动态模糊RBF神经网络焊接接头力学性能预测建模

建立动态模糊径向基神经网络RBF( Radial Basis Function,RBF)焊接接头力学性能预测模型,克服静态RBF和模糊神经网络( Fuzzy Neural Network,FNN)在结构辨识、动态样本训练及学习算法的不足。该模型的结构参数不再提前预设,在训练过程中动态自适应调整,适用动态样本数据学习,学习算法引入分级学习和模糊规则修剪策略,加速训练并使模型结构更加紧凑。利用三种厚度、不同工艺TC4钛合金TIG焊接试验数据对该模型进行仿真。结果表明：模型具有较高的预测精度,适用于预测焊接接头力学性能,为焊接过程在线控制开辟了新的途径。

Predictive Modeling of Mechanical Properties of Welded Joints Based on Dynamic Fuzzy RBF Neural Network

A dynamic fuzzy RBF neural network model was built to predict the mechanical properties of welded joints, and the purpose of the model was to overcome the shortcomings of static neural networks including structural identification, dynamic sample training and learning algorithm. The structure and parameters of the model are no longer head of default, dynamic adaptive adjustment in the training, suitable for dynamic sample data for learning, learning algorithm introduces hierarchical learning and fuzzy rule pruning strategy, to accelerate the train-ing speed of model and make the model more compact. Simulation of the model was carried out by using three kinds of thickness and different process TC4 titanium alloy TIG welding test data. The results show that the model has higher prediction accuracy, which is suitable for pre-dicting the mechanical properties of welded joints, and has opened up a new way for the on-line control of the welding process.