在油田生产过程中,油井受各种因素的影响容易发生结蜡。油井结蜡通常会降低油井产生,造成油井阻塞,甚至会造成停井及烧电机等现象,大大增加采油成本。对抽油井结蜡状态进行提前预测,实现抽油井设备预见性维护对油田降本增效及智能化管...在油田生产过程中,油井受各种因素的影响容易发生结蜡。油井结蜡通常会降低油井产生,造成油井阻塞,甚至会造成停井及烧电机等现象,大大增加采油成本。对抽油井结蜡状态进行提前预测,实现抽油井设备预见性维护对油田降本增效及智能化管理具有重要意义。针对基于不平衡数据集构建结蜡预测模型预测效果不理想的问题,文中提出了一种面向非平衡数据的集成学习方法SCRF(SMOTE CLUSTER RANDOM FOREST)。该方法首先使用SMOTE方法对原数据集中的少数类进行过采样以增加少数类的数量,缩小不平衡比例;然后对新的数据集采用CLUSTER聚类方法分层欠采样,生成训练数据集;最后采用基于bagging技术的随机森林算法对训练数据集进行集成学习,从而生成预测模型。实验结果表明,样本均衡后模型预测效果更佳,预测精度和效率都有一定程度的提高。展开更多
In order to conform to dimensional tolerances, an efficient numerical method, displacement iterative compensation method, based on finite element methodology （FEM） was presented for the wax pattern die profile desig...In order to conform to dimensional tolerances, an efficient numerical method, displacement iterative compensation method, based on finite element methodology （FEM） was presented for the wax pattern die profile design of turbine blades. Casting shrinkages at different positions of the blade which was considered nonlinear thermo-mechanical casting deformations were calculated. Based on the displacement iterative compensation method proposed, the optimized wax pattern die profile can be established. For a A356 alloy blade, substantial reduction in dimensional and shape tolerances was achieved with the developed die shape optimization system. Numerical simulation result obtained by the proposed method shows a good agreement with the result measured experimentally. After four times iterations, compared with the CAD model of turbine blade, the total form error decreases to 0.001 978 mm from the orevious 0.515 815 mm.展开更多
文摘在油田生产过程中,油井受各种因素的影响容易发生结蜡。油井结蜡通常会降低油井产生,造成油井阻塞,甚至会造成停井及烧电机等现象,大大增加采油成本。对抽油井结蜡状态进行提前预测,实现抽油井设备预见性维护对油田降本增效及智能化管理具有重要意义。针对基于不平衡数据集构建结蜡预测模型预测效果不理想的问题,文中提出了一种面向非平衡数据的集成学习方法SCRF(SMOTE CLUSTER RANDOM FOREST)。该方法首先使用SMOTE方法对原数据集中的少数类进行过采样以增加少数类的数量,缩小不平衡比例;然后对新的数据集采用CLUSTER聚类方法分层欠采样,生成训练数据集;最后采用基于bagging技术的随机森林算法对训练数据集进行集成学习,从而生成预测模型。实验结果表明,样本均衡后模型预测效果更佳,预测精度和效率都有一定程度的提高。
基金Project (2008ZE53042) supported by National Aerospace Science Foundation of China
文摘In order to conform to dimensional tolerances, an efficient numerical method, displacement iterative compensation method, based on finite element methodology （FEM） was presented for the wax pattern die profile design of turbine blades. Casting shrinkages at different positions of the blade which was considered nonlinear thermo-mechanical casting deformations were calculated. Based on the displacement iterative compensation method proposed, the optimized wax pattern die profile can be established. For a A356 alloy blade, substantial reduction in dimensional and shape tolerances was achieved with the developed die shape optimization system. Numerical simulation result obtained by the proposed method shows a good agreement with the result measured experimentally. After four times iterations, compared with the CAD model of turbine blade, the total form error decreases to 0.001 978 mm from the orevious 0.515 815 mm.
