جستجوی مقالات مرتبط با کلیدواژه "artificial neural networks" در نشریات گروه "مکانیک"

One of the main propositions of predictive maintenance is Prognostics and Health Management (PHM) which plays a unique role in identifying, diagnosing, and predicting the health status of physical assets. To that end, one of the fundamental solutions is to assess the condition of the equipment in the aviation sector in order to provide maintenance plans by determining the trend of deterioration or destruction. In this study, a developed model of an artificial neural network was presented, focusing on the concept of deep learning and its comparison with other conventional methods, in response to the limitations and uncertainties in traditional prediction methods in determining the deterioration process of the equipment. The comparative results revealed that the deep learning neural network method with a prediction accuracy of 94% had a high performance in determining the deterioration process in aircraft turbine engines compared to other conventional methods. The findings of this study can be used to predict the remaining useful life of aviation industry equipment and to provide appropriate maintenance programs.

The subject of helical springs, which are vital and widely used components in various industries, particularly gains significantly more importance in the aerospace industry. Helical springs are utilized for energy storage purposes in various applications, notably in spacecraft separator mechanisms. However, one of the major challenges in designing these components is reducing their mass, especially in space applications where every gram carries substantial significance. This paper delves into optimizing the mass of helical springs used in spacecraft separator mechanisms. Initially, we solve the mass optimization problem of helical springs using conventional mathematical equations and the genetic algorithm. Subsequently, we design a new model for the springs using artificial neural networks and conduct optimization using this new model. The obtained results demonstrate that using the proposed artificial neural network model offers considerable advantages over the initial method. This optimization using the proposed model enhances the accuracy of helical spring designs. Numerical results indicate that the output in the neural network related to shear stress is 88.545, differing by 0.83% from the finite element method, and the output related to deformation in the neural network is 41.76, differing by 0.58% from the finite element method. These findings underscore the significantly improved performance of the proposed model compared to previous methods.