Prediction of strength of adhesive bonded joints based on machine learning algorithm and finite element analysis

Aerospace, vehicle and civil engineering sectors have witnessed a quick swift from using single materials to multi-materials to develop components and structures to fulfill sophisticated functions. To bond multiple thin layers of different materials into a component (a substitute), there is a need to develop effective adhesive joining technologies that can exhibit excellent stress transfer behaviours and compatibility with a wide range of material types. For an adhesive joining technology, the failure load is critical to determine joint performance and important parameters, such as elastic modulus and fracture parameters of substitutes. Therefore, it is critical to effectively predict the maximum failure load and identify optimal parameters. Joints’ failure loads are usually acquired through lap shear tests on bundles of joint coupons. However, though the results of lap shear tests are usually reliable, experimental tests are highly sensitive to the material category of adherends and adhesive, making the experiments costly or even impractical to be conducted for each category of adherend and adhesive. To predict the failure load of joints and to reduce the number of experiments, the Finite Element Analysis (FEA) method has become a popular method. The advantages of FEA are:(i) boundary value problems can be resolved systematically;(ii) components with complex geometry can be resolved more accurately using FEA in comparison with experiments;(3) a FEA model just needs a single lap testing experiment for validation so that the number of required experiments is used to predict the performance of joints.