Skip to content
Ungewitter

All lessonsLesson 19

Thrust networks

From the line of thrust to the thrust network: form diagram, force diagram, and the least and greatest thrust of a vault.

Every line of thrust so far has lived in a plane: the arch of Lesson 1, the ring of Lesson 13, the flyer of Lesson 15. A vault is not a plane. Ungewitter crosses the gap the way the masons built the webs: he cuts each web into strips, treats every strip as a small arch resting on the ribs, and adds up what the strips deliver (p. 51). His Table I sums such strips for whole bays. Thrust network analysis, developed by Philippe Block and John Ochsendorf (2007), does the same job in one piece: it looks for a network of forces in three dimensions, in equilibrium with the loads, that lies entirely inside the vault.

From the Lehrbuch

„Die Zerlegung der Kappen in einzelne Streifen, die sich sonach nicht nach der Richtung der Steinschichten sondern nach der Gestalt der Kappen richtet, war sehr einfach durchzuführen, nunmehr handelt es sich darum, die zweckmässige Krümmung eines solchen Kappenstreifen festzustellen und den Widerlagsdruck, den er auf die ihn tragende Rippe ausübt.“Dividing the webs into single strips, which follow the shape of the webs rather than the direction of the courses, was very simple; what matters now is to find the suitable curvature of such a web strip and the abutment pressure it exerts on the rib that carries it.Ungewitter 1890, p. 51

One arch, one free number

Start with the plane. Ungewitter draws two figures for every arch (p. 52, Fig. 124): the arch with its line of thrust, and beside it the force plan, in which the weights of the voussoirs are laid off one below the other and a ray runs from a pole to every division. Each ray is parallel to the line of thrust in one joint. Today the first figure is called the form diagram and the second the force diagram. For a symmetric arch the whole pair hangs on one number, the horizontal thrust H, which is the distance of the pole from the load line.

OH = 32 kN
32 kN

H = 32 kN: the line of thrust fits inside the arch. Every thrust from 27.6 to 36.2 kN has such a line. admissible H: 27.6–36.2 kN; weight of the arch: 226 kN; free numbers: 1.

admissible H
27.6–36.2 kN
weight of the arch
226 kN
free numbers
1
Figure 1Left: a drop arch of 8 m span, R : s = 0.75, 0.6 m deep, with the line of thrust for the chosen H. Right: its force diagram, the pole O at the distance H from the load line, one ray per joint; ticks mark the least and greatest thrust. Computed with @ungewitter/statics (thrustRange).

Move the pole away and the rays flatten, and so does the line: more thrust, a flatter arch of forces. Between the least and the greatest thrust there is a line inside the masonry for every value; outside that range there is none. That is Lesson 13 again, now with the two diagrams side by side: the form diagram says where the forces run, the force diagram how large they are.

A network has many

In a vault the forces run in many directions at once: along the ribs, across the webs, round the boundary arches. The form diagram becomes a network in plan, and every node of it must be in horizontal equilibrium: the forces of its bars must close into a polygon. Put all those polygons together and they form the force diagram, a figure in which each bar of the network has a parallel line whose length is its horizontal force. James Clerk Maxwell called such a pair reciprocal figures. Unlike the arch, a network has many force diagrams: the forces can be shared out between ribs and webs in many ways, and each sharing can be scaled up or down. The scale is the thrust. Once the horizontal forces are fixed, the loads give the heights of the network, the thrust surface, which must stay between the soffit and the back of the vault.

The figure shows a bay of Table I, row II b: a square quadripartite vault of 8 m span, rise one third of the span, with a web of half a brick. The network was solved once when this page was built, at its least and at its greatest thrust.

Thrust network

Force diagram

State

Least thrust: 33.1 kN at each springer. The network hangs as deep as the vault allows and enters the springers 0.16 m above the springing. thrust per springer: 33.1–40.5 kN; H₀ by TNA: 203–249 kg/m²; H₀ in Table I, row II b: 220–240 kg/m².

thrust per springer
33.1–40.5 kN
H₀ by TNA
203–249 kg/m²
H₀ in Table I, row II b
220–240 kg/m²
Figure 2Left: the thrust network (red) over the ribs and arches of the bay (grey), in isometric view; crosses mark where the thrust enters the springers. Right: the force diagram of the same network, at the scale of the greatest thrust, so its size is the thrust. Computed with @ungewitter/statics (tna), after Block & Ochsendorf 2007.

At the least thrust the network hangs as deep as the vault allows, at the greatest it is as flat as it can be. The force diagram keeps its shape and grows. The two values bound what the vault can do to its supports, exactly as the flyer’s range bounded its push in Lesson 15. For this bay the network range in kilograms per square metre of plan overlaps the value of the book’s Table I, which Ungewitter built up from strips. For quick estimates the book condenses the table into a rule of thumb (p. 136).

From the Lehrbuch

„bei mittelhohen spitzbogigen Kreuzgewölben von etwa ⅔ Pfeilhöhe ist also ein Schub zu erwarten, der ungefähr gleich ⅓ des betreffenden Gewölbgewichtes (einer Hälfte) ist“for pointed cross vaults of middling height, with a rise of about 2/3, a thrust is therefore to be expected roughly equal to 1/3 of the weight of the vault in question (one half)Ungewitter 1890, p. 136

One caution belongs with every such number. The solver searches the force distributions of one pattern of ribs and courses; it explores them but does not prove that nothing lies further out. The true least and greatest thrust can only be wider than the range shown, which is the honest way round for a safety argument: a vault that admits one network stands.

Challenge

The essay recomputes every row of Table I with thrust networks. Find the row where the book and the networks disagree most. Is the book’s value too high or too low there, and what in the vault (rise, web, backing) might explain the difference?

Read Table I, recomputed →