Hierarchical knowledge guided fault intensity diagnosis of complex industrial systems

TL;DR

The paper tackles fault intensity diagnosis in complex industrial systems where target-class dependencies are essential but often neglected by traditional chain-of-thought approaches. It proposes Hierarchical Knowledge Guided (HKG), a two-stream, end-to-end framework that combines deep feature learning with a global hierarchical classifier built via graph convolutional networks, guided by a re-weighted hierarchical knowledge correlation matrix. Key contributions include the global hierarchical classifier, the Re-HKCM for structured, noise-robust inter-class guidance, and extensive validation on four real-world datasets across different domains, with ablation studies confirming the importance of hierarchical knowledge and correlation weighting. The results demonstrate significant improvements in accuracy and robustness, suggesting practical impact for industrial monitoring and potential applicability to other hierarchical multi-label diagnostic tasks.

Abstract

Fault intensity diagnosis (FID) plays a pivotal role in monitoring and maintaining mechanical devices within complex industrial systems. As current FID methods are based on chain of thought without considering dependencies among target classes. To capture and explore dependencies, we propose a hierarchical knowledge guided fault intensity diagnosis framework (HKG) inspired by the tree of thought, which is amenable to any representation learning methods. The HKG uses graph convolutional networks to map the hierarchical topological graph of class representations into a set of interdependent global hierarchical classifiers, where each node is denoted by word embeddings of a class. These global hierarchical classifiers are applied to learned deep features extracted by representation learning, allowing the entire model to be end-to-end learnable. In addition, we develop a re-weighted hierarchical knowledge correlation matrix (Re-HKCM) scheme by embedding inter-class hierarchical knowledge into a data-driven statistical correlation matrix (SCM) which effectively guides the information sharing of nodes in graphical convolutional neural networks and avoids over-smoothing issues. The Re-HKCM is derived from the SCM through a series of mathematical transformations. Extensive experiments are performed on four real-world datasets from different industrial domains (three cavitation datasets from SAMSON AG and one existing publicly) for FID, all showing superior results and outperform recent state-of-the-art FID methods.

Figures (9)

• Figure 1: Schematic diagrams for visualising different thoughts. Each rectangular box represents a thought, which is an intermediate step towards the problem. Different coloured rectangular boxes of the same colour scheme indicate their associations.
• Figure 2: Different cavitation states of acoustic signals.
• Figure 3: Overall framework of the HKG. The T-F domain spectrograms are fed into feature representation learning module for extracting deep features ${\boldsymbol{F}'}$. Meanwhile, the target classes are used to generate word embeddings ${\boldsymbol{E}_w}$ and re-weighted hierarchical knowledge correlation matrix $\tilde{\boldsymbol{A}}$, which are input to a GCNs to generate a set of interdependent global hierarchical knowledge classifiers $\mathcal{C}$. Finally, the deep features ${\boldsymbol{F}'}$ and the hierarchical knowledge classifier $\mathcal{C}$ are obtained predicted scores.
• Figure 4: Visualisation of the learned deep feature distribution of vanillia ResNet34 and HKG+ResNet34 via t-SNE van2008visualizing on Cavitation-Short.
• Figure 5: Different evaluation metrics results of various ablation experiments on Cavitation-Short. (a)-(b) and (d) are all performed with a window size of 466944 and a step size of 466944.