Inventory simulation and optimization using system dynamics, structural modeling equations and genetic algorithms in the drivetrain division of an automotive manufacturer

Strategic planning and control are among the most critical activities that modern enterprises require to succeed in the global economy. This research is an original study that investigated the combination of tools and methodologies in order to apply them to a midwestern tractor manufacturer. The current study identified the constraints applicable to a polishing line in the Drivetrain Division of a major tractor manufacturer interested in exploring alternative techniques to improve its worldwide manufacturing operations. The specific questions that this project tried to respond are stated as follows: What were the most important variables that affected inventory levels of an assembly line of an automotive manufacturer? What were the significant effects of the causal relationships identified in order to determine an initial model structure? What constrains restrict the behavior and improvement of the selected variables? What levels of the selected variables could be used in order to improve production levels? The current research explored the impact of a series of variables (work-in process, process utilization, cycle time, queue size, utilization of work centers, capacity, and others) in order to examine their impact in the overall performance of the polishing line. Two main models were developed based on two algorithms created for each of the selected part families (PTO and Covers), and in combination determined material flow, resource utilization, and sequencing within and outside the automatic polishing line. The two computer models combined both dynamic and discrete simulation to establish a reference to be used for improvement of similar processes within the company using structural equations modeling, path analysis, scatter plot diagrams, and eigen value plot. Besides, the results of this research indicated that: (a) cycle time can be improved with the addition of a new transporter in order to reduce the moving time within and between work centers; (b) the queue sizes of the polishing line were not improved significantly using either genetic algorithms (GA) and full factorial designs because of the low initial variability of the system; (c) the structural modeling equations model allowed to identify possible material flow errors based on its relationships, in this way it is possible to have a benchmark to compare both the results of the current study and the outcomes of similar studies developed by the company. In summary, a new methodology has been developed in order to study and optimize manufacturing systems, and avoid cost reductions without any statistical significance that might affect the strategic position of the company in the long run. The current study did not give a simple answer to the complexity of the discussed problem, but an alternative to many of the current academic and industrial solutions that can have more than one correct answer