Research On The Reliability Analysis Method In Optimal Design Considering Uncertain Factors

During the machining,assembling and operating processes of the electrical device,there are uncertain factors inevitably,such as manufacturing tolerance,uneven material property and temperature variation.The existence of uncertainty usually makes the design parameters deviate from their nominal values,and results in performance degradation or even failure of the electrical device.To reduce the impact from uncertain factors and properly balance reliability with other performance indicators of electrical devices,the uncertainty in engineering should be considered at the stage of design optimization.Firstly,to measure the reliability of electrical devices affected by uncertain factors,research is conducted on a scientific reliability calculation method: adaptive Kriging-assisted weight index Monte Carlo simulation(AK-WMCS).Furthermore,the adaptive Kriging part is constructed to approximate the expensive constraint function,which is usually obtained by finite element analysis in engineering problems.To improve the accuracy of Kriging for reliability calculation,the adaptive sampling process is guided by a learning function.A weight index is introduced into Monte Carlo simulation to develop a new reliability calculation method: weight index Monte Carlo simulation,which cuts down the number of trials in traditional Monte Carlo simulation and enhances the efficiency of reliability analysis.In addition,the adaptive Kriging-assisted weight index Monte Carlo simulation is verified by two analytic functions with different non-linearity and a benchmark problem used for testing electromagnetic analysis methods.The results of reliability calculation are compared with their counterparts from reliability index approach,first order sensitivity-assisted Monte Carlo simulation,second order sensitivity-assisted Monte Carlo simulation and traditional Monte Carlo simulation method.The comparison shows the satisfying efficiency and accuracy of adaptive Kriging-assisted weight index Monte Carlo simulation when conducting reliability calculation.Finally,the adaptive Kriging-assisted weight index Monte Carlo simulation is combined with particle swarm optimization to compose AK-WMCS based reliability analysis and design optimization.To deduce the cogging torque of a permanent magnet synchronous motor,reliability based design optimization is fulfilled by AK-WMCS based reliability analysis and design optimization.Under the influence of uncertain factors,the optimization can find a solution which minimizes the cogging torque,simultaneously keeps the stator-tooth magnetic flux density in a reasonable design range with a target reliability.Without improving the production and assembly techniques,the AK-WMCS based reliability analysis and design optimization can guarantee the same property and reliability of electric machines in batch manufacturing.With the present situation that safety factor or empirical formula is usually utilized to guarantee reliability in traditional electric machine design,AK-WMCS based reliability analysis and design optimization balances reliability with other performances properly.AK-WMCS based reliability analysis and design optimization could serve as a reference method and that is significant to the subtle design and production of high-performance electric machines in the future.