A distributed flexible flowline scheduling problem with unrelated parallel machines at each stage is studied. Considering sequence-dependent setup times and job dynamic arrival times, an integer programming model is established with the objective of minimizing total weighted earliness and tardiness penalty. A hybrid discrete artificial bee colony algorithm is proposed combined with the discrete differential evolution algorithm, the variable neighborhood descent algorithm and local search so as to obtain near optimal solutions. In this algorithm, factory-job number based encoding is applied and a dynamic decoding mechanism with the earliest machine idle time is designed. The initial factory-job sequence group is then generated by using a random rule and an average assignment strategy. In the leading bee phase, the discrete differential evolution algorithm is introduced to yield factory-job sequences with high quality. In the following bee phase, a variable neighborhood descent algorithm is used to search around the selected sequences in order to gain neighborhood sequences. Local search based on insertion between critical/non-critical factories is designed to enhance algorithm search ability in the scout bee phase. Simulation experiments are performed on different scale problems, and testing results demonstrate the proposed hybrid discrete artificial bee colony algorithm has a better resolution performance.