A single machine group scheduling problem was studied by incorporating learning and forgetting effects and maintenance into classic scheduling model. For the learning effects, two different models are formulated: one is to describe learning effect for the same product in a job; the other is to describe learning effect for different products in a family. And, because jobs are different in a family, so accumulated learning effects are partially transmitted, namely there exists forgetting effects from product to product, which is described by similarity in this paper. For maintenance policy, perfect preventive maintenance can be performed after one kind of products is finished, and when the machine fails due to degradation, minimal repair is performed to restore the machine to be “as bad as old”. Based on these policies, a more practical scheduling model within a family was formulated, whose objective is to minimize the completion time. Finally, a numerical example was given to show that the model is reasonable.