Joint model based on RMK-ASSA and DBSKNet

Blast furnace iron-making process (BFIP) is one of the most critical procedures in the iron and steel industry, in which timely detection and accurate classification of faults have always been of core focus. Nevertheless, due to the coupling effects of complex nonlinear and nonstationary characteristics hidden among the data, the consistent underlying information in the process cannot be accurately mined, hindering the establishment of the BFIP fault diagnosis model. Therefore, we propose a novel data-driven joint fault diagnosis strategy based on regularized mutual kernel analytic stationary subspace analysis (RMK-ASSA) and deep broad stationary kernel network (DBSKNet) to eliminate the above interference. First, considering that standard analytic stationary subspace analysis (ASSA) cannot cope with the complex process nonlinearity and leads to poor modeling accuracy, an RMK-ASSA method is constructed to address this issue. Here, local and global kernels are separately accepted to account for multiple nonlinearities of the process data. Then, the weights of different nonlinear data are obtained by regularized principal component analysis, and the main information is imported into ASSA to obtain more robust and accurate modeling results of the consistent underlying information by eliminating the interference of redundant noise. Moreover, design a DBSKNet-based classifier to implement fault diagnosis task. In this network, nonlinearities are further considered by boosting the kernel structure in depth and width, and the weights of different kernels are also determined to distinguish their respective contributions to fault diagnosis results. On this basis, a double-layer loop parameter optimization algorithm is used to optimize the fault diagnosis effect. Simulated case and practical BFIP validate that RMK-ASSA can eliminate the adverse effect caused by the nonstationary data, and the proposed joint fault diagnosis strategy produces a superior performance over other methods.

炼铁高炉生产工艺（BFIP）是钢铁工业中至关重要的工序之一，其中对故障的及时检测与精确分类始终是核心关注点。然而，由于数据中隐含的复杂非线性和非平稳特性耦合效应，该过程中的一致性潜在信息难以准确挖掘，阻碍了BFIP故障诊断模型的建立。因此，我们提出了一种基于正则化互核分析平稳子空间（RMK-ASSA）与深度广域平稳核网络（DBSKNet）的数据驱动联合故障诊断策略，以消除上述干扰。首先，鉴于标准的平稳子空间分析（ASSA）无法应对复杂过程非线性，导致建模精度不佳，我们构建了一种RMK-ASSA方法以解决这一问题。在此，局部和全局核分别被采纳，以应对过程数据的多种非线性。接着，通过正则化主成分分析获得不同非线性数据的权重，并将主要信息导入ASSA，通过消除冗余噪声的干扰，从而获得更为稳健和精确的建模结果。此外，设计了一种基于DBSKNet的分类器以实现故障诊断任务。在此网络中，通过增强核结构的深度和广度进一步考虑非线性，并确定不同核的权重，以区分它们对故障诊断结果的各自贡献。在此基础上，采用双层循环参数优化算法以优化故障诊断效果。模拟案例和实际BFIP验证了RMK-ASSA可以消除非平稳数据带来的不利影响，并且所提出的联合故障诊断策略相较于其他方法表现更为优越。