In the automation production packing system, appearance defects on the carton packaging films occur frequently. To solve this problem, a deep learning-based method is proposed in this work to identify film defects during the packaging process to ensure the production packaging quality. This method adopts the deformable convolution network (DCN) v2-C3 module to replace the C3 module in the Neck part of the You Only Look Once version 5 (YOLOv5s) network to extract the deep feature information of the defects in the production carton packaging, with the purpose to improve the spatial transformation ability of the detection model and the model generalization ability to different shapes of targets. Field data are used to evaluate the proposed method. The analysis results indicate that the recognition rate of the proposed method is 99.3 % for different carton packaging defects; and compared to the original YOLOv5s method, the detection accuracy of the proposed method increases by 2.7 % and the scrap rate is reduced by 1.3 %. As a result, the proposed method can meet the requirements for defect detection in the production carton packaging in practical applications.
The existing converter relies on a non-isolated transformer and lacks bidirectional support in the case of multiple and regenerative inputs. The proposed work focuses on an isolated multiport bidirectional DC-DC converter topology featuring a battery and a supercapacitor for an electric vehicle (EV) of 2 kW. The proposed topology can provide uninterrupted power to the EV’s electric motor and recover braking energy, allowing bidirectional power flow. As a result, the battery discharge time can increase due to the additional source, such as a supercapacitor and regenerative action. The system integrates two single H-bridge converters at the input and a single H-bridge converter at the output, with an existing switch. The converter enables the reduction of the rated voltages of the battery and supercapacitor packs. Additionally, a power-flow management scheme and its corresponding control scheme are proposed. The performance test has been conducted on traditional and PI control methods under different operating conditions with a switching frequency of 50 kHz, such as switching power from the battery or supercapacitor to the load during driving and from the load back to the source during energy generation. Power flow capabilities and efficiency values validate the viability and effectiveness of the proposed system of 2 kW.
This work introduces a novel intelligent security approach to improve the reliability of smart grid operations by protecting measurement data produced by the Phasor Measuring Unit. The increasing demand for power distribution requires continuous and accurate monitoring, which is dependent on reliable measurement data synchronized with strict timing constraints. To address the risks related to inaccurate measurements, unauthorized modification, and malicious disruption, this study develops a self-adaptive security analysis process using an artificial intelligence-based learning strategy. The system evaluates multiple operational factors and converts them into dynamic multivariate states to support accurate decision making. The proposed method integrates a learning mechanism based on nearest neighbor classification to detect abnormal measurement behavior and prevent false data events before operational failure occurs. An experimental evaluation in four scenarios demonstrates that the approach improves security and reliability metrics compared to existing solutions. The results confirm that the proposed method provides an effective pathway for reliable measurement data and resilient smart grid operation.
To protect offshore jacket platforms, it is essential to carry out the structure fault diagnosis. This paper proposes a new approach to identify structural faults in offshore jacket platforms, which is based on the integration of structural vibration data, a sophisticated data fusion process, and an intelligent diagnosis algorithm. In this new approach, firstly, the TCN is adopted to solve the efficiency bottleneck of the traditional recurrent neural networks in long time series signals, and significantly enhances the ability to capture long-distance dependent features in the structural response signals through the introduction of a dilated convolutional structure. Secondly, the bidirectional gated recurrent unit (BiGRU) network fuses forward and reverse gated recurrent units, which can effectively capture the forward and backward correlated timing features in the structure vibration data. The attention mechanism then further weights and optimises the BiGRU timing outputs so that the model can automatically focus on the signal pattern that is most discriminative for the fault identification. Furthermore, the artificial lemming algorithm (ALA) is used to optimise the hyperparameters of the TCN and BiGRU to improve the model efficiency and avoid the local optimum during the model training, thereby enhancing the generalisation performance of the proposed model. The validity of the proposed ALA-TCN- BiGRU model is substantiated through simulation and experimental validation. The results indicate that the proposed model can achieve an overall detection accuracy of over 98 % for the jacket structure, which is superior to several popular diagnosis methods.
Modern unmanned aerial vehicle (UAV) platforms predominantly rely on electric propulsion for controlled flight. Most current UAV propulsion systems employ brushless DC (BLDC) thrust motors. This paper demonstrates the passive detection of UAV electric propulsion systems using a dynamically changing alternating electromagnetic (EM) field created by BLDC thrust motors and power lines in the very low frequency and low frequency range. A high-performance low noise amplifier and a commercial Airspy HF+ SDR paired with a compact ferrite-core magnetic antenna, are used to capture near-field magnetic emissions from four different drones. The tested platforms include the mass-produced DJI Mini 2, DJI Mavic 2 Enterprise, DJI Air 3S, and the Eachine Tyro 109. The measured EM spectrum consistently exhibits motor phase commutation components and electronic speed controller PWM components with propulsion load-dependent sidebands, allowing extraction of time-, frequency-, and amplitude-dependent propulsion-related BLDC motor EM signatures for UAV detection and recognition. The results of the experiment show the capabilities of drone swarm detection and recognition of each flight mode of the propulsion system.
