A Probabilistic Time-Variant Corrosion Wastage Model for Seawater Ballast Tank

Probabilistic-based dynamic model of corrosion wastage has been a topic of interest in the past few decades due to the natural complexity of corrosion mechanism and the current tendency by structure owner to incorporate risk-based methodology into their existing maintenance scheme. This paper deals with the modelling of time-variant dynamic behaviour of corrosion pits in ship’s seawater ballast tanks. The proposed model is capable of projecting the likely growth pattern of corrosion pits in the future by eliminating the dependent factors governing the corrosion rate such as environmental factors, material properties and operational condition. A set of metal loss data from preceding published literatures was reused to establish the corrosion wastage model. To appropriately express the high variability of corrosion wastage which contributes to the uncertainties in corrosion assessment, the Weibull probability distributions with time variant-scale parameter were proposed as the probabilistic-based model to predict the growth of corrosion pit in seawater ballast tank. The deterministic linear regression equation of averaged pit depth was integrated into the probability density function equation to make a time-variant prediction model. Based on the results, it is obvious that the provided information from the ship inspections is full of uncertainties, due to the nature of marine corrosion as reflected by the poor correlation result of averaged depth against time. In spite of the drawback, the research is still able to purposely demonstrate the optimisation of the data for corrosion growth prediction. The predicted data as compared with actual data yield results with moderate accuracy based on the Root-Mean-Square-Error (RMSE), yet is still promising provided that more high-quality data can become available in the future. The proposed probabilistic models are intended to simplify the modelling process so that the available data can be fully utilised for prediction and structural evaluation purposes.