Electrodynamic suspension (EDS) has a broad application prospect in high-speed magnetic levitation transportation due to its advantages, such as strong self-stabilization ability and simple control. However, in high-speed application scenarios, flat-plate permanent magnet electric suspension has a high magnetic drag force and requires a large amount of weight of the installed permanent magnets. To ameliorate these problems, this article addresses the plate-type superconducting EDS system with a higher lift-to-drag ratio. First, the structure and principles of the superconducting EDS system are presented. Second, a 3-D analytical model of the electromagnetic force considering the transverse end effect is established by the magnetic vector potential equation and the boundary conditions at the end of the conductor plate. Among them, the source magnetic field of the superconducting magnet array required by the boundary conditions is solved by the coil discretization idea. Then, the reliability of the proposed analytical model is verified by comparing the computational results of the analytical model with the finite element simulation results. Finally, based on the 3-D analytical model, the suspension stiffness characteristics of the superconducting EDS system and the influence of specific parameters on the system's suspension performance are analyzed.

The track harmonic magnetic field acts as the excitation source of the linear generator and provides the power supply for superconducting electrodynamic suspension (EDS) trains. However, external disturbances can induce vibrations that deteriorate the operating conditions of the harmonic generator. Therefore, vibration suppression is essential to improve power supply conditions and operational safety. In this paper, the topology and equivalent circuit of a harmonic linear generator with onboard integrated coils are first introduced, in which power collection and electromagnetic damping functions are integrated. Afterwards, an electromagnetic model of the EDS system, including the harmonic linear generator, is established and verified to accurately characterize the electromagnetic interaction between the onboard integrated coils and null-flux coils. Finally, the dynamic characteristics of the suspension bogie are investigated, and a vertical-lateral synergistic vibration suppression strategy is carried out by additionally controlling d-axis current in the propulsion coils. The results show that the proposed strategy effectively makes up for the deficiency of existing control strategy by the onboard integrated coils, and the vibrations have been effectively suppressed under the excitations of vertical and lateral irregularities at different operating speeds. It lays a theoretical basis for the development of electromagnetic damping and contactless power supply technologies of superconducting EDS trains.

The embedded discrete-time repetitive active disturbance rejection control (EDTR-ADRC) is a promising solution for suppressing periodic disturbances in permanent magnet synchronous motor (PMSM) drives. However, conventional discrete-time implementations suffer from two inherent discretization challenges: the frequency mismatch caused by rounding fractional delay period, and the phase compensation error due to integer-order phase lead step. These discretization errors inevitably degrade the disturbance rejection capability and stability margin. To address these challenges, this article proposes a double fractional-order-based embedded discrete-time repetitive ADRC (DFER-ADRC) strategy. With Lagrange interpolation, a double fractional-order structure is designed to precisely reconstruct both the fractional delay period and the phase lead step. It approximates the ideal non-integer characteristics via polynomial fitting, effectively eliminating the quantization errors. This design guarantees precise alignment between the controller's resonant frequency and the actual harmonic frequency while achieving zero phase shift at arbitrary speeds. On this basis, the proposed strategy maximizes the equivalent open-loop gain at multiple targeted harmonic frequencies. As a result, the total harmonic distortion of the currents is significantly minimized, and the waveform quality is enhanced as well. Furthermore, a comprehensive stability analysis and parameter selection criteria are derived in the discrete-time domain. Finally, the proposed method is validated on a 1.0-kW PMSM platform.

To suppress torque ripple, stator tooth shaping and rotor auxiliary slot opening on a permanent magnet synchronous motor (PMSM) are conducted in this paper. Meanwhile, to avoid excessively affecting other performance characteristics of the motor, a multi-objective optimization algorithm based on an artificial intelligence model is employed to explore the most optimal scheme possible. First, finite element analysis (FEA) is used to establish models of the motor before and after optimization design. Subsequently, a parametric sweep is performed on the optimized motor model to calculate the average output torque and torque ripple for each design scheme, followed by a comparative analysis of the motor’s performance before and after optimization design. The results demonstrate that the stator tooth shaping and rotor auxiliary slot opening, along with the subsequent multi-objective optimization, significantly reduce the torque ripple of the motor. Moreover, the average output torque remains almost unaffected.

Permanent magnet synchronous motors (PMSMs) find extensive application across diverse domains thanks to their merits of high torque density and efficiency. An eccentric magnet-shaped design for a permanent magnet (PM) can effectively reduce the harmonic components in the air-gap magnetic field and suppress the torque ripple of the motor. The electromagnetic performance of the PMSMs is mostly analyzed by the finite element method (FEM), which requires a lot of time and computational resources. Therefore, a rapid optimization method for eccentric magnet-shaped PMSMs is proposed based on an accurate magnetic network model (MNM). The magnetic field characteristics of a six-phase eccentric magnet-shaped PMSM are analyzed based on the proposed model. Firstly, through an analysis of the motor’s magnetic circuit, an accurate MNM is established. Then, the established model is employed to achieve rapid and accurate calculation of the air-gap flux density (AGFD). Finally, the PM’s structure parameters are optimized. After the optimization by MNM, cogging torque (peak-to-peak) reduced from 2.2 N·m to 0.74 N·m; torque ripple (rated condition) from 0.14% to 0.05%; MNM time 20–30 s vs. FEM time 820 s. Compared with the experimental results, the MNM exhibits an error of 5.04% in EMF, 1.16% in output torque, and 0.67% in torque ripple. The effectiveness and accuracy of the MNM have been verified. The MNM can significantly shorten the design period while ensuring the calculation accuracy.

Aiming at the problems of low observation accuracy and serious chattering in traditional sliding mode observer (SMO), an adaptive exponential reaching law sliding mode observer (AERLSMO) is proposed in this paper. In the traditional SMO, the constant velocity reaching law is used to realize the observation of the back electromotive force (back-EMF) signal. However, the reaching speed of the constant velocity reaching law cannot change with the system state, so it will produce large chattering. Using exponential reaching law (ERL) can solve this problem. However, the performance of the traditional ERL cannot meet some occasions with high requirements for motor control performance. Therefore, this paper introduces adaptive gain to design adaptive ERL. Based on the designed adaptive ERL, a SMO is constructed to effectively achieve chattering suppression and observation performance improvement. Finally, the effectiveness and feasibility of the proposed AERLSMO are verified by experiments.

Cogging torque would be produced in the permanent magnet machine because of the stator teeth slotting and non-uniform air gap, and the generation of cogging torque can cause vibration and noise problems. To solve this problem, this paper adopts a semi-analytical method to reduce the cogging torque of permanent magnet motor, which is the combination of analytical method and finite element method. Firstly, based on the mechanism and calculation formula of cogging torque, the parameters of auxiliary slots on the stator teeth are introduced, and from the analytical formula, the effects of the parameters, such as the number, the width and the depth of auxiliary slots, on the cogging torque are analyzed. A 10-pole 12-slot IPMSM prototype is subjected to finite element analysis, and the simulation results show that the proposed structure design method effectively mitigates cogging torque and reduces torque ripple.

In the context of modern transportation advancing toward automation, electrification, and intelligence, there are increasingly stringent demands on drive motors, including high torque, high power, and lightweight design. This paper proposes an axial-radial hybrid flux permanent magnet synchronous motor (PMSM) with an asymmetric bidirectional skewed pole structure. By combining the advantages of axial flux motors and radial flux motors, an integrated salient stator is designed to achieve three-dimensional flux coupling, thereby effectively enhancing the motor's torque density and output performance. Meanwhile, by optimizing the pole arc coefficient, skew angle, and phase shift angle, cogging torque is significantly reduced, torque ripple is suppressed, and the motor's operational stability and efficiency are improved. After comprehensive optimization, the motor achieves a torque density of $16.92 ~\mathrm{N} \cdot ~\mathrm{m}$, with cogging torque peak-to-peak values reduced by 80.36 %, validating the effectiveness of the proposed structure and methods in the design of high-performance PMSMs.

For speed-sensorless induction motor (IM) drives (SSIMD), the adaptive magnetizing current control has already been presented to enhance the speed observability at low synchronous speeds. However, the rapid variations of magnetizing current reference would introduce inherent ripples during zero frequency crossing (ZFC). To solve this problem, this article proposes a variable-slope-based ripple reduction method (VSRRM) for SSIMD to relieve the ripples of speeds and currents during ZFC. First, the qualitative analyses are carried out to elaborate the origin of the ripples during ZFC. On this basis, the slope angle of VSRRM is derived to rearrange the operating point trajectory, reducing the ripples with rapid magnetizing current changes. Considering the introduced slope angle, the boundary points of ZFC are optimized by average current to reduce the ripples. Furthermore, the IM operating trajectory and output torque capability are analyzed with respect to magnetizing current limit. Finally, the effectiveness of the proposed method is verified on a 2.2 kW IM experimental setup.

The catenary dropper (CD) fault detection is an important technical means to ensure the train current collection quality and operational safety. The existing you only look once (YOLO) detection algorithms need improvement in terms of accuracy, especially in the detection of small objects. To address the problem, this article proposes a CD fault detection model based on improved YOLOv11s, named YOLOv11s-CD. First, a four-detection head structure DASFFHead is designed to achieve multiscale feature fusion by integrating a small object detection layer into the neck network and combining a dynamic adaptive spatial feature fusion (DASFF) module. Subsequently, the squeeze-excitation and attention module (SEAM) attention mechanism is embedded in the neck network layer to extract more small object features in occluded areas. In addition, combining the InnerIoU and CIoU methods, the InnerCIoU loss function is designed to enhance the small object detection ability. Finally, the effectiveness and accuracy of the proposed model are validated on the dataset, which is processed by the optimized contrast-limited adaptive histogram equalization (CLAHE) algorithm to enhance the contrast and clarity of the small object defects. Experimental results show that the proposed YOLOv11s-CD has superior performance compared with several other YOLO algorithms, whose mAP@0.5 has increased to 92.3% and AP of small object detection has significantly increased to 91.3%.

Permanent magnet synchronous linear motors (PM-SLM) are widely used in high-precision manufacturing and high-speed rail transit systems. Accurate acquisition of mover position is essential for reliable motor control. Traditional model-based position observers are constrained by their dependency on precise motor parameters, idealized modeling assumptions, and high design complexity. To overcome these limitations, this paper proposes an intelligent position observer based on the Long Short-Term Memory (LSTM) deep learning model. The observer is trained using measurable current and voltage signals collected from the motor drive system. A data-driven position estimation model is developed by leveraging the LSTM networks capability to capture temporal dependencies in sequential motor signals. Experimental results show that the proposed LSTM-based observer can accurately estimate both the position and velocity of the PMSLM mover. The effectiveness and robustness of the method under various operating conditions demonstrate its feasibility and performance advantages compared with conventional approaches.

For speed-sensorless induction motor drives (SSIMDs), the feedback gains design is one of the most significant techniques for adaptive full-order observer (AFO) to enhance the stability in low-speed regenerating region. However, the eigenvalues would also come to the right plane of eigenplane with the existing methods. To cope with this problem, this paper proposes an eigenvalue-distribution-based stability enhancement method to ensure stable operation in low-speed regenerating region. The drawback of the necessary condition is clarified for feedback gains design. Then, the sufficient condition for stability is derived. On the basis, the gain variable is introduced into feedback gains design, whose range of values is used to ensure that the real parts of eigenvalues are all negative. The effectiveness of proposed method is verified by experiments on a 2.2kW IM experimental setup.

[Objective]As construction error of the rigid cat-enary at 160km/h train speed affects the dynamic performance of the pantograph catenary,the technical parameters that meet the condition for the pantograph and rigid catenary system at 160 km/h are designed based on the on-site measurement data and the simulation analysis of rigid pantograph catenary sys-tem.[Method]Firstly,the guiding height(vertical distance from catenary to rail surface)parameters of Guangzhou Metro Sanbei Line are measured,and the probability distribution of guiding height is analyzed.It is concluded that the guiding height distribution is basically consistent with the static height distribution of the fixed points,both following the normal dis-tribution.Then,in consideration of the construction error of the guiding height,the DSA250 pantograph is selected and the dynamic simulation model of the pantograph is established ac-cording to TB/T 3271.Finally,the dynamic performance of the pantograph catenary under different construction errors is simulated and analyzed,and the technical parameters of the pantograph catenary system meeting 160 km/h requirements are designed.[Result & Conclusion]The construction error of rigid suspension device guiding height follows the probability distribution model of normal distribution.The standard devia-tion of the contact force with construction error is 2 to 3 times that without construction error,indicating that the construction error of the guiding height has significant influence on the dy-namic coupling performance of the pantograph catenary.For the construction error of rigid suspension device guiding height,the height of 6 fixed points at the anchor segment joints strictly follows the normal distribution with the mean being the nominal height of the overhead line and the standard deviation being 2 mm,and the height of other fixed points strictly fol-lows the normal distribution with the mean being the nominal height of the overhead line and the standard deviation being 3 mm.

In the task of text classification, the method based on large model has gradually replaced base model. However, in the case of insufficient computing resources, fine-tuning operations are still difficult to perform. In addition, the large model has the ability of general representation, and lacks flexibility in the face of domain data. Therefore, it is still a hot research direction to continue to mining the potential of base model. In this paper, a model SCLCNN based on self-supervised contrastive learning is proposed, which integrates the feature extraction ability of CNN model and the representation ability of SCL. The experimental results show that the SCLCNN model is better than the contrastive baseline model and that the classification precision is 97.7%. In this paper, the performance of the model is observed by fine-tuning the parameters of the model, and the high efficiency of the combination of SCL and CNN model is verified by ablation experiments.

When the permanent magnet assisted synchronous reluctance motor(PMaSynRM)runs at high-speed with flux-weakening control,the DC bus voltage utilization is not high,and the efficiency and torque output capacity of the motor are low. Therefore,a flux-weakening control strategy of permanent magnet assisted synchronous reluctance motor based on hexagonal trajectory was proposed. Firstly,based on the d-q axis equivalent circuit of the permanent magnet assisted synchronous reluctance motor,the voltage and current constraints of the flux-weakening process were derived,and the root cause of the flux-weakening control to improve the speed regulation ability of the motor was proved. Secondly,in order to give full play to the advantages of high-power density under high-speed operation of permanent magnet assisted synchronous reluctance motor,the over-modulation algorithm was derived. It was applied to the flux-weakening operation of permanent magnet assisted synchronous reluctance motor to achieve higher DC bus voltage utilization. Finally,the effectiveness of the proposed method was verified by simulation.

The active disturbance rejection controller (ADRC) can be integrated with advanced controllers to further improve its current disturbances suppression capability. However, the performance of existing methods would be deteriorated under the multiple specific periodic disturbances of current loop and the improper parameters of ADRC. To fill this gap, an embedded discrete-time repetitive ADRC (EDTR-ADRC) is proposed to suppress multiple frequency current disturbances in this article. Firstly, the DC and low-frequency disturbances can be eliminated by extended state observer (ESO) of ADRC. Then, the discrete-time repetitive controller is embedded into the control law and adopted to extract multiple specific periodic frequencies signals without phase shift. On this basis, the disturbances rejection capability can be improved by EDTR-ADRC. Meanwhile, the parameter selection can be further optimized by combining the amplitude and phase characteristics of ESO to various disturbances. Then, the stability analysis and anti-disturbance ability of the proposed method are elaborated in discrete-time domain. Finally, the effectiveness of the proposed method is verified on a 1.0 kW IPMSM setup.

To increase torque density, a combined-Halbach array for the axial flux permanent magnet machine (AFPMM) with yokeless and segmented armature (YASA) is presented in this paper. Firstly, the YASA machine models with conventional axial magnets and combined-Halbach array magnets are established by using 3-D finite-element method (FEM). Then, the torque, torque per magnet-volume and no-load back electromotive force (back-EMF) are calculated, and the performance of two machines are analyzed and compared. It is demonstrated that the combined-Halbach array can improve the torque density of the machine and the utilization rate of the magnets. Finally, the combined-Halbach array rotor is optimized, where the rotor with unequal-thickness magnets is established, and its torque density is calculated under different thickness. The results show that the combined-Halbach array with unequal-thickness magnets can decrease the amount of permanent magnet materials, thus improve the utilization of magnets and reduce the costs.

The recent advancements in neural network models and the availability of vast amounts of data, automatic summarization technology has become one of the primary solutions for dealing with information overload and pinpointing key information. Unlike narrative text, movie scripts consist of sequences of scene descriptions, and directly compressing the text may lead to truncation of plot-relevant content. Furthermore, movie script summarization tasks lack datasets that align script content with plot summaries. Therefore, this paper proposes a two-stage method T4S for generating plot summaries of movie scripts. First, the GraphTP model is employed as an extractor to extract key turning scenes from the scene text sequences. Second, an unsupervised text matching method is used to obtain text pairs that match scenes with plot summaries. Finally, a generator, utilizing an efficiently fine-tuned LLM, rewrites key scene text and concatenates it to form the final plot summary of the movie script. The results of the implementation show that the proposed method in this paper outperforms baseline methods.

Speed-sensorless induction motor drives (SSIMDs) are widely used in rail transportation applications, which use an adaptive full-order flux (AFO) observer a lot for speed estimation. When the motor operates in the low-speed regenerating region, the AFO has an unstable region. For this problem, there are a number of papers proposing different solutions. In contrast to previous papers, this paper proposes a parameter stabilization adjustment (PSA) method to enhance the SSIMD system robustness. First, the relationship between q-axis current error and speed error is derived. On the basis, the feedback gains are designed to stabilize the SSIMD system based on Routh-Hurwitz Criterion. Then, to complete the design of PSA, the gain coefficient is introduced and designed based on the difference between the actual and estimated rotor flux. Finally, the validity of PSA is confirmed by theoretical and experimental comparison with existing methods.