This work presents an innovative framework for thermoelastic-plastic reliability-based topology optimization, tackling challenges related to material uncertainties, geometric imperfections, and variations in volume fractions. An enhanced Bi-directional Evolutionary Structural Optimization (BESO) method is developed. It integrates thermoelastic-plastic finite element analysis with stochastic reliability constraints to achieve robust and efficient structural designs under combined thermal and mechanical loading. The framework incorporates advanced modeling techniques, including temperature-dependent material properties, elasto-plastic behavior, and eigenmode-based imperfection modeling. A key innovation lies in formulating reliability constraints by treating volume fraction as a random variable to model material usage uncertainty. This ensures compliance with target safety indices. The proposed methodology is verified through detailed numerical examples, including steel beam and shell structures subjected to temperatures up to 800 °C. Results show that the probabilistic designs achieved up to 30% higher load-bearing capacity compared to deterministic ones and demonstrated improved stress distribution and thermal resilience. These enhancements confirm the method’s effectiveness in achieving optimal layouts that balance material efficiency, structural stability, and reliability.