The SUMT algorithm of Fiaccoand McCormick (1968) is a powerful technique for the optimization of a multivariable, nonlinear function subject to both linear and nonlinear constraints. It has been used to solve optimization problems in many situations, but not very successfully in batch production problems. This is largely because, as with most algorithms using gradient search methods, SUMT does not cope very well with nonsmooth, discontinuous functions that are common in batch production. This paper reports on the modifications that have been made to ft class of batch production problems to permit the successful application of SUMT.