Using Adaptive Neuro-Fuzzy Inference System to Predict Strength for Different Orientation Laminated Composites

• Main Authors: Ming-Jen Ting, 丁明仁
• Other Authors: Bae-Muu Chang
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/57349166894106078345

碩士 === 建國科技大學 === 服務與科技管理研究所 === 103 === Laminated composites are cured panels formed by stacking several prepreg plies. Because fibers are directional, the laminates exhibit different strengths depending on the fiber direction. Designers arrange fibers by parallel to the force axes according to structural requirements. Fiber with directionality renders composites with substantial design diversity and flexibility. This study attempts to establish the relationship between material experimental data and strength prediction by using fuzzy theory. With output-input mapping of an adaptive neuro-fuzzy inference system (ANFIS), the basic material experimental data are transformed through failure criteria into training data. Observing MATLAB operations, this study constructs an ANFIS with four parameters of three numbers of 0°, 90° and θ∘plies and one selectable angle and predicts the strength of laminated composites. In the experiments, the ANFIS predictions, trained by using 696 training data entries, are highly accurate and systemically stable. The standard deviation of ANFIS prediction errors (4.42%) is lower than that of experimental material testing (5.13%). In addition, the established ANFIS can be used as a material database for structural design and coupon testing.