A Linguistic Information Granulation Model and Its Penalty Function-Based Co-Evolutionary PSO Solution Approach for Supporting GDM with Distributed Linguistic Preference Relations.