Using Design of Experiments to Minimize Variation – Case Study (Part 4)

By R. Michael Donovan (www.rmdonovan.com)

Result

.

Finally, a confirmation run was needed to verify that these low variation results would be repeated at the optimum process conditions. When these process conditions were implemented as part of the overall system upgrade, a stage time standard deviation near 0.5 was achieved (see Figure 4). Since the goal was reach a standard deviation of 1 or less, they had met their goal.

Conclusion

Using Minitab to design and analyze the experiment, Böwe Bell & Howell successfully identified the important variables and their corresponding settings to minimize the stage time variation. These findings were incorporated into a larger-scale project to redesign the full Böwe Bell & Howell equipment system. With only a small amount of time spent designing, performing and analyzing their experiment, the knowledge gained enabled them to reach their process improvement goal.

Using Design of Experiments to Minimize Variation – Case Study (Part 3)

By R. Michael Donovan (www.rmdonovan.com)

Analysis

After conducting the 16 runs of the experiment, they were ready to run their analysis. The response was the variation (standard deviation) of 500 staging time readings for each run. Because standard deviations are typically not normally distributed, they analyzed the results using the log of the stage time standard deviations. Minitab’s normal effects plot of the (log of the) standard deviations was used to determine which factors and interactions were significant (identified by the red symbols in Figure 2).

All four experimental variables had an effect on stage time variation, either through a main effect or an interaction. Once the significant factors were identified, additional graphs were used to determine the optimum settings. These graphs included a cube plot of the average standard deviations for all eight combinations of J-Plate Position, Stack Height, and Singulator Spacing. Keeping in mind that Stack Height cannot be controlled at a fixed level during true production, they needed to find settings for the other factors that minimized the stage time variation, regardless of the Stack Height. As shown in Figure 3, the cube plot indicated that the stage time standard deviation was minimized in a range of about 0.65 to 1.44 when the J-Plate was in the far position and the Singulator Spacing was set low, even when the Stack Height varied.

Using Design of Experiments to Minimize Variation – Case Study (Part 2)

By R. Michael Donovan (www.rmdonovan.com)

Experiment

The goal of the experiment was to identify the variables and corresponding settings that minimize variation in the staging time. Prior experiments indicated that J-Plate Height, J-Plate Position, Singulator Spacing, and Stack Height were important. Therefore, they were included in this experiment.

The experiment was conducted as follows:

1. High and low settings representative of a reasonable operating range for the four experimental variables were selected (see Table 1).

2. Using Minitab, a two-level full factorial design for the four variables was generated (see the first four columns in Table 2).

3. 500 readings at each of the 16 run conditions were performed.

4. The mean and standard deviation of the stage times were calculated for each of the 16 run conditions (see the last two columns in Table 2).

To protect the proprietary nature of the process and experiment results, the variable settings and units on the staging time results have been changed.

Table 1. Variable Settings

Factor	Low Setting	High Setting
J-Plate Height	Low	High
J-Plate Position	Near	Far
Singulator	Low	High
Stack Height	3 cm	12 cm

Table 2. Design of the Experiment

.

J-Plate Height	J-Plate Position	Singulator	Stack H Height	Stage Mean	Stage StdDev
Low	Far	High	12	30.9	1.9
Low	Near	High	3	29.9	1.1
Low	Far	Low	3	35.7	0.4
High	Far	Low	3	36.1	0.9
Low	Near	High	12	31.3	2.1
High	Near	Low	12	33.9	3.1
Low	Near	Low	12	30.9	2.8
High	Far	Low	12	33.4	1.5
High	Near	High	12	30.8	4.5
High	Near	High	3	30.4	3.1
High	Far	High	3	29.8	3.9
Low	Far	High	3	29.2	2.8
High	Far	High	12	29.5	3.1
Low	Far	Low	12	32.7	1.4
Low	Near	Low	3	31.6	0.6
High	Near	Low	3	33.2	1.3