Study On Predicting The Allowable Vertical Center Of Gravity Of Jack-up Platform Based On RBF Neural Network

The jack-up platform has a corresponding allowable vertical center of gravity under different working conditions.In order to facilitate the operator to grasp the stability of the platform under various actual working conditions,and to take effective measures in time when encountering sudden severe working conditions to avoid the platform overturning,the designer needs to provide the allowable vertical center curve of gravity.However,in the actual stability evaluation of the project,only the allowable vertical center of gravity under several typical working conditions is calculated usually.Taking the water draft as the x-coordinate and the allowable vertical center of gravity as the y-coordinate,the adjacent data points are connected by straight lines as the allowable barycenter height curve.The allowable vertical center of gravity outside the typical working conditions is obtained by linear interpolation method.The traditional method has obvious limitations and prediction error,so this paper proposes a method for predicting allowable vertical center of gravity of jack-up platform based on radial basis function neural network from the perspective of nonlinear approximation,in order to reduce the computational workload and improve the prediction accuracy.Firstly,Moses software was used to calculate the allowable vertical center of gravity of the jack-up wind power installation platform PT1 under different conditions of draught and wind speed.Then,factors influencing the allowable vertical center of gravity were studied and data were collected.Then,the influencing factors are normalized to eliminate the dimensional influence among the indexes.Finally using MATLAB software to build with a single hidden layer of RBF neural network,by changing the mean square error of the target and the expansion of constant both neural network parameters,different training sample set of PT1 fitting and simulation,to predict the corresponding test sample set as a result,obtained by neural network prediction mean square error(MSE),and compared with the linear interpolation method of MSE,get the best training sample set and neural network parameters.Results show that when the training sample set contains five draft working condition and working condition of five kinds of wind speed,the MSE of RBF neural network take 0.01 goals,extend the constant take 1.9 ? 3.7,and the sample set can more accurately reflect the input layer and output layer data mapping relationships between,and neural network prediction results than the linear interpolation method is closer to Moses for the allowable vertical center of gravity.Choosing the best training sample set based on the above method and parameters of RBF neural network application platform PT2 allowable in the prediction of the allowable vertical center of gravity,also had better than traditional linear interpolation prediction results.The results show that the RBF neural network method proposed in this paper can analyze the influence of multiple factors under complex working conditions and improve the accuracy of the prediction of the allowable vertical center of gravity.It can be used to calculate the emergency situation in the course of navigation quickly,which is convenient to take corresponding measures in time,and can be further used in engineering practice.