基于自适应稀疏表示的冷轧轧辊偏心诊断

Diagnosis of roll eccentricity for cold rolling via spare representation based on sparsity adaptation

轧辊偏心是冷连轧过程中导致带钢厚度偏差和质量缺陷的重要因素之一,针对轧辊偏心特征由于机架间的耦合作用和累计效应难以提取的问题,提出基于稀疏自适应匹配追踪(sparsity adaptive matching pursuit, SAMP)的偏心诊断方法。首先根据轧辊偏心模型构建偏心特征冗余字典,采用正交匹配追踪(orthogonal matching pursuit, OMP)算法将厚度数据进行稀疏分解,提取轧辊偏心特征。其次,采用匹配测试方法预测初始稀疏度,并通过迭代估计偏心和噪声的能量自适应地确定最优稀疏度,对偏心信号进行去噪。最后重构偏心信号并进行频谱分析,辨识各机架的轧辊偏心状态。经过仿真和实验验证,本方法比小波分析方法对早期偏心故障有更好的诊断效果,诊断结果为监控轧辊工作状态、优化换辊周期提供了可靠的依据。

Roll eccentricity is one of the important factors leading to strip thickness deviation and quality defects in tandem cold rolling process. Aiming at the problem that the feature of roll eccentricity was difficult to extract due to the inter-stand coupling and cumulative effects, an eccentricity diagnosis method based on sparsity adaptive matching pursuit(SAMP) algorithm was proposed. Firstly, a redundant dictionary containing eccentricity features was constructed based on the roll eccentricity model, and the orthogonal matching pursuit(OMP) algorithm was utilized for sparse decomposition of thickness data to extract the roll eccentricity features. Second, the initial sparsity was predicted by a matching test method, and the optimal sparsity was adaptively determined by iteratively estimating the energy of eccentricity and noise to denoise the eccentricity signal. Finally, the eccentricity of each stand was identified by reconstructing eccentricity signal and spectrum analysis. After simulation and experimental verification, the proposed method has better diagnostic effect than wavelet analysis method for early eccentricity faults. The diagnosis results provide a reliable basis for monitoring the roll condition and optimizing the roll change cycle.