Intelligent design of the bonded patch repairs for damaged composite laminates via conditional generative models

Bonded patch repair is a rapid and cost-effective technique for repairing the local damage of composite laminates. However, determining the repair solution is ‌usually time-consuming and relies on the experience of engineers due to the complex relations among an extensive set of parameters, like damage states, patch configurations, adhesive layers, repair performance, and lightweight demands. In this work, an intelligent design method integrating the conditional variational autoencoder-generative adversarial network is developed to autonomously generate repair solutions for damaged laminates under tension, compression, or both. Given the geometry and layup of damaged laminates, material properties of both adhesive layers and patches, our method can directly generate geometric parameters and layup of patches and adhesive thicknesses that meet the repair performance and lightweight demands under tension or compression. In addition, the balanced solutions for the damaged laminates under mixed compression and tension are also provided through additional criteria screening and overall repair performance evaluation. Initial and ultimate failure strains of the repaired laminates are calculated by the finite element method, showing the validity of the generated repair solutions.