29 January 2024 in IEEE Transactions on Industrial Electronics
Estimating contact force for a robot manipulator without an force/torque (F/T) sensor poses challenges due to uncertain torques, such as backlash and flexibility. To address this limitation, this article proposes a data-driven uncertain torque model and an overall gray-box structured approach. The contributions of this article are threefold. First, a joint domain unified neural networks (DUNNs)-based model is proposed to compensate for the uncertain torques. This model effectively captures uncertain torques beyond studies focusing solely on individual uncertainty. Second, the DUNNs model receives dynamic and joint domain information, enabling a single DUNNs model to estimate all joint uncertain torques through joint domain knowledge. This approach reduces the model size while maintaining performance. Third, the structure in which the DUNNs model works with conventional static friction models is introduced. This structure improves contact force estimation performance and enhances robustness against untrained data compared with the black-box model. Experimental results verify the method's effectiveness.
1 September 2023 in IEEE Transactions on Industrial Electronics
Electromagnetic navigation systems (eMNS) for medical applications, such as cardiovascular diseases, employ stationary coils and soft magnetic cores. Electrical currents in the coils synthesize magnetic fields to wirelessly control millimeter-sized microrobots, such as magnetic guidewires. Previous control methods of calculating the current references for the eMNS focus on either low conduction loss with pseudoinverse or wide magnetic-feasible workspace (MFW) with a linear combination of pseudoinverse and minimum infinity-norm. However, with the merits of both wide MFW and low conduction loss, this article proposes an iterative control method that modifies the initial pseudoinverse iteratively, while ensuring that the current limits of the coils are not exceeded. A detailed description of the control method, accompanied by a comprehensive analysis, is provided. The prototype of eMNS with eight coils is used for the experimental verification. It is shown that, in the current limits, MFW is increased by more than 45% compared with the pseudoinverse method, and the conduction loss is reduced by more than 2 kW compared with the linear combination method.
2 April 2024 in IEEE Transactions on Industry Applications
This study proposes a model for detecting anomalies in induction motors using audio signals. Compared to existing anomaly detection methods based on deep generative models, the model is presented in two main aspects. First, the proposed model consists of two generative adversarial network (GAN) modules, each processing the two types of mel-spectrograms with distinct time and frequency resolutions. These modules are trained to reconstruct input data effectively, enabling each module to extract the high-resolution time and frequency domain features. Secondly, the ensemble approach is proposed for calculating the anomaly score and loss function. This approach ensembles all the reconstruction losses and Comprehensive Feature Distances (CFDs) obtained from the two GAN modules. The CFD is calculated as the Maximum Mean Discrepancy (MMD) between the outputs of the encoder and decoder layers of the generator. The effectiveness of the proposed scheme is validated through experiments conducted with 0.75 kW induction motors.
31 December 2016 in Journal of Magnetics
1 August 2025 in IEEE Transactions on Industry Applications
This paper presents a data-driven method for position sensorless control of permanent magnet synchronous motors (PMSMs), specifically for improving performance in the low-speed region. Unlike previous data-driven approaches, this study introduces three key contributions. First, this study presents the design process of the neural network (NN)-based position estimator, with input feature selection through a grid search method. Second, an effective loss function and learning algorithm are proposed and utilized to improve the training performance of the NN-based estimator. Finally, a data augmentation method is introduced to efficiently construct the training dataset and reduce the data acquisition time. As a result, the proposed input features and training algorithm significantly improve the position estimation accuracy of the NN-based estimator when compared to existing data-driven methods. Furthermore, the developed model is implemented on the digital signal processor (DSP) to validate its real-time operation. The operable region in the four-quadrant operation is expanded by 49.80% in the low-speed region compared to the recently studied model-based method. Moreover, robustness under dynamic state operation is verified.