Lot streaming and batch scheduling : splitting and grouping jobs to improve production efficiency
This thesis deals with issues arising in manufacturing, in particular related to production efficiency. Lot streaming refers to the process of splitting jobs to move production through several stages as quickly as possible, whereas batch scheduling refers to the process of grouping jobs to improve the use of resources and customer satisfaction. We use a network representation and critical path approach to analyse the lot streaming problem of finding optimal sublot sizes and a job sequence in a two-machine flow shop with transportation and setup times. We introduce a model where the number of sublots for each job is not predetermined, presenting an algorithm to assign a new sublot efficiently, and discuss a heuristic to assign a fixed number of sublots between jobs. A model with several identical jobs in an multiple machine flow shop is analysed through a dominant machine approach to find optimal sublot sizes for jobs. For batch scheduling, we tackle the NP-hard problem of scheduling jobs on a batching machine with restricted batch size to minimise the maximum lateness. We design a branch and bound algorithm, and develop local search heuristics for the problem. Different neighbourhoods are compared, one of which is an exponential sized neighbourhood that can be searched in polynomial time. We develop dynamic programming algorithms to obtain lower bounds and explore neighbourhoods efficiently. The performance of the branch and bound algorithm and the local search heuristics is assessed and supported by extensive computational tests.