Otimização de produção de uma linha de montagem mista na indústria automotiva por meio de programação matemática

Abstract

This work uses Mixed Integer Linear Programming (MILP) to investigate productivity differences found in a Mixed-Model Assembly Line. The solution approach was based on construction and resolution of mathematical models for the Mixed-Model Assembly Line Balancing Problem (MALBP), followed by the Car Sequencing Problem (CSP). A discrete simulation was used to check the results obtained by the mathematical models. The study reports the real case of an automotive metal line that assemblies the closures of three different vehicles, sharing the same production line. The production line is characterized as low cadence (e.g., up to 15 vehicles/hour) and fully manual. The line is formed by roller tables in series with unpaced devices. The main objective is to investigate how to maintain productivity, while maintaining the flexibility characteristic to meet a variable demand in volume and product diversity. Results of this study are presented and indicate that some flexible balancing mixes are viable. Production time differences of no more than 6% were observed in such flexible choices. However, the checking of some possible combinations for the six mixes of evaluated products showed a production time difference up to 19%, depending on the sequence of products for a certain mix. Another factor that also influenced production time differences, 20% on average, was the allocation of buffers. The horizon of analysis encompasses a lot of production of one hour, to the mathematical model, and a week for discrete simulation. The obtained results indicate that it is possible to operate a flexible assembly line with a level of productivity similar to a single product line, in case of careful choices in balancing/sequencing. On the other hand, if the operating conditions of balancing and sequencing were not taken into account the cumulative loss in throughput may be significant (e.g., for some studied cases, production time exceed 30% the theoretical value).