The Accelerated Life Testing is used to model failure times for products with high reliability. The product is placed under high stress conditions, called accelerating variable, to accelerate the failure process. The failure times observed can then be used to estimate the failure times at normal conditions. Accelerated test can help to save time and money to study a product that will take a long time to reach failure at normal conditions.
You can use ProcessMA Accelerated Life Testing to model uncensored/right censored data with one of the seven distributions (Weibull, Exponential, Normal, Lognormal, Logistic, Loglogistic, Smallest Extreme Value). In order to perform accelerated tests, you need to know the relationship between the accelerating variable and failure times. You can choose one of the following relationships:
ProcessMA uses maximum likelihood estimates to model parameters.
At the Excel Menu (For Excel 2007, go to Add-ins first)
Choose ProcessMA > Reliability and Survival > Accelerated Life Testing
In Variable, select the column containing the data (Numeric)
In Frequency, select the column containing frequency counts (Optional, Positive integer)
In Assumed Distribution, choose the distribution to fit
In Accelerating Variable, select the column containing accelerating variable (Numeric)
In Relationship, choose a relationship for the accelerating variable
Under Censor Tab, in Censor, select the column containing censoring indicator (Optional, Exactly 2 distinct values - censored and uncensored)
Under Censor Tab, in Censor Value, choose the indicator you use to indicate censoring
Under Result Tab, in New Predictor Values, enter a new predictor value of the accelerating variable, usually the design value or normal running conditions (Optional, Numeric)
Under Result Tab, in Est. Percentiles for Percent, enter the percent values you want to calculate percentiles. Use this to calculate the time for x% of units to fail. (Optional, Numeric, >0 & <100, Separate multiple entries with commas)
Under Result Tab, in Est. Survival Prob for Times, enter the time value you want to calculate survival probabilities. Use this to calculate the probability that the product will survive past a specific period of time. (Optional, Numeric, >0, Separate multiple entries with commas)
Under Result Tab, in Confidence Level, enter the desired confidence level (Numeric, >0 & <1)
Under Graph Tab, check Relation Plot, if you want to display the relationship between the accelerating variable and failure time
Under Graph Tab, check Probability Plot (Individual), if you want to display a probability plot for each level of the accelerating variable based on individual fits
Under Graph Tab, check Probability Plot (Fitted), if you want to display a probability plot for each level of the accelerating variable based on the fitted model
Under Graph Tab, check Standardized Residuals, if you want to display a probability plot for standardized residuals
Under Graph Tab, check Cox-Snell Residuals, if you want to display a probability plot for Cox-Snell residuals
Under Graph Tab, check Show Confidence Interval, if you want to display confidence interval in the above plots
Click OK
You work in a manufacturing plant making widgets and you want to find out the deterioration of widgets. You tested the widgets at higher temperatures to accelerate deterioration and collected their failure times. You want to extrapolate to 50 degree Celsius, which are temperatures the widgets normally operate at. It is known that the relationship between temperature and failure time follows the Arrhenius relationship.
Open data worksheet by choosing ProcessMA > Tools > Data
Choose ProcessMA > Reliability and Survival > Accelerated Life Testing
In Variable, select CI - Times
In Assumed Distribution, choose Weibull
In Accelerating Variable, select CJ - Temp
In Relationship, choose Arrhenius
Under Censor Tab, in Censor, select CK - Cen
Under Censor Tab, in Censor Value, choose C
Under Result Tab, in New Predictor Values, enter 50
Under Graph Tab, check Probability Plot (Fitted)
Click OK
From the table of percentiles, you can see that at 50 degrees Celsius, widgets will last 150966 hours at 50% percentile.
Note: Y-axis of relation plot and X-axis of probablity plot is manually changed to log scale.
Copyright © ProcessMA