Testing statistical significance of trends in learning, ageing and safety indicators

A relatively new subject for probabilistic safety methodology is statistical analysis of trends in observed failures and other safety indicators reflecting ageing or learning in operational and maintenance experience at industrial facilities. Random variations of the indicators can mask real changes or cause false alarms. Methodology is proposed for testing statistical significance of apparent trends in safety indicators. Improved methods are developed for detecting both monotonic and non-monotonic trends, some demonstrated by simulation studies and real examples to be more powerful than those known so far. An effective way to use standard trend tests with transformed data for testing exponentiality of data is also demonstrated and found superior to a well-known Lilliefors’ goodness-of-fit test.