The paper deals with a transfer line optimal desing. In contrast to the known assembly line balancing problem, for considered transfer lines, operations are grouped into blocks of parallel operations. Each block is executed by one spindle head. Spindle heads allocated to each station can be activated both sequentially and simultaneously, i.e. in a mixed order. The set of all available spindle heads is given beforehand. In addition to the standard line balancing problem, precedence and cycle time constraints, blocks compatibility and parallelism constraints are taken into account. The objective function is defined by the sum of blocks costs and stations costs. The problem is to assign all operations grouped into blocks that all constraints are respected and line investment cost is minimum. This paper is focused on solving the problem by a branchand- bound algorithm. A new approach for obtaining a lower bound is offered. It is based on a reduction of the transfer line balancing problem to a set partitioning problem. Computational experiments show that the proposed approach is efficient to solve practical transfer line desing problems.