Abstract: Considering the structural characteristics of simply supported girder bridges (SSGBs) in Yunnan Province and develops a multi-parameter engineering demand model using a deep Multilayer Perceptron (MLP),followed by a comprehensive parameterized seismic fragility analysis.The structural parameters such as geometric,material,and construction of 196 SSGBs on national and provincial highways in Yunnan Province are statistically analyzed,and the parameter correlation analysis and significance test are innovatively introduced to determine the structural parameters included in the engineering demand model.Subsequently,to address the challenge of unequal levels between discrete and continuous parameters,a level-mixed uniform test design is proposed to establish a sample library of regional SSGBs.A total of 100 ground motions with magnitudes higher than 4 in Yunnan Province are selected and combined with structural parameter samples to obtain the input sample set of the MLP demand model.Each input sample undergoes incremental dynamic analysis to generate the output sample set of the engineering demand model.Following the training,validation,and testing phases of the MLP demand model,a multi-parameter probabilistic seismic demand substitution model is derived,reflecting the nonlinear mapping from structural multi-parameter inputs to engineering multi-demand outputs.Using Lasso-Logistic regression (LLR) to establish a parameterized seismic fragility function for bridges,followed by the construction of the regional probabilistic seismic damage assessment model for SSGBs in Yunnan Province.The study demonstrates the superior accuracy and reduced discreteness of the MLP demand model.The proposed methodology facilitates parameterized fragility analysis of regional SSGBs,allowing for swift evaluation of the seismic performance within the region.