The design and scheduling of chemical batch processes: building heuristics and probabilistic analysis

The number of industrial cases published in the design and scheduling of batch multiproduct plants is short, and the difficulties to solve large models of this kind are well known, since their modeling usually consider variables integrality and data uncertainty. One way to address such difficulties is to use analytical studies to obtain significant improvements in algorithms and problem structures. Several MILP models from the open literature are selected focusing the successive generalizations on the options set, namely: from single machine to multiple parallel machines (identified by S or M) in each stage; and from single product campaigns to multiple products campaigns (S or M too). Four models (hereby SS, MS, SM, MM) that consider zero wait operations are thus analyzed and compared, and several heuristics are developed in order to produce good approximations to the objective function’s value and to the binary solution. Then, the probabilistic analysis of the heuristics was performed: the deviations on the objective function values, the deviations on the binary solutions, and the computational times are evaluated. The analysis both allowed the certification of the modeling and the numerical implementation. The model MS, addressing multiple machines and SPC, is found to be the most promising model to further developments that aim the design and scheduling of batch processes in stochastic and robust frameworks.