基于贝叶斯网络的智能装配技术

摘要

为了弥补智能装配包含装配因素不完整的缺陷,在得出了常用零件装配推理算法的基础上,构建了基于装配推理的贝叶斯网络(Bayesian Networks,BN),通过概率推理对未知装配体进行推理装配约束.针对训练样本多而杂的问题,构建了零件识别库,使其对样本零件的特征进行识别和归类,从而提取出与BN相适应的样本,进而提高了BN推理结果的准确性.对UG进行二次开发,运用基于贝叶斯网络的装配推理模型和识别库构建了智能装配系统.实验证明:文中算法具有更高的准确性、高度的拟合性和很强的学习性.%In order to make up the defect of intelligent auxiliary assembly technology which has incomplete assembly factors, Bayesian Networks based on assembly reasoning is built after the assembly reasoning arithmetic of the general part is obtained. It uses probability reasoning for the assembly reasoning of unknown part. The Part Recognition Library is built for the miscellaneous training sample. It can recognize the parts of the sample and classify them, thereby extracting the sample adapted to BN and improving the veracity of the reasoning result. Intelligent assembly system which is based on the secondary development based on UG is built by combining assembly reasoning model based on BN with Part Recognition Library. The experiment show that the arithmetic of this article has great accuracy, high fitting and robust learning ability.