Over het archief
Het OWA, het open archief van het Waterbouwkundig Laboratorium heeft tot doel alle vrij toegankelijke onderzoeksresultaten van dit instituut in digitale vorm aan te bieden. Op die manier wil het de zichtbaarheid, verspreiding en gebruik van deze onderzoeksresultaten, alsook de wetenschappelijke communicatie maximaal bevorderen.
Dit archief wordt uitgebouwd en beheerd volgens de principes van de Open Access Movement, en het daaruit ontstane Open Archives Initiative.
Basisinformatie over ‘Open Access to scholarly information'.
| [ meld een fout in dit record ] | mandje (0): toevoegen | toon |
 |
| Validation of ship manoeuvring simulation models Gavrilin, S. (2017). Validation of ship manoeuvring simulation models. PhD Thesis. Norwegian University of Science and Technology (NTNU): Trondheim. viii, 49 + appendices pp.
|
| Beschikbaar in | Auteur |
| |
| Documenttype: Doctoraat/Thesis/Eindwerk | |
| Auteur | Top | |
|
| Abstract |
Normally only a single repetition of each type of the standard trial is performed, thus, making it impossible to estimate the experimental uncertainty. However, knowing the uncertainty is extremely important: if the uncertainty is too high, for instance, due to strong action of environmental effects such as wind, waves and current, the trials cannot be used for the validation. In this thesis, a practical method of the estimation of the uncertainty using the simulation model is proposed. The method is based on the Monte Carlo simulations and allows to propagate the uncertainty of the input parameters (such as environmental effects or control settings) through the model to estimate the uncertainty of the trial outcome. Thus, the method can be used to compensate for the lack of the results of the repeated tests. In addition, the procedure for global sensitivity analysis is proposed to reveal contributions to the total uncertainty of the individual input factors and their interaction effects. Contrary to the dedicated trials results, in-service data often contain several repetitions of similar non-standard trials with small variations of the control input and unknown environmental conditions. In the thesis, the method of “smart averaging” of such manoeuvres is proposed. The method is based on the identification of the simplified auxiliary model, called metamodel, using many similar trials simultaneously. Then, the metamodel is used to simulate a similar trial representing inherent ship behaviour. This simulation is later used as a benchmark for validation of the original simulation model. When a complex non-standard trial is available, for instance, recorded during normal operation of the ship, the question of comparison with the simulation model arises. In the thesis, an approach based on the calculation of residual force is proposed. According to this approach, experimental measurements of velocities and accelerations in chosen time instances are used to estimate all the forces acting on the ship according to the simulation model. Together with the inertial forces, the balance of the force should be zero. However, as the model is not ideal representation of the reality, this residual force will differ from zero and can be used as a validation metric. Such an approach has a number of benefits compared to the direct comparison of experimental and simulated time-series of velocities, for instance no integration is needed, different degrees of freedom are separated, parts of the time-series with strong noise can be excluded, and others. Finally, the problem of the model uncertainty is studied in the thesis. Contributions of the model coefficients to the uncertainty of the model predictions are studied both for the hull hydrodynamic coefficients and the interaction and steering coefficients. It is found that both groups can result in equally large uncertainty of the model predictions. Thus, to improve the quality of the manoeuvring simulation models, the same attention should be paid to the determination of the interaction coefficients as to the hydrodynamic coefficients. Multiple system identification with random initial conditions revealed that different sets of the model coefficients can result in similar model performance. Thus, it can be hard to define a single unique set of the model coefficients. |
| Top | Auteur |
