Speaker: Keming Yu (Brunel University)
Buried pipelines and wind turbine are important devices to convert or transfer energy. Buried pipelines are vulnerable to the threat of corrosion. Of interest is an estimate of the probability when or where an affected pipeline is likely to fail from the extreme growth of a corrosion defect. Wind turbine monitoring uses acoustic emission signal detection of damage processes in the structure. Peak signals are something under the concern of the industry. Of interest is an estimate of the probability of a signal beyond a threshold. Many factors involved need to be taken into consideration when building a probabilistic model for these extreme events. But some classical regression models such as the logistic regression whose response is a binary variable seems inef?cient for an observable continuous-response. Furthermore, the probit regression may face either massive data to process or small size to apply, depending on different cases. Whatever the case, estimation accuracy with the support of sound statistical theory and computational algorithm is expected. This talk will introduce a novel inference of probit regression for extreme events to cope with either massive streaming or small size data and show that this objective may be achievable.