A new rapid design method for agricultural machinery was proposed to promote agricultural design reuse technology. The method was implemented based on shape distribution model retrieval. Firstly, the CAD models in agricultural database were normalized. According to mesh area, all the triangular meshes that belong to each model were divided into Max, Mid and Min groups. The Sobol quasi-random sequence was used to sample feature point from each group. Then three points were defined as the basic sample unit, which was used to fuse the D2, D3, C1 and A3 features together. The variation coefficient of each feature was calculated and the proportion was regarded as sub-feature weight of fusion feature. Thereafter, the multi-feature histogram of shape distribution was formed by connecting different weighted features. The χ2 distance was selected to measure the similarity between feature histograms. Finally, the effectiveness of the proposed method was proved with selfbuilt agricultural model database using VS2010, Matlab 2016b and Open Cascade. The results showed that the retrieval precision of distance D2 feature was higher than that of curvature C1, angle A3 and area D3 features in agricultural model database. The proposed method of adaptive weighted shape distribution (AWSD) performed better than D2 when retrieval recall was ranged from 0 to 0.5, and as effective as D2 in the range of 0.5~1.0. Compared with D2 retrieval method, the comprehensive retrieval precision of AWSD method was increased by 8.5%. The retrieval cases of tractor wheel hub and harvester picking board demonstrated AWSD outperformance in the aspect of objective retrieval satisfaction. The new model retrieval method that could fuse multi-shape distribution features injected fresh energy to agricultural machinery rapid design.