盾构机是一种暗挖地下隧道工程的专用机械。 为了保证盾构机安全、高效掘进,必须控制 密封舱土压平衡。 盾构机掘进过程中密封舱土压受多系统、多场耦合和地质条件突变等诸多因素 影响,极易导致密封舱土压失衡而引起地表塌陷等安全事故。 针对难以建立有效的机理模型进行 密封舱土压预测并实施控制的问题,提出了基于多粒子群协同优化的并行支持向量机( PCPSO- PSVM)密封舱土压的数据驱动建模方法。 采用分而治之的原则将数据样本分成 4 个子集、3 个层 次进行并行学习,再采用交叉反馈的方式更新初始样本重新训练直至结束,得到支持向量;利用协 同粒子群并行计算优化支持向量机的参数,将粒子群并行分组寻优,在各自独立的进程内进行独立 搜索,最后各群体汇集到主进程,计算得到模型的全局最优参数 C 和 σ,得到密封舱土压预测模型。 基于盾构施工现场的实测数据进行了仿真实验,对密封舱内 4 个土压力监测点进行建模预测,结果 表明该方法具有较高的计算效率和预测精度,能够满足实时在线计算预测的要求。 因此,完全可以 通过这些压力监测点做出整个开挖面的土压力预测,以对密封舱土压失衡做出提前预警和决策。 该方法实现了基于数据的快速密封舱土压建模预测,能够为开挖面的稳定控制提供及时、准确的判 断依据,指导工程实践。

Shield tunneling machine is a kind of special machine for underground tunneling. To ensure the safety and efficiency of shield tunneling, the earth pressure in soil chamber must be controlled in balance. Earth pressure in chamber is affected by multi-system,multi-field coupling and abrupt geological conditions and so on during the tunne-ling process,so it easily leads to earth pressure imbalance in soil chamber which can cause safety accidents such as surface subsidence. It is difficult to establish an effective mechanism model to predict the earth pressure for balance control,therefore,a data-driven modeling method of earth pressure in soil chamber is proposed,which is based on par-allel support vector machine optimized by parallel cooperative particle swarm (PCPSO-PSVM). The data samples are divided into four subsets and they are studied in parallel according to three different hierarchies,then the initial sam-ples are updated for retraining by using feedback to obtain the support vector. The parameters of SVM are optimized by using the PCPSO. The particle swarms are grouped in parallel to be optimized and they search solution independently in the separate processes. Each swam assembles to the main process to compute the global optimal parameters and,and the earth pressure predictive model of soil chamber is finally obtained. The simulation experiments are carried out based on the measured data of shield machine construction site. The prediction models are established for the four pressure monitoring points in the chamber. The simulation results indicate that the method has high computing efficien-cy and prediction accuracy which can meet the requirement of real time online calculation and prediction. Therefore, the earth pressure change of entire excavation face can be predicted through the four monitoring points,so as to make early warnings and decisions for earth pressure imbalance in the chamber. The method achieves a quick modeling and prediction based on data,and it can provide timely and accurate judgement evidence for the stability of excavation face which can guide engineering practice.