Comprehensive Foundation Design Synthesis and Forensic Case Studies

Selection, safety cases and verification

Foundation selection is an engineering decision under uncertainty, not a contest between footing and pile calculators. It should identify governing load cases, predicted settlement ranges, construction access, water conditions, neighboring infrastructure, available materials and consequences of failure. Initial cost is only one dimension. A shallow foundation can be economical in concrete but expensive in groundwater control; piles may avoid deep excavation yet demand noise mitigation and substantial testing. Improvement may be efficient if it can be verified within the schedule.

A transparent decision matrix is useful when its limitations are explicit. Weights and ratings are expressions of stakeholder priorities and project assumptions. If the ranking changes drastically under small adjustments, the selection is fragile and requires further evidence. A defensible recommendation should state the information that would reverse the conclusion, not simply present the highest score as truth. The interactive laboratory illustrates this sensitivity with synthetic ratings intentionally unrelated to market costs.

The design basis should record the governing standard and its edition, soil and groundwater parameters, calculation conventions, load combinations, serviceability limits, project constraints and approvals. Mixing a resistance reduction factor from one design format with service actions from another can create a false impression of safety. Codes also change; this open textbook does not reproduce proprietary standards or claim to stay automatically current. Project engineers must check their jurisdiction and applicable editions.

Verification connects the report with the constructed work. Construction logs, inspection, test results, monitoring records, nonconformances and remedial actions should be part of a traceable record. Independent review is particularly important when geology is complex or failure consequences are high. A useful handover states both what has been demonstrated and what remains uncertain. The educational labs here can support that conversation, but none can approve a real foundation.

Case integration: an investment decision with uncertain geology. A developer compares spread footings, a raft and piles for a medium-rise building on partially investigated ground. The first cost estimate favors shallow foundations, but a possible deep soft layer creates substantial settlement uncertainty. A transparent decision requires the same loads, groundwater state and performance criteria for each alternative. A weighted score can document priorities such as cost, serviceability and schedule, but synthetic ratings must never be mistaken for market estimates or geotechnical evidence. Carry at least two ground scenarios through all alternatives and compute a range of plausible outcomes. Identify the price of a targeted site investigation and compare it with the potential cost of remediation or a design change if the soft layer exists. A preferred option is defensible only if it remains acceptable under the agreed credible scenarios or if a verification step is scheduled to resolve a critical unknown. The final design record should include rejected options with reasons, versioned assumptions, the adopted code basis and a clear handoff to construction testing and settlement monitoring.

Four lessons and their applications

Lesson 8.1 — Comparing foundation alternatives

Why this mechanism matters

The lowest unit price for concrete or piling is rarely the full installed project cost. Compare footing, raft, improvement and deep-foundation options by excavation, groundwater control, materials, equipment access, schedule, noise, carbon impact, testing and credible failure consequences. Uncertainty should be priced as investigations and contingency rather than hidden in an optimistic estimate.

Evidence, scales and the engineering model

The choice of model changes the interpretation of comparing foundation alternatives: a single observation establishes a condition at a place and time, while an engineering prediction requires a link from that observation to a repeatable physical mechanism. Establish the relevant load or environmental scenario, then identify whether the response is governed by strength, deformation, pore pressure, geometry, or construction sequence. The question to ask before entering numerical values is not simply which equation applies? but which field evidence could make the equation inapplicable? The answer often determines whether the apparent precision of an output is justified.

Resistance mobilization. Maximum resistance depends on compatibility of movement and mechanism. Different components of an embedded foundation may require different displacements to mobilize. For this lesson, make the assumed boundary conditions visible in a section drawing and distinguish measured inputs from correlations. A parameter that is merely convenient should not be presented as if a project investigation had confirmed it.

A useful field-to-model workflow is to identify an observed quantity, the geological or structural mechanism, the chosen constitutive assumption, the calculation, and a practical verification measurement. If the observation cannot distinguish two mechanisms, carry both possibilities into the sensitivity analysis. The reported result should preserve units, stress convention, drainage condition, stage and uncertainty bounds. It must be possible to trace a final design recommendation back to the underlying evidence.

Worked case 29 — method and interpretation

Given. Three foundation concepts receive hypothetical cost, serviceability and constructability scores: spread (6,7,9), raft (7,9,6), piles (8,5,7).

Solution. With weights 5,8,4, normalized scores are spread=(30+56+36)/17=7.18; raft=(35+72+24)/17=7.71; piles=(40+40+28)/17=6.35.

Interpretation. The ranking only reflects invented ratings; change inputs when site investigation yields real comparative evidence.

The numerical answer above uses deliberately narrow assumptions to make the dependency transparent. If any assumption is relaxed, the expression may need a different correction or an altogether different mechanical model. Before applying the result to another footprint, soil layer or construction state, ask whether the test procedure and the selected parameters represent the new condition. Check dimensional consistency and confirm that the direction of a predicted change agrees with the underlying physics.

Deeper technical application

Diagnostic study — alternatives require common comparison boundaries. A foundation selection exercise may favor the cheapest visible foundation component while ignoring excavation support, water control, testing and delay risk. The decision frame must include full installed cost, schedule, risk, disturbance, and performance for the same building loads and ground scenario. A numerical preference matrix is legitimate only if the weights, ratings and alternatives are auditable. If a raft option is judged for settlement under service loads but piles are judged only for resistance under ultimate loads, the comparison is inconsistent. An alternative design should also be revisited if a newly observed soft layer or unexpected groundwater condition changes the assumptions.

Decision exercise. Using the synthetic ratings in the book, serviceability receives weight eight, cost five and constructability four. The raft obtains the highest illustrative score, about 7.71. Increase cost priority substantially and reduce serviceability priority to see whether spread footings or piles move higher. This is not a market forecast: all ratings are invented. Instead of accepting the ranking, ask which real evidence would inform each rating. For cost, request a complete construction estimate; for serviceability, obtain deformation analyses consistent with the same ground model; for constructability, assess access, noise and water. Record critical uncertainties that could reverse the selection and assign an investigation or pilot test to the most consequential one.

Counterfactual test

In this lesson, begin by writing the input assumption in a calculation sheet and introduce one perturbation at a time. Consider a 20% increase in the most uncertain load, a plausible change in groundwater, and a different depth or stiffness of a critical layer. Do not automatically multiply every output by the same factor: soil response is often nonlinear and the mechanism itself can change. Record which sensitivity is purely numerical and which requires more field data. If the decision to use shallow versus deep foundations changes, that is evidence that additional investigation may have a higher value than further decimal places.

Field practice and failure-first review

Practical action. Use a transparent multi-criteria comparison with documented weights and sensitivity to weights. Prioritize direct observations whenever a correlation could conceal a critical stratum or groundwater path. Document who reviews anomalies, whether repeat tests are required, and which construction stage is exposed. Failure trap: A numerical ranking is a decision aid, not an automatic design approval.

Make a brief alternative hypothesis for the same observations. Ask whether a weak pocket, high water level, unrecognized load eccentricity, or installation defect might explain an unexpectedly large deformation. The right response is not always more conservative material strength; sometimes it is a different foundation system, a modified construction sequence, or a targeted test that resolves the uncertainty. Protect neighboring structures and utilities when the mechanism can propagate outside the footprint.

Study questions

1. Explain the difference between a measured site observation and an inferred parameter in the context of comparing foundation alternatives. Which information would you insist on before approving a preliminary calculation?
2. Rework the example after increasing the most influential input by 20%. Show units and explain whether a linear response is physically justified.
3. Sketch resistance and displacement, separating load-transfer components.
4. Give one mode of failure or serviceability loss that the worked equation cannot detect, and propose a corresponding observation or laboratory test.
5. With the two unit laboratories, Laterally Loaded Pile Decay and Foundation Alternatives Decision Matrix, describe one experiment that would reveal the limitations of an oversimplified model.

Lesson 8.2 — Serviceability and monitoring

Why this mechanism matters

A foundation assessment requires limits for total movement, differential movement, rotation, cracking and changes over time. Instruments may include precise survey monuments, inclinometers, piezometers and load cells, but each measurement has installation and interpretation limits. Monitoring should test design assumptions and lead to action, not simply produce plots after a problem becomes visible.

Evidence, scales and the engineering model

The choice of model changes the interpretation of serviceability and monitoring: a single observation establishes a condition at a place and time, while an engineering prediction requires a link from that observation to a repeatable physical mechanism. Establish the relevant load or environmental scenario, then identify whether the response is governed by strength, deformation, pore pressure, geometry, or construction sequence. The question to ask before entering numerical values is not simply which equation applies? but which field evidence could make the equation inapplicable? The answer often determines whether the apparent precision of an output is justified.

Testing and response. Define how field observations will validate model assumptions and what a failing test will trigger. Inspection without decision thresholds is documentation, not control. For this lesson, make the assumed boundary conditions visible in a section drawing and distinguish measured inputs from correlations. A parameter that is merely convenient should not be presented as if a project investigation had confirmed it.

A useful field-to-model workflow is to identify an observed quantity, the geological or structural mechanism, the chosen constitutive assumption, the calculation, and a practical verification measurement. If the observation cannot distinguish two mechanisms, carry both possibilities into the sensitivity analysis. The reported result should preserve units, stress convention, drainage condition, stage and uncertainty bounds. It must be possible to trace a final design recommendation back to the underlying evidence.

Worked case 30 — method and interpretation

Given. Monitoring shows settlements 2.0 and 6.0 mm at positions separated by 8 m.

Solution. The observed rotation proxy is (6−2)/8000 = 0.0005 radians.

Interpretation. Structural consequence depends on geometry, loading, existing cracking and movement rate, not just this scalar.

The numerical answer above uses deliberately narrow assumptions to make the dependency transparent. If any assumption is relaxed, the expression may need a different correction or an altogether different mechanical model. Before applying the result to another footprint, soil layer or construction state, ask whether the test procedure and the selected parameters represent the new condition. Check dimensional consistency and confirm that the direction of a predicted change agrees with the underlying physics.

Deeper technical application

Diagnostic study — movement limits need an owner and a mechanism. Survey monuments measure displacement only relative to their reference system. They may drift if reference benchmarks lie within the zone being deformed. Inclinometers provide lateral movement profiles, piezometers provide hydraulic head information, and load cells reveal selected support forces. None directly tells the team that a building is safe. A monitoring plan should map each instrument to a specific predicted mechanism and define trigger levels, measurement frequency, responsible reviewer and actions. For serviceability, relative settlement and angular distortion may matter more than a single absolute movement, but the structure's actual geometry and use determine the acceptable response.

Decision exercise. Two columns separated by 8 m settle 2 and 6 mm, creating an angular change of 0.0005 rad under the simple end-point interpretation. Imagine that a further 3 mm develops at one column over a day following heavy pumping. The extra rate changes the risk picture even if the absolute value remains modest. Specify survey validation, additional groundwater readings and temporary construction decisions. Build a trigger-action-response matrix with observation thresholds agreed before construction, not chosen after alarm readings appear. Ask which personnel can stop work and how false alarms are verified without delaying action when a significant hazard is plausible.

Counterfactual test

In this lesson, begin by writing the input assumption in a calculation sheet and introduce one perturbation at a time. Consider a 20% increase in the most uncertain load, a plausible change in groundwater, and a different depth or stiffness of a critical layer. Do not automatically multiply every output by the same factor: soil response is often nonlinear and the mechanism itself can change. Record which sensitivity is purely numerical and which requires more field data. If the decision to use shallow versus deep foundations changes, that is evidence that additional investigation may have a higher value than further decimal places.

Field practice and failure-first review

Practical action. Write trigger-action-response levels before construction and assign ownership for escalation. Prioritize direct observations whenever a correlation could conceal a critical stratum or groundwater path. Document who reviews anomalies, whether repeat tests are required, and which construction stage is exposed. Failure trap: Instrumentation without an agreed response is not risk management.

Make a brief alternative hypothesis for the same observations. Ask whether a weak pocket, high water level, unrecognized load eccentricity, or installation defect might explain an unexpectedly large deformation. The right response is not always more conservative material strength; sometimes it is a different foundation system, a modified construction sequence, or a targeted test that resolves the uncertainty. Protect neighboring structures and utilities when the mechanism can propagate outside the footprint.

Study questions

1. Explain the difference between a measured site observation and an inferred parameter in the context of serviceability and monitoring. Which information would you insist on before approving a preliminary calculation?
2. Rework the example after increasing the most influential input by 20%. Show units and explain whether a linear response is physically justified.
3. Write a hold point and a concrete action for a field result outside the assumed range.
4. Give one mode of failure or serviceability loss that the worked equation cannot detect, and propose a corresponding observation or laboratory test.
5. With the two unit laboratories, Laterally Loaded Pile Decay and Foundation Alternatives Decision Matrix, describe one experiment that would reveal the limitations of an oversimplified model.

Lesson 8.3 — Codes, factors and professional responsibility

Why this mechanism matters

Geotechnical resistance methods differ among jurisdictions: allowable-stress, limit-state and load-and-resistance-factor formats cannot be mixed casually. Loads, groundwater, durability and environmental conditions need the same reference framework as resistance calculations. Assumed data, versioned standards, design basis and review signatures should be traceable. The educational equations in this book intentionally omit many code-specific corrections.

Evidence, scales and the engineering model

The choice of model changes the interpretation of codes, factors and professional responsibility: a single observation establishes a condition at a place and time, while an engineering prediction requires a link from that observation to a repeatable physical mechanism. Establish the relevant load or environmental scenario, then identify whether the response is governed by strength, deformation, pore pressure, geometry, or construction sequence. The question to ask before entering numerical values is not simply which equation applies? but which field evidence could make the equation inapplicable? The answer often determines whether the apparent precision of an output is justified.

Mechanism of improvement. Treat water, soil fabric and load paths as separate levers. A technique that accelerates drainage may not reduce ultimate compression. For this lesson, make the assumed boundary conditions visible in a section drawing and distinguish measured inputs from correlations. A parameter that is merely convenient should not be presented as if a project investigation had confirmed it.

A useful field-to-model workflow is to identify an observed quantity, the geological or structural mechanism, the chosen constitutive assumption, the calculation, and a practical verification measurement. If the observation cannot distinguish two mechanisms, carry both possibilities into the sensitivity analysis. The reported result should preserve units, stress convention, drainage condition, stage and uncertainty bounds. It must be possible to trace a final design recommendation back to the underlying evidence.

Worked case 31 — method and interpretation

Given. An illustrative nominal pile resistance is 2,400 kN, and a teaching exercise applies resistance reduction factor 0.55.

Solution. Factored illustration: 0.55×2,400=1,320 kN, to be compared only with actions in the same declared design basis.

Interpretation. This factor is arbitrary for the exercise and must not be presented as a governing code requirement.

The numerical answer above uses deliberately narrow assumptions to make the dependency transparent. If any assumption is relaxed, the expression may need a different correction or an altogether different mechanical model. Before applying the result to another footprint, soil layer or construction state, ask whether the test procedure and the selected parameters represent the new condition. Check dimensional consistency and confirm that the direction of a predicted change agrees with the underlying physics.

Deeper technical application

Diagnostic study — coherent resistance and load formats. Allowable-stress design, limit-state design and load-and-resistance-factor design use different combinations of actions, safety factors and resistance parameters. Transferring a resistance factor from one document into a load combination from another can destroy the intended reliability calibration. Groundwater assumptions, long-term and temporary conditions, structural material factors and geotechnical characteristic values must also be consistent with the adopted standard. This textbook avoids presenting any particular numerical factor as a universal requirement because foundation codes and legal responsibilities differ by region and project.

Decision exercise. If a nominal resistance were 2,400 kN and an explicitly hypothetical factor 0.55 were applied, the illustrative factored resistance would be 1,320 kN. The arithmetic is trivial; the real engineering question is whether that factor and its associated actions belong to the same governing framework. Make a design-basis register listing jurisdiction, project class, adopted standards and editions, load cases, water conditions, testing requirements and required sign-offs. Include temporary works and existing infrastructure protection. If two disciplines use different units or stress conventions, add a reconciliation check. A reviewer should be able to trace every resistance number to a test, correlation or documented assumption instead of finding unexplained values scattered across software exports.

Counterfactual test

In this lesson, begin by writing the input assumption in a calculation sheet and introduce one perturbation at a time. Consider a 20% increase in the most uncertain load, a plausible change in groundwater, and a different depth or stiffness of a critical layer. Do not automatically multiply every output by the same factor: soil response is often nonlinear and the mechanism itself can change. Record which sensitivity is purely numerical and which requires more field data. If the decision to use shallow versus deep foundations changes, that is evidence that additional investigation may have a higher value than further decimal places.

Field practice and failure-first review

Practical action. Create a design-basis register naming jurisdiction, limit state, assumptions and required reviewer. Prioritize direct observations whenever a correlation could conceal a critical stratum or groundwater path. Document who reviews anomalies, whether repeat tests are required, and which construction stage is exposed. Failure trap: No model in this textbook is a substitute for licensed professional design.

Make a brief alternative hypothesis for the same observations. Ask whether a weak pocket, high water level, unrecognized load eccentricity, or installation defect might explain an unexpectedly large deformation. The right response is not always more conservative material strength; sometimes it is a different foundation system, a modified construction sequence, or a targeted test that resolves the uncertainty. Protect neighboring structures and utilities when the mechanism can propagate outside the footprint.

Study questions

1. Explain the difference between a measured site observation and an inferred parameter in the context of codes, factors and professional responsibility. Which information would you insist on before approving a preliminary calculation?
2. Rework the example after increasing the most influential input by 20%. Show units and explain whether a linear response is physically justified.
3. Show which input to the foundation model changes after treatment and which does not.
4. Give one mode of failure or serviceability loss that the worked equation cannot detect, and propose a corresponding observation or laboratory test.
5. With the two unit laboratories, Laterally Loaded Pile Decay and Foundation Alternatives Decision Matrix, describe one experiment that would reveal the limitations of an oversimplified model.

Lesson 8.4 — Verification and an auditable design record

Why this mechanism matters

A convincing design report allows another engineer to reconstruct the ground model, choose governing parameters and follow every calculation to a clear decision. Independent hand calculations, second models and observations can catch correlated computational errors. Risks should remain visible where additional field evidence could change the solution. Close the loop between investigation, design, construction testing and monitoring.

Evidence, scales and the engineering model

The choice of model changes the interpretation of verification and an auditable design record: a single observation establishes a condition at a place and time, while an engineering prediction requires a link from that observation to a repeatable physical mechanism. Establish the relevant load or environmental scenario, then identify whether the response is governed by strength, deformation, pore pressure, geometry, or construction sequence. The question to ask before entering numerical values is not simply which equation applies? but which field evidence could make the equation inapplicable? The answer often determines whether the apparent precision of an output is justified.

Limits and communication. Use explicit design basis, sensitivity study and independent check. Charts are decision aids whose input assumptions should be inspectable. For this lesson, make the assumed boundary conditions visible in a section drawing and distinguish measured inputs from correlations. A parameter that is merely convenient should not be presented as if a project investigation had confirmed it.

A useful field-to-model workflow is to identify an observed quantity, the geological or structural mechanism, the chosen constitutive assumption, the calculation, and a practical verification measurement. If the observation cannot distinguish two mechanisms, carry both possibilities into the sensitivity analysis. The reported result should preserve units, stress convention, drainage condition, stage and uncertainty bounds. It must be possible to trace a final design recommendation back to the underlying evidence.

Worked case 32 — method and interpretation

Given. A design team records predicted total settlement values from two independent idealizations: 18 mm and 32 mm.

Solution. The absolute model spread is 14 mm; relative to their mean 25 mm it is 56%.

Interpretation. Disagreement signals modeling uncertainty; the response is to examine soil stiffness, strata and boundary assumptions, not average the numbers blindly.

The numerical answer above uses deliberately narrow assumptions to make the dependency transparent. If any assumption is relaxed, the expression may need a different correction or an altogether different mechanical model. Before applying the result to another footprint, soil layer or construction state, ask whether the test procedure and the selected parameters represent the new condition. Check dimensional consistency and confirm that the direction of a predicted change agrees with the underlying physics.

Deeper technical application

Diagnostic study — disagreement is useful evidence. Independent models of the same foundation may produce different settlement results because they emphasize different mechanics. An elastic half-space may use an average modulus, while a layered calculation accounts for varied compressibility and drainage. A finite-element model may add interaction and nonlinear behavior but also introduces parameter and boundary-condition choices. Agreement between two calculations is valuable only if their errors are not perfectly correlated. Duplicating the same mistaken groundwater assumption in two software packages does not create independent verification. A competent design record preserves discrepancy and traces it to the underlying mechanism.

Decision exercise. Compare settlement predictions of 18 and 32 mm, a 14 mm spread and roughly 56% of their 25 mm mean. Rather than taking the average, perform a reconciliation audit: check loads, foundation geometry, net versus gross pressure, groundwater, modulus definitions, drainage assumption and construction stage. Identify which site observations could distinguish the models. Next assess the consequence of selecting the lower value when the structure has an assumed movement threshold near 25 mm; additional investigation may be more rational than a neat average. Maintain calculation versions and a change log so that later construction observations can be compared with both prior predictions. Independent peer review should challenge the ground model as well as the algebra.

Counterfactual test

In this lesson, begin by writing the input assumption in a calculation sheet and introduce one perturbation at a time. Consider a 20% increase in the most uncertain load, a plausible change in groundwater, and a different depth or stiffness of a critical layer. Do not automatically multiply every output by the same factor: soil response is often nonlinear and the mechanism itself can change. Record which sensitivity is purely numerical and which requires more field data. If the decision to use shallow versus deep foundations changes, that is evidence that additional investigation may have a higher value than further decimal places.

Field practice and failure-first review

Practical action. Use reproducible input tables, units, version histories and explicit unresolved issues. Prioritize direct observations whenever a correlation could conceal a critical stratum or groundwater path. Document who reviews anomalies, whether repeat tests are required, and which construction stage is exposed. Failure trap: A polished graph is never a replacement for a justified ground model.

Make a brief alternative hypothesis for the same observations. Ask whether a weak pocket, high water level, unrecognized load eccentricity, or installation defect might explain an unexpectedly large deformation. The right response is not always more conservative material strength; sometimes it is a different foundation system, a modified construction sequence, or a targeted test that resolves the uncertainty. Protect neighboring structures and utilities when the mechanism can propagate outside the footprint.

Study questions

1. Explain the difference between a measured site observation and an inferred parameter in the context of verification and an auditable design record. Which information would you insist on before approving a preliminary calculation?
2. Rework the example after increasing the most influential input by 20%. Show units and explain whether a linear response is physically justified.
3. State the observation that would make the current recommendation unacceptable.
4. Give one mode of failure or serviceability loss that the worked equation cannot detect, and propose a corresponding observation or laboratory test.
5. With the two unit laboratories, Laterally Loaded Pile Decay and Foundation Alternatives Decision Matrix, describe one experiment that would reveal the limitations of an oversimplified model.

Unit 8 synthesis

This unit connects its four lessons through one overarching design decision: the right model follows from the mechanism and evidence, not from a convenient formula. Prepare a one-page design-basis table listing assumptions, parameter origin, uncertainty, sensitivity, and the condition requiring professional escalation. Pair that table with a labeled cross-section. For the investigation or structure in your own scenario, justify whether the controlling concern is ultimate resistance, unacceptable movement, unexpected groundwater, temporary construction loading or interaction with neighbors. Do not confuse a numerical check with an approval; identify the external standard and reviewer required for a real project.

Paired simulations: [Lab 15](../labs/lab-15.html) and [Lab 16](../labs/lab-16.html).

Assessment: [Unit 08 questions](../assessments/unit-08.html).

Project: Foundation selection and design-basis memorandum — Develop a weighted selection matrix for three foundation options and demonstrate when the ranking changes. Add a consistent limit-state design basis, groundwater assumptions, construction constraints, monitoring thresholds and evidence required for third-party review. State why the numerical matrix is not an automatic sign-off.