Analysis Of The Structure And Properties For The Superalloys By Applying Artificial Neural Network

Increasing in the properties of the superalloys has been continuous a request of the development of aviation industry, and it is also concerned form many researchers. In order to improve the properties, it is necessary to proceed the prediction before experiment. And then researchers can adjust the parameter to lower the research cost and shorten the developing time according to the prediction result.A lot of factors will influence on the superalloys during their manufacturing process. So that, it is difficult to model the correlation between processing parameters and properties of superalloys in common method. However, the artificial neural network technique shows its obvious advantage to model the complicated system. In this thesis, we analysis the process parameters which influence the structure and properties for the single crystal and fine grained superalloys though the artificial neural network technique, and then model the structure and properties prediction though the process parameters.The inputs of the neural network are the parameters that influence the properties of the alloy. The outputs of the neural network are the most important mechanical properties. The model is based on multiplayer feedforword neural network. We use the BP algorithm and update the model to apply the issue.We observe and study the influence of the temperature gradient and withdraw rate on the structure and properties for the single crystal superalloys CMSX-2 and DD3. We also analysis the influence of the cast temperature for the structure and properties for the fine grained superalloy K4169. The neural network is trained with the comprehensive dataset and a satisfied performance is achieved. The error between the predictions and actual measured numerical value is less than 5%. We use the prediction of the network to simulate the influence curve when the parameters changed and find disciplinarian and optimize the parameters. This offers the new method to study the correlation among process parameters, structure and mechanical properties.The establishment of the prediction model makes an active sense for new materials and new processing, and establishes the foundation for experimental study.