COMBINATION GENETIC/TABU SEARCH ALGORITHM FOR HYBRID FLOWSHOPS OPTIMIZATION

The paper describes an algorithm for solving the scheduling problem of a hybrid o wshop (o wshop with multiple processors, FSMP). The algorithm is a combination of genetic algorithm and tabu search, and the batch processes it is applied on are modeled by a model of time requirements. The paper describes the algorithm and compares its performance with other optimization techniques. 1. Introduction. The highest portion of production of chemical commodities comes from continuous processes, but batch processes also have their place in the chemical industry. Batch processes are often found in low-volume manufacture of products such as pure chemicals, specialty chemicals, pigments, pharmaceuticals, etc. The advantages of batch processes include high exibilit y that allows to quickly re- act to changes in demands, to change technology based on current situation, and to quickly introduce new or modied products. Batch processes also allow easier shar- ing of some of the resources (e.g. production units, storage capacities, manpower). Because a new product is likely to be manufactured on existing equipment, the empha- sis is on process planning and control rather than on sizing the equipment. Current trends in abovementioned industries are towards products with shorter life cycles and higher functionality that are tailored to specic market niches, and consequently pro- cess development problems are encountered very frequently. The development of new or modied products \from scratch" would be too costly and time-consuming, and so would be a purchase of new equipment, and for these reasons new products utilize existing equipment. This means that the problem of production scheduling in these plants is one of high importance. Two categories of chemical batch plants are widely recognized: multi-product plants and multi-purpose plants. In a basic serial multi-product plant, called o wshop, the production line consists of a single set of m processing stages, the plant has only one path for all products, and this path consists of a chain of stages where no branches or loops exist. Hybrid o wshops can be derived from the classical multistage o wshop, each stage being composed of one or more identical parallel machines (see Fig. 1.1). Each machine is able to process one job at a time, and at least one stage consists of more than one machine. This paper addresses the problem of nding optimum schedule for a set of n jobs on such a conguration, and applies combination genetic/tabu search algorithm to solve the problem. The objective function is the makespan of a schedule.