For this week’s data analysis, I used publicly available data from the Challenger Space Shuttle O-ring disaster. The model developed in Python involved determining the number of O-rings that would disintegrate at 31 Farenheight (-0.555 Celsius). The Challenger Space Shuttle incident of 1986 was later connected to the failure of the rocket’s O-rings (integral for lift-off) which disintegrated as a result of low temperatures.
I was interested in observing whether there was a correlation between temperature and pressure and if in fact, low temperatures directly impact O-ring disintegration. Data used for this analysis was obtained from the UCI Machine Learning Repository and converted into an excel file. The data included the total number of O-rings (6), the number of rings experiencing thermal distress, launch temperature and leak-check pressure. The excel file loaded onto a data frame and the data was printed out into a table:
|Number of O-rings||Distress||Temperature (F)||Pressure (psi)|
To study the correlation between pressure and temperature, a regression analysis was completed, with the independent variables (Temperature and Pressure) and the dependent variable (Distress O-rings). These variables represented the crux of the regression model.
The model used for this analysis, ordinary least squares regression, estimates the unknown parameter in a linear regression model (assuming that there is a linear relationship). The model.summary() provided an R-square value of 0.354, which indicates that the data does not fit a linear relationship. With this model we can determine the number of O-rings that would disintegrate under a specific temperature (31F) and pressure conditions (0,50,100,200 psi):
From this analysis, we can conclude that O-rings can disintegrate (~2) at 31F, at any change in pressure (0-200), further indicating the importance of monitoring temperature and ensuring that take-off does not occur at freezing temperatures. It’s important to note that not all data sets will depict a linear relationship, but identifying the relevant variables will assist in identifying the type of analysis needed.