Application Of Artificial Intelligence In Typical Ocean Engineering Components Features Recognition

In the field of ocean engineering,the traditional research methods include theoretical research,experimental research and numerical simulation,but in the research process,the traditional research methods have some problems,such as simplification of principle,insufficient experimental coverage,non-convergence of numerical simulation.At present,artificial intelligence has been used in many fields due to its advantages of fast calculation speed,stable results,less consumption of resources.Artificial neural network method is used as data processing tool to explore its application in specific ocean engineering problems in this paper.For the common data types of ocean engineering(discrete point data and image data),this paper mainly focuses on three specific issues:(1)Particle impact plane target and cause erosion.A large number of previous experimental data was collected to establish a basic database.The traditional back propagation neural network was optimized by changing activation function,initialization mode and optimization algorithm.The prediction results of the proposed optimization algorithm were compared with those of the existing formulas,and the parameters are analyzed.Compared with the existing formulas,the proposed method has higher prediction accuracy,wider application scope,and is more safe and reliable.(2)Intelligent health monitoring of offshore platforms.Based on the linear superposition principle,the finite element software was used to establish macrocell matrix calculation method about internal force of upper components of platform.And construct neural network base digital database by digital twin method,to optimize neural network.Then,macrocell and neural network calculation method were embedded in visual interface,to set up Intelligent health monitoring system for offshore platform.The system can realize unmanned,real-time health monitoring and early warning.(3)Image recognition of concrete.Image data was obtained by conducting properties experiment of lightweight aggregate concrete.The depth convolution neural networks were used to construct image feature recognition model of lightweight aggregate concrete.The optimal parameters of network and pictures were explored.The extension of the model was verified by applying the optimal method to self-compacting concrete image recognition.This method has good prediction effect and can be further popularized.