بهينه سازي چند هدفه رفتار استاتيكي ورق چند لايه اي FGM بر پايه حل دقيق با استفاده از روش عصبي - ژنتيك

در اين مقاله، ابتدا روابطي جديدي براي خمش ورق چند لايه اي داراي لايه هاي FGMبه صورت تغيير خواص طبق قانوني تواني، به روش حل دقيق الاستيسيته سه بعدي مواد ناهمگن، ارايه مي شود. با استفاده از تغيير خواص طبق قانون تواني و نمايي براي يك ورق 6 لايه اي تحت بار خمشي سينوسي و با استفاده از متغيرهايي شامل ثابت هاي رابطه تغيير خواص تواني و نمايي در هر لايه و همچنين ضخامت لايه ها، بر پايه حل دقيق، تابع هدف بهينه سازي توسط كد نويسي در نرم افزار متلب تعريف مي شود. تابع هدف داراي دو هدف جرم و حداكثر تغيير شكل قايم ورق مي باشد كه بايد حداقل شوند. قيد بهينه سازي در تابع هدف، حداقل ضريب اطمينان تنشي موضعي است كه بايد همواره بزرگتر از يك باشد. سپس تابع هدف به وسيله شبكه عصبي مدل سازي مي شود. در نهايت با استفاده از روش الگوريتم ژنتيك چند هدفه با مرتب سازي غيرمغلوب بهينه سازي چند هدفه براي تابع هدف مذكور انجام مي پذيرد. نتايج نشان مي دهد كه ورق با هسته نرم و رويه هاي سخت به طوريكه تغيير خواص در راستاي ضخامت ورق پيوسته باشد، اهداف مورد نظر را بهينه ترمي نمايد. In this paper, a new formulation for exact bending analysis of FGM multi-layered plate with power-law distribution of properties in the each layer based on three-dimensional inhomogeneous elasticity is presented. Multi objective optimization is implemented for the plate under sinusoidal and uniform transverse load. Variables are the thickness of each layer and exponential or power law inhomogeneous parameters in the each layer. Objectives of optimization are taken to be mass and maximum transverse deflection of the plate. Constraint of optimization is the local safety factor based on stress criteria, that must always be larger than 1. Objective functions are coded in MATLAB software. The plate has 6 layers with power-law distribution of properties at faces and exponential properties at layers of the core. Objective functions is simulated with neural network because of the computational cost involved. Also, in this thesis, non-dominated sorting genetic algorithm method is used for multi objective optimization. Results show that the use of faces with high stiffness an core with low stiffness instead of homogeneous plate reduces the mass and the maximum normal deformation significantly.

In this paper, a new formulation for exact bending analysis of FGM multi-layered plate with power-law distribution of properties in the each layer based on three-dimensional inhomogeneous elasticity is presented. Multi objective optimization is implemented for the plate under sinusoidal and uniform transverse load. Variables are the thickness of each layer and exponential or power law inhomogeneous parameters in the each layer. Objectives of optimization are taken to be mass and maximum transverse deflection of the plate. Constraint of optimization is the local safety factor based on stress criteria, that must always be larger than 1. Objective functions are coded in MATLAB software. The plate has 6 layers with power-law distribution of properties at faces and exponential properties at layers of the core. Objective functions is simulated with neural network because of the computational cost involved. Also, in this thesis, non-dominated sorting genetic algorithm method is used for multi objective optimization. Results show that the use of faces with high stiffness an core with low stiffness instead of homogeneous plate reduces the mass and the maximum normal deformation significantly