面向流程工业任务分派的工时估计方法

Man-hour estimation method for task allocating procedure of process industry

为了解决流程工业中定额工时准确度低的问题,提出了基于决策树和模型树的作业工时预估方法。对混合类型属性的训练集,首先用标称属性完成部分树的构建;然后在各分枝上采用模型树算法完成子树的构建,在叶节点处给出线性模型。此外,提出方法可基于数据集给出较优训练参数。以某炼油企业的实际生产数据对该方法进行验证,结果证明提出方法能更准确地预估实际任务量,显著缩小计划与执行之间的偏差,提高计划的可执行性。

This paper proposed a man-hour estimation method which was based on decision tree and model tree to solve the problem of inaccurate time quota in process industry. For the train set of mixed attribute types, firstly, the method used the nominal attributes to build the part of tree, and then used model tree method to construct each sub-tree. In addition, the proposed method could dynamic adjust training parameters when data-set changed. Finally, this paper used the actual production data of an oil refining enterprise to evaluate the method. The result demonstrates the proposed method can accurately estimate the actual tasks load, reduce the discrepancy between plan and execution, and improve the performance of production plan.