基于CART回归树及模型树算法的钢结构整体提升调平技术

Lifting and leveling technology of steel structure based on CART regression tree and model tree algorithm

高层或超高层建筑的多塔互联、大跨度悬臂等建筑造型常采用钢结构技术来完成,对于大体量的钢结构桁架体系的安装,通常采用整体提升方法来进行。由于初始地面高度和提升行程偏差等原因,被提升的钢桁架在提升过程中并非总是处于理想的水平状态,导致构件受力复杂,增加了提升的危险性。首先通过SAP2000模拟计算桁架整体提升过程中,吊点不同位移差状态时构件上各测点的微应变,用API读取并导入到Python中,然后基于CART回归树及模型树算法对计算结果进行训练,建立起吊点相对位移行程与测点微应变之间的对应关系,并验证算法的有效性。将某实际工程提升过程时的实时应变监测数据作为输入,实时获知了钢桁架的提升状态,正确调整各液压提升器的位移行程,使被提升钢桁架始终处于水平位置,保证了提升过程的安全性。

The multi-tower interconnection and long-span cantilever of high-rise or super-high-rise buildings are usually finished by steel structure technology, and the hoisting of large steel truss system is usually carried out by integral lifting.Because of the initial ground height and the deviation of lifting stroke, the steel truss is not in the ideal horizontal state,which leads to the complex stress of members and increases the danger of lifting. In this paper, the micro-strain of each measuring point in different displacement difference states of lifting truss is simulated and calculated by SAP2000. The measuring points are arranged on many members of the whole structure, and the micro-strain of each working state is read by API, then, the micro-strain results are trained based on CART regression tree and model tree algorithm, and the relative displacement stroke of the lifting point is obtained and compared with the actual results. Finally, based on the real-time monitoring data of a practical project during the lifting process, a technical method was developed to know the hoisting state of steel truss in real time, adjust the displacement stroke of each hydraulic hoist and keep the steel truss in horizontal position.