Workpackages

Workpackage 4: Stochastic ship response modelling

Objectives

Establish requirements and uncertainty bounds on methods for prediction of the time it takes a ship to capsize or sink after damage. The requirements must list and categorise importance of key variables to be accommodated by the methods used, e.g. how the damage is described, is the wind effect accounted for, how accurately is the wave impact represented, how is ship manoeuvrability accounted for, how to address geographical location, etc. The requirements must also put forward uncertainty bounds to be assigned to such methods and input variables.

Description of work

Task 4.1 Benchmark data on time to capsize, ttc
(Responsible: SSPA, Participants: SSRC)

Prepare and execute a set of physical model experiments aiming to characterise stochastic process of the time it takes the vessel to capsize/sink after hull breach event. A RoPax vessel or cruise ship model is used. A set of two different types of damages will be modelled and a series of tests at stationary beam-on-to-waves, as well as at-speed conditions and in waves will be performed. The series of tests will comprise repetitions in order to create sufficiently consistent relative frequency distribution for time to capsize at each of the damage conditions. Finally for establishing the character of the random variable of time to capsize, a set of tests at longer experiment lengths will be performed. The internal geometry will comprise the damaged compartments as well as at least one deck above subdivision deck. All floodable spaces in the damage cases selected will be modelled. Transient and progressive flooding process in waves will be considered.

Task 4.2 Test/develop analytical time to capsize model
(Responsible: SSRC, Participants: SaS, NTUA)

At least one simplified analytical model will be put forward as an alternative for modelling of the stochastic behaviour of time to capsize. Detailed description of the method, input information and its sensitivity to the accuracy of input information will be developed. The validation will be performed on the basis of the experimental tests performed in Task 4.1. The range of the sensitivity studies parameters must allow for quantitative uncertainty quantification to be undertaken in Task 4.5.

Task 4.3 Test/develop numerical time to capsize model
(Responsible: NTUA, Participants: SSRC, SSPA, SaS)

At least one comprehensive numerical model will be put forward as an alternative for modelling of the stochastic behaviour of time to capsize. As in Task 4.2, detailed description of the method, input information and its sensitivity to the accuracy of input information will be developed. The validation will be performed on the basis of the experimental tests performed in Task 4.1. The range of the sensitivity studies parameters must allow for quantitative uncertainty quantification to be undertaken in Task 4.5.

Task 4.4 Test/develop hybrid time to capsize model
(Responsible: SSRC, Participants: SaS, NTUA)

By hybrid model is meant a technique of combining some more sophisticated approaches, such as e.g. the numerical simulation techniques, with regression post processing, case-based reasoning, Bayesian learning, neural-networks or other inference techniques. At last one such model will be put forward as an alternative for modelling of the stochastic behaviour of time to capsize, and similarly as in Task 4.2 and Task 4.3, detailed description of the method, input information and its sensitivity to the accuracy of input information will be developed. The validation will be performed on the basis of the experimental tests performed in Task 4.1. The range of the sensitivity studies parameters must allow for quantitative uncertainty quantification to be undertaken in Task 4.5.

Task 4.5 Establish uncertainty bound on ttc models
(Responsible: SSRC, Participants: BMT, SaS, NTUA, MCA)

It is proposed here that non-intrusive schemes based on ensemble methods, such as Markov Chain - or Differential Evolution - Monte Carlo, or Bayesian Inference be applied for quantification of uncertainty associated with prediction of the stability deterioration process by either of the approaches put forward in Tasks 4.2, 4.3 and 4.4. The deviations of the predictions from the experimentally established bench-test cases as well as the spread of the results between different approaches will be quantified and based on these deviations, bounds on the acceptable deviations will be proposed for common approval. Expectedly, different bounds would be applicable depending on the sophistication and sensitivity of the model to the resolution and quality of the input information. A standard uncertainty test scheme will be devised for any method to be applied to document its predictive capability before use in decision support or design systems.

Deliverables

• D4.1 Report on physical model experiments with ship model
• D4.2 Report on validation and sensitivity testing of an analytical method for characterising ttc
• D4.3 Report on validation and sensitivity testing of a numerical method for characterising ttc
• D4.4 Report on validation and sensitivity testing of a hybrid method for characterising ttc
• D4.5 Report on the method for assigning of uncertainty bounds for methods for characterising of ttc