数据驱动的正则化载荷重构方法及其应用  

Regularized load reconstruction method based on datadriven and its application

作　　者：任春平 李瑞 REN Chunping;LI Rui(School of Mechanical Engineering,Heilongjiang University of Science and Technology,Harbin 150022,Heilongjiang,China)

机构地区：[1]黑龙江科技大学机械工程学院,黑龙江哈尔滨150022

出　　处：《中国工程机械学报》2025年第1期152-156,162,共6页Chinese Journal of Construction Machinery

基　　金：国家自然科学基金资助项目(52204131)。

摘　　要：针对传统的TSVD和Tikhonov正则化随机载荷重构方法,通常会淹没或放大核函数矩阵的奇异值,使其重构结果呈现光滑或出现虚假峰值现象。提出一种TSVD-Tikhonov正则化相结合的新型滤波函数因子构造的正则化载荷重构方法,重新赋予滤波函数因子一种新的表示,采用分段形式表征滤波函数因子,目的是抑制奇异值较大时被无限放大,奇异值较小时也可被有效地截断,将其核函数矩阵由不适定性转为良性适定性,突破了TSVD和Tikhonov正则化方法载荷重构精确程度低的限制。根据所提出正则化方法、TSVD和Tikhonov正则化方法,依次对截齿截割煤岩载荷进行了重构应用并对比分析。实验结果表明:所提出方法能够克服重构载荷平滑及虚假峰值问题,其在载荷重构的整个时间历程的抗噪性及其鲁棒性优于TSVD正则化以及Tikhonov正则化方法。For the traditional TSVD regularization and Tikhonov regularization random load reconstruction methods,the singular values of the kernel function matrix are usually swamped or amplified,making the reconstruction results appear smooth or with spurious peaks.Taking into account the filter function factors of TSVD regularization and Tikhonov regularization,A novel regularized load reconstruction method with combined TSVD-Tikhonov regularization for filter function factor construction is proposed,and the filter function factors are given a new representation,and the filter function is represented in subsection form,in order to suppress the infinite amplification when the singular value is large,and the singular value can be effectively truncated when the singular value is small,and the kernel function matrix is changed from ill-posed to well-posed,which breaks through the bottleneck of low load reconstruction accuracy of TSVD regularization and Tikhonov regularization methods.The proposed regularization method is compared with TSVD regularization and Tikhonov regularization in the application of load reconstruction in coal and rock cutting using the pick.The experimental results show that the proposed method can surmount the problems of load smoothing and false peak,and its anti-noise and robustness in the whole time course of load reconstruction are better than TSVD regularization and Tikhonov regularization methods.

关 键 词：正则化 良性适定性 载荷重构 新型滤波函数因子 数据驱动 

分 类 号：TD421[矿业工程—矿山机电]