I think Berta gave a very correct answer. But it may be that an approximate control of focal length would be sufficient. In the diagram below a pair of elements form a lens. A Source Ray launches a single ray parallel the optical axis and through the lens to strike a detector. The merit function targets the y centroid location to be y = 0, meaning on the axis. The detector is therefor targeted to be the image plane for an object at infinity.

Strictly speaking, the focal length is measured from the principle plane. We can see by inspection that that is located somewhere inside the second element. With the detector in a fixed position, and for small variations of the design, where the principle plane does not move much, this part of the merit function provides approximate control of the focal length. (It actually requires the design to be such as to make the detector the image surface for distant objects.)

I say this part of the merit function because I assume other parts of the merit function will employ other sources and perhaps a different detector for controlling the main objectives in the design.