Simulated annealing vs. genetic algorithms applied using a new cost function for the car sequencing problem

It is well known that in the automobile industry there is a need to maintain a certain order in the vehicles as they pass through the assembly line. Sequences have to be built according to each vehicle's 'options', each one requiring different resources and production time, the objective being to avoid exceeding the maximum human and facility potential. The problem resides in the complexity of ordering, at the same time, the presence or absence of each and every truly restrictive option. For this type of problem there are no efficient polynomial resolution algorithms, and heuristic methods are the most widely used. This paper uses two global heuristic optimisation methods such as simulated annealing and genetic algorithms applied to the specific problem of finding the optimum sequence for unbalanced car assembly lines. The best optimisation parameters are calculated using the experimental design method. The paper also proposes a new cost function to better represent car scheduling problem constraints. This cost function and the optimisation methods have proved their efficiency in the scheduling of real production data for a highly flexible car manufacturing assembly line (PSA Peugeot Citroen car assembly line at Vigo, Spain).