HiIf a cable is carrying low-level analog signals, the shield drain wires should be grounded at only one end. Grounding at one end only prevents any current from flowing between none equipotential ground points, because ground current flowing in the drain wire or shield can couple into the small-signal wires and corrupt the signal. When the drain/shield is connected at only one end, it serves as a Farady cage to isolate the low level signal(s) from noise/interference.
This is now not the current best practice, due to the increasing prevalence of high-frequency noise in industrial plant.
The earthing of the screen/shield at one end turns the shield into an antenna, and the frequencies at which the characteristic impedance start making this effect noticeable are now very relevant in a modern plant, being somewhere in the 1-10Mhz range.
The best guys to draw down on the knowledge and experience for this is RF and circuit board designers, some quite bizarre things can occur with resonance, terminating impedances etc etc.
Vendors such as Siemens have installation recommendations for the S7-1500 series PLC to earth the shield at both ends, on the basis that the return impedance thru structure etc is 10 times greater than the instrument cable drain wire impedance. If this can't be known then they suggest to run a Parallel Earth Conductor (PEC) out with the instrument cable, sized to be 10 times less impedance than the drain wire and effectively taking an earth out with you. (There are also less preferred alternative configs where the filed earth is via a capacitor etc.)
This has two impacts, one is to make the shield more like a Faraday cage and less like an antenna - a shield as an antenna has the possibility of induced currents being impressed on the active cable cores, but it also it acts as a current splitter for surge, lessening the surge current (and therefore voltage via V=IR at inputs) by a factor of 10.
In Functional Safety Service (IEC 61511) the undetected failure rates quoted on these S7-1500 controllers are phenomenally low, but Siemens tell me that a surge event can cause partial failures of silicon devices such that it is very difficult to predict likely and /or possible failure modes, so the diagnostics that give you the low failure rates cannot be as complete if surge is not managed.
The relevant Siemens publication is A5E03461486-AB - Designing Interference Free Controllers.
Further general detailed technical information is available in ABB publication:
Technical Application Papers No. 25
Electromagnetic compatibility: theory and
application measures in MV environments.
Plus there is some IEC standards etc.