When DOE Wisdom software is opened for a new experiment, only two folders appear; the message log folder and the design folder. The message log folder includes any error message information that occurs during use of the software. This information is only important if you need to call technical support. The design folder opens the design definition screen. To open this screen, click the design button. The design definition screen will appear. DOE Wisdom supports 5 types of experimental objectives; screening, robust design, modeling, D- optimal, and user defined designs. For our statapult experiment, we will use a screening design.
We need to define three factors for our experiment. To add the first factor, simply choose Add in the factors section. The factor definition window appears. Our first factor is pull-back angle. Type pull-back in the name section. Tab to the unit section and type degrees. Our low level is 440 degrees. Tab to the low-level section and type 440. Our high level is 490 degrees. Tab to the high level section and type 490. We can adjust the angle in 1 degree increments. Tab to the precision section and enter 1. It is easy to adjust this factor so our ease of adjustment is easy. Click on the OK button. You now see the pull-back factor listed in the factors section of the design definition window. The same procedure is used for the other factors.
We now need to define the response for our experiment. In this case the response is distance. To add the response, simply choose add in the response section. The response definition window appears. Type distance in the name section. Tab to the units section and type inches. Tab to the Min section and enter the minimum expected value for the response. In this case we will enter 30. Our maximum expected value is 230. Click the arrows at the right of the replicates section to indicate the number of replicates desired. In this experiment, we will run three replicates. Since there is only one response in this experiment, we do not need to use the desirability function. Click on the OK button. You now see distance listed in the response section of the design definition window.
When factors and responses have been entered, DOE Wisdom displays the permissible design types for the objective selected. The recommend design type is highlighted. You can change this design type by positioning the mouse pointer over the design type and clicking. In this experiment, we will use a factorial design.
The base runs, resolution, design types and total runs sections are all interactive. If you select a different design types, the number of runs and the resolution will change. DOE Wisdom allows you to choose either standard or random order for your worksheet. Standard order creates a worksheet in the textbook order of the design. Random order creates a worksheet with all the runs in randomized order. This assists in averaging out the effects of extraneous, uncontrolled factors that may be present. Click on the order button to switch back and forth
between standard and random order. We will use standard order for this experiment. NOTE: When DOE Wisdom provides a resolution of V, it implies resolution V or higher. In our experiment, there will be no center point.the number of design runs is 8. Since we have three replicates the number of total runs is 24, or 3 times 8. Since this is a full-factorial design, there is no alias pattern to view. Once the factors, responses, design type, and order have been defined, simply choose the OK
command button to accept this design.
Three new folders will now appear in the project window; the design report folder, the worksheet folder, and the data folder.double-click on the design report folder. The DOE Wisdom design report folder will appear. This summarizes all factor, response, and design information. The next step is to view our worksheet folder. Double-click the worksheet folder. The worksheet window will appear. DOE Wisdom will create a worksheet listing the runs to perform in the order we are to perform them. We will print out a copy of this worksheet and use it during our
experimental runs. Now it is time to perform our experiment. For the first run we will set the pull-back angle to 440, the hook position to 1, and the peg to 1. We will then launch the ball and measure the distance is traveled. In this case it traveled 32.5 inches. We record this distance on the worksheet. We will launch the ball two additional times with these settings and record the distances. The distances were 36 and 38.5 inches.
The worksheet shows that the second setup should be as follows; pull-back angle 440, hook position 1, peg position 4. After setting the statapult to these settings we now launch a ball and measure the distance. In this case the ball traveled 69 inches. This is recorded on the worksheet. All additional runs are completed as outlined on the worksheet. The final worksheet is shown here.
It is now time to enter our data. Double-click the data folder in the project window. The data definition window will open. Enter the results listed on the worksheet in the appropriate columns of the data entry screen. Select file and save to save your data.
Now it is time to analyze our data. This is where the software saves us a huge amount of time and effort, as trying to analyze the data by hand would take days.
Once the data is entered, two additional folder appear in the project window; the graphs folder and the stats folder. Double-click on the stats folder. The analysis of variance screen will appear. Remember our previous discussions regarding the ANOVA? The first thing we want to review is the P2-Tail value for each term. If the P2-Tail value is greater than 0.1, we should remove the term from our model. In this case, the P2-Tail value for the BC interaction is 0.326 and the value for the ABC interaction is 0.286. We need to remove these two terms from our model. To remove the terms, select edit, and report variables. In the factor folder all highlighted terms will remain in the ANOVA. Use the mouse pointer to select the BC term and click on the left-mouse button to remove the term. Follow the same procedure to remove the ABC term. Select OK. The ANOVA now appears with both BC and ABC terms removed.
The following terms had a significant impact on the distance the ball traveled; pullback, hook, peg, the pullback-hook interaction or AB, and the pullback-peg interaction or AC. Our multiple-r value is 0.988208. This is very close to 1. Our squared multiple R value is 0.976555. 97.65% of the variance in the data was due to the fact the response varied for different levels of the factors. The F-Ratio is 149.951. This is definitely greater than 6 and indicates that the change in the response at different settings didn t happen by chance. The p value is 0.000. We can be very confident that the model detects a shift in the data.
Now let s have the software automatically tell us how to hit a target. Click on the prediction equation button. The prediction equation screen will appear. Click on the find target button. The Find target value screen will appear. Select the target button. We would like to hit a target of 120 inches. Type this number in the target screen. Click on the search button. The software searches for the factor settings necessary to hit a distance of 120 inches. In this case, the recommended settings are; pull-back 459, hook 3, peg 4.
Let s look at some of the graphs we discussed in previous modules. In the project window, double-click the graphs folder. The Pareto Chart will appear. The Pareto Chart shows vertical bars with heights proportional to the average delta over 2 values for each factor interaction. Since the ABC and BC interactions were not statistically significant, we would like to remove them from our Pareto Chart. Select edit, and graph. The edit graph tabs appear. Click on the factors tab. Remove the ABC and BC terms. The Pareto Chart now appears without these terms.
We can also edit the Pareto Bar colors and style, select edit, and graph. Click on the factor display tab. Highlight pull-back. Click on fill style and select solid. Click on fill color and select red. Click on the OK button. The Pareto Chart now appears as shown here.
Click on the scatter-plot button to view the scatter plot. It initially displays only pull-back. Let s add the other two factors. Select edit and graph, select the factors tab, and add the hook and peg factors. The graph now appears as shown here.
Click on the main-effects button. The main-effects graph appears. Notice that pull-back, hook, and peg have the greatest slope. The BC, and ABC interactions have very little slope.
Click on the interactions button.the pullback-hook interaction appears. To add the other interactions, select edit and graph. Click on the factors tab and add the other two interactions. Select OK. The graph will now appear as shown here. Click on the contour plot button. The contour plot appears. Remember the hit a target feature predicted we could hit a target of 120 inches by setting pullback to 459, hook to 3, and peg to 4. To set the peg to 4, click on the edit model button. Slide the peg value to 4. Select file and exit. The peg value will change to 4. If we set pullback to 459 and hook to 3, the contour plot shows a predicted value of 120 inches.
Click on the response surface button. The response surface graph appears. You can rotate the graph by clicking on the horizontal and vertical arrows. This allows you to see a 3-dimensional view of your model.
This concludes module 6, in this lesson we have shown how to use design of experiments software. The use of software is vital to the successful implementation of experimental design. In the next lesson we will show some examples of how others have used these powerful tools to save time and money.