2019/12/22 by Mojtaba Farrokh, Mehrdad Shafiei Dizaji, Farrokh, Mojtaba +5
Engineering · Materials Science · Physics and Astronomy · #FOS: Computer and information sciences #FOS: Electrical engineering #Force Microscopy Techniques and Applications #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Magnetic Properties and Applications #Neural and Evolutionary Computing (cs.NE) #Piezoelectric Actuators and Control #Signal Processing (eess.SP) #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2001.01559
openalex publication_date 2019/12/22 · openalex created_date 2022/07/26 · openalex updated_date 2026/07/28
Hysteresis phenomena have been observed in different branches of physics and\nengineering sciences. Therefore, several models have been proposed for\nhysteresis simulation in different fields; however, almost neither of them can\nbe utilized universally. In this paper by inspiring of Preisach Neural Network\nwhich was inspired by the Preisach model that basically stemmed from Madelungs\nrules and using the learning capability of the neural networks, an adaptive\nuniversal model for hysteresis is introduced and called Extended Preisach\nNeural Network Model. It is comprised of input, output and, two hidden layers.\nThe input and output layers contain linear neurons while the first hidden layer\nincorporates neurons called Deteriorating Stop neurons, which their activation\nfunction follows Deteriorating Stop operator. Deteriorating Stop operators can\ngenerate non-congruent hysteresis loops. The second hidden layer includes\nSigmoidal neurons. Adding the second hidden layer, helps the neural network\nlearn non-Masing and asymmetric hysteresis loops very smoothly. At the input\nlayer, besides input data the rate at which input data changes, is included as\nwell in order to give the model the capability of learning rate-dependent\nhysteresis loops. Hence, the proposed approach has the capability of the\nsimulation of both rate-independent and rate-dependent hysteresis with either\ncongruent or non-congruent loops as well as symmetric and asymmetric loops. A\nnew hybridized algorithm has been adopted for training the model which is based\non a combination of the Genetic Algorithm and the optimization method of\nsub-gradient with space dilatation. The generality of the proposed model has\nbeen evaluated by applying it to various hysteresis from different areas of\nengineering with different characteristics. The results show that the model is\nsuccessful in the identification of the considered hystereses.\n