input doubles, when stiffness increases, and when a support is relaxed or stiffened. Explain whether the observed changes follow from algebraic scaling or from a change in load path. Keep the original units in the file even if a graph displays a converted unit.

8.1 Plastic hinge and mechanism concepts

Learning target and physical interpretation

A fully plastic section may carry a plastic moment capacity Mp if sufficient ductility and stability are available. A collapse mechanism forms when enough hinges develop to turn the ideal structure into a kinematic mechanism.

Governing relation

For an ideal fixed-fixed beam with central point load, a symmetric three-hinge mechanism gives Pcollapse=8Mp/L by virtual work.

Worked engineering interpretation

For Mp=200 kN·m and L=8 m, the ideal fixed-fixed central-load mechanism reaches 200 kN, whereas a simple-span central hinge mechanism reaches 100 kN.

What the idealization omits

The rigid-perfectly-plastic result does not contain a load–deflection curve: a plotted straight line to collapse is only illustrative.

Verification and interpretation of results

Write the virtual displacement and hinge rotation pattern, then confirm external work equals internal plastic dissipation.

Applied case and decision boundary

Use Lab 13 to compare ideal collapse mechanisms under different boundary conditions. Before drawing a design conclusion, identify which operating condition governs and which variation in span, load, support restraint or stiffness most affects the answer. If the calculation is to be used beyond teaching, a competent engineer must apply the relevant local code, material specification, site conditions, partial factors and independent review. This book's interactive demonstrations do not implement those obligations. Use the result to formulate better questions about the structure, not to replace a complete engineering assessment.

Practice and reflection

1. Derivation: Starting from the stated model, identify the equilibrium equation and the additional compatibility or constitutive equation required. Rewrite the derivation with explicit coordinate signs.

2. Numerical perturbation: Repeat the worked interpretation after increasing the principal load parameter by 10%, keeping the other inputs fixed. Predict the response before calculating it and state whether proportionality is theoretically expected.

3. Boundary-condition test: Replace one ideal support or restraint with a more flexible or more restrictive condition. Explain qualitatively which actions and movements are likely to change and what additional data are required.

4. Error diagnosis: Construct a plausible incorrect solution based on the warning in this lesson: The rigid-perfectly-plastic result does not contain a load–deflection curve: a plotted straight line to collapse is only illustrative. Explain which physical check exposes the error.

5. Professional judgment: Using the case Use Lab 13 to compare ideal collapse mechanisms under different boundary conditions., identify one observable site measurement that would be needed to validate the idealization and one design-code question the classroom model cannot answer.

Model-check prompt: State in one sentence what would falsify the predicted response. State separately what numerical residual could be tested without any field data.

8.2 Sensitivity and parameter uncertainty

Learning target and physical interpretation

Calculated responses are conditional on selected loads, support stiffness and material properties. Parameter sensitivity identifies which inputs most affect deformation and which are relatively benign.

Governing relation

For a simple-span beam with central force, δ=PL³/(48EI); logarithmic sensitivities are +1 to P, +3 to L, −1 to E and −1 to I.

Worked engineering interpretation

A 1% span increase changes the leading-order elastic deflection by approximately 3% when all other variables are held fixed.

What the idealization omits

Sampling an elastic formula with arbitrary normal distributions does not provide a calibrated failure probability, confidence interval or design reliability index.

Verification and interpretation of results

Compare the first-order sensitivity estimate against an explicit +5% finite difference for span and stiffness.

Applied case and decision boundary

Use Lab 14 to distinguish numerical spread from real measured population variability. Before drawing a design conclusion, identify which operating condition governs and which variation in span, load, support restraint or stiffness most affects the answer. If the calculation is to be used beyond teaching, a competent engineer must apply the relevant local code, material specification, site conditions, partial factors and independent review. This book's interactive demonstrations do not implement those obligations. Use the result to formulate better questions about the structure, not to replace a complete engineering assessment.

Practice and reflection

1. Derivation: Starting from the stated model, identify the equilibrium equation and the additional compatibility or constitutive equation required. Rewrite the derivation with explicit coordinate signs.

2. Numerical perturbation: Repeat the worked interpretation after increasing the principal load parameter by 10%, keeping the other inputs fixed. Predict the response before calculating it and state whether proportionality is theoretically expected.

3. Boundary-condition test: Replace one ideal support or restraint with a more flexible or more restrictive condition. Explain qualitatively which actions and movements are likely to change and what additional data are required.

4. Error diagnosis: Construct a plausible incorrect solution based on the warning in this lesson: Sampling an elastic formula with arbitrary normal distributions does not provide a calibrated failure probability, confidence interval or design reliability index. Explain which physical check exposes the error.

5. Professional judgment: Using the case Use Lab 14 to distinguish numerical spread from real measured population variability., identify one observable site measurement that would be needed to validate the idealization and one design-code question the classroom model cannot answer.

Model-check prompt: State in one sentence what would falsify the predicted response. State separately what numerical residual could be tested without any field data.

8.3 Beam element interpolation and convergence

Learning target and physical interpretation

Finite-element convergence is a property of a sequence of models and quantities of interest. Beam elements with cubic displacement interpolation may reproduce particular nodal values accurately but still miss interior curvature or displacement.

Governing relation

Hermite beam field v(x)=N1v_i+N2θ_i+N3v_j+N4θ_j; exact uniform-load deflection for simple span is v(x)=−w x(L³−2Lx²+x³)/(24EI).

Worked engineering interpretation

At x=0.37L, an interior coordinate that does not always coincide with a node, varying the mesh reveals interpolation error that would be hidden by examining only a nodal midspan value.

What the idealization omits

A converged displacement is not proof of converged extreme shear or stress, which may exhibit stronger sensitivity to loading representation or discontinuities.

Verification and interpretation of results

Plot absolute error against element count and compare the numerical field with a closed-form reference at the same physical coordinate.

Applied case and decision boundary

Use Lab 15 as a regression test when changing a beam-element calculation engine. Before drawing a design conclusion, identify which operating condition governs and which variation in span, load, support restraint or stiffness most affects the answer. If the calculation is to be used beyond teaching, a competent engineer must apply the relevant local code, material specification, site conditions, partial factors and independent review. This book's interactive demonstrations do not implement those obligations. Use the result to formulate better questions about the structure, not to replace a complete engineering assessment.

Practice and reflection

1. Derivation: Starting from the stated model, identify the equilibrium equation and the additional compatibility or constitutive equation required. Rewrite the derivation with explicit coordinate signs.

2. Numerical perturbation: Repeat the worked interpretation after increasing the principal load parameter by 10%, keeping the other inputs fixed. Predict the response before calculating it and state whether proportionality is theoretically expected.

3. Boundary-condition test: Replace one ideal support or restraint with a more flexible or more restrictive condition. Explain qualitatively which actions and movements are likely to change and what additional data are required.

4. Error diagnosis: Construct a plausible incorrect solution based on the warning in this lesson: A converged displacement is not proof of converged extreme shear or stress, which may exhibit stronger sensitivity to loading representation or discontinuities. Explain which physical check exposes the error.

5. Professional judgment: Using the case Use Lab 15 as a regression test when changing a beam-element calculation engine., identify one observable site measurement that would be needed to validate the idealization and one design-code question the classroom model cannot answer.

Model-check prompt: State in one sentence what would falsify the predicted response. State separately what numerical residual could be tested without any field data.

8.4 Verification versus validation

Learning target and physical interpretation

Verification asks whether the equations are solved correctly. Validation asks whether the selected equations and assumptions describe the physical system sufficiently for the intended decision. Both are needed.

Governing relation

A useful error metric is ε_rel=|Q_num−Q_ref|/max(|Q_ref|,Q_floor), with a declared floor to avoid division by near-zero reference values.

Worked engineering interpretation

A numerical model can pass a reaction-balance check perfectly while still be invalid for a deep beam if shear deformation is neglected.

What the idealization omits

Reporting a visually smooth graph as evidence of physical correctness confuses graphical polish with conservation laws and governing mechanics.

Verification and interpretation of results

Use manufactured solutions, independent benchmarks, limiting cases, unit checks and browser interaction tests as separate items in the audit trail.

Applied case and decision boundary

Create a model-review sheet comparing a hand calculation, an independent solver and experimental observations. Before drawing a design conclusion, identify which operating condition governs and which variation in span, load, support restraint or stiffness most affects the answer. If the calculation is to be used beyond teaching, a competent engineer must apply the relevant local code, material specification, site conditions, partial factors and independent review. This book's interactive demonstrations do not implement those obligations. Use the result to formulate better questions about the structure, not to replace a complete engineering assessment.

Practice and reflection

1. Derivation: Starting from the stated model, identify the equilibrium equation and the additional compatibility or constitutive equation required. Rewrite the derivation with explicit coordinate signs.

2. Numerical perturbation: Repeat the worked interpretation after increasing the principal load parameter by 10%, keeping the other inputs fixed. Predict the response before calculating it and state whether proportionality is theoretically expected.

3. Boundary-condition test: Replace one ideal support or restraint with a more flexible or more restrictive condition. Explain qualitatively which actions and movements are likely to change and what additional data are required.

4. Error diagnosis: Construct a plausible incorrect solution based on the warning in this lesson: Reporting a visually smooth graph as evidence of physical correctness confuses graphical polish with conservation laws and governing mechanics. Explain which physical check exposes the error.

5. Professional judgment: Using the case Create a model-review sheet comparing a hand calculation, an independent solver and experimental observations., identify one observable site measurement that would be needed to validate the idealization and one design-code question the classroom model cannot answer.

Model-check prompt: State in one sentence what would falsify the predicted response. State separately what numerical residual could be tested without any field data.

8.5 Ethical reporting and engineering decisions

Learning target and physical interpretation

Engineering reports should separate educational demonstrations from design deliverables. Inputs, output units, coordinate conventions, load factors, code versions and unresolved limitations belong alongside figures.

Governing relation

Factor-of-safety or utilization ratios are meaningful only when the load and resistance bases are compatible and the applicable design methodology is explicit.

Worked engineering interpretation

A 10 mm educational deflection model may teach mechanics, but does not establish compliance with serviceability limits unless span limits, support conditions, load combinations and code clauses are checked.

What the idealization omits

Presenting an unvalidated lab as a certified design engine can expose users to unsafe decisions despite correct algebra in selected demonstrations.

Verification and interpretation of results

List what must be independently checked before a model could inform real structural assessment, including site measurements and licensed engineering review.

Applied case and decision boundary

Compose a concise peer-review request that includes model files, assumptions, test evidence and open issues. Before drawing a design conclusion, identify which operating condition governs and which variation in span, load, support restraint or stiffness most affects the answer. If the calculation is to be used beyond teaching, a competent engineer must apply the relevant local code, material specification, site conditions, partial factors and independent review. This book's interactive demonstrations do not implement those obligations. Use the result to formulate better questions about the structure, not to replace a complete engineering assessment.

Practice and reflection

1. Derivation: Starting from the stated model, identify the equilibrium equation and the additional compatibility or constitutive equation required. Rewrite the derivation with explicit coordinate signs.

2. Numerical perturbation: Repeat the worked interpretation after increasing the principal load parameter by 10%, keeping the other inputs fixed. Predict the response before calculating it and state whether proportionality is theoretically expected.

3. Boundary-condition test: Replace one ideal support or restraint with a more flexible or more restrictive condition. Explain qualitatively which actions and movements are likely to change and what additional data are required.

4. Error diagnosis: Construct a plausible incorrect solution based on the warning in this lesson: Presenting an unvalidated lab as a certified design engine can expose users to unsafe decisions despite correct algebra in selected demonstrations. Explain which physical check exposes the error.

5. Professional judgment: Using the case Compose a concise peer-review request that includes model files, assumptions, test evidence and open issues., identify one observable site measurement that would be needed to validate the idealization and one design-code question the classroom model cannot answer.

Model-check prompt: State in one sentence what would falsify the predicted response. State separately what numerical residual could be tested without any field data.

End-of-unit self-check

1. Which unknown in the five lesson models is governed principally by equilibrium, and which requires a deformation condition? Explain for two different examples.
2. Identify a boundary-condition assumption in this unit that could change if a real bearing loses contact or a joint begins to slip.
3. Form a dimensionless or unit-consistent ratio from a model response and a characteristic load or span, then explain how that ratio changes with structural stiffness.
4. Explain how a model could satisfy overall force balance and nevertheless predict the wrong deformation field.
5. Design a two-point numerical verification plan that does not reuse the same solver for both checks.
6. Identify the limitation of treating code design or material yielding as if it were included in these ideal elastic analyses.

Short answer guidance: A complete answer should use a named force path or deformation coordinate rather than generic phrases. A verification answer should specify the numerical quantity being compared and the expected reference; a validation answer should identify a relevant observation or tested physical condition.

Appendix A — Advanced readings

Advanced reading — Structural Analysis II

These original technical essays extend the eight units. They distinguish underlying mathematics from the limitations of the ideal numerical laboratories.

Extension 1 — Flexibility matrix energy and reciprocity

For multiple redundants, a force-method model is most transparent when generalized coordinates are defined before releasing a support. Consider two unknown reactions X1 and X2, paired with two displacements Δ1 and Δ2. Apply one unit reaction at coordinate 1 to the released primary structure and measure movement at both coordinates. That gives the first column of the flexibility matrix. Apply a unit reaction at coordinate 2 to obtain the second column. For conservative, linear elastic behavior with well-matched force and displacement coordinates, Maxwell–Betti reciprocity predicts f12=f21. The matrix is often symmetric positive definite for a stable set of independent coordinates, because the internal strain energy under nonzero generalized loading must be positive. If it is nearly singular, one redundant may not be independent or the primary structure may be nearly a mechanism.

Physical compatibility requires the displacement induced by real loads plus the displacement due to each redundant to equal the actual constrained movement. This rule distinguishes settlement from an external applied load. A spring support imposes neither zero displacement nor an arbitrary known force; it contributes a reaction proportional to its own movement. A very stiff spring approaches rigid-restraint behavior only in a limiting sense. If the spring has a gap or a one-sided bearing, the force law is piecewise and a conventional symmetric linear formulation may no longer apply over all loading cases.

An independent energy check can use U=(1/2)XᵀfX. Under consistent units, this energy must be nonnegative for an elastic stable primary structure. If one redundant is a moment and another is a force, the matrix entries carry different physical units, and changing coordinate scales may improve numerical conditioning. However, rescaling cannot repair a mistaken structural release. Another useful observation is that the response to twice a load is twice the response only while material and support conditions remain linear. No amount of matrix algebra can justify extending a constant-stiffness prediction through contact loss or yielding.

In the parallel-spring laboratory, compatibility is simply equality of movement. For the propped cantilever, compatibility is zero final tip movement. These appear to be different calculations because one uses a spring law and the other a beam-flexibility expression, but both are instances of the same force–displacement relationship. An engineering report should draw that conceptual link and state which physical displacement is actually being restrained. A clean numerical reaction is not enough unless the assumed bearing movement is plausible.

Extension 2 — Deriving beam stiffness from energy

A two-node beam element can be developed from a displacement interpolation rather than memorizing the 4×4 matrix. Let the local coordinate ξ=x/L run from zero at the left node to one at the right node. The transverse displacement is represented by the cubic Hermite field v=N1v1+N2θ1+N3v2+N4θ2, where N1=1−3ξ²+2ξ³, N2=L(ξ−2ξ²+ξ³), N3=3ξ²−2ξ³ and N4=L(−ξ²+ξ³). These functions ensure that the prescribed nodal translations and slopes are reproduced exactly. Differentiating the field twice gives curvature, and the Euler–Bernoulli bending energy is U=(1/2)∫EI(vʺ)²dx. Differentiation of U with respect to nodal displacements produces the beam stiffness matrix.

The origin of the terms 12EI/L³, 6EI/L², 4EI/L and 2EI/L is consequently mechanical rather than arbitrary. Translational stiffness scales as L⁻³, coupling translation to rotation scales as L⁻², and purely rotational stiffness scales as L⁻¹. This explains why a long beam may be considerably more flexible in transverse displacement even when its local rotational properties change more modestly. A uniform distributed load must be integrated against the same shape functions to form consistent nodal actions. Using only half the total load at each end reproduces total vertical force but misses the generalized moments and can therefore change the solution.

The exact deflection of a simply supported beam under a uniform load is quartic in x. A single cubic element cannot reproduce that exact interior field for arbitrary x, even if some nodal quantities are recovered accurately with a consistent load vector. The distinction between nodal accuracy and interior-field accuracy explains why refinement matters. In a continuous beam, adjacent elements share the interior rotation; the common joint rotation enforces displacement continuity and no relative kink at an ideal rigid connection. If the joint is a real hinge, the rotational continuity constraint must be released at that member end, with care to preserve translation compatibility.

Slope–deflection and moment distribution emerge from the same stiffness relations after vertical displacements are fixed and only rotations remain free. A hand analysis might be faster for one interior joint. A matrix solver is more reliable for many supports, settlements or changing loads, provided DOF indexing, end releases and consistent nodal loads are checked independently. When comparing the two methods, specify which member end moment is positive and draw both clockwise action directions. A result cannot be called wrong solely because one convention reverses the displayed end-moment sign.

Extension 3 — Frame transformation and rigid-body consistency

A frame element is evaluated most naturally in local coordinates aligned with its own axis. Axial extension appears in the local longitudinal displacement components, while transverse bending and end rotation appear in the remaining degrees of freedom. For an element at angle α to the global x-axis, the direction cosines c=cosα and s=sinα map global translations into local axial and transverse components. An appropriate block-diagonal transformation maps the two ends consistently and leaves the scalar in-plane rotation unchanged. The transformed global element stiffness is Tᵀk_local T, not Tk_local Tᵀ when T is defined as global-to-local.

This distinction is easy to test. A horizontal member subjected to a global horizontal relative movement should respond in axial stiffness EA/L. Rotating the same member by ninety degrees makes a global vertical relative movement axial. A global rigid translation of both ends must cause zero internal strain and force. For a frame, a small rigid-body rotation should likewise produce zero internal deformation in a consistent linearized kinematic representation when all nodal translations and rotations follow that rigid-body motion. If the element reports a large internal force under such a test, the transformation or DOF order may be incorrect.

The assembled structure must also be stable after applying supports. A planar frame node has three degrees of freedom; a base fixity removes two translations and its rotation. A pin support removes its translations but not its rotation. A moment release belongs to a member end and is not always equivalent to releasing a whole joint coordinate shared by other members. For example, three beams meeting at one physical pin may still share translations while allowing appropriate individual end rotations. Misrepresenting the releases can introduce spurious stiffness or a mechanism. The resulting matrix singularity is a diagnostic clue, not something to hide by adding enormous stiffness to a random diagonal element.

In the portal-frame laboratory, both fixed bases carry reactions that balance the lateral roof load. Because the roof beam joins both column tops rigidly, it also carries moments. Equal and opposite joint actions should reconcile when every local member force is transformed back to the global system. An analyst who checks only horizontal force equilibrium may miss a wrong sign in a member moment. It is better to test global reaction sums, moment equilibrium, strain-energy positivity, rigid-body modes and one simplified hand case. When all these hold, the solver has stronger evidence of correct implementation, though real boundary fixity and member properties still need validation.

Extension 4 — From settlement to contact and construction staging

A prescribed support settlement creates a special kind of loading: the restraint moves, while the structure is forced to follow at the contact. The associated reaction is not known in advance. After the global stiffness matrix is partitioned into free and constrained coordinates, the imposed displacement vector contributes −K_fc u_c to the free-coordinate right-hand side. This is why simply assigning a vertical point load in place of settlement is a different problem. For a continuous beam, a lowered interior support may induce both positive and negative end moments that change the bending pattern even if applied external loading remains unchanged.

The linear calculation assumes that all prescribed support constraints stay active. A roller or bearing capable of compression only cannot transmit arbitrary tension. If the solver predicts tensile contact pressure at a bearing, the active set of supports may change. Removing that bearing changes both the stiffness matrix and the deformation compatibility conditions. A separate contact analysis would alternate between calculating force under a selected set of active restraints and checking that all reactions obey contact inequalities. A smooth graph from a fixed-support model conceals such a change if the program has no contact logic.

Temperature introduces another compatibility problem. For a segment that expands freely by αΔTL, restraining its actual extension produces an axial force. For two serial segments with fixed end-to-end distance, the same force passes through both members and their free extensions add. Their mechanical extensions, force times axial flexibility L/(EA), sum with the free thermal extensions to meet the prescribed overall displacement. A bar that is warmed uniformly but remains completely free to expand has no axial force in this idealization. This difference is important when explaining why expansion joints and bearings can reduce thermal restraint actions.

Construction sequencing complicates both examples. A beam may be supported at one elevation during casting and at another after removal of temporary props. Some dead load may be carried before continuity is established, whereas later live loads act on the final continuous structure. Likewise, concrete shrinkage and creep depend on age and restraint history. A final geometry alone cannot capture these histories. A staged analysis must document the time each element and support becomes active, the loads applied at that stage, and the material properties appropriate to it. For every calculated moment redistribution, ask whether it comes from imposed movement, a new load path, changed stiffness or altered contact. Without that distinction, the numerical values have little diagnostic meaning.

Extension 5 — Structural stability beyond a single Euler member

The elegant Euler formula describes an ideal linear bifurcation of a perfectly straight prismatic member. In a real frame, load-induced internal forces can alter the global tangent stiffness before any individual member reaches its isolated pinned-pinned critical load. Geometric stiffness depends on axial force distribution and deformation mode. An ideal linearized stability analysis searches for a load multiplier at which det(K_elastic−λK_geometric)=0, with sign convention chosen so compression reduces restoring stiffness. The eigenvector describes the mode that loses incremental stiffness. This is a frame property, not merely a collection of independent column formulas.

A sway frame can exhibit global instability because the beam-column joints move laterally together. Bracing may dramatically increase the lowest critical multiplier by providing a stiff lateral load path, even when it barely changes individual member EI. Conversely, a nominally fixed connection that is flexible in practice can reduce rotational restraint and effective member buckling capacity. Effective-length factors approximate these end restraints under particular ideal conditions; they should not be guessed from a diagram. The inverse-square dependence on effective length means that even modest uncertainty in bracing length can have substantial consequences.

Second-order analysis ordinarily distinguishes local member curvature (P–δ) from larger frame story translation (P–Δ). Both can matter. In an iterative equilibrium solution, the force acting through displaced geometry alters the moment, which changes deformation, which further changes moment. A first-order correction M+PΔ computed once may underestimate the fully converged effect when the applied axial load is near a stability limit. It is useful to carry out increasing-load steps and monitor not only drift but also tangent stiffness and convergence. Loss of numerical convergence is not, by itself, proof of physical collapse, because solver tolerances and constitutive modeling affect it.

The educational buckling lab presents Euler Pcr and a simple amplifier 1/(1−P/Pcr). This is intentionally a sensitivity illustration, not an implementation of geometric stiffness analysis or a certified beam-column strength method. Initial crookedness, residual stress, inelasticity, connection slip and member imperfections all influence real behavior. To turn an ideal stability benchmark into a design assessment requires selecting appropriate analysis procedures and resistance provisions from the governing standards, with independent checking by a qualified engineer.

Extension 6 — Moving loads: response influence and envelopes

A bending-moment diagram shows moment along the structure for one specified loading arrangement. An influence line instead chooses one response quantity—such as left bearing reaction or bending moment at a fixed section—and varies the position of a unit load. This difference matters when examining vehicles crossing a bridge. For a simply supported span of length L, the right reaction under a unit load at coordinate x is x/L. The left reaction is (L−x)/L. A moment influence line for section a is triangular with maximum ordinate a(L−a)/L when the unit load lies at the section. Multiplying an ordinate by a real axle force gives that axle's contribution to the chosen response.

An axle group can be treated as a sum of moving point loads as long as the assumed linear superposition is valid. When the first axle is on the span but the second has not yet entered, the sum must include only the first. The same care is needed when the vehicle exits. For each vehicle position, moment equilibrium determines support reactions; the section moment is then assembled from the loads located to one side of the section. Because the relative axle positions are fixed by the vehicle geometry, extreme moment is not necessarily produced when the heaviest axle alone sits at the section. Searching the vehicle position over the allowed range is a straightforward numerical envelope calculation.

For an indeterminate continuous beam, support moments and reaction influence lines are no longer the simple triangular forms associated with determinate spans. A point force acting inside a beam finite element should be distributed to generalized nodal load coordinates through the beam shape functions, not rounded to the nearest node. Nearest-node allocation creates nonphysical jumps as the axle crosses the mesh. Even with correct consistent loads, the chosen mesh must be refined until the specific response envelope is sufficiently stable for the purpose of the calculation.

Real highway bridge design adds complexities absent from the classroom train: truck and tandem arrangements defined by a bridge code, multiple loaded lanes, load factors, dynamic amplification, wearing surface, lateral deck distribution, fatigue spectra, influence surfaces for two-dimensional systems and potentially nonlinear bearings. The lab's value is that it makes response superposition and equilibrium visible. A proper output report should identify the axle placement yielding the extreme, the fixed section being evaluated, whether each axle is on or off the bridge, and whether the output is a reaction, bending moment or shear. These quantities should not be mixed.

Extension 7 — Modes, frequency response and numerical integration

A two-degree-of-freedom shear building provides a small matrix eigenproblem with interpretable mode shapes. Its mass matrix represents floor inertia; its stiffness matrix reflects resistance to relative story drifts. Solving det(K−ω²M)=0 yields two circular frequencies. The corresponding vectors express relative floor movement, not absolute displacement amplitude. Any nonzero multiple of an eigenvector is the same mode. A customary normalization is to set the roof component to one, although mass normalization is often more useful in modal superposition. For linear undamped systems with suitable mass and stiffness properties, distinct modes are orthogonal under the mass matrix: φ_iᵀMφ_j=0 for i≠j.

The lower frequency commonly represents an in-phase sway-like motion, whereas the upper mode exhibits relative reversal between floors. How much each mode contributes under a particular excitation depends on the excitation direction and modal participation. Higher modes can make a small contribution to roof displacement while still influencing acceleration or story shear. Therefore, a statement that the first mode dominates every response is too broad. A full structural vibration model also needs rotational inertia, torsion and possibly three-dimensional interactions.

For one generalized coordinate, the governing damped motion is m ü+c u̇+ku=p(t). The circular natural frequency is √(k/m), and viscous damping ratio is c/(2√(km)). A harmonic input may be amplified near resonance; in steady state the dimensionless displacement amplification is 1/√((1−r²)²+(2ζr)²), where r is forcing frequency divided by natural frequency. At resonance r=1, this ratio becomes 1/(2ζ) for ζ>0. That result characterizes a linear steady-state solution, while a numerical simulation initiated from rest also contains transient motion.

Newmark average acceleration calculates the next displacement, velocity and acceleration from the previous step while enforcing dynamic equilibrium. It is unconditionally stable for the classical linear system under standard assumptions but not infinitely accurate. If the time increment is too large, the oscillation phase and amplitude may be distorted. The correct numerical experiment is to refine the time step and compare the result for a defined output such as peak displacement, phase delay or energy balance. Browser animation alone is not evidence of convergence. An actual floor-vibration study would additionally require a credible forcing spectrum, damping estimate, comfort criteria and mode identification from physical or calibrated model data.

Extension 8 — Engineering model auditing and epistemic limits

Verification is the process of checking that the mathematical model has been implemented and solved consistently. A stiffness matrix should have the expected symmetry and rigid-body modes before supports are applied. Its restrained solution should reproduce force and moment equilibrium. A small exactly solvable beam or truss can serve as a benchmark. Mesh refinement tests whether the reported quantity approaches a stable value. Browser interaction tests identify broken sliders and graphs but cannot independently establish the mechanical correctness of the solver.

Validation is a different question. A model of a deep shear-sensitive beam might solve Euler–Bernoulli equations perfectly and yet be inappropriate for the physical beam. An ideal fixed column base can strongly underpredict sway if the actual foundation rotates. A selected Young modulus might not represent a cracked reinforced-concrete section, and a flexible diaphragm may not behave as a rigid plane. These are physical-model questions that need measurement, test evidence, design-standard guidance and independent expert judgment, not just smaller numerical residuals.

Uncertainty studies should declare which quantities are random and why their distributions are credible. Varying a nominal load and modulus with arbitrary Gaussian coefficients of variation demonstrates mathematical sensitivity but does not establish a probability of structural failure. Real reliability analysis needs evidence-based tails of distributions, correlations, resistance and demand models, limit-state definitions and decision thresholds. A reproducible random seed makes an educational simulation easier to audit; it does not make its prior assumptions scientifically measured.

An engineering submission needs traceable units, versions of input geometry and software, analysis and load cases, support conditions, material properties, solver tolerances, verification examples, independent peer review, applicable standards and an explicit acceptance criterion. Where a result is unusually sensitive to small changes in material properties or restraints, that sensitivity should be highlighted rather than averaged away. If an educational model is taken outside its tested assumptions, it should be withdrawn from decision-making until the new case is validated. The purpose of these laboratories is to teach structural reasoning and the discipline of checking a model—not to create a shortcut around qualified structural design.

Appendix B — Laboratory procedures

Structural Analysis II — Interactive Laboratory Manual

These fifteen programs are offline teaching models. They are neither design-code checks nor physical test data. Keep a written record of the units, signs, governing idealizations, and predicted trend before changing the controls. In each lab export its numerical curve as CSV and compare with the equations in the associated unit. Numerical outputs are only as meaningful as the structural idealization that generated them. Browser controls adjust live values; the reset button restores the baseline condition.

Lab 01 — Indeterminate axial load sharing

Linked unit: 1. Open [labs/lab-01.html](labs/lab-01.html).

Before interaction. Change the ratio k1/k2 while holding total load constant. Estimate the common displacement before moving the controls.

Expected mechanical insight. Force carried by each branch scales with its stiffness; force fractions must add to 100%.

Independent or limiting-case check. Set equal spring stiffnesses and confirm each carries half the load. A zero-load limit would yield zero deformation.

Report evidence. Record your initial prediction and initial input values. Include one table of input modifications, screenshots or plots of response curves, the relevant balance or compatibility equation, and a short explanation of how the numerical answer changes with the altered parameter. Distinguish the direction of the change from its magnitude. State at least one parameter not modeled and explain how ignoring it could matter in a physical structure.

Reflection questions. (1) Which physical quantity is controlled, and which quantity is calculated? (2) Does the program use a closed-form equation, a small stiffness matrix, an eigenproblem or time integration? (3) Which check can be performed independently by hand? (4) What decision would be unsafe to make solely from this model?

Lab 02 — Propped cantilever compatibility

Linked unit: 1. Open [labs/lab-02.html](labs/lab-02.html).

Before interaction. First predict the direction of prop reaction under a positive downward UDL; then calculate its magnitude.

Expected mechanical insight. The released cantilever deflects downward and the prop enforces zero final tip displacement.

Independent or limiting-case check. Double span length while holding w and EI: the prop force doubles, but the unpropped tip deflection grows sixteenfold.

Report evidence. Record your initial prediction and initial input values. Include one table of input modifications, screenshots or plots of response curves, the relevant balance or compatibility equation, and a short explanation of how the numerical answer changes with the altered parameter. Distinguish the direction of the change from its magnitude. State at least one parameter not modeled and explain how ignoring it could matter in a physical structure.

Reflection questions. (1) Which physical quantity is controlled, and which quantity is calculated? (2) Does the program use a closed-form equation, a small stiffness matrix, an eigenproblem or time integration? (3) Which check can be performed independently by hand? (4) What decision would be unsafe to make solely from this model?

Lab 03 — Two-span continuous beam FEM

Linked unit: 2. Open [labs/lab-03.html](labs/lab-03.html).

Before interaction. Compare equal and unequal spans and then remove the UDL from one span. Track the sign of the interior support moment.

Expected mechanical insight. The rotational DOF at the interior support remains free but is shared by the adjacent beam elements.

Independent or limiting-case check. Check support reactions sum to the combined UDL. Interior end moments at the support should balance when consistent signs are used.

Report evidence. Record your initial prediction and initial input values. Include one table of input modifications, screenshots or plots of response curves, the relevant balance or compatibility equation, and a short explanation of how the numerical answer changes with the altered parameter. Distinguish the direction of the change from its magnitude. State at least one parameter not modeled and explain how ignoring it could matter in a physical structure.

Reflection questions. (1) Which physical quantity is controlled, and which quantity is calculated? (2) Does the program use a closed-form equation, a small stiffness matrix, an eigenproblem or time integration? (3) Which check can be performed independently by hand? (4) What decision would be unsafe to make solely from this model?

Lab 04 — Moment distribution and joint rotation

Linked unit: 2. Open [labs/lab-04.html](labs/lab-04.html).

Before interaction. Use identical spans and equal UDLs, then perturb only the second UDL by 10%.

Expected mechanical insight. Distribution factors depend on rotational stiffness, so length imbalance changes relative correction even if EI is unchanged.

Independent or limiting-case check. At a fixed exterior end the rotation is zero, but the bending moment need not vanish.

Report evidence. Record your initial prediction and initial input values. Include one table of input modifications, screenshots or plots of response curves, the relevant balance or compatibility equation, and a short explanation of how the numerical answer changes with the altered parameter. Distinguish the direction of the change from its magnitude. State at least one parameter not modeled and explain how ignoring it could matter in a physical structure.

Reflection questions. (1) Which physical quantity is controlled, and which quantity is calculated? (2) Does the program use a closed-form equation, a small stiffness matrix, an eigenproblem or time integration? (3) Which check can be performed independently by hand? (4) What decision would be unsafe to make solely from this model?

Lab 05 — Planar three-bar truss matrix

Linked unit: 3. Open [labs/lab-05.html](labs/lab-05.html).

Before interaction. Increase cross-sectional area with geometry fixed and compare member forces with displacements.

Expected mechanical insight. For a truss whose every area changes by a common scale factor, load sharing may remain constant while displacement scales inversely with axial stiffness.

Independent or limiting-case check. With symmetric geometry and vertical loading, lateral top displacement should remain extremely close to zero.

Report evidence. Record your initial prediction and initial input values. Include one table of input modifications, screenshots or plots of response curves, the relevant balance or compatibility equation, and a short explanation of how the numerical answer changes with the altered parameter. Distinguish the direction of the change from its magnitude. State at least one parameter not modeled and explain how ignoring it could matter in a physical structure.

Reflection questions. (1) Which physical quantity is controlled, and which quantity is calculated? (2) Does the program use a closed-form equation, a small stiffness matrix, an eigenproblem or time integration? (3) Which check can be performed independently by hand? (4) What decision would be unsafe to make solely from this model?

Lab 06 — Planar rigid portal-frame stiffness

Linked unit: 3. Open [labs/lab-06.html](labs/lab-06.html).

Before interaction. Sweep roof load F and EI separately. Record drift, base reaction and member end moments.

Expected mechanical insight. For linear elastic small-displacement models, doubling F doubles every displacement and member action.

Independent or limiting-case check. Horizontal support reactions must sum to the opposite of applied horizontal force even if columns have unequal stiffness.

Report evidence. Record your initial prediction and initial input values. Include one table of input modifications, screenshots or plots of response curves, the relevant balance or compatibility equation, and a short explanation of how the numerical answer changes with the altered parameter. Distinguish the direction of the change from its magnitude. State at least one parameter not modeled and explain how ignoring it could matter in a physical structure.

Reflection questions. (1) Which physical quantity is controlled, and which quantity is calculated? (2) Does the program use a closed-form equation, a small stiffness matrix, an eigenproblem or time integration? (3) Which check can be performed independently by hand? (4) What decision would be unsafe to make solely from this model?

Lab 07 — Continuous-beam support settlement

Linked unit: 4. Open [labs/lab-07.html](labs/lab-07.html).

Before interaction. Apply only support settlement and no gravity load. Reverse the sign of the settlement using the model description as a prediction exercise.

Expected mechanical insight. Prescribed movement can create reactions that sum to zero externally, because all internal member forces are self-equilibrating.

Independent or limiting-case check. When settlement is zero the unloaded beam should have zero reactions and zero displacement throughout.

Report evidence. Record your initial prediction and initial input values. Include one table of input modifications, screenshots or plots of response curves, the relevant balance or compatibility equation, and a short explanation of how the numerical answer changes with the altered parameter. Distinguish the direction of the change from its magnitude. State at least one parameter not modeled and explain how ignoring it could matter in a physical structure.

Reflection questions. (1) Which physical quantity is controlled, and which quantity is calculated? (2) Does the program use a closed-form equation, a small stiffness matrix, an eigenproblem or time integration? (3) Which check can be performed independently by hand? (4) What decision would be unsafe to make solely from this model?

Lab 08 — Restrained thermal elongation

Linked unit: 4. Open [labs/lab-08.html](labs/lab-08.html).

Before interaction. Heat the steel-like and aluminum-like segments separately; inspect the sign of axial force.

Expected mechanical insight. A free-expanding fixed-fixed assembly is compressed under positive net thermal growth, but cooling can induce tension.

Independent or limiting-case check. Compare computed total actual extension against the prescribed overall zero displacement.

Report evidence. Record your initial prediction and initial input values. Include one table of input modifications, screenshots or plots of response curves, the relevant balance or compatibility equation, and a short explanation of how the numerical answer changes with the altered parameter. Distinguish the direction of the change from its magnitude. State at least one parameter not modeled and explain how ignoring it could matter in a physical structure.

Reflection questions. (1) Which physical quantity is controlled, and which quantity is calculated? (2) Does the program use a closed-form equation, a small stiffness matrix, an eigenproblem or time integration? (3) Which check can be performed independently by hand? (4) What decision would be unsafe to make solely from this model?

Lab 09 — Euler buckling and moment amplification

Linked unit: 5. Open [labs/lab-09.html](labs/lab-09.html).

Before interaction. Compare K=0.5,1 and 2 for equal EI and span. Then sweep load ratio close to 0.9.

Expected mechanical insight. Pcr depends on the inverse square of effective length. Moment amplification is increasingly sensitive to uncertainty near the critical state.

Independent or limiting-case check. At half effective-length factor, ideal critical load becomes four times larger.

Report evidence. Record your initial prediction and initial input values. Include one table of input modifications, screenshots or plots of response curves, the relevant balance or compatibility equation, and a short explanation of how the numerical answer changes with the altered parameter. Distinguish the direction of the change from its magnitude. State at least one parameter not modeled and explain how ignoring it could matter in a physical structure.

Reflection questions. (1) Which physical quantity is controlled, and which quantity is calculated? (2) Does the program use a closed-form equation, a small stiffness matrix, an eigenproblem or time integration? (3) Which check can be performed independently by hand? (4) What decision would be unsafe to make solely from this model?

Lab 10 — Moving two-axle load and moment influence

Linked unit: 6. Open [labs/lab-10.html](labs/lab-10.html).

Before interaction. Move the two axles through both entry and exit events. Compare section moments before any axle enters, while both are on the span, and after the vehicle exits.

Expected mechanical insight. The real moment diagram is piecewise linear between point loads, whereas the influence line is a response ordinate versus load position.

Independent or limiting-case check. For every vehicle position, test reaction sum against active axle weights only.

Report evidence. Record your initial prediction and initial input values. Include one table of input modifications, screenshots or plots of response curves, the relevant balance or compatibility equation, and a short explanation of how the numerical answer changes with the altered parameter. Distinguish the direction of the change from its magnitude. State at least one parameter not modeled and explain how ignoring it could matter in a physical structure.

Reflection questions. (1) Which physical quantity is controlled, and which quantity is calculated? (2) Does the program use a closed-form equation, a small stiffness matrix, an eigenproblem or time integration? (3) Which check can be performed independently by hand? (4) What decision would be unsafe to make solely from this model?

Lab 11 — Two-floor shear-building modes

Linked unit: 7. Open [labs/lab-11.html](labs/lab-11.html).

Before interaction. Increase story 2 stiffness with story 1 stiffness fixed. Compare first and second natural periods.

Expected mechanical insight. The two mode shapes differ in the relative signs or sizes of floor displacements; this distinction is invisible in a one-DOF approximation.

Independent or limiting-case check. Substitute each computed squared frequency into the two-by-two characteristic determinant as an independent check.

Report evidence. Record your initial prediction and initial input values. Include one table of input modifications, screenshots or plots of response curves, the relevant balance or compatibility equation, and a short explanation of how the numerical answer changes with the altered parameter. Distinguish the direction of the change from its magnitude. State at least one parameter not modeled and explain how ignoring it could matter in a physical structure.

Reflection questions. (1) Which physical quantity is controlled, and which quantity is calculated? (2) Does the program use a closed-form equation, a small stiffness matrix, an eigenproblem or time integration? (3) Which check can be performed independently by hand? (4) What decision would be unsafe to make solely from this model?

Lab 12 — Damped SDOF harmonic response

Linked unit: 7. Open [labs/lab-12.html](labs/lab-12.html).

Before interaction. Compare forcing-frequency ratio 0.5, 1 and 1.5 under low and high damping. Observe the initial transient.

Expected mechanical insight. The near-resonant response can grow over several cycles before a damped steady-state condition is approached.

Independent or limiting-case check. If the damping ratio increases, resonance peaks should generally decrease; transient phase and peak timing will also change.

Report evidence. Record your initial prediction and initial input values. Include one table of input modifications, screenshots or plots of response curves, the relevant balance or compatibility equation, and a short explanation of how the numerical answer changes with the altered parameter. Distinguish the direction of the change from its magnitude. State at least one parameter not modeled and explain how ignoring it could matter in a physical structure.

Reflection questions. (1) Which physical quantity is controlled, and which quantity is calculated? (2) Does the program use a closed-form equation, a small stiffness matrix, an eigenproblem or time integration? (3) Which check can be performed independently by hand? (4) What decision would be unsafe to make solely from this model?

Lab 13 — Plastic collapse mechanism

Linked unit: 8. Open [labs/lab-13.html](labs/lab-13.html).

Before interaction. Keep plastic moment Mp fixed while varying span L. Discuss why a longer span lowers ideal collapse load.

Expected mechanical insight. Collapse load is a mechanism result, not an allowable service load; serviceability and elastic strength may govern well below it.

Independent or limiting-case check. For a fixed-fixed central-load mechanism, collapse load is twice the ideal simple-span value for the same Mp and L.

Report evidence. Record your initial prediction and initial input values. Include one table of input modifications, screenshots or plots of response curves, the relevant balance or compatibility equation, and a short explanation of how the numerical answer changes with the altered parameter. Distinguish the direction of the change from its magnitude. State at least one parameter not modeled and explain how ignoring it could matter in a physical structure.

Reflection questions. (1) Which physical quantity is controlled, and which quantity is calculated? (2) Does the program use a closed-form equation, a small stiffness matrix, an eigenproblem or time integration? (3) Which check can be performed independently by hand? (4) What decision would be unsafe to make solely from this model?

Lab 14 — Load and stiffness uncertainty

Linked unit: 8. Open [labs/lab-14.html](labs/lab-14.html).

Before interaction. Increase the coefficient of variation of load and note the effect on the upper deflection percentile.

Expected mechanical insight. These results show sensitivity to selected distributions, not real evidence about failure probability or project safety.

Independent or limiting-case check. When both CV values are zero, every sample must yield the same nominal deflection up to rounding.

Report evidence. Record your initial prediction and initial input values. Include one table of input modifications, screenshots or plots of response curves, the relevant balance or compatibility equation, and a short explanation of how the numerical answer changes with the altered parameter. Distinguish the direction of the change from its magnitude. State at least one parameter not modeled and explain how ignoring it could matter in a physical structure.

Reflection questions. (1) Which physical quantity is controlled, and which quantity is calculated? (2) Does the program use a closed-form equation, a small stiffness matrix, an eigenproblem or time integration? (3) Which check can be performed independently by hand? (4) What decision would be unsafe to make solely from this model?

Lab 15 — Beam mesh convergence at x=0.37L

Linked unit: 8. Open [labs/lab-15.html](labs/lab-15.html).

Before interaction. Compare interior deflection at x=0.37L for meshes of 2,3,4,6,8,12,16 and 24 beam elements.

Expected mechanical insight. The closed form provides a reference; a smooth graph can still differ from the exact within-element displacement.

Independent or limiting-case check. Check error progression with mesh refinement rather than assuming monotonic error reduction for every arbitrary sample point.

Report evidence. Record your initial prediction and initial input values. Include one table of input modifications, screenshots or plots of response curves, the relevant balance or compatibility equation, and a short explanation of how the numerical answer changes with the altered parameter. Distinguish the direction of the change from its magnitude. State at least one parameter not modeled and explain how ignoring it could matter in a physical structure.

Reflection questions. (1) Which physical quantity is controlled, and which quantity is calculated? (2) Does the program use a closed-form equation, a small stiffness matrix, an eigenproblem or time integration? (3) Which check can be performed independently by hand? (4) What decision would be unsafe to make solely from this model?

Appendix C — Numerical interpretation notebook

Numerical interpretation notebook — Structural Analysis II

Each discussion below explains a baseline simulation from the teaching software and identifies a physical check. The examples use idealized mechanical conditions; figures are not specifications for real structural work.

Lab 01 — interpretation and independent reasoning

The common displacement of two parallel springs can be derived without solving a matrix. Every branch connects the same node to an immovable support, so both branches experience the same change in length. Multiply that change by each spring stiffness and sum the member reactions. In the default example, a 100 kN force acting against 40 kN/mm aggregate stiffness produces 2.5 mm displacement. This is enough information to recover both branch forces independently. If the first spring is softened while the second stays unchanged, the total stiffness falls, increasing the common displacement, but the softer branch also attracts a smaller fraction of the applied force. These two effects must be distinguished. The branch-force sum remains 100 kN because no other load path is modeled. For a real elastomeric bearing, stiffness depends on geometry, confinement, aging and sometimes loading rate. The ideal spring model intentionally omits those complexities.

Lab 02 — interpretation and independent reasoning

The propped cantilever exposes the power of a carefully selected redundant. Remove the roller at the right tip to obtain a stable cantilever. A uniformly distributed downward load produces downward free-tip movement wL^4/(8EI). A hypothetical upward tip reaction produces opposite movement RL^3/(3EI). Compatibility requires their sum to vanish because the actual prop prevents vertical tip motion. The result is R=3wL/8, leaving 5wL/8 at the fixed support. The fixed support moment is not zero; it must close moment equilibrium for the full load and prop. At w=12 kN/m and L=6 m, R=27 kN and the fixed end takes 45 kN of vertical force. In a nonuniform girder, piecewise EI changes the flexibility integrals, so substituting a single average rigidity can be a poor shortcut. In a lift-off situation a roller cannot pull downward, so a negative calculated contact reaction would require a new unilateral contact model.

Lab 03 — interpretation and independent reasoning

A continuous beam is statically indeterminate because a middle vertical support adds a redundant reaction. With three simple supports, the rotations remain possible but are shared across member ends because the beam itself is continuous. The two Euler–Bernoulli beam elements contribute their stiffness to one common interior rotation. Computing each span as an independent simply supported beam destroys continuity and erases the hogging moment over the middle support. In the default 5 m and 7 m example, asymmetric loading produces an interior rotation and three support reactions that must sum to 106 kN of gravity load. The plot depicts only cubic displacement interpolation across each element; a uniform load produces a quartic exact deflected shape between nodes. Thus nodal rotations and reactions can be useful while the drawn displacement curve may still need mesh refinement. Comparing those different accuracy levels is part of the lab, not a reason to hide the limitation.

Lab 04 — interpretation and independent reasoning

Moment distribution can be understood as solving the same joint equilibrium as a small stiffness matrix. Consider a middle joint B joining two members whose outer ends A and C are rotationally fixed. Each connected member contributes rotational stiffness 4EI/L at B. Fixed-end loading moments initially leave a nonzero sum at B. A balancing rotation generates additional end moments proportional to the individual stiffnesses, and the net correction must cancel the unbalanced moment. The well-known distribution factors are simply each stiffness divided by their sum. When both spans have the same EI and length, the correction splits evenly; different span lengths shift the split even before changing distributed loads. Far-end moments also change because rotation at one end affects the other end through bending continuity. A physical drawing should label near and far ends and moments using one explicit convention. Do not interpret equal and opposite joint actions as separate externally applied moments.

Lab 05 — interpretation and independent reasoning

A three-member pin-jointed truss provides a transparent matrix example with only two unknown displacements. The top node can move horizontally and vertically; each bar resists the projection of that motion onto its own axis. A bar inclined by angle θ contributes horizontal stiffness proportional to cos²θ and vertical stiffness proportional to sin²θ, with coupling terms proportional to sinθ cosθ. The central vertical member mainly adds vertical resistance. In a symmetric arrangement, the horizontal coupling of left and right diagonal bars cancels, so the top displacement under purely vertical load has essentially zero horizontal component. The reported diagonal member axial forces need not equal half the applied vertical force because only their vertical components contribute to vertical equilibrium. The middle member also carries load. Summing raw axial force magnitudes, instead of their resolved components, would incorrectly imply a violation of equilibrium. The lab assumes pinned nodes, no flexural rigidity, and common linear elastic behavior.

Lab 06 — interpretation and independent reasoning

The rigid-jointed portal frame extends truss analysis by adding bending rotation at each node. Its three members connect a left column, a roof beam and a right column, with their bases fixed. Each node has horizontal displacement, vertical displacement and in-plane rotation. Axial stiffness EA/L acts along the member; bending stiffness enters the transverse and rotational coordinates. Local element matrices are rotated to global axes before assembly so a vertical column contributes correctly to horizontal sway resistance. At the baseline horizontal roof load of 80 kN, the two base horizontal reactions sum to −80 kN. This check is necessary but not sufficient: a swapped local axis may still accidentally preserve total force while corrupting the member end moments. Compare both roof joint displacements and the complete set of end moments when testing the transformation. The plotted polyline traces displacements along an unfolded frame perimeter and is not a geometrically correct visualization of the bent frame.

Lab 07 — interpretation and independent reasoning

Settlement converts what might look like an ordinary loading problem into one governed by known displacement. The middle beam support is assigned a downward movement while end supports remain vertically fixed. Even with the external distributed load set to zero, an elastic continuous beam can develop shears and bending moments because its supported profile is no longer straight. In stiffness terms, the vector of prescribed displacements is multiplied by the constrained-to-free stiffness block and transferred into the reduced equilibrium equation. The resulting support reactions should sum to zero when no external vertical load acts. A positive reaction at one support can be paired with a negative reaction at another, but in a real one-sided contact support a negative reaction might indicate uplift and the need to revise the model. To avoid a misleading conclusion, distinguish a support forced to move with the structure from a roller that may separate as the deck deforms. These represent different boundary conditions.

Lab 08 — interpretation and independent reasoning

Thermal expansion in a fixed-fixed assembly is resisted by uniform internal axial force when axial equilibrium requires the same force through serial segments. A steel-like first bar and aluminum-like second bar can have different Young moduli, cross-sectional areas and thermal expansion coefficients. Their unrestrained changes in length are added, not their stresses. The total axial compliance equals L1/(E1A1)+L2/(E2A2). Dividing the negative free extension by that compliance gives the internal force needed to preserve the fixed endpoint spacing. At the default heating inputs, the total free growth is 3.32 mm and the model predicts about −57.62 kN, or compression. Internal stress varies between segments because force divided by area differs. The displacement plot shows the actual accumulated extension of each segment rather than the free extension that would occur if unrestrained. Joint slip, transverse curvature from temperature gradient and nonlinear properties with temperature are intentionally absent.

Lab 09 — interpretation and independent reasoning

The Euler critical load is an eigenvalue benchmark for an ideal perfectly straight prismatic member. The effective length factor changes the wavelength of the admissible buckling mode. For EI=5 MN·m² and L=5 m with pinned-pinned conditions, Pcr is approximately 1974 kN. Doubling the unbraced member length divides Pcr by four. Varying K from 1 to 2 produces the same ideal reduction because Pcr depends on (KL)⁻². The second graph shows the multiplier 1/(1−P/Pcr), but this is only an illustrative one-mode sensitivity relationship. It does not incorporate measured initial crookedness, material yielding, residual stress, connection nonlinearities or the frame-level geometric stiffness eigenproblem. A high value of the multiplier is a warning that assumptions matter greatly; it must not be reported as a verified reserve of design strength. Stable geometry and credible end-restraint selection are prerequisites.

Lab 10 — interpretation and independent reasoning

An axle train is more subtle than a single moving point load because load presence changes at bridge entry and exit. The first axle position is controlled directly; the second is offset behind it by the chosen spacing. Only axles whose positions lie between zero and bridge length contribute to the reactions and internal moment. With front load 80 kN at x=9 m and rear load 110 kN at x=5 m on a 20 m simple span, the support reactions are 126.5 kN at the left and 63.5 kN at the right. At the midspan section the bending moment is 635 kN·m, after subtracting the moment effects of both loads lying left of the section. As the train moves, this moment is not necessarily maximized when the front axle is exactly centered. It is useful to compare the section moment with an influence-line ordinate formulation to ensure both methods use the same axle positions and force units.

Lab 11 — interpretation and independent reasoning

A two-story shear model has two lateral displacements and therefore two natural modes. The mass matrix is diagonal when all story masses are lumped directly at the floors. The first story spring contributes stiffness to the ground-to-first-floor motion, while the second spring contributes coupling between floor displacements. Consequently the global stiffness matrix contains off-diagonal −k2 terms. Solving det(K−ω²M)=0 produces two positive squared circular frequencies if the assumed masses and spring stiffnesses are positive. In the default demonstration, the frequencies are about 1.995 Hz and 4.461 Hz, with first-mode period about 0.501 s. If the eigenvector is normalized to unit roof displacement, the second mode may exhibit opposite relative floor motions. Sign reversal of a whole eigenvector changes no physical prediction. A real building requires more modes, rotational DOFs, torsional stiffness, distributed mass and potentially soil interaction.

Lab 12 — interpretation and independent reasoning

The damped single-degree oscillator is the simplest model that includes inertia, elastic resistance and dissipation. Applied force is harmonic, F sin(ωt), and the displacement starts from rest. The code integrates m ü+c u̇+k u=F sin(ωt) using the average-acceleration Newmark method. Near the natural frequency, a small damping ratio can permit large oscillatory displacement compared with the static force-over-stiffness response. At the default 4 tonne mass and 2 MN/m spring, the circular natural frequency is √(k/m), so cyclic frequency is approximately 3.56 Hz. A 20 kN load gives a static displacement of 10 mm; the near-resonant transient may be many times larger. The plotted interval is finite and includes transient response, so its maximum is not identical to the theoretical infinite-time steady-state amplitude. Time-step refinement and a separate analytical frequency-response function are helpful independent checks.

Lab 13 — interpretation and independent reasoning

An ideal plastic collapse mechanism is a work balance, not a nonlinear finite-element load–deflection calculation. A fixed-fixed beam with a concentrated force at midspan can develop plastic hinges at the two fixed ends and at the center, provided ductile sections can sustain the required rotations without local or lateral buckling. If each hinge resists the fully plastic moment Mp, internal plastic work is the sum of moment times hinge rotation. For the symmetric mechanism this yields central-load collapse P=8Mp/L. At Mp=220 kN·m and L=8 m, the ideal result is 220 kN. A simply supported beam requires only one central hinge and its corresponding ideal collapse load is 4Mp/L, or 110 kN for the same section. The response line shown before the ideal collapse threshold is schematic rather than a physical stiffness or load–deflection curve. Serviceability or connection capacity can govern well before this theoretical ultimate mechanism.

Lab 14 — interpretation and independent reasoning

A reproducible uncertainty demonstration illustrates how assumed distributions change output spread without claiming measured structural reliability. The baseline beam has a central point force on a simply supported span and is treated as linearly elastic, so δ=PL³/(48EI). The force and Young modulus are varied through deterministic pseudo-random Gaussian perturbations with chosen coefficients of variation; span and second moment of area remain fixed during each draw. A finite sample of 350 calculations produces a sorted empirical distribution and selected percentiles. At the baseline values, the nominal deflection is around 8.93 mm, with a particular sample mean near 8.87 mm. These numbers are outputs of the invented distributions and random seed, not predictions of a measured population of real beams. Negative perturbations are clipped to positive values, which further modifies the distribution. A reliability calculation would require evidence-based material and load statistics, resistance definitions, failure limit states, correlations and calibration.

Lab 15 — interpretation and independent reasoning

The convergence lab is intentionally evaluated at x=0.37L instead of at the center or at a node shared by many meshes. A simply supported beam under uniform load has an exact Euler–Bernoulli quartic displacement field. Two-node beam elements use cubic Hermite interpolation within each element, even though their nodal rotations and displacements are solved from consistent equivalent loads. A mesh can therefore give very accurate selected nodal quantities while the within-element curve remains imperfect. At the chosen coordinate, the lab compares the interpolated finite-element displacement with v(x)=−wx(L³−2Lx²+x³)/(24EI). It plots absolute interpolation error versus number of elements for a sequence including odd and even mesh sizes. Because a plotted error can become smaller than the displayed rounding precision, a zero shown to five decimals is not proof of exact mathematical equality. This experiment is about convergence of a specific point response, not every possible member shear or stress output.

Appendix D — Integrative engineering projects

Engineering projects — Structural Analysis II

All project quantities are teaching examples and require independent design verification in real applications. For each project submit a model sketch, governing equation, parameter sheet, one hand-check, annotated graphs and limitation statement.

Project 1 — Warehouse roof reactions

A warehouse roof has a redundant bearing installed after a temporary stage. Determine the bearing load under dead load and under a subsequent live load, taking the actual support displacement to be zero. Begin with a released-beam cantilever, determine the unrestrained tip displacement and apply a compatibility correction. A useful sensitivity question is whether a soft bearing of finite stiffness meaningfully changes peak reaction. To avoid disguising uncertainty, compare an ideal rigid prop with a spring bearing and state why the result is conditional on bearing stiffness.

Deliverables: 1. A reference geometry and ideal boundary condition; 2. a traceable calculation in SI-derived units; 3. three parameter variants that isolate load, span or restraint influence; 4. a side-by-side plot or table with clear output units; 5. a limiting-case benchmark; 6. a written explanation of which modeling assumptions must be reconsidered before using results in design.

Technical oral examination: Explain why the calculated response is compatible with the chosen supports. Propose one check not built into the code. Identify a potential nonlinearity, and explain whether increasing the level of model sophistication will necessarily improve accuracy without better physical input data.

Project 2 — Two-span classroom girder

A classroom floor girder crosses one interior column and carries different distributed live loads on each span. Estimate support moments by slope–deflection and by the direct stiffness method. Explain any sign differences from end-action conventions. Test three load patterns: full-span dead load, live load on the left span only and live load on the right span only. For each pattern, record joint equilibrium, total support reaction and a deformation sketch. Use at least two stiffness values to illustrate that under linear, uniformly scaled EI, moment distribution depends more on relative EI than absolute EI.

Deliverables: 1. A reference geometry and ideal boundary condition; 2. a traceable calculation in SI-derived units; 3. three parameter variants that isolate load, span or restraint influence; 4. a side-by-side plot or table with clear output units; 5. a limiting-case benchmark; 6. a written explanation of which modeling assumptions must be reconsidered before using results in design.

Technical oral examination: Explain why the calculated response is compatible with the chosen supports. Propose one check not built into the code. Identify a potential nonlinearity, and explain whether increasing the level of model sophistication will necessarily improve accuracy without better physical input data.

Project 3 — Industrial portal roof sway

A rectangular portal supports a lateral wind action and roof gravity action. First build a two-column and one-beam stiffness model with fixed bases. Then examine the effects of connection release, frame height and member EI on roof drift. Draw the local member end moments using a consistent sign convention. Comment on why a correct global reaction sum does not guarantee that bending and axial force recovery are correct. In a professional analysis, wind pressure distribution, second-order effects, bracing, diaphragm action and load combinations must be considered separately.

Deliverables: 1. A reference geometry and ideal boundary condition; 2. a traceable calculation in SI-derived units; 3. three parameter variants that isolate load, span or restraint influence; 4. a side-by-side plot or table with clear output units; 5. a limiting-case benchmark; 6. a written explanation of which modeling assumptions must be reconsidered before using results in design.

Technical oral examination: Explain why the calculated response is compatible with the chosen supports. Propose one check not built into the code. Identify a potential nonlinearity, and explain whether increasing the level of model sophistication will necessarily improve accuracy without better physical input data.

Project 4 — Bridge bearing replacement

During bearing replacement, the central support of a two-span steel beam is temporarily lowered by several millimeters. Model settlement as prescribed displacement, not an external force. Determine the change in support reactions and end moments, and contrast the result with the unchanged-support baseline. Identify which assumptions are violated if the bearing loses contact and the support reaction cannot become tensile. Such a unilateral contact problem requires updated constraints and is not solved by simply accepting a negative bearing reaction from a fixed-support linear model.

Deliverables: 1. A reference geometry and ideal boundary condition; 2. a traceable calculation in SI-derived units; 3. three parameter variants that isolate load, span or restraint influence; 4. a side-by-side plot or table with clear output units; 5. a limiting-case benchmark; 6. a written explanation of which modeling assumptions must be reconsidered before using results in design.

Technical oral examination: Explain why the calculated response is compatible with the chosen supports. Propose one check not built into the code. Identify a potential nonlinearity, and explain whether increasing the level of model sophistication will necessarily improve accuracy without better physical input data.

Project 5 — Temporary construction support stability

A temporary steel strut holds a wall while excavation proceeds. Compare nominal Euler critical loads for idealized pinned, fixed and cantilever end conditions. Explain why the real construction sequence may change effective length, alignment and lateral support. Perform a sensitivity analysis on actual unbraced length. An educational instability calculation should trigger a more detailed design review, not be translated directly into a permissible excavation load.

Deliverables: 1. A reference geometry and ideal boundary condition; 2. a traceable calculation in SI-derived units; 3. three parameter variants that isolate load, span or restraint influence; 4. a side-by-side plot or table with clear output units; 5. a limiting-case benchmark; 6. a written explanation of which modeling assumptions must be reconsidered before using results in design.

Technical oral examination: Explain why the calculated response is compatible with the chosen supports. Propose one check not built into the code. Identify a potential nonlinearity, and explain whether increasing the level of model sophistication will necessarily improve accuracy without better physical input data.

Project 6 — Bridge truck crossing

A two-axle vehicle enters and exits a simply supported deck span. Compute end reactions and midspan bending moment at several axle positions, including a position with one axle off the span. Confirm the piecewise moment function and note the section of maximum curvature under the assumed simple-span linear model. Describe the additional factors required for traffic design: dynamic amplification, multiple vehicle lanes, load distribution between girders and prescribed load patterns.

Deliverables: 1. A reference geometry and ideal boundary condition; 2. a traceable calculation in SI-derived units; 3. three parameter variants that isolate load, span or restraint influence; 4. a side-by-side plot or table with clear output units; 5. a limiting-case benchmark; 6. a written explanation of which modeling assumptions must be reconsidered before using results in design.

Technical oral examination: Explain why the calculated response is compatible with the chosen supports. Propose one check not built into the code. Identify a potential nonlinearity, and explain whether increasing the level of model sophistication will necessarily improve accuracy without better physical input data.

Project 7 — Office-floor vibration

A two-floor idealized shear structure has different floor masses and lateral story stiffnesses. Compute its two modal frequencies, explain mode shape normalization and compare a time-history forced vibration of a single ideal mode. Differentiate displacement amplitudes from acceleration amplitudes when discussing human comfort. A clean classroom oscillation does not establish performance under real wind, occupant walking or earthquakes without realistic forcing spectra and structural damping.

Deliverables: 1. A reference geometry and ideal boundary condition; 2. a traceable calculation in SI-derived units; 3. three parameter variants that isolate load, span or restraint influence; 4. a side-by-side plot or table with clear output units; 5. a limiting-case benchmark; 6. a written explanation of which modeling assumptions must be reconsidered before using results in design.

Technical oral examination: Explain why the calculated response is compatible with the chosen supports. Propose one check not built into the code. Identify a potential nonlinearity, and explain whether increasing the level of model sophistication will necessarily improve accuracy without better physical input data.

Project 8 — Model-audit exercise

A contractor submits a continuous-beam analysis with a smooth deflected-shape plot and a single stated maximum moment. Demand the stiffness matrix, support assumptions, input load cases and a reaction summary. Compare an interior deflection against the exact uniform-load beam formula, and run mesh refinement at a fixed nonnodal coordinate. Add a deterministic parameter-uncertainty plot while warning that arbitrary probability distributions cannot establish failure risk. Finally document which results have passed verification and which assumptions remain to be validated physically.

Deliverables: 1. A reference geometry and ideal boundary condition; 2. a traceable calculation in SI-derived units; 3. three parameter variants that isolate load, span or restraint influence; 4. a side-by-side plot or table with clear output units; 5. a limiting-case benchmark; 6. a written explanation of which modeling assumptions must be reconsidered before using results in design.

Technical oral examination: Explain why the calculated response is compatible with the chosen supports. Propose one check not built into the code. Identify a potential nonlinearity, and explain whether increasing the level of model sophistication will necessarily improve accuracy without better physical input data.

Appendix E — Numerical conditioning and force recovery

Additional technical note — numerical conditioning and force recovery

A small linear residual does not always imply a small error in the displacement solution. When a stiffness matrix is ill-conditioned, a tiny perturbation to a support spring or a nearly redundant restraint can produce a comparatively large change in displacement. The matrix condition number summarizes how solution error may amplify under perturbation, but it depends on DOF scaling and the units assigned to different displacement coordinates. A rotation in radians and a translation in meters should not be compared as if they had the same numerical magnitude without physical interpretation. When diagnosing a nearly singular frame, inspect connectivity and released DOFs first, rather than adding arbitrary diagonal stiffness until the solver stops warning.

Member force recovery deserves a separate check from nodal displacement recovery. For a beam with distributed loading, the local element end forces require subtracting or adding consistent fixed-end load actions according to the chosen convention. Simply multiplying the local stiffness by the end-displacement vector may omit member actions induced directly by applied UDL. This can generate a bending-moment chart that is smooth but missing its proper fixed-end contribution. A sound diagnostic compares the sum of all member end shears with the equivalent applied load and verifies that the two end moments and shears satisfy member equilibrium about one end.

An engineering model should expose the solver tolerance, the residual norm, the physical units and at least one independent benchmark. The comparison ought to be made for a clearly identified output, such as middle support reaction, roof lateral drift or an interior point deflection, rather than for an unspecified statement that two graphs look similar. If a load is moved through a beam finite-element model, convergence is required for the full response envelope—not only for one arbitrarily chosen axle location. These checks make model results more trustworthy as educational calculations, though they still cannot substitute for validation of the physical structure.