Welding quality directly affects the welding structure’s service performance and life. Hence, the effective monitoring welding defects is essential to ensure the quality of the weld structure. Owing to the non-uniformity of the shape, position and size of welding defects, it is a complicated task to analyze and evaluate the acquired welding defects images manually. Fortunately, deep learning has been successfully applied to image analysis and target recognition. However, the use of deep learning to identify welding defects is time-consuming and less accurate due to the lack of adequate training data samples, which easily cause redundancy into the classifier. In this situation, we proposed a new transfer learning model based on MobileNet as a welding defect feature extractor. By using the ImageNet dataset (non-welding defect data) to pre-train a MobileNet model, migrate the MobileNet model to the welding defects classification field. This article suggested a new TL-MobileNet structure by adding a new Full Connection layer (FC-128) and a Softmax classifier into a traditional model called MobileNet. The entire training process of TL-MobileNet model has been successfully optimized by the DropBlock technology and Global average pooling (GAP) method. They can effectively accelerate the convergence rate and improve the classification network generalization. By testing the proposed TL-MobileNet on the welding defects dataset, it turned out our model prediction accuracy has arrived at 97.69%. The experimental results show that in several aspects, TL-MobileNet have better performance than other transfer learning models and traditional neural network methods.

In order to guarantee safe and reliable operation of electric vehicle batteries and to optimize their energy and capacity utilization, it is indispensable to estimate their state-of-charge (SoC). This study aimed to develop a novel estimation approach based on the grey model (GM) and genetic algorithms without the need of a high-fidelity battery model demanding high computation power. A SoC analytical model was established using the grey system theory based on a limited amount of incomplete data in contrast with conventional methods. The model was further improved by applying a sliding window mechanism to adjust the model parameters according to the evolving operating status and conditions. In addition, the genetic algorithms were introduced to identify the optimal adjustment coefficient λ in a traditional grey model (1, 1) model to further improve the source estimation accuracy. For experimental verification, two types of lithium-ion batteries were used as the device-under-test that underwent typical passenger car driving cycles. The proposed SoC estimation method were verified under diverse battery discharging conditions and it demonstrated superior accuracy and repeatability compared to the benchmarking GM method.

The lithium-ion battery plays a crucial role in the power supply of the electric vehicles (EVs). Battery remaining useful life (RUL) is critically vital to ensure the vehicles' safety and reliability. Due to the complicated aging mechanism, predicting RUL for the battery management systems is challenging. In this paper, a novel degradation indicator was constructed using the information extracted from the discharge voltage. The indicator reflected the complete and effective energy information from the voltage signals to reveal battery degradation characteristics. Additionally, an innovative FRGM-UPF framework was developed for RUL prediction in this study. To improve the accuracy and traceability of prediction, the framework adopted a novel fractional grey model (FRGM) to update the state transition equation in UPF. Meanwhile, the UPF was employed to extrapolate trends of the indicator and achieve the RUL prediction. The performances of FRGM-UPF with the degradation indicator were synthetically verified by the data from various types of batteries under different aging tests. The experimental results indicated that the proposed method could achieve precise prediction results and had a wide range of practicability and universality. The developed technologies could be incorporated with the other control algorithms for application in BMS of EVs.