Abstract
This paper presents the results of ASTM G124 promoted ignition of titanium metal, with two new data sets covering high pressure oxidation at concentrations of 10% v/v O2 and 20% v/v O2 and a total pressure of 800 psi(g), and one data set covering medium - low pressure oxidation at a concentration of 40% v/v O2 and a total pressure of 100 psi(g). This data is then combined with promoted ignition test data from previous investigations, with the objective of developing a statistical model for titanium flammability.
Logistic regression is identified as an appropriate method for modeling binary "burn/no-burn" test data. The regression model is used to predict the probability of a burn over a total pressure range of 100 - 800 psi(g) and oxygen concentrations between 10% - 50% v/v. Where sufficient data sets exist in both the non-flammable and flammable domains, confidence intervals are also determined, allowing the uncertainty associated with burn probabilities to be quantified. The relationship between the number of burn/no-burn results and confidence levels is examined. The current criterion used to determine exemption pressures for materials in EIGA standard IGC 13/02/E and CGA standard G-4.4 is re-examined in light of the confidence intervals for titanium, and recommendations made to improve the current standard(s).