Determination of the compositions of NiMnGa magnetic shape memory alloys using hybrid evolutionary algorithms

Abstract

Magnetic shape memory (MSM) alloys are a new class of actuator materials with high actuation frequency, energy density and strain. MSM effect occurs in alloys, which exhibit a martensitic transformation and are ferromagnetic. It involves, under effect of magnetic field, a high strain achieved via reorientation of twinned martensite plates. The major problem is that even a slight change in the alloy’s composition causes drastic changes in the martensitic transformation temperature (MTT) and MSM effect is only possible in the martensitic region. Therefore, it is crucial to be able to predict the MTT of any NiMnGa alloy. Artificial neural networks (ANN) with their learning and generalization ability may act as a suitable tool to predict the MTTs of NiMnGa alloys. ANNs are generally used when the problem cannot be explicitly described by an algorithm, a set of equations, or a set of rules. A genetic algorithm (GA) is a search technique used in computing to find true or approximate solutions to optimization and search problems. In a previous study, in order to predict the MTT, the performance of a multi-layer perceptron has been studied. Training and validation stages of the approach are performed by using data sets from many separate analysis results and our chemical analysis results were used for testing. In this paper, as an inverse design approach, we concentrated on finding the composition of any NiMnGa alloy by using the MTT as the input in our ANN model. To build an advanced solution, we used GA and ANN in a hybrid manner to obtain the composition values. In order to compare the performance of the candidate solutions obtained from alternative methods, the required fitness function for the MTT was determined by the ANN developed in the previous study. Solution quality was used as the crosscheck parameters for the comparison of results obtained from either method.