Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Xiyang Liu has a total of 15 co-authors. Here are the top co-authors by number of shared publications:

  • Guo Chen

    College of General Aviation and Flight, Nanjing University of Aeronautics and Astronautics, Liyang, China

    2 shared publications
  • Hao Wang

    Beijing Aeronautical Technology Research Center, Beijing, China

    2 shared publications
  • Xunkai Wei

    Beijing Aeronautical Technology Research Center, Beijing, China

    2 shared publications
  • Bao Huaiqian

    College of Mechanical and Electrical Engineering, Shandong University of Science and Technology, Qingdao, China

    1 shared publication
  • Chen Guo

    College of Civil Aviation, Nanjing University of Aeronautics and Astronautics, Nanjing, China

    1 shared publication
  • Han Baokun

    College of Mechanical and Electrical Engineering, Shandong University of Science and Technology, Qingdao, China

    1 shared publication

Latest Publications

  • Open Access

    Condition-decoupled reliable multisource fusion for aero-engine fault diagnosis

    by Xiyang Liu, Jiayi Liang, Ruiqi Song

    Aero-engine monitoring signals reflect both operating conditions and health states, while the diagnostic value of individual sources varies across samples. This coupling complicates feature learning and multisource fusion under changing operating conditions. This study proposes a Condition-Decoupled Reliable Fusion Network (CDRF-Net) for multisource fault diagnosis. Preprocessed condition observations serve as a fixed reference for source-specific gated residual correction. Dimension-wise gates regulate the correction magnitude, and a within-class decorrelation constraint suppresses linear associations between residual features and operating conditions within each class. Joint optimization with the classification objective promotes class discrimination. The condition representation and residual features jointly generate sample-dependent fusion weights, while availability masks handle known missing sources. The method is evaluated on degradation-stage identification using N-CMAPSS and fault classification using the HIT intershaft bearing dataset. On N-CMAPSS, CDRF-Net achieves an accuracy of 94.51% and a Macro-F1 of 93.32%; removing the decoupling module reduces Macro-F1 by 11.13 percentage points. On HIT, Macro-F1 reaches 98.67% when Gaussian noise with a standard deviation of 1.0 is added to the standardized displacement source. Ablation and feature correlation analyses support the roles of residual correction and within-class decorrelation. The results support the method’s effectiveness in the evaluated tasks, with relative advantages depending on the held-out operating condition and perturbed source.

  • Open Access

    Local soft-correspondence-guided cross-signal evidence fusion of vibration–current measurements for bearing fault diagnosis under operating-condition drift and class imbalance

    by Xiyang Liu, Shenghui Qiu, Yongbo Li

    To address fault-evidence degradation and insufficient cross-signal coordination between vibration and current measurements under operating-condition drift and class imbalance, this paper proposes LSCF-Net, a local soft-correspondence fusion network for cross-signal evidence fusion in rolling bearing fault diagnosis.Specifically, a local structure-anchored gated residual injection mechanism is first developed to enhance single-signal fault representations of vibration and current signals. Subsequently, cross-signal conditional modulation and local bidirectional Cross-Attention are introduced to enable selective interaction between complementary fault evidence, thereby suppressing weakly correlated channels and mismatched responses. Finally, Fusion Token-based global readout and a channel-graph GCN are integrated for adaptive decision-level fusion.Experiments on the independently partitioned PU bearing dataset show that, under the fixed-speed 100:1 extreme imbalance condition, LSCF-Net achieves an Accuracy of 99.53 ± 0.12%, improving HE2024 by 18.09 percentage points. Under the cross-speed 10:1 target-domain imbalance condition, LSCF-Net achieves an Accuracy of 91.00 ± 0.35% and a Macro-F1 of 90.71 ± 0.51%, exceeding IRDDDA by 12.47 and 12.39 percentage points, respectively. These results indicate the effectiveness of LSCF-Net in local soft-correspondence-guided cross-signal evidence fusion and robust fault identification under operating-condition drift and class imbalance.

  • Open Access

    A Siamese CNN-BiLSTM-based method for unbalance few-shot fault diagnosis of rolling bearings

    by Xiyang Liu, Guo Chen, Hao Wang, Xunkai Wei

    Small and imbalanced fault samples have a profound impact on the diagnostic performance of a model in the process of locating and quantifying the rolling bearing damage of aeroengines in practice. Therefore, a Siamese Convolutional Neural Network-Bidirectional Long Short-Term Memory (CNN-BiLSTM) model was proposed in this paper. Random selection and cross combination methods were used to augment and balance sample sizes at first. Then, two weight-sharing CNN-BiLSTM models were used for adaptive extraction and distance measurement of weak fault features. Finally, the fault classification was performed based on feature distance. Model performance was verified using simulated fault test data of rolling bearings. The results showed that the Siamese CNN-BiLSTM model could achieve an accuracy of up to 96.0% for quantitative diagnosis and 98.0% for location diagnosis. This model was also capable of solving the imbalanced classification of samples and made it possible to transfer between different rotating speeds and working conditions.


  • Open Access

    Convolution neural network based particle filtering for remaining useful life prediction of rolling bearing

    by Xiyang Liu, Guo Chen, Zhenjie Cheng, Xunkai Wei, Hao Wang

    Aiming at the problem of remaining useful life prediction of rolling bearing in aero engine, a data-driven prediction method based on deep learning and particle filter is proposed. Initially, only the vibration data of rolling bearing in normal stage are trained by the deep convolution neural network. According to the feature distance between normal and degraded samples, the evolution features during the whole lifetime are extracted adaptively, and the health index of rolling bearing is constructed. Then, the alarm and failure threshold are determined by unsupervised clustering algorithm. Combined with the extracted feature, remaining useful life of rolling bearing is tracked and predicted by particle filter algorithm based on four parameter exponential model. Finally, the effectiveness of the proposed method is verified by three groups of whole lifetime test data of rolling bearings. Results show that the degradation feature extracted by deep learning method has higher prediction accuracy of 2.19%, 0.93%, and 1.43% respectively than RMS values, and has more stable performance and less influenced by the number of particles or resampling methods, which can better reflect the evolution trend of rolling bearing than the traditional feature.


  • A combined deep learning model for damage size estimation of rolling bearing

    by Liu Xiyang, Chen Guo, Hao Tengfei, Pan Wenping

    The identification of damage sizes based on vibration monitoring data is significant in the study of fault evolution, prediction and diagnosis of rolling bearing. There are inherent restrictions in traditional identification models such as high dependence on prior knowledge and insufficient feature extraction, aiming at the limitation, a prediction method of rolling bearing damage size based on deep learning is proposed. In this paper, a combined model of deep convolutional long-short-term memory network is developed, which can sufficiently extract the multi-dimensional and time-series characteristics of bearing vibration signal, and realize the intelligent and efficient diagnosis of bearing fault. On the basis of theoretical analysis, the rolling bearing fault tests under various damage sizes and rotational velocities are carried out by using the accelerated fatigue testing machine for rolling bearings, and the traditional and novel methods are compared based on the test data. The results show that the prediction accuracy of the combined network can reach 99.94% and 98.67%, respectively under normal and noisy conditions, which is higher than the single deep convolution network, long-short-term memory network and other models. The comparison results amply demonstrate the superiority of the proposed method.


  • Open Access

    Study on acoustic source characteristics of gas pipeline leakage

    by Han Baokun, Liu Xiyang, Liu Bing, Bao Huaiqian, Jiang Xiangguang

    The characteristics of acoustic source of gas pipeline leakage determine the accuracy and adaptability of leak detection for gas pipelines based on acoustic method. In order to explore the characteristics of leakage acoustic source of gas pipeline, a physical model of gas pipeline leakage was developed. Initially, generation mechanism of acoustic source of gas pipeline leakage is studied in order to identify the basic principle of leak detection of gas pipeline based on acoustic method. After that, the simulation model is developed according to the basic principle of aeroacoustics, the flow field, and sound source characteristics of the gas pipeline leakage. Then, the experimental platform for the study on acoustic source characteristics of gas pipeline leakage is set up using the basic principle of the leak detection based on acoustic method and simulation results. The characteristics of acoustic source of leakage are obtained through experiments. Finally, the simulation and experimental results of acoustic source of leakage under different operating conditions are compared, and the accuracy of the simulation results is verified, which provides a theoretical basis for the engineering application of leak detection for gas pipeline based on acoustic method. The results show that the acoustic source of gas pipeline leakage is mainly quadrupole source and the leakage acoustic energy is mainly in the range of less than 100 Hz. The mean value of sound pressure level is increased with the pressure in pipeline and the size of leak aperture.