I would return a list of "records", where each record is implemented as a hash. E.g.

  my $infh = IO::File->new("< data") or die;

  local $_ = $infh->getline;
  s/LV /LV_/g; # :-)
  my @column_names = split;

  my @records;
  while ( $infh->getlines ) {
    my %rec;
    @rec{@column_names} = split;
    push @records, \%rec;
  }

  return @records;

# or simply

  return map { my %r; @r{@c} = split; \%r } $infh->getlines;
[download]

...just for example. A better approach would parse fixed-width columns. A more robust extension of that would attempt to infer the column widths.

Update. If one of the columns is a "unique key", then it probably makes sense to load the set of records into a hash, using those values as the hash keys. E.g.:

  return map {
      my %r;
      @r{@c} = split;
      ( $r{'LV_NAME'} => \%r )
  } $infh->getlines;
[download]

(That doesn't really return a hash, it returns a list... but that list can sensibly be assigned to a hash and it will DWIM.)