For modelling superconductors, interpolation and analytical formulas are commonly used to consider the relationship between the critical current density and other electromagnetic and physical quantities. However, look-up tables are not available in all modelling and coding environments, and interpolation methods must be manually implemented. Moreover, analytical formulas only approximate real physics of superconductors and, in many cases, lack a high level of accuracy. In this paper, we propose a new approach for addressing this problem involving artificial intelligence (AI) techniques for reconstructing the critical surface of high temperature superconducting (HTS) tapes and predicting their index value known as n-value. Different AI models were proposed and implemented, relying on a public experimental database for electromagnetic specifications of HTS tapes, including artificial neural networks (ANN), eXtreme Gradient Boosting (XGBoost), and kernel ridge regressor (KRR). The ANN model was the most accurate in predicting the critical current of HTS materials, performing goodness of fit very close to 1 and extremely low root mean squared error. The XGBoost model proved to be the fastest method, with training computational times under 1 s; whilst KRR could be used as an alternative solution with intermediate performance.
13 May 2024 in Machine Learning: Science and Technology
Index-value, or so-called n-value prediction is of paramount importance for understanding the superconductors’ behaviour specially when modeling of superconductors is needed. This parameter is dependent on several physical quantities including temperature, the magnetic field’s density and orientation, and affects the behaviour of high-temperature superconducting devices made out of coated conductors in terms of losses and quench propagation. In this paper, a comprehensive analysis of many machine learning (ML) methods for estimating the n-value has been carried out. The results demonstrated that cascade forward neural network (CFNN) excels in this scope. Despite needing considerably higher training time when compared to the other attempted models, it performs at the highest accuracy, with 0.48 root mean squared error (RMSE) and 99.72% Pearson coefficient for goodness of fit (R-squared). In contrast, the rigid regression method had the worst predictions with 4.92 RMSE and 37.29% R-squared. Also, random forest, boosting methods, and simple feed forward neural network can be considered as a middle accuracy model with faster training time than CFNN. The findings of this study not only advance modeling of superconductors but also pave the way for applications and further research on ML plug-and-play codes for superconducting studies including modeling of superconducting devices.
5 May 2022 in Superconductor Science and Technology
In rotating flux pumps, a rectified voltage, with non-zero DC component, is obtained at the terminals due to the combined effect of the distributed AC electromotive force, produced by one or more permanent magnets in circular motion and the non-linear resistivity of the superconductor. Overcritical currents are continuously induced in the tape during operation, giving rise to the DC voltage and producing, at the same time, dissipation. In this paper, the energy behavior of the flux pump is numerically investigated. It is shown that induced currents interact with the rotating magnet(s), producing a resistant torque that is little affected by the output current of the flux pump. Due to this interaction mechanism, a significant part of the mechanical power supplied to the rotor is converted into Joule heating within the tape. The paper also explores the operational limits of the flux pump, showing that the generator operation, involving an electric power delivered to the load combined with a mechanical power supplied to the rotor, can only be achieved in a restricted range of current and voltage at the terminals and that the maximum power transfer and efficiency are reached at the middle of the generator range. Under no conditions, the mechanical torque produced on the rotor can be reversed, reaching the motor mode involving an electric power absorbed at the terminals combined with a mechanical power produced on the rotor. A revised equivalent circuit comprising, besides the effective resistance reported in the literature, a further intrinsic resistance is proposed in the paper for taking all the dissipation mechanisms into account. It is shown that this equivalent circuit can predict the energization of an RL load both concerning the final steady values and the full time-domain behavior of the current (including ripples).
7 June 2024 in Superconductor Science and Technology
The advent of superconducting bulks, due to their compactness and performance, offers new perspectives and opportunities in many applications and sectors, such as magnetic field shielding, motors/generators, NMR/MRI, magnetic bearings, flywheel energy storage, Maglev trains, among others. The investigation and characterization of bulks typically relies on time-consuming and expensive experimental campaigns; hence the development of effective surrogate models would considerably speed up the research progress around them. In this study, we first produced an experimental dataset containing the levitation and lateral forces between different MgB2 bulks and one permanent magnet under different operating conditions. Next, we have exploited the dataset to develop surrogate models based on Artificial Intelligence (AI) techniques, namely Extremely Gradient Boosting, Support Vector Regressor (SVR), and Kernel Ridge Regression. After the tuning of the hyperparameters of the AI models, the results demonstrated that SVR is the superior technique and can predict levitation and lateral forces with a worst-case accuracy scenario 99.86% in terms of goodness of fit to experimental data. Moreover, the response time of these models for the estimation of new datapoints is ultra-fast.
High temperature superconductivity is emerging as a solution for lightweight, cost-effective and high-power wind generators. Current injection and maintainment/sustainment in the field winding are obtained by metal current leads which, due to persistent heat conduction and joule loss, are responsible for a large part of the total cryogenic heat load. Slip rings, which further reduce the overall performance and reliability of the system, are also required. In this paper we assess the viability of the HTS dynamo and the rectifier flux pumps for energizing the field coils of the EcoSwing 3.6 MW HTS wind generator. Both a “warm” solution, with the rectifier at room temperature, and a “cold” solution, in which the latter is integrated into the cryostat, are investigated with regard to the rectifier flux pump. A comparison with the actual, state-of-the-art, system of the EcoSwing machine is carried out in terms of the total required cooling power and the ability to charge the HTS field winding up to the rated current. It is found that the dynamo flux pump, beside avoiding the need of slip rings, allows the reduction in the required cooling by about 74% with respect to the conventional current-leads-based solution.
11 January 2024 in Superconductor Science and Technology
This rapid communication highlights a remarkable achievement obtained with a 120 mm diameter large cylindrical MgB2 bulk: the measurement of magnetic levitation forces up to 700 N at 20 K, levitation being stable up to levitation forces exceeding 400 N. The investigated sample, which has one of the largest dimensions reported in the literature, was fabricated using the unconventional spark plasma sintering technique. This method is known as one of the most efficient methods for producing dense MgB2 samples. This breakthrough holds significant potential for advancing high-performance levitation systems and superconducting applications, including high-speed transportation systems and magnetic bearings.