Related Topics: | ||
This topic describes how to use the ALTA standard folio control panel to perform accelerated life testing data analysis. It includes the following sections:
Accelerated life testing analysis involves careful planning and requires many tasks. To make things easy to find, the ALTA standard folio control panel is arranged into groups of related tasks. Each group has its own page on the control panel.
Tip: There is a horizontal splitter bar directly above the Main button. If you drag it as far up as it will go, all of the pages of the panel will be accessed by large buttons. If you drag it all the way down, all of the pages will be accessed by small icons. Positions in between allow you to use some large buttons and some small icons. See Control Panels.
The Main page is the most frequently used page because it contains most of the tools you would need to analyze a data set.
The Model drop-down list is used to select the life-stress relationship and life distribution that will be used in your analysis.
Directly underneath the drop-down list are options for selecting the stress columns and (when applicable) the stress transformations that will be used in the analysis.
The Analysis Settings area displays "MLE" and "FM" to indicate that the maximum likelihood estimation (MLE) method will be used to estimate your model’s parameters and the Fisher matrix (FM) method will be used to solve for the confidence bounds. In ALTA, these settings cannot be changed. This area also displays the number of failures and suspensions in your data set (e.g., "F=17/S=23" means there are 17 failures and 23 suspensions).
Click Set Use Stress directly underneath the Analysis Settings area to define the normal stress level for every stress that is selected to be used in the analysis.
The Analysis Summary area displays the calculated parameter values and other values for your selected model.
The folio tools are arranged on the left side of the panel. Use these tools to manage data and experiment with the results of your analysis. Click a tool's icon in the picture above to learn more about it.
Calculate estimates the parameters of the chosen model, based on the current data set and the specified analysis settings. This option is also available by choosing ALTA > Analysis > Calculate.
Plot creates a new sheet in the folio that provides a choice of applicable plot types. For standard folios, this includes plots such as Use Level Probability, Reliability vs. Time, Life vs. Stress, etc. You can also create a Plot sheet by choosing ALTA > Analysis > Plot.
QCP opens the Quick Calculation Pad, which allows you to calculate results, such as the mean life and the probability of failure, based on the currently active data sheet. This tool is also available by choosing ALTA > Analysis > Quick Calculation Pad.
Distribution Wizard opens the Distribution Wizard, which helps you select the life distribution that will best fit your data. This tool is also available by choosing ALTA > Analysis > Distribution Wizard.
Alter Data Type opens the Data Sheet Setup window, which allows you to change the data entry columns in the current data sheet. Note that when you change the setup, you may lose information already entered in the data sheet. For example, switching the data sheet from one that supports interval data to one that supports only right censored data will remove all the inspection times. This tool is also available by choosing ALTA > Format and View > Alter Data Type.
Add or Remove Columns allows you to add/remove columns for stress levels and unit IDs.
Change Units opens the Change Units window, which allows you to change the units for the time values in the current data sheet.
Auto Group Data opens the Auto Group Data window, which allows you to group together data points with the same failure/suspension times. Grouping data significantly speeds up calculations when you have a very large data set. You can also group data by choosing ALTA Data > Options > Auto Group Data.
Alter Parameters allows you to alter the values of the calculated parameters for what-if analysis. In probability plots, the position of the data points remain the same but the solution line and all subsequent analyses made via the QCP and other tools are based on the modified parameter values. Therefore, the analysis may be appropriate only within the context of your specific what-if scenario. This tool is also available by choosing ALTA > Options > Alter Parameters > Alter Parameters (w/o Recalculation).
Likelihood Ratio Test opens the Likelihood Ratio Test tool that allows you to assess the assumption of a common shape parameter across all the stress levels used in the test. This assumption underlies the use of any model that includes the Weibull or lognormal life distribution. You can also open this tool by choosing ALTA > Options > Likelihood Ratio Test.
Transfer Data to Weibull++ allows you to transfer data from the ALTA standard folio to a Weibull++ standard folio.
Accelerated life testing data analysis makes use of life distributions and life-stress relationships to extrapolate failure behavior at normal operating conditions from data obtained at accelerated stress levels. The selected life distribution is fitted to the data obtained at each accelerated stress level used in the test. The selected life-stress relationship is then used to extrapolate the probability density function (pdf) at the use stress level from the pdfs at the accelerated levels.
While any distribution can be used to analyze a data set, if the distribution does not fit the data well, your results may contain too much uncertainty. Moreover, each of the available life-stress relationships is designed to work with specific stress types (e.g., the Arrhenius relationship is designed for data obtained from accelerated temperatures). Therefore, to ensure accurate predictions for the product's use level reliability, the appropriate life distribution and life-stress relationship must be selected.
What’s Changed? In previous versions of ALTA, the life-stress relationship and life distribution were selected separately, in two different drop-down lists. Now, a single drop-down list is used to select a single model that combines both.
The first step is to select a life-stress relationship. Click the Model drop-down list on the Main page of the control panel. There are five life-stress relationships to select from in ALTA Standard, and there are four additional relationships in ALTA PRO:
The five life-stress relationships in ALTA Standard are:
Arrhenius: a single-stress model typically used when temperature is the accelerated stress.
Eyring: a single-stress model typically used when temperature or humidity is the accelerated stress.
Inverse power law (IPL): a single-stress model typically used with a non-thermal stress, such as vibration, voltage or temperature cycling.
Temperature-humidity (TH): a double-Arrhenius model that is typically used when temperature and humidity are the acceleration variables.
Temperature-nonthermal (TNT): a combination of the Arrhenius and IPL relationships that is typically used when one stress is temperature and the other is non-thermal (e.g., voltage).
In addition, ALTA PRO offers four other models for more advanced applications:
Generalized Eyring (GER): a variation of the Eyring relationship that is typically used when both temperature and humidity are accelerated.
Proportional hazards (PPH): analyzes data with up to 8 simultaneous stress types using the exponential relationship for each stress. It also allows the use of zero as a stress value, which enables the analysis of data with indicator variables (e.g., 0 = on/off and 1 = continuous operation).
General log-linear (GLL): supports the analysis of data with up to 8 simultaneous stress types and provides the flexibility of specifying the life-stress relationship (Arrhenius, IPL or exponential) for each stress. Like the PPH model, this model also allows for the analysis of data with indicator variables.
Cumulative damage (CD): analyzes data with up to 8 stress types and/or situations where the stress varies with time.
The ReliaWiki resource portal has more information on these relationships at: http://www.ReliaWiki.org/index.php/Accelerated_Life_Testing_Data_Analysis_Reference.
After choosing a life-stress relationship, choose a life distribution in the submenu. For example, if you select Arrhenius-Weibull from the submenu, your analysis will use a combination of the Arrhenius life-stress relationship and the Weibull life distribution.
The three available life distributions are 2-parameter Weibull, lognormal (not available with the proportional hazards relationship) and 1-parameter exponential. The ReliaWiki resources portal has more information on these distributions at: http://www.ReliaWiki.org/index.php/Distributions_Used_in_Accelerated_Testing.
Tip: If you are not certain which life distribution would provide the best fit for your data, you can use the Distribution Wizard to guide you through the selection process. The choice of the life-stress relationship depends on various factors, including the types and number of stresses used in the test.
The next step is to estimate the parameters of your chosen model, as described in the next section.
You can calculate the parameters by choosing ALTA > Analysis > Calculate or by clicking the icon on the Main page of the control panel.
The control panel will display the calculated values of the parameters in the Analysis Summary area, as shown next.
Note: If the model and parameters are associated with a published model, then the published model's name will appear as a link at the bottom of the Analysis Summary area. Click the link to view the model's properties. For details on how to publish the results as a model, see Control Panel Publishing page.
Depending on your settings, the following information may also be displayed:
The Ea (activation energy) value represents the magnitude of the effect that the applied stress will have on the product. A large activation energy indicates that the applied stress has a large effect on the life of the product.
The Scale Parameter value represents the product's characteristic life.
For the Weibull distribution, the characteristic life is equal to the value of the eta parameter (i.e., the time at which unreliability = 63.2%).
For the lognormal distribution, it is equal to Exp(Log-mean) (i.e., the time at which unreliability = 50%).
For the exponential distribution, it is equal to the mean life.
The LK Value (likelihood function value) is the peak value of the likelihood function, at the estimated parameter values.
Click anywhere within the Analysis Summary area to open the Results window, which displays the calculated results in a worksheet. From the Results window, you can edit, copy or print the results.
Accelerated life testing results are estimates based on observed data points from a limited sample size and obtained at increased stress levels. For this reason, there is always some uncertainty in the results. It is therefore important to evaluate how well the chosen model fits the data set. There are several ways to evaluate the model’s fit and compare it to other models. To make a sound assessment, the following tools may be used in conjunction with your own engineering knowledge about the model:
The Use Level Probability and Standardized Residuals plots show you how well the solution line tracks the plotted data points. This allows you to visually assess the fit of the selected model.
The Likelihood Ratio Test is used to confirm the assumption of a common shape parameter among the data obtained at the various accelerated stress levels. This assumption underlies the use of any model that includes the Weibull or lognormal life distribution.
Confidence Bounds allow you to quantify the amount of uncertainty in the parameter estimates and other functions, such as reliability predictions, due to sampling error. The narrower the confidence bounds at a given confidence level, the closer the estimates are to the true value of the parameters and the better the precision on the reliability predictions.
You can visualize the confidence bounds by using plots, or you can view the calculated values of the confidence bounds by using the Quick Calculation Pad.
The LK Value (likelihood function value) is the peak value of the likelihood function, at the estimated parameter values. It is influenced by the sample size and the variability of the data set. This means that the LK value varies across different data sets and cannot be used by itself to make a judgment about the fit of the model. However, this value becomes useful when it is used to compare how different models fit a particular data set. For a given set of data, the model with the highest LK value is considered to have the best fit statistically.
The LK value for the selected model is automatically displayed on the Main page of the control panel.
To view a 3D plot of the likelihood surface, choose ALTA > Analysis > 3D Plot, and then choose Likelihood Function from the Plot Type drop-down list.
ALTA's Distribution Wizard uses LK values to automatically compare life distributions. The Wizard assumes the life-stress relationship selected on the control panel.
© 1992-2013. ReliaSoft Corporation. ALL RIGHTS RESERVED.