The aircraft thermal management system functions to provide suitable working conditions for pilot, crew, passengers, and the other aircraft systems. The additional weight, drag and power consumption caused by it greatly influences the performance of the aircraft. However, due to rising heat load of emerging novel aircraft concepts, traditional design approaches which rely on data and empirical equations may not apply to the future thermal management systems. Many existing literature which tried to identify the optimal thermal management system architectures only considered limited architecture space where the candidates were pre-selected in terms of experience or intuition. Therefore, viable but non-intuitive architectures may not be included in the design space. To fill this gap, this paper proposes a behavior-based backtracking methodology to systematically populate the architecture space by enumerating both intuitive and non-intuitive architectures. Thermal management requirements for traditional and novel configurations are used to generate the architectures. By comparing the generated architectures with existing ones, this paper validates that the proposed methodology is capable of generating both intuitive and non-intuitive architectures.