The free body is a choice, and choosing it well is the whole skill
Every structural calculation begins with an act of violence. Something is cut away from everything it touches, and the forces that were doing the touching are drawn on the cut.
The cut is imaginary and the forces are not. That combination is what makes the free-body diagram strange, and it is worth being precise about, because the technique is usually presented as a step and is actually the entire subject.
What a cut creates
Before the cut, the beam has a load and two reactions. After the cut, the left-hand piece has a load, one reaction, and two new quantities on its right-hand face: a shear force and a bending moment.
Those two did not exist as separate objects a moment earlier. They are what the material at that station was doing, made visible by removing everything on the other side of it and asking what has to replace it.
The logic runs backwards from the answer. The left-hand piece is not moving. Therefore the sums on it must vanish. Therefore whatever the right-hand piece was providing must have been exactly enough to make them vanish — and that is the definition of the internal forces, not a calculation of them.
Newton’s third law then guarantees consistency: the same actions appear on the right-hand piece with opposite signs. Either piece gives the same answer, so the choice between them is purely one of convenience. Cutting to the left of a complicated set of loads and working with the simple side is not a shortcut; it is the same calculation done with fewer terms.
Where to cut
The working rule is short: cut through what is wanted, and through as little else as possible.
A cut that passes through three unknown member forces in a plane frame leaves three unknowns and three equations, which is solvable but awkward. A cut through two leaves a choice of moment centre that kills one of them, and the remaining unknown falls out in a line.
Comparing the two cuts makes the point that the internal forces are not properties of a location in isolation. They are properties of a location given everything to one side of it, and moving the cut past a load changes what that everything contains.
That is why the shear diagram jumps at a point load and the moment diagram merely kinks. The free body gains a whole force at once, so the force sum steps; it gains that force at zero distance from the cut, so the moment sum is continuous.
The joint as a free body
The same technique applied to a single joint of a truss gives the method of joints, and the free body is a point rather than a piece.
A joint is the smallest useful free body and the most restricted: with all forces passing through one point, the moment equation is satisfied automatically and only two equations survive. That limitation is the reason the method has an order — start at a support where only two members meet, solve it, and use the answers as knowns at the next joint.
The whole-truss solver behind the truss figures does the same thing without the ordering, by writing all the joint equations at once and solving the system. The answers are identical; what disappears is the need to find a joint simple enough to start at.
The boundary decides what is external
A force is external or internal depending on where the boundary was drawn, and nothing else.
Take the whole beam as the free body and the shear at mid-span does not appear anywhere in the equations — it is internal, and internal forces cancel in pairs. Take half the beam and the same quantity is external, drawn on the cut face, and it is the unknown being solved for.
This is the deepest thing about the technique and the easiest to lose. The question “is this force internal or external?” has no answer until the boundary is stated. A column’s axial force is internal to the building and external to the column. The reaction at a foundation is external to the structure and internal to the structure-plus-ground.
Drawing the boundary badly usually shows up as a diagram with forces that cannot be determined. Drawing it well means the boundary crosses exactly the things that are wanted and nothing else.
What has to be drawn on it
A free-body diagram is complete when four things are on it, and a diagram missing any of them will give an answer that is wrong without looking wrong.
Every applied load, including self-weight where it matters.
Every reaction the supports can supply, and only those. A roller drawn with a horizontal component invents a force; a pin drawn without one loses an equation.
Every internal action on every cut face — for a beam that is three quantities, since a cut can transmit axial force as well as shear and moment.
Nothing else. The commonest error is drawing a force that is already accounted for, usually by including both a load and the reaction it produces in a body that contains only one of them.
The diagram is then a complete statement of the problem, and the arithmetic that follows is mechanical. Nearly every error in statics is an error in the diagram rather than in the algebra after it, which is why the diagram is worth drawing even when the answer seems obvious.
Cutting a whole structure open
The same technique applied to a frame rather than a beam is the method of sections, and it answers one question without solving the rest.
Choosing the cut is the skill. A cut through three members, with moments taken about the point where two of them meet, gives the third in one line. That is why the method of joints is used to solve a whole truss and the method of sections to interrogate one member of it.
Read that way, a shear diagram is not a new idea at all. It is the free-body calculation performed at every station and plotted, and its shape records what each successive free body contained.
Where the model stops
Rigid bodies. The cut is made on the undeformed shape, and the pieces are assumed not to change shape as the forces are applied. That is a first-order assumption, and it fails when deflections are large enough to move the lines of action.
Statics only. The technique gives the internal forces of a determinate structure completely. For a structure with a redundant restraint it gives a family of answers, and picking among them needs the stiffness of the members — information the free body does not contain.
No local effects. The forces on a cut face are drawn as a shear, a moment and an axial force — three numbers standing in for a distribution of stress: three numbers standing in for a distribution of stress across the whole face. Saint-Venant’s principle says the substitution is safe a short distance away from where the load is applied, and unsafe close to it. Bearing failures, bolt tear-out and web crippling all live in the region where three numbers are not enough.
Two dimensions. A plane free body has three equations. A real one has six, and the three that were dropped are assumed to be looked after by something outside the picture.
The figures have a limit too, and it is one this site cannot get round. A cut face is drawn with a gap so the internal actions can be seen, and there is no gap — the material is continuous, and the shear and moment are distributions of stress rather than an arrow and a curved arrow. Every free-body diagram ever drawn commits this, and it is a good habit to remember that the arrow is a summary of something spread across an area.
The ladder from here
Later rungs on this anchor: the method of sections, and answering one question without solving the whole truss. Free bodies with internal releases. Sign conventions, and why the sagging-positive convention exists at all. Saint-Venant’s principle. Three-dimensional free bodies. Superposition, which is a statement about free bodies added together. The free body in a moving frame, where an inertia force is added and statics is recovered. And the free body of a piece of material rather than a piece of structure, which is where stress analysis begins.
Newton drew free bodies without naming them. The phrase and the teaching convention are twentieth-century, and the practice of insisting on the diagram before the algebra is younger still than that.