Switched reluctance machines (SRMs) are of great interest because of their simplicity, low-cost, reliability, robustness, fault-tolerance and extended-speed constant-power operation. However, conventional SRMs suffer from high torque ripples. There exist several methods, which have been proposed to reduce torque ripples. One of the proposed methods is to change the geometric structure of the machine. However, analysis of the state-of-the-art designs show that, despite achieving favourable results in applications, the moulding pins of the machines are normally neglected. A motor that gives positive results may get affected negatively by its random moulding during its manufacturing. In this paper, mitigation of torque ripples in short-pitched SRMs (SPSRMs) and fully-pitched SRMs (FPSRMs) are investigated. Three-phase SPSRM and FPSRM are chosen for this study and the effects of the geometric points of moulding pins in the machines are studied comparatively. Maxwell 2D program is used for the analysis and two different models are compared for both SPSRM and FPSRM. The obtained results show that the torque ripples of the two machines are lower when moulding pins are closer to the rotor position. It is reduced by 2.56% at 10 Amps in the proposed SPSRM and 12% at 6 Amps in the proposed FPSRM. It is also observed that the applied method is more effective in reducing torque ripples of FPSRMs than SPSRMs.
In first stage, a machine learning (ML) was performed to predict in-cylinder pressure using both fuzzy logic (FL) and artificial neural networks (ANN) depending on the results of experimental studies in a spark ignition (SI) engine. In the ML phase, the experimental in-cylinder pressure data of SI engine was used. SI engine was operated at stoichiometric air–fuel mixture (φ = 1.0) at 1200, 1400, and 1600 rpm engine speeds. Six different ignition timings, ranging from 15 to 45 °CA, were used for each engine speed. Correlations (R2) between data from in-cylinder pressure obtained via FL and ANN models and data form experimental in-cylinder pressure were determined. R2 values over 0.995 were obtained at an ML stage of ANN model for all test conditions of the engine. However, R2 values were remained between range of 0.820–0.949 with the FL model for different engine speeds and ignition timings. In the second stage, in-cylinder pressure prediction was performed by using an ANN model for engine operating conditions where no experimental results were obtained. Furthermore, indicated mean effective pressure (IMEP) values were calculated by predicting in-cylinder pressure data for different engine operation conditions, and then compared with experimental IMEP values. The results show that the in-cylinder pressure and IMEP results estimated with the trained ANN model are fairly close to the experimental results. Moreover, it was found that using the trained ANN model, the ignition timing corresponding to the maximum brake torque (MBT) used in the engine management systems and engine studies could be determined with high accuracy.
In this study, modeling MCS RM (Mutually Couple S witched Reluctance Machine) which is produced through modifications in wrap around structure of S RM with Feed Forward Back Propagation ANN (Artificial Neural Network) is performed. Data obtained from angle, current, flux and torque components obtained through FEM analysis of MCS RM has been used in ANN training.In the course of literature research, no use of ANN in MCSRM modeling is detected and it is seen that algorithms consisting of analytical methods are preferred. It is established that, in modeling studies which are based on such algorithms, the structure consists of thousands of loops and that these loops extend time needed for simulation; besides, it is seen that installation of loops in modeling become rather di fficult. The data obtained from dynamic analysis of the model are compared with the data obtained from motor tests in the literature and it is witnessed that the model produces similar torques in similar voltage and current forms.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.