It is pretty easy to do.
Step 1:
Find the bad board from the GUI, either by the following BitMaxz's method (keep in mind that his picture is for S9, where it labels the chains as 6,7,8 which is different from the new models that show 1,2,3) or by searching for these words in the kernel log
copy the exact words, hit ctrl+F and paste it in the kernel log, let's say you get the following
2020-04-08 09:32:51 driver-btm-api.c:1010:check_asic_number_with_power_on: Chain[0]: find 30 asic, times 0
2020-04-08 09:33:01 driver-btm-api.c:1010:check_asic_number_with_power_on: Chain[1]: find 0 asic, times 0
2020-04-08 09:33:11 driver-btm-api.c:1010:check_asic_number_with_power_on: Chain[2]: find 30 asic, times 0
This means Chain 1 is bad.
Step 2:
Allocate the bad board in the miner, when you stand in front of the miner, facing the ethernet port they go 1,2,3 (left to right)
They can double-check the numbering by reading the label near the socket where the ribbon cable goes, disabling the board is an easy process too, of course, step 2 must be done while the miner is powered-off.