Open-Set Fault Diagnosis in Multimode Processes via Fine-Grained Deep Feature Representation

TL;DR

Open-set fault diagnosis in multimode processes is addressed with FGCRN, a deep framework that learns fine-grained representations for each health state and uses EVT-based rejection to identify unknown faults. The feature extractor integrates multiscale depthwise convolution (MSDC), BiGRU, and temporal attention (TAM) to capture discriminative temporal features, while BN+SAIN stabilizes statistics. Fine-grained representations are built by clustering correctly classified samples per health state with K-means++, modeling distances to cluster centers via Mahalanobis distance and fitting a Weibull tail to obtain rejection probabilities. A distance-based loss term coupled with cross-entropy tightens intra-class cohesion, and extensive experiments on TE, CSTR, and IPCTF demonstrate superior open-set identification and reduced false acceptance/rejection compared to existing methods.

Abstract

A reliable fault diagnosis system should not only accurately classify known health states but also effectively identify unknown faults. In multimode processes, samples belonging to the same health state often show multiple cluster distributions, making it difficult to construct compact and accurate decision boundaries for that state. To address this challenge, a novel open-set fault diagnosis model named fine-grained clustering and rejection network (FGCRN) is proposed. It combines multiscale depthwise convolution, bidirectional gated recurrent unit and temporal attention mechanism to capture discriminative features. A distance-based loss function is designed to enhance the intra-class compactness. Fine-grained feature representations are constructed through unsupervised learning to uncover the intrinsic structures of each health state. Extreme value theory is employed to model the distance between sample features and their corresponding fine-grained representations, enabling effective identification of unknown faults. Extensive experiments demonstrate the superior performance of the proposed method.

Figures (12)

• Figure 1: Visualization of the TE process.
• Figure 2: EVT Modeling for unknown fault identification.
• Figure 3: Overall architecture of FGCRN.
• Figure 4: Forward and backward flows of BiGRU.
• Figure 5: Application workflow of FGCRN.