An Integrated Human–AI Design Decision Algorithm for Chinese Heritage Folding Screens with Acoustic Screening

Connecting culturally informed concept design to engineering assessment requires explicit links between generated images, human judgments, and physical-model inputs. This study presents an integrated human–artificial intelligence (AI) decision algorithm for Chinese heritage folding screens that records the actor, input, transformation, output, and evidential scope at each stage. Stable Diffusion 1.5 with low-rank adaptation is trained on 76 heritage images and generates 240 candidates under fixed prompt and seed conditions. The first author selects 30 candidates for five domain-informed evaluators to rate against five criteria. Intuitionistic fuzzy VIKOR (multi-criteria optimization and compromise solution) converts these ratings into a compromise set, followed by analytical and semi-empirical acoustic screening of assumed engineering representations. Under the reference decision settings, S12 and S20 form a two-member compromise set because their ranking separation falls below the acceptable-advantage threshold. Both representations yield mean predicted insertion losses of approximately 10.72 dB. The algorithm connects human design decisions and reproducible calculations through a forward-only evidence record and explicit continuation, abstention, and failure rules, demonstrated in this folding-screen case.