从全面考虑机床的加工能力、加工中心偏移等各种加工因素的角度出发 ,提出一种由机械加工能力反求公差的新方法 ,能较正确地反映零件加工的真实过程。给出了其数学分析模型 ,并运用神经网络和遗传算法相结合的方法实现之。最后给出实例并分析了其计算结果 From the perspective of considering various processing factors such as machine tool’s machining ability and machining center offset, a new method to reverse tolerance of machining ability is proposed, which can accurately reflect the real process of machining parts. Given its mathematical analysis model, and the use of neural networks and genetic algorithms to achieve a combination of methods. Finally, an example is given and its calculation results are analyzed