You can view logfiles with journalctl. Use the --boot=-1 option to view the log from the previous boot. So on the next start after a failed shutdown
journalctl --boot=-1 --no-pager
will output the whole log from the previous boot. Look for anything suspicious towards the end where it’s shutting down. Or upload it to one of these pastebin services and post the link here. E.g.
journalctl --boot=-1 --no-pager | curl -F 'file=@-' https://0x0.st
You could also try
journalctl --boot=-1 --no-pager | grep watchdog
to see any messages about watchdog timers as that seems to have been the issue for some people.