Silvio Simani

In aerospace engine systems, reliable fault diagnosis is essential for ensuring operational safety, yet the scarcity of labelled fault data remains a major obstacle. Although massive amounts of operational signals are routinely collected, expert annotation is costly and limited, leading to small-sample learning challenges, where only a limited number of labelled samples are available for model training. To address this issue, we propose a diagnostic framework tailored for such label-scarce scenarios. Specifically, a Multi-Scale Attentive Convolutional Neural Network is constructed to enhance discriminative representation by integrating multi-scale feature extraction with squeeze-and-excitation attention. More importantly, a Gap-regularized loss function is introduced to jointly enforce intra-class compactness and inter-class separability, thereby improving generalisation when only limited labelled data are available. Extensive experiments on two aerospace engine–related datasets demonstrate that the proposed method achieves superior diagnostic accuracy in limited-labelled-data settings, highlighting its potential for intelligent health monitoring in safety-critical aerospace applications.

In the aerospace field, equipment failures can lead to substantial economic losses and pose significant safety risks, making effective fault diagnosis crucial. Traditional fault diagnosis methods typically require large, precisely labeled datasets, which are challenging to obtain in aerospace applications due to the rarity and unpredictability of faults. To overcome these limitations, this article proposes a novel conditioned multiscale generative adversarial networks (GANs) approach designed to enhance fault diagnosis performance under small-sample conditions. Initially, raw vibration signals undergo preprocessing using the short-time Fourier transform, which expands frequency-domain features while preserving essential time-frequency characteristics. Subsequently, conditioned multiscale GANs are trained on these limited datasets, employing multiscale convolutional kernels to extract and fuse rich features, thus generating high-quality synthetic samples. Finally, these synthetic samples are combined with the original dataset to train a convolutional neural network offline, which can subsequently perform real-time online fault diagnosis. Extensive validation on two aerospace-related datasets demonstrates that the proposed method significantly enhances fault diagnosis accuracy and efficiency, even when the available training data is severely limited.

Effective process monitoring and fault diagnosis (PMFD) are essential for safe and efficient operation in large-scale industrial processes. However, many existing methods still suffer from limited interpretability, insufficient exploitation of process knowledge, and difficulty in representing causal relationships in strongly coupled systems. This article addresses these challenges by proposing a causality-aware spatiotemporal adversarial learning framework for fault diagnosis (FD) that integrates both process knowledge and data. First, a two-stage procedure for constructing a causal graph is developed. In the first stage, a temporal neural network with attention is used to discover candidate spatiotemporal causal relations from historical data. In the second stage, perturbation-based validation guided by domain expertise filters spurious links and yields an interpretable causal graph consistent with the underlying process mechanisms. The resulting global causal graph is then partitioned into subgraphs associated with physical subsystems. On this basis, a causality-aware spatiotemporal adversarial model is designed to extract fault-sensitive features under subgraph constraints and to build monitoring statistics at both subsystem and plant-wide levels. Finally, an interpretable fault root-cause diagnosis strategy is introduced, which combines Shapley additive explanation (SHAP)-based variable contribution analysis with causal path inspection to identify fault sources and propagation routes. Experiments on a real hot strip mill process (HSMP) demonstrate that the proposed framework improves fault detection performance over representative existing methods while providing physically meaningful explanations that support engineering decision-making.