Artificial Neural Networks Based Prediction And Validation Tool (Annpvt) For Ship Model Powering Tests In Towing Tanks

The most important step in accurately accomplishing the powering performance tests performed in towing tanks is the post-test data analysis. Microsoft Excel sheets are normally used in towing tanks for this purpose. Artificial Neural Networks (ANN) are the real advancement towards simplifying complex calculations and fitting of non-linear functions. In this research paper, the MATLAB programs based Artificial Neural Networks based Prediction and Validation Tool (ANNPVT) have been developed to use neural networks in the ship powering test analysis of model of 45,000 DWT Bulk Carrier. These neural networks are trained with data from towing tank test results and different unknown parameters are obtained without any cumbersome calculations. These neural networks worked on the basis of back propagation and Levenberg Marquardt algorithms to approximate the input-output data relationships. These neural networks have yielded results with aggression values of more than 99%. Parameters and results obtained from these neural networks are validated from the results obtained from post-test data analysis. The predicted powers (both effective and delivered power at the design speed) from neural networks were quite accurate with errors ranging from 0.03-2%. The conclusion from these validation studies is that the results from ANNPVT are reliable and can be used in early design stages.