טכניון מכון טכנולוגי לישראל
הטכניון מכון טכנולוגי לישראל - בית הספר ללימודי מוסמכים  
M.Sc Thesis
M.Sc StudentShay Hatzbany
SubjectParallel Mixed-Model Assembly Lines for Efficiency
Improvement
DepartmentDepartment of Industrial Engineering and Management
Supervisors Professor Gilad Issachar
Professor Emeritus Shpitalni Moshe


Abstract

Serial assembly lines are used to assemble different modules in a product or to execute its final assembly when a high production rate is needed. A factory that produces a number of similar products or models can simultaneously assemble them on a Mixed-Model assembly line with relative proportion to their demand. Although Mixed-Model assembly lines have a number of advantages over Single-Model lines, such as reduced inventories, lowered model transfer costs and improved flexibility in meeting ever-changing customer demands, they suffer from several drawbacks, such as increased assembly complexity and greater work flow fluctuations, caused mainly due to the differences between the models assembled in the line .

We used a new approach to cope with those disadvantages; accordingly models are clustered into several assembly lines, meaning each line assembles different models. Our research presents a Model Partitioning and Clustering Algorithm (MPCA), which determines similarity between models based on their characteristics and assigns them to different parallel assembly lines using a classification process. The algorithm is completed with a Genetic Algorithm, which solves the Mixed-Model Assembly Line Balancing problem for each parallel line, according to the partition established.

An empirical analysis conducted shows that MPCA finds a better solution than the optimal solution for one Mixed-Model line, which assembles all the models simultaneously. Meaning, MPCA provides a platform to gain better solutions than any algorithm, optimal or heuristic, for balancing Mixed-Model assembly lines. Moreover, further analysis of the algorithm shows that MPCA completely outperforms Lehman's algorithm, the only known algorithm to deal with this type of problems, completely .

In conclusion, one can think of parallel lines as an organizational approach for planning and designing the assembly system. Other organizational approaches are Mixed-Model configuration, Single-Model configuration and Batch configuration. While the last three provide a rigid planning methodology, parallel lines offer a flexible one. This methodology examines different configurations of the assembly system, each combines Mixed-Model lines and Single-Model lines in diverse forms, and so enjoying the advantages of all methods, such as lowered costs, high efficiency, reduced inventories and flexibility in meeting ever-changing customer demands.