Mat and Raft Foundations: Soil-Structure Interaction

Mat stiffness and soil–structure interaction

A raft foundation shifts the engineer's attention from a single loaded pad to the coupled problem of superstructure, slab and soil. The raft's bending stiffness influences the spatial distribution of base pressure and movement. A rigid idealization may offer one tractable limit; a flexible plate on elastic soil offers another. Neither should be accepted automatically, because real contact is nonlinear and real soil has shear coupling between adjacent positions. A useful preliminary analysis compares these idealizations and identifies where they diverge.

Global equilibrium remains mandatory even when the model is numerical. The sum of reactions must match the imposed vertical loading, with corresponding moment balances. Local column actions create bending and punching demands; using an average pressure to describe a raft does not make those local effects disappear. When a slab bridges a weak patch, contact stresses can shift toward stiffer soil, but the amount depends on both slab stiffness and the geometry of the anomaly. An uncalibrated spring bed can generate precise-looking displacement contours that have little physical authority.

Construction affects the reference state. Excavation unloads soil and may allow swelling or heave, followed by staged reloading as the raft and building are placed. Groundwater can produce uplift on basement rafts. Adjacent pile installation or dewatering may change stresses after the design analysis is complete. Accordingly, the monitoring baseline should be defined around the actual construction stages rather than assumed to begin at first occupancy.

A sensible selection argument should compare isolated footings, a raft, a compensated raft and hybrid or improved ground where appropriate. The objective is to reduce unacceptable movement and construction risk at reasonable total cost. Raft design cannot be reduced to a slab-thickness calculator; understanding the source of settlement is more important than selecting a dramatic visualization.

Case integration: basement excavation and a compensated raft. A developer plans a basement whose excavation removes considerable soil before a reinforced raft and superstructure are built. The net pressure applied to the founding layer may differ from gross superstructure pressure because excavation unloads the site. During temporary exposure, fine-grained soils may heave, pore pressures may change and nearby structures may move due to groundwater drawdown. After the raft is placed, different parts of the site load at different rates as the building rises. An analysis based only on a final uniform pressure can conceal the governing construction condition. Draw a stage diagram from pre-excavation through excavation, formation, raft construction and completion. Track applied stress increments and expected movement direction in each stage, and note what a simple Winkler spring model leaves out. Compare a raft spanning a deep soft pocket with isolated spread footings. Require monitoring that can distinguish excavation-induced displacement from loading settlement. Do not assume that the smallest final total settlement means the smallest damaging differential movement during construction. If groundwater produces uplift on the basement slab, its resistance and water-pressure assumptions should appear as a separate check.

Four lessons and their applications

Lesson 3.1 — When a raft changes the problem

Why this mechanism matters

A raft couples multiple columns through a relatively stiff slab, redistributing reactions and potentially limiting differential movements. It does not erase weak soil or guarantee a smaller total settlement. A soil-structure analysis should conserve forces and moments while representing slab bending and ground deformation. Rafts are attractive when separate footings nearly overlap, but construction tolerance, basement buoyancy and groundwater control still govern.

Evidence, scales and the engineering model

The choice of model changes the interpretation of when a raft changes the problem: 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.

Decision and evidence. An investigation should resolve what changes the foundation selection, not maximize tests per hectare. Ground interpretation must be linked to an explicit load and performance objective. 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 9 — method and interpretation

Given. Two column loads of 1,000 and 1,600 kN act 6 m apart on a 12 m long raft.

Solution. The two-load resultant relative to the first column is x=(1000×0+1600×6)/2600=3.69 m.

Interpretation. Reaction and slab moment models must equilibrate this resultant; replacing the loads by one pressure without moment is incomplete.

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 — force equilibrium in a raft. Suppose a raft is represented by a continuous line of ground reaction under two discrete column loads. The total area beneath the pressure curve must equal the total applied force per corresponding transverse width, and its first moment must match the applied moment. A Gaussian load-spreading sketch can be normalized to satisfy the vertical force integral, which is valuable for teaching equilibrium. However, this does not guarantee correct deformation compatibility or moment balance unless the load centers and boundary assumptions are treated consistently. The real distribution depends on slab stiffness, contact conditions, soil variability and column locations. A numerical plot that simply looks smooth is not a verification method.

Decision exercise. With a 1,000 kN left load and a 1,600 kN right load separated by 6 m, determine the resultant 3.69 m from the left column. Try shifting the larger load farther toward the raft edge and explain how the contact pressure must redistribute to maintain global moments. Compare the same system with soft soil at the right edge and with uniform support: the contact forces are not expected to remain identical because compatibility changes. Check total integrated load, resultant position, and plausible nonnegative contact pressure before examining detail reinforcement. Explain why the interactive line-load model is a conceptual representation rather than a two-dimensional reinforced-concrete plate or a continuum finite-element computation.

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. Compare a raft against spread footings by serviceability, constructability and robustness, not area alone. 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: Uniform average contact pressure does not imply uniform settlement.

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 when a raft changes the problem. 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. Compare an optimistic continuous stratum with a discontinuous weak lens; which foundation system is sensitive to the contrast?
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, Elastic Settlement Sensitivity and Raft Reaction Redistribution, describe one experiment that would reveal the limitations of an oversimplified model.

Lesson 3.2 — Elastic foundation idealization

Why this mechanism matters

A Winkler representation treats soil as independent vertical springs with pressure proportional to local deflection. It is a useful teaching approximation, yet real soil couples neighboring points and changes stiffness with depth, stress and loading history. Spring stiffness should be consistent with discretization; a spring assigned to a node represents the tributary area rather than an intrinsic soil constant.

Evidence, scales and the engineering model

The choice of model changes the interpretation of elastic foundation idealization: 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.

Constitutive and hydraulic assumptions. Strength and stiffness must match the stress path, drainage regime and level of deformation being checked. A single number cannot represent a layered or nonlinear profile across all load combinations. 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 10 — method and interpretation

Given. A simplified Winkler model assigns modulus k = 20 MN/m³ under a 100 kPa contact pressure.

Solution. The local spring settlement is s=q/k = 100/20000=0.005 m or 5.0 mm.

Interpretation. This is only the prescribed local-spring response; it is not a settlement prediction calibrated to the same foundation footprint.

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 — calibrating Winkler springs. In a Winkler idealization local pressure equals a subgrade reaction modulus times local displacement. If one spring represents a tributary area A, its force stiffness is kA, so refining the grid requires adjusting individual spring stiffness rather than copying the same spring at every node. More importantly, the assigned modulus of subgrade reaction is not a fundamental material property: it depends on footing geometry, soil profile, state and the calibration method. Neighboring patches in real ground transmit shear and may not behave as independent columns of springs. Comparing a spring model with an elastic-continuum settlement estimate is therefore more revealing than arbitrarily increasing mesh density.

Decision exercise. With 100 kPa and 20 MN/m³ the local spring relation yields 5 mm. Refine a 2 m × 2 m conceptual patch into four equal-area nodes. The force in each should sum to the original total if displacement and pressure remain uniform; giving each new node the old full-area force stiffness would multiply resistance by four. Document the tributary area rule and a simple mesh convergence check. Next replace the homogeneous bed with a stiffer central strip and explain how a bending raft couples the nodes that the uncoupled spring sketch treats independently. The aim is to understand what the element assembly represents, not to select a stiffness until a preferred output appears.

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. Conduct mesh and spring-stiffness sensitivity studies, and compare with an elastic-continuum interpretation. 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 refined mesh does not cure a wrong soil 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 elastic foundation idealization. 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. Vary the groundwater model independently of geometry and identify which conclusions reverse.
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, Elastic Settlement Sensitivity and Raft Reaction Redistribution, describe one experiment that would reveal the limitations of an oversimplified model.

Lesson 3.3 — Contact, bending and redistribution

Why this mechanism matters

Column loads create local moments and punching demands; ground reactions must close the global vertical-force and moment balances. A stiff raft tends to spread local load while a flexible slab follows soil displacement more closely. Soil tension cannot develop at an unanchored contact. Settlement-induced frame forces may feed back into raft demand. Construction sequencing can matter when excavation and groundwater recovery precede loading.

Evidence, scales and the engineering model

The choice of model changes the interpretation of contact, bending and redistribution: 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.

Equilibrium and load distribution. Global force and moment equilibrium are necessary conditions even for sophisticated nonlinear models. A model that balances actions can still be wrong in local deformation and contact assumptions. 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 11 — method and interpretation

Given. Under the same contact pressure 120 kPa, left soil support has k=24 MN/m³ and right has k=40 MN/m³ over an effective length of 10 m.

Solution. End spring movements are 5.0 and 3.0 mm, with illustrative angular distortion |5−3|/10000=0.0002 rad.

Interpretation. A real raft bends; a full plate-on-soil model is necessary to predict actual differential movement.

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 — relative movement and spatial variability. Two raft corners can move the same average distance and still create very different structural outcomes if their movements have different spatial shapes. A linear tilt may be tolerable for some geometries, while localized curvature can create high frame distortion even with the same end-to-end displacement. A spring gradient is useful for plotting sensitivity, but real soil stiffness may change abruptly where a buried channel cuts a foundation. The raft's bending stiffness redistributes pressure into adjacent stiffer zones; an uncoupled spring plot overstates the independence of adjacent settlements. A reliable study therefore plots both ground reaction and raft deformation.

Decision exercise. Compare a 10 m span with 5 mm settlement at one end and 3 mm at the other. The end-to-end tilt proxy is 0.0002 rad. Now imagine a 10 mm localized dip around the center while end movements remain unchanged; the same tilt ratio would conceal a significant deformation concentration. Sketch three profiles with the same end values: straight tilt, smooth curvature and local soft pocket. Discuss which observations or finite-element model outputs distinguish them. Place survey markers at structural load-transfer points rather than only at widely separated building corners. Finally, define trigger thresholds in terms of the actual building's movement tolerance rather than adopting an arbitrary universal angular distortion limit.

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. Plot column reactions, contact pressure and slab displacement together for at least two stiffness choices. 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: Assumed uniform reactions can conceal governing local reinforcement requirements.

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 contact, bending and redistribution. 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. Quantify the reactions, then test whether the assumed local stress path is physically possible.
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, Elastic Settlement Sensitivity and Raft Reaction Redistribution, describe one experiment that would reveal the limitations of an oversimplified model.

Lesson 3.4 — Differential settlement control

Why this mechanism matters

Differential settlement produces angular distortion and redistributes actions in the superstructure. Localized soft pockets, adjacent excavation, and variable fill can matter more than average bearing strength. Monitoring should include benchmarks outside the moving zone, repeated surveys and trigger levels connected to responsive actions. Predicted settlement profiles should show plausible bounds.

Evidence, scales and the engineering model

The choice of model changes the interpretation of differential settlement control: 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.

Serviceability and sequence. Movement develops with loading, unloading, water changes and consolidation. Record construction stages and timescales before selecting a stiffness model. 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 12 — method and interpretation

Given. A survey benchmark moves 2, 4 and 7 mm across three observations at equally spaced one-week intervals.

Solution. Incremental movements are 2 and 3 mm per week; the increase is not linear and suggests looking at load history and measurement uncertainty.

Interpretation. Do not extrapolate indefinitely from three measurements; verify datum stability, instrument repeatability and structural relevance.

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 — trends are evidence, not forecasts. Foundation monitoring data are affected by instrument precision, temperature, reference datum stability, repeated loading and the actual soil movement process. A progressive settlement record should be plotted against construction events, rain and water-level changes. The rate may fall under consolidation or rise when new loading or pore-pressure changes occur. A single line drawn through three measurements cannot establish long-term behavior. Differential movement between nearby points is more informative for many superstructures than absolute movement alone, but both should be retained. Measurements without agreed action thresholds offer limited protection.

Decision exercise. For observations of 2, 4 and 7 mm at equal weekly intervals, calculate 2 and 3 mm/week increments. If the next reading is 12 mm, the increment increases to 5 mm and demands a different response from a projected flattening trend. Before inferring acceleration, check instrument setup, measurement repeatability and benchmark movement. Then compare the observed history with construction loads and groundwater records; a sudden increase during dewatering suggests a testable mechanism. Design a dashboard that displays absolute displacement, relative settlement between supports, rate of change and the dates of work stages. Assign a role responsible for responding to a threshold exceedance, with an alternative foundation support or work stoppage plan where justified.

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. Evaluate the same raft with uniform and locally weakened ground; report both tilt and absolute movement. 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: The acceptable distortion depends on the actual structure and finishes.

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 differential settlement control. 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. Identify one early-stage condition that could exceed the final-state response.
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, Elastic Settlement Sensitivity and Raft Reaction Redistribution, describe one experiment that would reveal the limitations of an oversimplified model.

Unit 3 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 05](../labs/lab-05.html) and [Lab 06](../labs/lab-06.html).

Assessment: [Unit 03 questions](../assessments/unit-03.html).

Project: Raft stiffness and localized soft zones — Model a 20 m × 12 m raft with three columns and a soft patch near one corner. Provide overall force and moment balance, spring sensitivity, a sketch of predicted settlement profiles and a monitoring concept. Explain why the toy load-spreading simulation is not a validated plate solution.