基于神经网络的机械运动方案多级模糊综合评价模型

 为了更加有效地对机械运动方案实施定量评价，利用神经网络具有的非线性映射特征，提出了一种神经网络评价方法，建立了一种多层次多目标的方案评价模型。该方法将神经网络与模糊数学相结合，首先利用方案评价样本对根据评价指标体系构建的神经网络进行训练，使得评价模型可以较好地反映评价属性值和评价结论间的关系以及各评价指标的权重，然后，将备选方案的各项指标值模糊量化，输入该模型，即可获得评价结果，从而可有效地利用专家经验代替评价群体对运动方案进行评价，简化了评价过程。讨论了基于神经网络运动方案评价的一些关键问题，并通过一个实例验证该模型是合理可行的，从而为解决运动方案评价提供了一种新的思路。

Study on multi-level fuzzy synthetic evaluation model for mechanical kinematical scheme based on neural network

To implement quantificational evaluation for a mechanical kinematical scheme effectively, a multi-level and multi-objective evaluating model was established by applying the nonlinear characteristic of neural network. Integrating neural network and fuzzy mathematics, firstly, this method trained the neural network, which was constructed according to evaluation index system through scheme evaluation samples and could make evaluation model reflect the relation between evaluation attribute values and evaluation conclusions as well as the weights of evaluation index better. Then, after completing fuzzy quantification of index values of candidate schemes and inputting these values into the neural network model, evaluation result could be obtained. This method could utilize expert knowledge more effectively and simplify evaluation process. Moreover, some key problems of kinematical scheme evaluation based on neural network were discussed. An illustration had demonstrated that this model was feasible and could be regarded as a new idea for solving kinematical scheme evaluation.