In the automotive industry, production of car bodies is one of the most capital-intensive processes. Once a production technology has been chosen, it is maintained in the long term, which causes the importance of capacity planning over 10 to 15 years. The parts of car bodies, e.g. front hoods, back doors, and roofs, are manufactured from sheet metal by press lines. These press lines form and cut the sheet using toolkits. By assembling the resulting parts, the so-called bodyshell is produced.
Task of the utilization planning is the assignment of toolkits to press lines considering technical and organizational restrictions. The great number of variants in combination with complicated contraints makes utilization planning intractable for humans. Likewise from a mathematical point of view, the problem features a high complexity: it belongs to the class of assignment problems, which in turn are proven to be NP-hard.
The project MBOpt comprehends the development of a software which computes a month-wise assignment of toolkits to press lines for the whole time horizon considering all constraints. The objective is to minimize overall costs, which comprise costs of manufacture, logistics, external procurement, and change of machine. The computed assignments feature a high and even load of machines as well as little need of expensive external production capacity. The specified restrictions assure a certain degree of continuity in the manufacturing process, e.g. a minimum period before change of machine or change of shift model.
The software is already successfully in use with our project partner BMW. It makes possible a considerable reduction of planning work and thus a higher degree of efficiency in the planning process, particularly with regard to sensitivity analysis.
The software combines longtime approved optimization methods with modern techniques based on the results of the latest research on assignment problems.
The algorithms join elements of tabu search, hillclimbing, and greedy techniques adapted to the specific problem of long-term machine allocation.
An enhancement of the software includes the early phase of product design, when different production concepts are conceivable for each stamping. Thus it is necessary to choose between different toolkits. With the aid of our software, the long-term development of capacity in the press shops is considered in the investigation of the optimal production concept.
The problem of assigning tasks to agents not only arises in connection with manufacturing processes. For example, transportation problems bear resemblance to assignment problems: given a set of warehouses offering a single product and a set of customers demanding a certain amount of the product, customers must be assigned to warehouses. The objective is to minimize transportation costs.
There are some related projects which deal with the modelling and simulation of production and logistic concepts.
Those or similar fields of research could be subjects of optimization projects at SCAI in the future.
Fraunhofer Institute for Algorithms and Scientific Computing SCAI