Effects of the Breakdown of a Mixer

During production a breakdown of a mixer may occur. This can result in not being able to use the mixer for an entire day. When this happens, a lot of waiting time may occur in the current situation. We experiment how our planning cycle would perform in such a situation. To run this experiment, we take a production day where this

breakdown occurred and compare this situation with the results from our experiment. The day we simulate is the 25th _{of March 2016. Mixer 5 was broken down on this day. Press 1, 2, and 5 were operational and producing tiles} with a height of 60 mm, 80 mm, and 50 mm respectively. We simulate this day using our planning cycle solution in order to determine whether our solution would lower the waiting time. We use the same mixing orders that are used on that day.

In the current situation the total waiting time for top- and bottom-layer was more than 2 hours. Using our planning cycle solution, no improvement is made in the waiting time. In order to determine the cause of the waiting time, we use a Gantt chart to visualise the resulting schedule from a planning cycle iteration (see Figure 5.4). The Gantt chart shows the scheduled window of processing each job for each resource. Mixer 5 has no jobs scheduled as the mixer is broken down. However, Mixer 4 is continuously processing jobs as it is the only mixer available for creating bottom-layer mixtures. Therefore, the capacity of the mixer is, in this case, the major cause for waiting time of the bottom-layer.

Figure 5.4: Gantt Chart of One Planning Cycle Iteration With the Breakdown of Mixer 5

SVI stated it may be possible to reduce the processing time of a job on the mixer. Therefore, it may be beneficent for SVI to perform a thorough analysis on the mixing process in order to reduce the processing time. In order to determine the effect of reducing the processing time, we perform two experiments. The current processing time of a job on Mixer 4 is 160 seconds. We perform experiments with a processing time of 145 and 130 seconds in order to analyse the effect of the processing time on the waiting time for material. Table 5.14 contains the results for the reduced processing time. This shows that by reducing the processing time on Mixer 4 by 15 seconds, the waiting time for raw materials at the press is reduced by 95.62%. By reducing the processing time even further to 130 seconds, the waiting time of almost 3 hours is reduced to just 5 seconds. This shows that reducing the processing time of Mixer 4 would largely reduce the waiting time for this scenario.

Table 5.14: Total Waiting Time with Reduced Processing Times for Mixer 4

Processing Time Mixer 4 Waiting Time (sec) Reduction (%)

160 sec. (current situation) 10,672 -

145 sec. 467 95.62%

130 sec. 5 99.95%

5.4

Conclusions

In this chapter we performed multiple experiments in order to assess the performance of our planning cycle solution. First, we obtained the optimal parameters for the planning cycle algorithm using a 2k _{factorial design.} This resulted in a planning horizon of 26.5 minutes, a planning interval of 100 seconds, and a fixed zone of 250 seconds. We used the optimal parameters for assessing the performance of the new planning cycle solution. We assessed the performance using 3 experiments.

In the first experiment, we compared the planning cycle with the current situation by running experiments that represent the production days at SVI for the year 2015. This resulted in a savings of 128.4 hours of waiting time and € xx,xxx.xx in cost compared to the year 2015 excluding Press 3 and a savings of 137.5 hours of waiting time and € xx,xxx.xx in cost including an estimation of Press 3. The results from the planning cycle solution do not include the waiting time for breakdowns, while these are included in the current situation. Therefore, the actual savings would turn out lower. In order to determine the actual savings, a distinction should be made in the future in the measurement between the different causes of waiting time.

In the second experiment, we researched the effect of different mixer assignment and tile height setup for each press on the performance of the planning cycle solution. This resulted in a list of unfavorable combinations of mixer assignments given the tile height setup for a day (see Appendix C for an example of this list). This list can support the planning of production orders by the management for a day as it shows possible unfavorable order combinations. Finally, we concluded that Press 5 should never be combined with Press 1 or Press 2 on the same bottom-layer mixer when both are producing tiles with a tile height of 80 mm.

In the third experiment, we experimented with the breakdown of a mixer (Mixer 5). The breakdown of the mixer showed the planning of jobs was not the cause for the waiting time on the day Mixer 5 was broken down. The visual representation showed the capacity of Mixer 4 was the bottleneck. We experimented with the reduction of processing time of this mixer and concluded that the waiting time on this day could be reduced by 95.62% by reducing the processing time on Mixer 4 from 160 seconds to 145 seconds.