The gearbox is one of the most fragile parts of a wind turbine (WT). Fault diagnosis of the WT gearbox is of great importance to reduce operation and maintenance (O&M) costs and improve cost-effectiveness. At present, intelligent fault diagnosis methods based on long short-term memory (LSTM) networks have been widely adopted. As the traditional softmax loss of an LSTM network usually lacks the power of discrimination, this paper proposes a fault diagnosis method for wind turbine gearboxes based on optimized LSTM neural networks with cosine loss (Cos-LSTM). The loss can be converted from Euclid space to angular space by cosine loss, thus eliminating the effect of signal strength and improve the diagnosis accuracy. The energy sequence features and the wavelet energy entropy of the vibration signals are used to evaluate the Cos-LSTM networks. The effectiveness of the proposed method is verified with the fault vibration data collected on a gearbox fault diagnosis experimental platform. In addition, the Cos-LSTM method is also compared with other classic fault diagnosis techniques. The results demonstrate that the Cos-LSTM has better performance for gearbox fault diagnosis.

Bearing, as a vital component in electric powertrains, is increasingly used globally such as in electric vehicle (EV). Their damages and faults may bring huge cost loss to the industry and even threaten personal safety. This paper proposes an inferable deep distilled attention network (IDDAN) method which is a self-attention mechanism and transfer learning-based method to diagnose and classify multiple bearing faults in various motor drive systems efficiently and accurately. Compared with convolutional networks, the self-attention-based network can better extract the global feature information and easier to benefit from large amounts of pre-training data. Its significance is to accurately classify various faults of the target machine when the labeled data of the target machine is not enough to directly train the diagnosis model. Firstly, this paper attempt to apply the self-attention-based network to build an advanced fault diagnosis model. Secondly, this paper optimizes the structure of networks through knowledge distillation (KD) technique to require a lighter and fast model. Thirdly, this paper proposes a new data augmentation strategy for 1-D vibration signals to provide large-scale pre-training samples for IDDAN. Experiments show that the self-attention mechanism-based model is more likely to benefit from large-scale data. After testing, compared with many methods and other exist similar methods, the proposed method achieves higher classification accuracy and better performance.

The state of health (SOH) prediction of lithium-ion batteries (LIBs) is of crucial importance for the normal operation of the battery system. In this paper, a new method for cycle life and full life cycle capacity prediction is proposed, which combines the early discharge characteristics with the neural Gaussian process (NGP) model. The cycle data sets of commercial LiFePO₄(LFP)/graphite cells generated under different operating conditions are analyzed, and the power characteristic P is extracted from the voltage and current curves of the early cycles. A Pearson correlation analysis shows that there is a strong correlation between P and cycle life. Our model achieves 8.8% test error for predicting cycle life using degradation data for the 20th to 110th cycles. Based on the predicted cycle life, capacity degradation curves for the whole life cycle of the cells are predicted. In addition, the NGP method, combined with power characteristics, is compared with other classical methods for predicting the remaining useful life (RUL) of LIBs. The results demonstrate that the proposed prediction method of cycle life and capacity has better battery life and capacity prediction. This work highlights the use of early discharge characteristics to predict battery performance, and shows the application prospect in accelerating the development of electrode materials and optimizing battery management systems (BMS).

In this work, we propose a novel coarse-to-fine method for object pose estimation coupled with admittance control to promote robotic shaft-in-hole assembly. Considering that traditional approaches to locate the hole by force sensing are time-consuming, we employ 3D vision to estimate the axis pose of the hole. Thus, robots can locate the target hole in both position and orientation and enable the shaft to move into the hole along the axis orientation. In our method, first, the raw point cloud of a hole is processed to acquire the keypoints. Then, a coarse axis is extracted according to the geometric constraints between the surface normals and axis. Lastly, axis refinement is performed on the coarse axis to achieve higher precision. Practical experiments verified the effectiveness of the axis pose estimation. The assembly strategy composed of axis pose estimation and admittance control was effectively applied to the robotic shaft-in-hole assembly.

Based on the scattering attenuation theory of loaded polycrystalline metals, a bolt axial stress measurement method using ultrasonic echo attenuation is proposed for the problem that the conventional time of flight (TOF) method is difficult to measure the stress of high-strength short bolts. Firstly, the attenuation coefficients of body-centred cubic crystalline materials in the Rayleigh scattering range are described, and the mathematical model between the ratio of the peak of the two bottom echoes and the axial stress of the bolt is further derived. Then, the effects of the transducer wafer diameter, mounting eccentricity and bolt thread on the ultrasonic signal were analysed by finite element simulation, to provide a strategy for transducer selection and mounting. In addition, a magnetic ring clamping device was designed to ensure a tight fit between the transducer and the bolt, and to improve the alignment of both. Finally, a bolt axial stress ultrasonic measurement system was built, and a comparison experiment between TOF method and attenuation method was conducted to predict the axial stress of bolts. The experimental results shown that the average percentage measurement error of the attenuation method proposed in this paper is 4.42%–5.76% lower than that of the conventional TOF method, which is more suitable for measuring the axial stress of bolts.