Prediction of remaining useful life of multi-stage aero-engine based on clustering and LSTM fusion

In industry, prognostics and health management (PHM) is used to improve the system reliability and efficiency. In PHM, remaining useful life (RUL) prediction plays a key role in preventing machine failure and reducing operation cost. Recently, with the development of deep learning technology, long short-term memory (LSTM) and convolutional neural networks (CNNs) are adopted into many RUL prediction approaches, which shows impressive performances. However, existing deep learning-based methods directly utilize raw signals. Since noise widely exists in raw signals, the quality of these approaches’ feature representation is degraded, which degenerates their RUL prediction accuracy. To address this issue, we first propose a series of new handcrafted feature flows (HFFs), which can suppress the raw signal noise and thus improve the encoded sequential information for the RUL prediction. In addition, to effectively integrate our proposed HFFs with the raw input signals, a novel bidirectional LSTM (Bi-LSTM)-based two-stream network is proposed. In this novel two-stream network, three different fusion methods are designed to investigate how to combine both streams’ feature representations in a reasonable way. To verify our proposed Bi-LSTM-based two-stream network, extensive experiments are carried out on the commercial modular aero propulsion system simulation (C-MAPSS) dataset, showing superior performances over state-of-the-art approaches.

Bi-LSTM-Based Two-Stream Network for Machine Remaining Useful Life Prediction

One of the primary objectives of the brain-computer interface (BCI) is to obtain a command with higher classification accuracy within the shortest possible time duration. Therefore, this study evaluates several stimulation durations to propose a duration that can yield the highest classification accuracy. Furthermore, this study aims to address the inherent delay in the hemodynamic responses (HRs) for the command generation time. To this end, HRs in the sensorimotor cortex were evaluated for the functional near-infrared spectroscopy (fNIRS)-based BCI. To evoke brain activity, right-hand-index finger poking and tapping tasks were used. In this study, six different stimulation durations (i.e., 1, 3, 5, 7, 10, and 15 s) were tested on 10 healthy male subjects. Upon stimulation, different temporal features and multiple time windows were utilized to extract temporal features. The extracted features were then classified using linear discriminant analysis. The classification results using the main HR showed that a 5 s stimulation duration could yield the highest classification accuracy, i.e., 74%, with a combination of the mean and maximum value features. However, the results were not significantly different from the classification accuracy obtained using the 15 s stimulation. To further validate the results, a classification using the initial dip was performed. The results obtained endorsed the finding with an average classification accuracy of 73.5% using the features of minimum peak and skewness in the 5 s window. The results based on classification using the initial dip for 5 s were significantly different from all other tested stimulation durations (p < 0.05) for all feature combinations. Moreover, from the visual inspection of the HRs, it is observed that the initial dip occurred as soon as the task started, but the main HR had a delay of more than 2 s. Another interesting finding is that impulsive stimulation in the sensorimotor cortex can result in the generation of a clearer initial dip phenomenon. The results reveal that the command for the fNIRS-based BCI can be generated using the 5 s stimulation duration. In conclusion, the use of the initial dip can reduce the time taken for the generation of commands and can be used to achieve a higher classification accuracy for the fNIRS-BCI within a 5 s task duration rather than relying on longer durations.

Most favorable stimulation duration in the sensorimotor cortex for fNIRS-based BCI.

Accurate remaining useful life (RUL) prediction is important in industrial systems. It prevents machines from working under failure conditions, and ensures that the industrial system works reliably and efficiently. Recently, many deep learning based methods have been proposed to predict RUL. Among these methods, recurrent neural network (RNN) based approaches show a strong capability of capturing sequential information. This allows RNN based methods to perform better than convolutional neural network (CNN) based approaches on the RUL prediction task. In this paper, we question this common paradigm and argue that existing CNN based approaches are not designed according to the classic principles of CNN, which reduces their performances. Additionally, the capacity of capturing sequential information is highly affected by the receptive field of CNN, which is neglected by existing CNN based methods. To solve these problems, we propose a series of new CNNs, which show competitive results to RNN based methods. Compared with RNN, CNN processes the input signals in parallel so that the temporal sequence is not easily determined. To alleviate this issue, a position encoding scheme is developed to enhance the sequential information encoded by a CNN. Hence, our proposed position encoding based CNN called PE-Net is further improved and even performs better than RNN based methods. Extensive experiments are conducted on the C-MAPSS dataset, where our PE-Net shows state-of-the-art performance. 

Position Encoding Based Convolutional Neural Networks for Machine Remaining Useful Life Prediction

Transfer learning‐based platform for detecting multi‐classification retinal disorders using optical coherence tomography images

Machine remaining useful life (RUL) prediction is vital in improving the reliability of industrial systems and reducing maintenance cost Recently, long short-term memory (LSTM) based algorithms have achieved state-of-the-art performance for RUL prediction due to their strong capability of modeling sequential sensory data In many cases, the RUL prediction algorithms are required to be deployed on edge devices to support real-time decision making, reduce the data communication cost, and preserve the data privacy However, the powerful LSTM-based methods which have high complexity cannot be deployed to edge devices with limited computational power and memory To solve this problem, we propose a knowledge distillation framework, entitled KDnet-RUL, to compress a complex LSTM-based method for RUL prediction Specifically, it includes a generative adversarial network based knowledge distillation (GAN-KD) for disparate architecture knowledge transfer, a learning-during-teaching based knowledge distillation (LDT-KD) for identical architecture knowledge transfer, and a sequential distillation upon LDT-KD for complicated datasets We leverage simple and complicated datasets to verify the effectiveness of the proposed KDnet-RUL The results demonstrate that the proposed method significantly outperforms state-of-the-art KD methods The compressed model with 128 times less weights and 462 times less total float point operations even achieves a comparable performance with the complex LSTM model for RUL prediction

KDnet-RUL: A Knowledge Distillation Framework to Compress Deep Neural Networks for Machine Remaining Useful Life Prediction

Automatic detection of retinopathy via computer vision techniques is of great importance for clinical applications However, traditional deep learning based methods in computer vision require a large amount of labeled data, which are expensive and may not be available in clinical applications To mitigate this issue, in this paper, we propose a semi-supervised deep learning method built upon pre-trained VGG-16 and virtual adversarial training (VAT) for the detection of retinopathy with optical coherence tomography (OCT) images It only requires very few labeled and a number of unlabeled OCT images for model training In experiments, we have evaluated the proposed method on two popular datasets With only 80 labeled OCT images, the proposed method can achieve classification accuracies of 0942 and 0936, sensitivities of 0942 and 0936, specificities of 0971 and 0979, and AUCs (Area under the ROC Curves) of 0997 and 0993 on the two datasets, respectively When comparing with human experts, it achieves expert level with 80 labeled OCT images and outperforms four out of six experts with 200 labeled OCT images Furthermore, we also adopt the Gradient Class Activation Map (Grad-CAM) method to visualize the key regions that the proposed method focuses on when making predictions It shows that the proposed method can accurately recognize the key patterns of the input OCT images when predicting retinopathy

Automatic detection of retinopathy with optical coherence tomography images via a semi-supervised deep learning method

Contrastive adversarial knowledge distillation for deep model compression in time-series regression tasks

Deep learning based retinopathy classification with optical coherence tomography (OCT) images has recently attracted great attention. However, existing deep learning methods fail to work well when training and testing datasets are different due to the general issue of domain shift between datasets caused by different collection devices, subjects, imaging parameters, etc. To address this practical and challenging issue, we propose a novel deep domain adaptation (DDA) method to train a model on a labeled dataset and adapt it to an unlabelled dataset (collected under different conditions). It consists of two modules for domain alignment, that is, adversarial learning and entropy minimization. We conduct extensive experiments on three public datasets to evaluate the performance of the proposed method. The results indicate that there are large domain shifts between datasets, resulting a poor performance for conventional deep learning methods. The proposed DDA method can significantly outperform existing methods for retinopathy classification with OCT images. It achieves retinopathy classification accuracies of 0.915, 0.959 and 0.990 under three cross-domain (cross-dataset) scenarios. Moreover, it obtains a comparable performance with human experts on a dataset where no labeled data in this dataset have been used to train the proposed DDA method. We have also visualized the learnt features by using the t-distributed stochastic neighbor embedding (t-SNE) technique. The results demonstrate that the proposed method can learn discriminative features for retinopathy classification.

Qing Xu

Papers

KDnet-RUL: A Knowledge Distillation Framework to Compress Deep Neural Networks for Machine Remaining Useful Life Prediction

Automatic detection of retinopathy with optical coherence tomography images via a semi-supervised deep learning method

Contrastive adversarial knowledge distillation for deep model compression in time-series regression tasks

Cross-domain retinopathy classification with optical coherence tomography images via a novel deep domain adaptation method.