目的 为实现在满足用户偏好前提下的产品形态创新设计，以保证新形态的市场接受度和降低产品开发风险，从而探索一种神经网络算法和象元理论相结合的轮毂形态设计方法。方法 首先运用KJ法建立目标轮毂样本库，并采用李克特量表和三角模糊法，对样本的用户模糊偏好值进行量化，其次引入象元理论中的本体象元对轮毂形态特征因子进行提取和归类，构建轮毂形态本体象元空间，并进行轮毂本体象元组合编码处理;再次通过BP神经网络算法探析轮毂形态本体象元组合编码和用户模糊偏好值之间的映射关系，用于预测和输出用户模糊偏好值最佳的本体象元组合方案;最后对最佳本体象元组合方案进行拓展，构建喻体象元空间，并运用形状文法规则对最佳本体象元组合方案进行形态创新设计。结论 BP神经模型测试结果显示MSE为0.005 5，表明模型精度良好，且最终4个设计方案的实际评价值和预测值的MSE为0.005 3，验证了该设计方法的有效性，能够为相关设计人员提供了一定的理论参考。

The work aims to explore a hub form design method combining neural network algorithm and pixel theory, so as to realize the innovative design of product form under the premise of satisfying user's preference, ensure the market acceptance of the new form, and reduce the risk of product development. Firstly, the KJ method was used to establish the target hub sample library, and the Likert scale and the triangular fuzzy method were used to quantify the user's fuzzy preference value of the sample. Secondly, the ontology pixel in the pixel theory was introduced to extract and classify the morphological characteristic factors of the hub, construct the hub form ontology pixel space, and process the hub ontology pixel combination code. Then, the BP neural network algorithm was adopted to analyze the mapping relationship between the hub form ontology pixel combination code and the user's fuzzy preference value, which was used to predict and output the optimal ontology pixel combination scheme of the user's fuzzy preference value. Finally, the optimal ontology combination scheme was expanded to construct a metaphorical element space, and the optimal ontology pixel combination scheme was subject to innovative design of form by the rules of shape grammar. The test results of the BP neural model show that the MSE is 0.005 5, indicating that the model has good accuracy, and the MSE of the actual evaluation value and predicted value of the final four design schemes is 0.005 3, which verifies the effectiveness of the design method and can provide relevant designers with certain theoretical reference.