The reason for this thesis’ initialisation was that Siemens Energy Trollhättan wanted to take their manufacturing production a step further into Industrie 4.0. This process was to begin with a simulated model of part of their factory being built, which could then later be turned into part of a Digital twin with an alarm functionality that could help optimise production. This thesis will focus on these first steps taken to have a Digital twin of the factory, made together with a theoretical research part regarding how this alarm functionality can be realized. A solution found by the author was sending data from the program Plant Simulation to the computational platform MATLAB, wherein statistical process control is utilised to trigger an alarm when the monitored mean goes beyond acceptable limits. The theoretical research regarding requirements of implementing real time data in a Digital twin is handled in the partnering work to this one written by Nyberg [1].

The simulation model built during this degree work in the end one included of the initially six suggested productions regarding the BK800 combustion chamber, the scope of which had to be decreased due to time constraints. Data has been analysed and structured so that further expansion of the model has been prepared and simplified for the future modeller tasked with continuing work on the thesis’ model. The model has been validated to have within five percent throughput as compared to current data and it was agreed that the model is behaving within expected limits when compared to what Siemens sees in their factory today.

Once validation was confirmed it was time to perform experiments on the simulation model. Of these experiments, one of the most interesting parameters tested was increasing how many source parts are initialised into the system at once. This initially began with increased model throughput, but once new parts rate reached a certain value this led to a decrease in model throughput. This is thought to be because of too many products in the flow leading to bottlenecks forming. Other parameters that were tested and changed were storage sizes, scheduling of workers, sick leave percentage and number of workers. The parameters with the greatest effect on model throughput were increased incoming parts and changed scheduling for workers. The other parameters had minimal effect on model throughput. One of these, sick leave percentage, had minimal effect on the modelled factory’s efficiency. This was surprising since the Siemens factory has a majority of manual labour in the factory today, which would lead to a more noticeable decrease in production if even one worker was sick in the actual factory. Sick leave’s minimal effect is however thought to be due to the model encompassing a fairly small part of the factory when these experiments were performed. A solution for implementing the requested alarm has been researched and a concept of how this would be extended is outlined in this thesis, though the implementation of the solution itself fell outside of the timeframe of this degree work.