A novel algorithm called free search (FS) is applied to solve the scheduling problems on a single batch processing machine with non-identical job sizes (NSBM) in this paper. The concept of sensibility was introduced in FS, with which the algorithm had no restriction of the probability to zero for the parts of the search space and thus it could avoid premature convergence. In order to make it feasible to NSBM problem, we modify the parameters settings and also provide coding by vector method which makes FS proper to discrete problem. Then the algorithm of batching based FS (BFS) is proposed. In the experiment, all levels of instances are simulated and the results show that BFS has better performance on global search and outperforms other algorithm in the literature.