As Corion mentioned, your best bet is probably one of the many fine CSV parsing modules.

If you want to code your own solution, here is my suggestion. It will, I think (but this is untested), produce the same output as the code in wilsond's reply, but introduces the map and grep built-ins and a little more use of regexes (see perlre, perlretut and perlrequick), all of which I think you will find rewarding.

(BTW: I, also, am afraid I do not clearly understand your requirements, but perhaps this example may yet be helpful.)

use warnings;
use strict;


my $splitter = '|';
my $joiner1  = $splitter;
my $joiner2  = ' ';

while (my $record = <DATA>) {

    my ($head, @body) =
        grep ! m{ \A [\nSQ] \z  }xms,
        split  m{ \Q$splitter\E }xms,
        $record
        ;

    print
        join $joiner2,
        map  join($joiner1, $head, $_),
        @body
        ;

    print "\n";

    }


__DATA__
AAAA|Q|dfadfsfaf|S|asfddsfasfdasdfaf|werwqr345435|
BBBB|Q|ccccccccc|S|ddddddddddddddddd|eeeeee111111|
FFFF|Q|ggggggggg|S|hhhhhhhhhhhhhhhhh|iiiiii222222|
[download]

Output:

AAAA|dfadfsfaf AAAA|asfddsfasfdasdfaf AAAA|werwqr345435
BBBB|ccccccccc BBBB|ddddddddddddddddd BBBB|eeeeee111111
FFFF|ggggggggg FFFF|hhhhhhhhhhhhhhhhh FFFF|iiiiii222222
[download]