针对传统偏最小二乘(PLS)模型的在线更新问题,提出了带有自适应遗忘因子的块式递推PLS建模方法.通过Hotelling-T2和Q统计量确定遗忘因子的大小,并且进行模型递推更新,确保模型跟踪过程特性的变化.将所提出的方法应用于管坯斜轧穿孔能耗过程,表现出较强的模型在线更新能力.测试结果表明,带有自适应遗忘因子的块式递推PLS方法的性能优于传统的迭代偏最小二乘方法的性能. In order to solve the problem of on-line update of traditional partial least squares (PLS) model, a block recursive PLS modeling method with adaptive forgetting factor is proposed. The size of the forgetting factor is determined by Hotelling-T2 and Q statistics, Recursive updating to ensure the change of the characteristics of the model tracking process.The proposed method is applied to the energy consumption process of the tube piercing and rolling perforation and shows a strong model online update capability.The test results show that the block with adaptive forgetting factor The performance of the recursive PLS method is superior to that of the traditional iterative PLS method.