Reliability Analysis Method Of Crane Metal Structure Based On Dynamic Kriging Surrogate Model

As an indispensable lifting equipment in modern manufacturing and logistics transportation,cranes are widely used in construction,metallurgy,machining and other fields.With the continuous increase in the number of cranes in service and considering the working characteristics of the cranes,higher requirements are placed on the safety and reliability of the cranes.With the growing maturity of various computer simulation software,reliability analysis is mostly achieved by means of finite element simulation tools,and this process often requires repeated calls to the finite element model in the iterative process to find the most reasonable design.With the increase of various modeling complexity and the improvement of designers’ requirements for analysis accuracy,the efficiency improvement of computer simulation analysis is very limited,followed by huge computational cost and low efficiency.In addition,in practical engineering,it is more difficult to obtain accurate reliability calculation results because of the high dimension and small failure probability problems.Consequently,how to reduce calculation costs while improving efficiency and taking into account the correctness of results has become the focus of scientific research workers.In response to the aforementioned issues,this paper proposes a reliability computation method of crane metal structure based on dynamic Kriging surrogate model combined with importance sampling method.The specific methods are as follows : Firstly,apply the Latin hypercube sampling method to evenly acquire the initial sample points in the design space,and the simulation response value is acquired based on the finite element model to construct the initial Kriging surrogate model.Then,two learning functions are used to ensure the fitting accuracy of the surrogate model of the interested part of the model with the least calculating cost.Meanwhile,the improved first-order second-moment method combined with Kriging surrogate model iteration is used to find the maximum possible failure point.Thirdly,taking the maximum possible failure point as the sampling center,construct an important sampling density function in the form of Gaussian distribution.Finally,based on the established surrogate model,the reliability is calculated using the importance sampling method,and the feasibility and effectiveness of the method in this paper are verified by engineering cases.In addition,this paper constructs a new stopping criterion and combines it with the learning function,and uses numerical examples and engineering examples to verify the effectiveness of the new stopping criterion in reducing the number of finite element calls.Through engineering examples,it is verified that compared with the traditional learning function,the learning function with the new stopping criterion can save about 20 % to 46 % of the calculation cost while making the dynamic Kriging surrogate model achieve similar accuracy.The proposed method works well in balancing the fitting accuracy and computational cost of the surrogate model.