This is possible and I would recommend the following scheme:



This scheme assumes the use of a 'safety class II' mains transformer in the power supply unit, where there's no need to connect the signal ground to safety earth. But mostly it's needed to have the enclosure being connected to safety earth, because mains wire routing inside enclosure could touch the walls. This situation can be solved by "hard" (means direct) earthing of enclosure but "soft" earthing of signal ground. Soft earthing is done by connecting signal ground to safety earth via a capacitor, which yields a quasi direct connection at high frequencies, but prevents DC and low frequency currents from flowing. To protect the operator from floating of signal ground a transzorb is connected in parallel to this capacitor.
The soft earthing connection point between signal ground and (hard earthed) enclosure must sit where the cable enters the enclosure. If more than one cable enters the enclosure, then they must enter it at the same point!! So, the mains cable must enter the enclosure just where the relay driver cable does it.

If you make the relay driver cable removable, then ESD protection at the terminals is needed. For this the two SMBJ24A transzorbs are used. (One of them works as flywheel diode at the same time.) Take care, that the soft earthing components are also involved with the ESD protection. This means, that all the involved parts must be located just where the cable enters the enclosure!! ESD protection means nothing else than that the electrostatic discharge is shunted to the enclosure walls, away from sensible electronics inside. So, at the left module the SMBJ24A, SMBJ12CA and 100n capacitor must sit directly where the cable enters the enclosure. At the right modul the SMBJ24A, SMBJ12CA and 100n capacitor must also sit where the cable enters the enclosure. But here even the 100µF electrolytic and its associated 100n capacitor are part of the ESD protection scheme and must also be located just where the cable enters the enclosure!

Latter components and the 68Ohm resistor form an effective low pass filter reducing noise on the +24V supply line and the common ground routing. Think about how the current will flow when turning-on the relay: At the very first moment, the very first current, which contains some unsane harmonics, will flow from positive terminal of 100µF electrolytic, through the relay, via cable to BC817 and back to negative terminal of 100µF. The 68Ohm resistor prevents any current from flowing outside this loop. So, the first portion of relay current is limited to this loop. Other regions of ground routing are not involved by this current and are kept quiet!

When using above scheme, then not much interference is produced when the relay is turned-on or -off: The rather slowly BC817 turns-on and -off in more than 1µsec. This creates rather long current transition times and by this minimizes electrical, magnetic and electromagnetic interference. The use of twisted pair cable and the location of components (directly where the cable enters the enclosure) helps herin by consequently reducing the loop area of current to the absolute minimum.

If you use two metallic enclosures and connect them by a shielded cable with the shield connected to enclosure at both ends, as shown above, then the Faraday cage is extended over your whole application. This makes it very difficult for interference to enter or leave.

Make no mistake, relays can produce lots of interference at their contacts, especially if capacitve or inductive loads are switched. Here only snubbers and similar can help. But this noise has nothing to do with the relay driver section.

Kai