Open research questions in Machine Fault Diagnosis Techniques

What the literature leaves open

• This study develops a CT-BiGRU architecture for fault diagnosis of aero-engine inter-shaft bearings under noisy and varying operating conditions. A two-layer large-kernel 1DCNN is employed to enlarge the receptive field and improve hierarchical feature extraction from long vibration sequences. The Transformer encoder is used to model dependencies among the extracted features, while the BiGRU network captures bidirectional temporal variation patterns. The three components are arranged in a progressive manner to integrate local feature extraction, feature dependency modeling, and temporal information extraction. Experimental results on the HIT aero-engine inter-shaft bearing dataset show that the proposed architecture achieves favorable diagnostic performance under different operating conditions and maintains stable classification accuracy in the presence of noise interference. The feasibility of the proposed CT-BiGRU architecture is evaluated using a real aero-engine inter-shaft bearing dataset containing multiple rotational speeds. The current study focuses on normal conditions, inner ring faults, and outer ring faults. Additional fault types, including rolling element faults and cage faults, should be considered in future work to provide a more comprehensive evaluation. The Transformer encoder also increases computational cost, which may limit real-time deployment on resource-constrained devices. Model compression methods, such as attention pruning and knowledge distillation, will be investigated to reduce computational complexity while preserving diagnostic performance.

  A fault diagnosis method for aero-engine inter-shaft bearings based on 1DCNN-Transformer-BiGRU · 2026 · DOI

• Future work will focus on extending the framework to more complex operating conditions, larger datasets, and multi-class fault identification. A key finding of this work is that anomaly-based modelling of healthy operation enables reliable fault detection, particularly in situations where labelled faulty data are limited.

  Hybrid Machine Learning Framework for Direction-Dependent Gearbox Fault Detection · 2026 · DOI

• Future research may address these limitations by evaluating the proposed workflow across multiple da- tasets, exploring adaptive threshold selection strategies, and investigating hybrid configurations that incorporate additional feature extraction mechanisms while preserving the overall architectural simplicity of the current design. The presented methodology sup- ports detection of statistically anomalous vibration pat- terns without requiring pre-labeled fault categories, mak- ing it particularly suitable for early-stage anomaly detec- tion or monitoring scenarios where labeled fault examples are scarce.

  Simplified convolutional model for detecting vibration anomalies in helicopters operation · 2026 · DOI

• Future work will focus on detecting bearing faults in induction motors supplied by Pulse Width Modulated (PWM) voltage source inverters to assess the robustness of the proposed method under inverter-driven conditions. For future work, rotor and stator faults should be investigated using the proposed methodology, and additional evaluation techniques should be considered.

  A comparative study of machine and deep learning models for time-series-based bearing fault diagnosis of induction motors · 2026 · DOI

• 19 International Journal of Information Technologies and Systems Approach Volume 19 • Issue 1 • January-December 2026 The existing research is still limited by the deviations between public datasets and actual production lines in terms of working condition complexity and equipment differences, and the impacts of uneven federated node quality and label noise on collaboration effects have not been systematically characterized.

  An Empirical Study on Federated Learning-Based Predictive Maintenance Method for Equipment · 2026 · DOI

• Future work will focus on getting a Bayesian deep learning model for probabilistic characterization in prognostics, using domain adaptation and transfer learning techniques for cross-fleet generalization, designing high-performing architectures for real-time edge implementation, and inte- grating physics-based constraints to make the degradation model more consistent and interpretable.

  An integrated GAN–VAE–CNN–BiLSTM framework with lion optimization for remaining useful life prediction of aeroengines · 2026 · DOI

• It exhibits superior accuracy, stability, fitness degree, and reliability as- sessment capability under limited data conditions and success- fully applies to the evaluation of turbine disks. The results indicate that the MPC-MCNSO outperforms all competing algorithms in estimat- ing the three-parameter Weibull distribution and performing reli- ability assessment, particularly with limited data conditions. Specifically, informed by finite element analysis of the turbine disk, fatigue tests are designed for Inconel 718 alloy specimens, generating four da- tasets with limited data. In conclusion, this study presents an effective solution that in- tegrates artificial intelligence to improve Weibull parameter esti- mation performance in the reliability assessment of aero-engine turbine disk with limited data. This extension enhances the algorithmic di- versity, enabling it to address more complex scenarios such as mixed or competing failure mechanisms and non-Weibull distri- butions under limited data, thereby significantly broadening its applicability in the aerospace and energy sectors.

  MPC-MCNSO: Artificial intelligence algorithm for Weibull parameter estimation with limited data-application to aero-engine turbine disk reliability assessment · 2026 · DOI

• By explicitly mapping these gaps, this review contrib- utes a structured foundation for guiding future investiga- tions toward underexplored combinations, particularly those involving edge intelligence, AIoT architectures, sys- tem-level validation, and long-term operational assessment.

  Artificial intelligence for prognostics and health management in critical machinery: a systematic review · 2026 · DOI

• Although deep learning-based methods have achieved promising performance, they usually require sufficient labeled data, which is difficult to obtain in practical industrial scenarios where fault samples are scarce and data sharing across sites is restricted by privacy and confidentiality constraints.

  ProtoFed: Prototype-Enhanced Federated Meta-Learning for Few-Shot Rolling Bearing Fault Diagnosis · 2026 · DOI

• Existing trend-based signal processing approaches employ various forms of temporal and spectral analysis to identify characteristic defect frequencies of the REBs but are often limited by their reliance on monotonic trend evolution, retrospectively recognising defect onset once the trend has developed sufficiently.

  UNFIT monitoring of roller bearing degradation: A new event-based concept for early defect detection · 2026 · DOI

• Future research will address these issues by conducting on-site verification at higher rotational speeds and under more complex working conditions, introducing complete nonlinear bearing force models into the simulation frame- work, and transferring the proposed method to public datas- ets (e. While this study focuses on verify- ing the identification performance of the proposed model, future work will investigate the incorporation of interpret- ability techniques (e.

  Deep learning-based multi-parameter unbalance identification for turbine rotor systems · 2026 · DOI

• As the current architecture still poses certain requirements for hardware computational resources, future research could focus on lightweight techniques such as model pruning, deep quan- tization, or knowledge distillation to further reduce the parameter scale while preserving core anti-noise charac- teristics, thereby meeting the deployment needs of edge computing devices.

  C-SwinNet: hybrid CNN-Swin transformer integration for enhanced fault diagnosis of rolling bearings under noisy conditions · 2026 · DOI

• Despite extensive research in this field, PdM remains an open area with several challenges identified across multiple sur- veys and reviews [4, 23]. Although significant progress has been achieved, further research is needed to fully resolve these issues. However, issues such as scalability and adaptability remain open challenges.

  Predictive maintenance in cyber-physical systems: a comprehensive review of applications, approaches, and challenges · 2026 · DOI

• Gao et al. Drone-based disaster management analysis (case Demonstrated drone’s effectiveness for rapid Focuses on operational disaster assessment rather (2017) study framework) situational awareness and response than engineering system health monitoring; no coordination during the 2024 Noto Peninsula ML-based explainability or condition diagnosis…

  Explainable machine learning for condition monitoring of aircraft electromechanical actuators under variable loads: health state diagnosis from multi-sensor dynamic responses · 2026 · DOI

• Recent studies emphasize intelligent learning and artificial techniques, intelligence, which improve fault detection accuracy and enable early diagnosis. IoT-based monitoring systems have gained attention for providing real-time data acquisition, predictive maintenance remote capabilities. These sensors, cloud integrate platforms, and communication technologies to enhance operational efficiency and reduce downtime. Additionally, advancements in soft-start and energy-efficient motor drives have demonstrated significant improvements in reducing and inrush optimizing these developments, most studies focus on individual aspects such as monitoring, control, or efficiency improvement rather than a unified solution. Overall, the literature indicates a shift toward intelligent, connected, and energy-efficient motor systems for modern industrial applications. current, minimizing mechanical and systems consumption.

  Implementation of an Advanced Soft-Start Induction Motor System with Integrated IoT-Based Monitoring Technique: A Comprehensive Review · 2026 · DOI

• Since the frequency error has time-varying characteristics, the constraint term of DCMTE is no longer limited to the overall weighting of the residual signal by a single fixed value parameter.

  Dynamic Chirp Mode Tracking Extraction with Frequency-Error-Guidance for Bearing Fault Diagnosis · 2026 · DOI

• Performance consistency varies significantly across four different flight dates (July 12, 13, 21, 23) with multi-classification F1-scores ranging from 0.7880 to 0.8867, indicating potential seasonal, environmental, or aircraft-specific factors affecting fault diagnosis. Systematic evaluation is needed to test the model's generalization across different UAV platforms, flight conditions, sensor degradation patterns, and long-term temporal drift in sensor data.

  Low-sample supervised fault diagnosis for fixed-wing UAVs based on multi-scale adaptive state-aware sequence learning · 2026 · DOI

• The hyperparameter sensitivity analysis (Table 14) mentions loss rate, batch size, MHSA heads count, Mamba-MSTN blocks, and hidden layers as critical parameters, but does not report the specific optimal values identified or their sensitivity ranges. Concrete optimization guidelines are needed for practitioners configuring these hyperparameters under different low-sample regimes (90 vs. 1440 samples) for UAV fault diagnosis.

  Low-sample supervised fault diagnosis for fixed-wing UAVs based on multi-scale adaptive state-aware sequence learning · 2026 · DOI

• While the model achieves 0.000510 seconds inference time on July 21 flight data, the relationship between sample size (90 to 1440 samples tested) and computational efficiency on resource-constrained UAV autopilot systems has not been characterized. Specific investigation is needed on inference latency, memory footprint, and real-time processing constraints for embedded deployment on fixed-wing UAV platforms.

  Low-sample supervised fault diagnosis for fixed-wing UAVs based on multi-scale adaptive state-aware sequence learning · 2026 · DOI

• The ablation study demonstrates that MHSA module importance varies dramatically with data scarcity (6.56% performance drop with 90 samples on binary classification, versus 30.63% drop on multi-classification with 90 samples), but the interaction mechanisms between MHSA, Mamba, and MSTFE components across different sample sizes and fault severity levels remain unexplored for fixed-wing UAV fault detection.

  Low-sample supervised fault diagnosis for fixed-wing UAVs based on multi-scale adaptive state-aware sequence learning · 2026 · DOI

• The model exhibits systematic confusion between mild faults (d2 = 0.4 to 0.9) and normal operational states in multi-classification tasks, with F1-scores as low as 0.4538 for d2 = 0.4 on July 21 flight data. Future improvements should specifically implement contrastive learning or fine-grained feature extraction techniques to amplify interclass differences and resolve misclassifications between normal and mild fault states in UAV fault diagnosis.

  Low-sample supervised fault diagnosis for fixed-wing UAVs based on multi-scale adaptive state-aware sequence learning · 2026 · DOI

• A discrepancy was observed between high AUC-ROC values (0.98-1.00 for all fault classes) and lower overall classification accuracy (0.928), indicating a gap between class separability in feature space and final label assignment. Future work should investigate whether this gap results from suboptimal decision thresholds, class boundary definition, or feature representation issues specific to rolling element bearing fault detection networks.

  Comparative analysis of deep learning algorithms for rolling element bearing fault classification under variable loads and speeds · 2026 · DOI

• The paper observed an abrupt loss climb between epochs 8-12 during CNN training (loss reaching 20-25 with accuracy dropping to 0.2-0.3), followed by stabilization, with this behavior varying significantly across different fold splits. Future investigation should characterize the causes of this non-monotonic training behavior in bearing fault classification networks and develop regularization or optimization strategies to prevent or mitigate these training instabilities.

  Comparative analysis of deep learning algorithms for rolling element bearing fault classification under variable loads and speeds · 2026 · DOI

• Vision Transformer (ViT) demonstrated unpredictable and non-monotonic performance degradation across noise levels (0.830 baseline, 0.690 at 5 dB, 0.744 at 3 dB, 0.733 at 1 dB SNR) compared to CNN-based architectures, with significant accuracy fluctuation at intermediate SNR conditions. Research is needed to identify why transformer-based architectures exhibit this erratic behavior in rolling element bearing fault detection under noisy conditions and how to stabilize their performance.

  Comparative analysis of deep learning algorithms for rolling element bearing fault classification under variable loads and speeds · 2026 · DOI

• The study identified inconsistent class-wise performance across k-fold validation splits, with specific confusions observed between Combined Fault (CF) and Healthy Baseline (HB), as well as between Inner Race Fault (IRF) and Outer Race Fault (ORF) in certain folds. Future work should investigate class-imbalance mitigation strategies and fault-category-specific augmentation techniques for deep learning bearing fault classifiers to address these persistent misclassifications in particular fault pair combinations.

  Comparative analysis of deep learning algorithms for rolling element bearing fault classification under variable loads and speeds · 2026 · DOI

Most-cited papers in Machine Fault Diagnosis Techniques

• Domain generalization for cross-domain fault diagnosis: An application-oriented perspective and a benchmark study · Reliability Engineering & System Safety · 2024 · 399 citations
• Artificial Intelligence for Predictive Maintenance Applications: Key Components, Trustworthiness, and Future Trends · Applied Sciences · 2024 · 351 citations
• An optimized CNN-BiLSTM network for bearing fault diagnosis under multiple working conditions with limited training samples · Neurocomputing · 2024 · 225 citations
• Evolvable graph neural network for system-level incremental fault diagnosis of train transmission systems · Mechanical Systems and Signal Processing · 2024 · 206 citations
• Fault Diagnosis of Rolling Bearings Based on an Improved Stack Autoencoder and Support Vector Machine · IEEE Sensors Journal · 2020 · 204 citations
• Rotating Machinery Fault Diagnosis Under Time-Varying Speeds: A Review · IEEE Sensors Journal · 2023 · 198 citations
• Meta-learning with elastic prototypical network for fault transfer diagnosis of bearings under unstable speeds · Reliability Engineering & System Safety · 2024 · 192 citations
• A hybrid prognosis scheme for rolling bearings based on a novel health indicator and nonlinear Wiener process · Reliability Engineering & System Safety · 2024 · 179 citations
• Domain generalization for rotating machinery fault diagnosis: A survey · Advanced Engineering Informatics · 2024 · 174 citations
• Few-Shot Cross-Domain Fault Diagnosis of Bearing Driven by Task-Supervised ANIL · IEEE Internet of Things Journal · 2024 · 169 citations

Most recent work

• Conformer-PhyFaultNet: Physics-Informed Spectral Attention Conformer for Generalizable Bearing Fault Diagnosis · IEEE Internet of Things Journal · 2026
• Trend-guided deep physics-informed transfer learning method for remaining useful life prediction of rolling bearings across machines · Reliability Engineering & System Safety · 2026
• A multistage transfer learning framework for intelligent fault diagnosis of rotating machinery under variable operating conditions · Scientific Reports · 2026
• Early-warning industrial fault detection based on physics-guided residual learning and calibrated CRNNs · Scientific Reports · 2026
• Fault Diagnosis of Rotating Machinery Using Supervised Machine Learning Algorithms with Integrated Data-Driven and Physics-Informed Feature Sets · Sensors · 2026
• A time–frequency representation guided deep domain adaptation method for aircraft EMA fault diagnosis under class imbalance and variable working conditions · Measurement Science and Technology · 2026
• A novel rotating machinery fault diagnosis method based on multi-channel correlation strategy image information enhancement · Engineering Applications of Artificial Intelligence · 2026
• A dual-condition cost-sensitive optimization framework for intelligent bearing fault diagnosis · Journal of the Brazilian Society of Mechanical Sciences and Engineering · 2026
• A Physics-Informed Deep Encoding-Parsing Network for Cross-Domain Bearing Fault Diagnosis Under Noisy Sensor Data · Journal of Computing and Information Science in Engineering · 2026
• Domain-augmented open-set generalization framework for fault diagnosis in rotating machinery · Engineering Applications of Artificial Intelligence · 2026

Related topics in Engineering