This paper presents a quantitative framework oriented to computer analysis for evaluating the influence of stiffness matrix properties on the stability of tensegrity structures during structural design. A data set containing 3600 configuration samples is generated around the component topology, prestress level, node coordinates, material parameters and boundary conditions. The global stiffness matrix is calculated for each set of samples, and statistics such as eigenvalue distribution, condition number, energy curvature, modal sensitivity, and geometric coupling are extracted. Subsequently, graph-aided regression and classification modules are constructed to estimate the stability margins and identify stable, critical and unstable states under different load paths. The experimental results show that the proposed framework achieves 95.4% classification accuracy and 0.918 F1 value in steady state recognition, the mean absolute error of stability margin prediction is 0.037, and the inference time for a single sample is 0.42s. The research results provide a computable basis for the design of tensegrity structures considering stability in engineering, and reflect a relatively stable engineering adaptability and screening ability.
