Using the ALT Design Platform

1.	To launch the ALT Design platform, select DOE > Accelerated Life Test Design.

Initial ALT Design Window

2.	Select one or two accelerating factors.

For two factors, you can choose to include the interaction between the factors.

3.	Select either Monitoring at Intervals or Continuous Monitoring.

Choose Monitoring at Intervals if you can estimate when the experiment will fail. Enter the number of inspections, the time of the first inspection, and the time between inspections.

Choose Continuous Monitoring if you are unsure of when the experiment will fail.

4.	Click Continue.

A window appears for specifying details of the accelerating factor or factors.

Accelerating Factor Details Window

5.	Fill in these options:

Factor Name

Enter a name for the accelerating factor.

Number of Levels

Enter the number of levels of the factor to include in the experiment.

Factor Transformation

Select a transformation for the factor. The options are Arrhenius Celsius, Reciprocal, Log, Square Root, and Linear.

Low Usage Condition

Enter a value for the low usage conditions.

High Usage Conditions

Enter a value for the high usage conditions.

6.	Click Continue.

A window appears for specifying additional information about the assumed distribution and desired experimental conditions.

Distribution Details

7.	Fill in these options:

Factors

Enter the values of the accelerating factor levels.

Distribution Choice

Select the failure distribution, either Weibull or LogNormal.

Prior Mean

Enter the acceleration model parameters. These values can be either a best guess or the estimates of the parameters from an analysis of previously obtained data.

Prior Variance Matrix

Enter variances and covariances for the acceleration model parameters. These values can be either a best guess or the estimated variance matrix from an analysis of previously obtained data.

Ignore prior variance

Select this option to ignore variances and covariances for model parameters. When the variances and covariances are ignored, the design is created for the specific fixed parameters entered under Prior Mean. The resulting design is said to be a locally optimal design. This design will be good if the prior mean parameters are close to the true values. However, the design will not be robust to mis-specified parameters. When the variances and covariances are used, a multivariate normal prior distribution is assumed for the acceleration model parameters. This is useful when the values entered under Prior Mean are estimates or guesses, and you want the design to reflect the associated uncertainty.

Diagnostic Choices

Enter values for the following:

Time range of interest are values for which you want an estimated probability of failure. For example, if you are interested in the probability of failure by 100,000 hours, then enter 100,000 for both the lower and upper ranges.

Probability of interest is the value for which you want an estimated time of failure. For example, if you are most interested in obtaining the time until 10% of units fail, then enter 0.1.

Design Choices

Enter values for the following:

Length of Test is the length of time to run the experiment.

Number of Units Under Test is the number of units in the experiment. If augmenting a previous experiment, enter the number of units from the previous experiment plus the number of units that you want to run for the next experiment. If designing an initial experiment, enter the number of units that you want to run.

8.	Click Continue.

New outline nodes appear as shown in Additional Outline Nodes.

Additional Outline Nodes

Candidate Runs

Enter the minimum and maximum number of runs allowed at each level of the acceleration factor. If augmenting a previous experiment, enter the number of units already run at each level for the Minimum Units.

Parameter Variance for Balanced Design

Gives the expected variances and covariances for the acceleration model parameters after running and analyzing the balanced design (same number of runs at each level of the acceleration factor). These values are valid under the assumption that the values for the prior mean and variance are correct.

These values can be compared to the ones entered under Prior Variance Matrix, to see whether the balanced design can improve the variances of the parameter estimates. They can also be compared to the final parameter variances for the optimal design, after clicking Make Design.

Distribution Profiler

Use the profiler to visualize the probability that a unit will fail at different values of the acceleration factor and time.

9.	Click Update Profiler to update the profiler if changes are made to the distribution choice, means, variances, design choices, or candidate runs.

10.	Click Make Design to create the optimal design and display the results.

Design Results

The information below describes the results you get after clicking Make Design.

The Design report gives the expected number of failures for each level of the acceleration factor. Also given is the probability that none of the units at this setting will fail.

The Parameter Variance for Optimal Design report gives the variances and covariances for the acceleration model parameters for the optimal design. These values are valid under the assumption that the values for the prior mean and variance are correct. These values can be compared to those under Parameter Variance for Balanced Design to determine whether the optimal design is able to reduce the parameter variances more than the balanced design.

The Optimality Criteria report gives the values of the optimality criterion for the optimal design. For more information about the optimality criterion, see Platform Options.

The Make Design button updates the optimal design if any changes are made to the distribution choice, prior means or variances, design choices, or candidate runs.

The Make Test Plan button creates a data table with the acceleration factor levels and the number of units to include in the experiment for each level.

The Make Table button creates a table that can be used for data collection during the experiment.