In this session, we will revisit Chapter 20 and continue where we left off, trying to improve the process.
The findings in Section 20.4.6 were not sufficient to explain all defects, but we found a correlation between elevated pressure during the epoxy application step and excessive epoxy application (NOK_1). Regarding the uneven distribution of applied epoxy (NOK_4), the operators provided the information that they suspect that this is caused by an insufficiently cleaned nozzle. The cleaning process happens at the end of the previous processing step, where the nozzle moves away from the product and blows out residual epoxy.
Exercise 22.1 (Investigating the blow-out step)
Investigate the pressure during the blow-out step (process step index 3 of the previous iteration) and see if you can find a correlation to NOK_4 defects.
Having more information about NOK_1 and NOK_4, we can now try to fine-tune the machine’s parameters to reduce the number of defects.
Exercise 22.2 (Fine-tuning the machine parameters)
The accompanying readme.md file shipped with the industrial_datascience_sim_machine_knifes repository contains information on how to run the machine again with different parameters. Try to improve the parameters to reduce the number of defects. Keep the night shift and NUM_PIECES at the default values.
You experience new issues or want to further fine-tune the parameters? Feel free to have a look at the new csv files that the machine is producing.
Hint: The process is stochastic, so you can not expect to get rid of every single defect.
How do the results of the new parameters compare to the original ones?
The original dataset had 5000 pieces with the following distribution:
| OK |
4628 |
92.56% |
| NOK_1 |
70 |
1.40% |
| NOK_2 |
1 |
0.02% |
| NOK_3 |
5 |
0.10% |
| NOK_4 |
296 |
5.92% |
Great job? Everything looks good on the night shift?
Exercise 22.3 (Try the machine on day shift)
Run the machine again with your parameters, but this time during day shift.
How do the results of the new parameters compare to the ones from the night shift?
Investigate the root cause of this difference by looking at the produced csv files.
Exercise 22.4 (Fine-tune the parameters to work well for both shifts)
The machine is not only telling you a defect summary, but also the duration it took to produce the parts. Try to fine-tune the parameters to achieve a good compromise between quality (few defects) and speed (short duration).
Hints:
- The machine’s temperature depends on the ambient temperature, the processing speed and the pressure settings.
- The duration it takes to produce knifes is determined by the processing speed.