Benchmark instances for the flexible job shop scheduling problem with dual-mode parallel batch processing (FJSP-DMPBP)

Published: 9 October 2026| Version 1 | DOI: 10.17632/gdm7nkn6b3.1
Contributors:
Shengping Lv,
, zicong Ou, hong jin

Description

This dataset contains the 35 benchmark instances used in the manuscript "Exact CP and MILP formulations for flexible job shop scheduling with dual-mode parallel batch processing" (FJSP-DMPBP), together with the preprocessing scripts, the preprocessed solver input files in IBM ILOG OPL format, and the Gantt-chart outputs of the schedules illustrated in the paper. In FJSP-DMPBP, three operation types coexist: solo operations; MPBPOs (mandatory parallel-batch operations that must be processed together in one p-batch task); and FPBPOs (flexible parallel-batch operations that may be grouped into one or several p-batch tasks, or processed individually). A p-batch task starts and finishes simultaneously and its processing time equals the maximum processing time of its members on that machine; the objective is to minimise the makespan. The 35 instances are organised into three groups (gsfjs01–gsfjs10, gmfjs01–gmfjs10, gmk01–gmk15) and were generated from Fdata (Bagheri et al., 2010) and BRdata (Brandimarte, 1993). For each instance, solver input files are provided for five exact formulations: three CP models (Baseline-CP, Sync-CP, USR-CP) and two MILP models (Baseline-MILP, EMILP). The two Python scripts that generate the solver input files from the source workbooks are included.

Files

Steps to reproduce

Run the preprocessing scripts in code/ on the workbooks in instances/ to regenerate preprocessed/. See section 5 of README.txt for the exact commands: the CP families (Baseline-CP, Sync-CP, USR-CP) are produced by code/tools_gen_dat.py, and the MILP families (Baseline-MILP, EMILP) by code/emilp_dat_gen.py. Solve the .dat files with IBM ILOG CPLEX Optimization Studio 22.1.1.0 (CP Optimizer for the CP models, CPLEX Optimizer for the MILP models) using the primary protocol: Threads = Workers = 1, RandomSeed = 0, wall-clock limit 300 s per formulation-instance pair. Plot any schedule with gantt/plot_gantt_from_csv.m; see section 4.3 of README.txt for the CSV format.

Categories

Computer Science Applications, Manufacturing Engineering, Management Science Methods

Licence