Reliability Optimization Design Based On Non-probabilistic Uncertain Mechanical Structure

In practical engineering problems,mechanical structures have many uncertain factors that affect their structural performance,such as material characteristics,load environment,structure size,etc.Small fluctuation of any of these uncertain factors may lead to large deviation of structural performance.To optimize the design of mechanical structure with uncertainty is to take full account of the existence of uncertain factors so as to obtain a stable and efficient structural design scheme.Due to the influence of error and application environment,the actual parameter value of mechanical structure can hardly reach the design value.Small deviations from the actual situation may lead to deterioration or even failure of mechanical structure.The existing research mostly adopts the method of model transformation to transform the optimization design model of uncertain structure into a deterministic model.In the process of model transformation,it is easy to cause information loss,and various parameters need to be introduced,so that the calculation results are greatly deviated.In addition,the existing research often ignores the difference between the design value and the actual value of the mechanical structure,and ignores the influence of the disturbance on the structure.Therefore,the reliability optimization design based on improbability uncertain mechanical structure is studied.The disturbance of the design vector and the fluctuation of the objective function are fully considered to avoid the disturbance causing the fluctuation of the target value to be too large,and at the same time,other constraints are still satisfied within the disturbance range,so that the mechanical products still meet the use requirements even in the harsh environment.The method of solving the model directly without model transformation is proposed,which provides a new idea for the optimization design of uncertain structure.In this thesis,radial basis function(RBF)neural network and BP neural network are used as approximation models.RBF neural network was chosen because it was necessary to predict the related parameters of mechanical structure,and the requirements of prediction accuracy,optimal approximation performance and global optimal characteristics were considered.The BP neural network is used to obtain the explicit mathematical expression.Genetic(GA)algorithm and particle swarm optimization(PSO)algorithm were used to optimize the initial weights and thresholds of RBF neural network and BP neural network respectively,and the prediction effects of the two optimized neural networks were compared.The results showed that the optimization effect of PSO was better,so the optimized neural network was adopted in this paper.Considering the disturbance of the design vector,the interval based structural improbability robust design model was established,and the three-layer nested genetic algorithm and PSO improved RBF neural network were combined,and the design vector was ranked according to the constraint violation vector value to achieve the direct solution of the model.Numerical examples of cantilever beam and application examples of crane plate are designed,respectively,and compared with original values and interval based robust design methods,to verify the effectiveness and superiority of the proposed method.The interval reliability is introduced and the perturbation of the design vector is fully considered.According to the formula 1 model,the uniform calculation method of interval reliability was summarized,and the design vector was ranked.The reliability design model was solved by the combination of three-layer nested genetic algorithm and PSO improved RBF neural network.With different reliability requirements,the reliability design of the cantilever beam in the numerical example and the plate in the application is carried out to verify the flexibility and effectiveness of the model.The reliability design of uncertain structure based on probabilistic and improbabilistic mixture is proposed.By using BP neural network,the function relation is fitted and the explicit mathematical expression of the performance function is obtained.The reliability design model of uncertain structure is solved by genetic algorithm and improved first-order second-moment method.The accuracy of the model was verified by calculating examples from other literatures and comparing the results.The model is applied to design the reliability of triple tooth disk.