Research On The Ship Dynamic System Model Based On Fuzzy Cognitive Maps

With the development of technology,people can rely on the useful information obtained from the ship dynamic system to make decisions on the system to ensure the safety of the ship and environmental protection.Traditionally,the research scope of ship dynamic system is classified into the engineering field,and the human factors in the model are eliminated as far as possible.However,in fact,the cognition of ship dynamic system has a variety of perspectives,which should be based on the view of systems engineering,integrating human factors as well as others various useful information.Based on the knowledge,experience,model and data of the shipping industry,this paper studies the model of the ship dynamic system based on multiblocks of an FCM.The main work is as follows.To solve the influence of slope parameters on fuzzy cognitive map,when the range of slope parameters is known,the least square identification algorithm of an FCM with optimization of slope parameters is constructed,which is a trade-off between computational consumption and accuracy during learning.When the slope parameter is unknown,the least square identification algorithm of an FCM with gradient descent of slope parameters is constructed,which speeds up the search speed of slope parameters and reduces the calculation burden.These two methods can effectively realize the learning of slope parameter and improve the accuracy of model.A steady turning motion model based on multi-blocks of an FCM is proposed to solve the problem of interaction between variables in ship dynamic system.The model considers not only the measuring instrument but also the experience of the expert.Firstly,physical variables are used as conceptual nodes,and the structure of the cognitive map is established by analyzing the causal relationship between concepts.Then,combined with the normalized data,the loss function was optimized by the least square method.Under the condition of setting the accuracy,the automatic data split of the training data and the parameters estimation of the FCM in each data block were completed simultaneously.The learned data block is the ship dynamic subsystem.Finally,the ship-FCMs system is obtained by combining all the block systems.The simulation data of "Yulong" proves the effectiveness of the algorithm,which makes up for the deficiency of the traditional model.For the engineering applicability of the algorithm,a motion test platform for a real boat was built.The reason of missing sample was analyzed,and a data filling preprocessing model based on ARIMA and Kalman filter was built to improve the utilization rate of sampled data.The validity of the algorithm is verified by sampling data from sea trials.Based on the filling data,the dynamic system model of a real boat based on multi-blocks of an FCM model is realized,which shows its good application prospect in engineering.The ship dynamic system model proposed in this paper is not only valid for itself,but also can be extended to other similar dynamic systems to better serve the national economy.