按既定单节段循环工序已下料生产的斜拉桥预制构件，如何在调整后的双节段循环工序下装配协调，往往是服从总体进度决策时面临的关键难题。融合无应力状态法和容差分配法，本文首先构建了基于机器学习的斜拉桥装配容差区间反演方法，并依据先验误差储备反演了某斜拉桥各部件的被动装配容差区间；融合工程前行节段实测数据，构建了面向主动容错的斜拉桥装配容差设计框架。结果表明：区间反演方法可在保证结构安全性和设计最优性的前提下，有效提升施工现场的可操作性；与被动容差设计相比，主动容差设计方法可使主梁拼接角度容差区间提升1.4倍（G2梁段），使斜拉索无应力索长容差区间提升2.1倍（斜拉索S16）；主动容差分析框架可通过调整后续部件的容差范围实现装配失效场景的自适应调整，减少停工、返工现象的出现。

For the prefabricated components of cable-stayed bridges manufactured according to established procedures, how to achieve coordination in assembly under adjusted new processes is often a challenge faced when making decisions based on overall progress. By integrating the stress-free state method and tolerance allocation method, this paper first constructs an inverse method for the assembly tolerance interval of cable-stayed bridges based on machine learning, and inversely calculates the passive assembly tolerance intervals for each component of a certain cable-stayed bridge based on prior error reserves. Integrating the actual measurement data from previous engineering sections, a framework for active fault-tolerant assembly tolerance design of cable-stayed bridges is constructed. The results show that the interval inversion method can effectively improve the operability of the construction site while ensuring structural safety and design optimality. Compared with passive tolerance design, the active tolerance design method can increase the main beam splicing angle tolerance interval by 1.4 times (G2 beam section) and the stress-free cable length tolerance interval by 2.1 times (cable S16). The active tolerance analysis framework can adaptively adjust the tolerance range of subsequent components to assemble failure scenarios, reducing the occurrence of work stoppages and reworks.