Norris-Landzberg-Exponential Model

This example compares the results for the Norris-Landzberg relationship (can be done using the GLL model in ALTA with proper transformation for each stress).

The data set is from Example 7.2 on page 257 in book Life Cycle Reliability Engineering by Dr. Guangbin Yang, John Wiley & Sons, 2007.

The thermal cycling profiles and test results for chip-scale package solder joints are given below. A thermal cycling profile can be represented by 3 independent stresses in Norris-Landzberg relationship. They are the maximum temperature ($$\,\!T_{Max}$$), the temperature difference ($$\,\!\Delta T$$) and cycling frequency ($$\,\!f$$).

For this data set, a multiple linear regression model is used in the book. The model parameters are estimated using the least squared estimation. The regression model is shown next:


 * $$\begin{align}\\

=&ln\left ( L \right )=\alpha _{0}+\alpha _{1}ln\left ( \Delta T \right )+\alpha _{3}ln\left ( \frac{1}{T_{MAX}} \right )\\ \\ =&9.517-2.0635\times ln\left ( \Delta T \right )+0.3452\times ln\left ( f \right )+2006.4\times \left ( \frac{1}{T_{MAX}} \right ) \end{align}\,\!$$

In ALTA, the General Log-linear model is used with the proper transformation for each stress. The failure time distribution is Exponential.



In ALTA, the maximum likelihood estimation is used. Therefore, the estimated model parameters are not exactly the same as the one given in the book where least squares method is used. The model result in ATLA is:



The above results are close to what the book has.