A European Automotive company is producing two different models and applies small trains to load its assembly lines with components stored on pallets. These trains are filled at a central location in the plant, after which the pallets are transported to the assembly lines. The trains are filled in sequence based on the production sequence or order planning, the BOM and a load algorithm, which calculates when the pallets are empty and due to be replaced.
We were requested to reduce logistical traffic of these trains to and from the assembly lines. This had to be achieved within the existing IT structure. Key condition was to reduce the number of trains and keep the load per train as close as possible to the maximum of 4 pallets.
We started with an analysis of the existing process by dissecting and understanding the load algorithm. This analysis resulted in a number of parameters with which we could influence the process. It appeared that almost all parameters were set at their default values. One parameter was set to always fully load the train with 4 pallets independent of their assembly line destination.
The load algorithm works as follows: it finds the most urgent (first empty) pallet and completes the train with 3 more pallets due to be brought to the line. Pallet content and number of pallets at the line determine how well the load algorithm can replenish the line in time.
By dividing the assembly lines into zones and by balancing between them, pallets were classified in clusters. Trains were now directed to zones of the line, resulting in large reduction of distance and time travelled per train. Less trains could now bring the same amount of pallets to the line.
The analysis has revealed the complexity of the present load algorithm. As an exercise we have written a new load algorithm in the programming language Python to see if we could simplify and improve it. The advantages of this newly created algorithm was that we achieved better overview and now could anticipate loads. As a result the trains could be loaded more efficiently. In short we created a better dashboard with clear KPI’s and a much easier maintainable and understandable load algorithm.