THE COULOMB-PONCELET ultimate earth-pressure theory
1. Assumptions
Consider a rigid retaining wall of a massive type, generally not vertical, its contact surface AB is described by a certain angle β in Fig.1. Assume what follows.
1) There is a plane state of displacements typical for long objects; only a representative 1m section is considered; the wall can move outwards the soil (to the left in Fig.1, f < 0) or towards the soil (to the right in Fig.1, f > 0),
2) the backfill parameters are c = 0 kPa, ϕ > 0, γ > 0; the backfill top boundary is generally not horizontal, usually ε ≥ 0,
3) ultimate shearing (soil sliding) happens if
• τ = σ⋅ tg ϕ , for a sliding inside the soil, ϕ = internal friction angle,
• τ = σ⋅ tg δ , for a sliding on the contact wall-soil, δ = external friction angle at the contact, 4) when the wall moves infinitesimally, a rigid wedge of soil is observed (triangle ABC of the
weight G
ABC[kN/m]) which a little bit slides downwards in limit equilibrium along the wall AB and along a narrow shearing zone BC in the soil interior; three separate parts - the wall, the backfill triangle ABC and the rest of the backfill - are considered as rigid bodies,
5) there are used general symbols δ
i, because the signs (orientations) are very important, but of course the values are different δ
1= -ϕ, δ
2= δ,
6) by assumption, the BC curve is linear, so determined by an angle χ .
The details are presented in Fig.1, also the angles β, ε, χ, δ
2which are positive here (but δ
1< 0).
Fig.1. Denotations.
The force polygon is presented on the right:
vertical vector G
ABCis known (if a value is assigned to the angle χ), directions of both vectors E
a[kN/m] and R [kN/m] are known, too.
The triangle can be solved; the value of E
ais of interest (the value of R is not).
+
χ A’
H A
B
C
β
ε E
aR G
ABCδ
1δ
2L
∠ = χ+δ
1∠ = 90
o-β-δ
2f < 0
2. The Coulomb-Poncelet concept
The Coulomb idea is to decompose the vertical vector G
ABCinto two vectorial components E
aand R, along two given action lines, Fig.1. The value E
a= E
a is to be found, the reaction R
= Ris less important. Mathematically, this is a simple geometrical solving of the force triangle in which all angles are known, as well as one side G = G
ABC ;
the theorem called law of sines can be used, which yields E
a= G ⋅ sin( χ + δ
1) / sin(90
o- β - δ
2).
For the allowed horizontal translation f < 0 (Fig.1), a virtual downward sliding of the triangle ABC is expected. This simply remark has very important consequences – it means that shearing components of both forces E
aand R have a – more or less – downward orientation, along AB and BC respectively; both forces happen above the normal line perpendicular to the surface.
Therefore, the signs δ
2≥ 0 and δ
1< 0 are justified.
The BC line is entirely included in the soil mass interior, so for the limit equilibrium there is δ
1= - ϕ as usually for shearing in soils; the situation on AB depends on roughness of the soil- concrete contact, 0 ≤ δ
2≤ ϕ . Some comments on realistic δ
2values are discussed in Lecture 5.
In particular:
• min δ
2= 0
ocorresponds to a perfectly smooth surface (no shearing component, no friction),
• max δ
2= +ϕ (still greater value at the contact is impossible, because this is the maximal value in the soil interior, maybe only one milimeter next to the contact).
Note that hypothetically opposite δ
2sign is also possible in such a sense that the Poncelet approach is still effective. In practice, this could happen for large settlements of the wall (hitherto only horizontal translation f < 0 was assumed causing a dropping of the ABC triangle); this case looks much less probably and is not considered here.
Following conclusion is of the special importance:
the right selection of the angle signs depends
on the relative kinematics of the sliding block ABC and the wall itself (Lecture 5).
The convention of δ signs should be used only in conjunction with the normal vectors (dashed lines) outer to the wall-surface AB. This can be a source of misunderstandings, because two walls in Fig.2 are physically the same in the sense of the soil-pressure value E
a; one should not think that δ
2angles to the normal are opposite here.
We focus on the left case.
3. The Coulomb-Poncelet solution for the active pressure
Neither the wall height H nor the vertical axis z but the AB-interval length L in Fig.1 is the leading parameter of the model; all intervals are going to be expressed using L along AB, not the height H nor z for example.
The ABC-triangle area equals S = h⋅AC/2:
• for the triangle height h = A’B = L ⋅ sin( ∠ CAB) = L ⋅ cos( ε - β ),
• for the triangle base AC = L ⋅ sin( β - χ +90
o) / sin( χ - ε ) using again the law of sines.
The triangle weight is equal to G = S ⋅γ [kN/m] therefore from the force G decomposition:
Fig.2. Two equivalent design situations of δ
2> 0.
+ +
E
aE
a( χ ) = ½⋅γ⋅ ⋅ sin (χ − ϕ)
sin (χ − ϕ + 90 − β − δ ) ⋅ cos (ε − β) ⋅ sin (β − χ + 90 ) sin (χ − ε)
Value of one parameter is missing in this solution and this is the sliding wedge angle χ in Fig.1.
Coulomb proposed – as every good engineer would had to do – that the unknown variable χ takes the worst value as possible, maximal in this case E
a= max{E
a(χ)} = E
a(χ
a), for which dE
a/dχ = 0 if d
2E
a/d
2χ > 0. This way, the trigonometric equation dE
a/dχ = 0 yields an “ultimate”
angle χ
aand finally E
a= E
a( χ = χ
a).
This is J.V.Poncelet who first solved this general situation.
The solution is usually presented in the following form:
=
( )γ⋅ γ ⋅ /2 for
( )
= ( ϕ − β )
cos( β + ) ∙
1
1 + !sin( ϕ + ) ∙ sin(" − #) cos($ + ) ∙ cos(# − $)%
Note again that these formulae can be different sometimes if another sign convention is used.
4. Comments
1. The general concept comes from Ch.Coulomb (1773) but he solved in detail only the
simplest case of “The Coulomb wall”, i.e. with β = 0, ε = 0, δ
2= 0 for which χ
a= 45
o+ ϕ /2 and next K
aγ= (1-sin ϕ )/(1+sin ϕ ) = tg
2(45
o- ϕ /2).
2. The Coulomb concept is a prototype of a much more general variational approach, where generally the shape, of the “ultimate” sliding line BC is of interest, i.e. a certain functional equation z(x) for BC has to be found; in the family of linear functions, the problem is
algebraized, because such lines are fully described by the only one numerical parameter χ.
3. Note that the simply assumption of the linear sliding surface (triangular sliding wedge ABC) finds a strong justification in experiments; for Coulomb walls, this assumption and the theory are exact, worse agreement happens for situations being far from Coulomb walls.
4. The reader should appreciate a great skilfulness of J.V.Poncelet (mathematician in fact), who managed to overcome this toilsome derivation as early as just before 1840 (moreover, in pre-computer times he invented a useful graphical method supporting the earth-pressure calculations).
5. About one hundred years ago, Műller-Breslau preferred H coordinate instead of the length L and he got the similar solution =
(&'()γ⋅ γ ⋅ ) /2 .
Simply substitution of L = H/cos( β ) from Fig.1 expresses
(&'()γ=
( )γ⋅ /cos (β) , so this is nothing original.
Just
(&'()γvalues are presented in the Polish National Code PN-83/B-03010; nevertheless,
the use of the Poncelet original notation and his coefficients
( )γis strongly recommended
because this relates earth pressure to the physical contact surface AB, not to the abstract
variable H.
5. Uniform load q = const on the boundary
The idea for q > 0 is similar but the vertical weight force G
ABCis completed by q-action
integrated over AC, i.e. Q
AC= q⋅AC [kN/m]. The weight G
ABCis always the vertical vector but Q
ACis not necessarily so; the vectorial sum G
ABC+ Q
ACmust be decomposed into E
aand R – as previously.
F ig.3. Case of q > 0.
The situation becomes very simple if q > 0 is vertical. Indeed, Q
ACalgebraically increases the weight of the triangle ABC. Instead of the real triangle loaded by q, it suffices to consider the unloaded triangle ABC increasing its unit weight from the real γ to a virtual γ* > γ; note that the
“ultimate” angle χ
adoes not depend on the unit weight γ thus the solution will be correct.
To keep the equivalence of forces (weights), there must be G* = G + Q
ACor G* = S ⋅γ* = γ*⋅ h ⋅AC/2 = G + Q
AC= γ⋅ h ⋅AC/2 + q⋅AC.
Finally, γ * = γ + 2q/h = γ + 2 ⋅ q/(L ⋅ cos( ε - β )) = const( χ ).
The general formula =
*∙ +
∗∙ ∙ is true for every unit weight, thus making the substitution of γ * results in:
= 1
2 ∙ γ ∙ ∙ + - ∙ ∙
.where
.=
4 5(6'7)/0123.
6. Earth pressure as a continuous loading
Since the resultant earth pressure = 8 9 ( ):
<;, so there is 9 ( ) = : ( ) : ⁄ and finally:
9 ( ) = γ ∙ ∙ + - ∙
.This is a trapezoidal shape. For a homogeneous cohesionless backfill, the earth pressure e
a[kPa] increases linearly along the wall and has constant angle δ
2to the normal vector.
7. Layered soils
Layered backfills are not in use (exception for local drainage layers) but such a case can happen for slurry walls or sheet-pile walls embedded in natural soils, in their upper, not anchored part. Simple engineering approach is as follows.
+
χ q A
B
C
β
ε E
aR Q
ACδ
1δ
2L
∠ = χ+δ
1∠ = 90
o-β-δ
2f < 0
Simplify (replace) the boundary of the lower soil “2” in Fig.4 by a straight line at B
1which is parallel to the ground surface. For the two layers the angles ε and β are the same but the angles δ
2generally are not, being dependent on different ϕ
i. On the segment AB
1the solution is the same as previously (starting with L = 0 at A), on the segment B
1B
2the solutions starts anew for q
2= const instead of q
1for “1” (starting with L = 0 at B
1). Therefore, the earth pressure e
ais bilinear, having discontinuities in both the angle δ
2and the value e
aat B
1. In Fig.4 there must be ϕ
2> ϕ
1(explain why – two reasons). The same idea is effective for 3, 4 and more layers.
Note that the method is approximate and it overestimates a little bit the earth pressure, because q
2is locally overestimated on the separation level near the point B
1; indeed, a part of vertical loadings from q
1and γ
1is locally transmitted along AB
1due to friction.
8. Multi-linear wall profiles
For multi-linear wall profiles, like the one in Fig.5, the situation from Fig.4 can be adopted.
The backfill is the same, here only the angles differ β
1≠ β
2.
The earth pressure e
ais bilinear due to discontinuity of the angle β and therefore also
discontinuity of the value e
aat B
1occurs. In Fig.5 there is β
1> β
2. The same idea is effective for 2, 3 and more corners on the wall; the analogous overestimation of e
atakes place.
9. Passive earth pressure
Ch.Coulomb did not analyse walls pushed into the soil but J.V.Poncelet did. Since f > 0 in Fig.1, so the relative kinematics is different – the rigid triangular wedge ABC moves virtually upwards;
this way, the ultimate shearing stresses τ have signs opposite to the ones in the active case (δ
1= +ϕ and -ϕ ≤ δ
2≤ 0; as previously, there can be some exceptionsJ).
q
1A
B
1B
2„2”
„1” q
2h
1Fig.4. Two-layer system.
The upper layer has an assumed design thickness h
1;
for “1”: q
1 ≥ 0, γ1, ϕ
1, c
1= 0, h
1for “2”: q
2= q
1+ γ
1⋅h1⋅cos(ε), γ2, ϕ
2, c
2= 0 and next calculate as if the wall started at B
1q
1A
B
1q
2h
1Fig.5. Two-layer system.
The upper virtual layer has an assumed thickness h
1(see two parallel lines, it is a vertical distance).
First solve along AB
1ignoring B
1B
2for “1”: q
1≥ 0, γ
1, ϕ
1, c
1= 0, h
1is known.
Next solve along B
1B
2ignoring AB
1for “2”: q
2= q
1+ γ
1⋅ h
1⋅cos(ε), γ
1, ϕ
1, c
1= 0.
And so on, if there are still further, deeper intervals:
always focus only on the lower one.
B
2For q = 0 [kPa] the following result can be derived basing on E
p= min{E
p(χ)} = E
p(χ
p):
>
= 1
2 ∙ + ∙ ∙
>( )where
>( )