A study on the prediction of mechanical properties of titanium alloy based on adaptive fuzzy-neural network

An important trend in material research is to predict mechanical properties for a new titanium alloy before committing experimental resources. Often the prediction of mechanical properties of these alloys changes depending on their chemical composition and processing methods. Therefore, modeling the relationship between composition and property is crucial to the engineering. This study employs an adaptive fuzzy-neural network approach to predict the mechanical properties of titanium alloys. In adaptive fuzzy-neural network, to reduce the complexity of fuzzy models while keeping good model accuracy, a fuzzy clustering algorithm and a back-propagation learning algorithm are introduced to improve the accuracy of the simple model. For purpose of constructing this model, experimental results for 57 specimens with 14 different chemical compositions were gathered from the literature. The chemical composition contents were employed as the inputs while yield strength, tensile strength, elongation and reduction of area, which were employed as the outputs. Thus, the model can be trained by using the prepared training set. After training process, the testing data were used to verify model accuracy. It is found that there is insignificant difference between predict results and experimental value and the maximum relative error is less than 9%. It proved that the predictive performance of the clustering-based adaptive fuzzy-neural network modeling is available and effective in simulating the composition content and predicting the mechanical properties of titanium alloys.

► A novel method was developed to predict mechanical property.
► A fuzzy C-means (FCM) was introduced in pattern recognition.
► Linguistic “IF–THEN” fuzzy rules are describing the input-output mapping relationship.
► Predicted results are much better agreement with the experimental results.