Sequential Optimization Research And Its Application Based On Performance Degradation Process

The material properties,stress state and environmental factors in the mechanical equipment system will affect the performance of the mechanism degradation,in which the internal and external factors present features of being time-variant,uncertain,complex,coupling interactional and so on.To characterize the stochastic degradation process by establishing the effective time-variant model of high precision is an important method to understand the performance of the system,which also exposes the model complexity and further relevant design problem.Many performance degradation features should be considered fully at the beginning of the structure design,which is significant to ensure the large complex equipment system is reliable,safe and stable in operation for a long time.Therefore,this paper intends to combine with the performance degradation of stochastic process abstracted from the limited experimental data,and implement the related time-variant reliability-based sequential optimization design.With the necessary equipment performance conditions,the efficiency of model optimization and the reasonable degree of accuracy can be improved.By following the simple economic concept,the structure size and complexity can then be reduced and the optimization is completed.At the same time,this paper demonstrates the two improved sequential optimization methods based on the performance degradation applied to the design optimization of the harmonic reducer,one of the typical space mechanism,which further proves the feasibility and effectiveness of the methods.The main research contents in this paper include:(1)The time-variant reliability-based sequential optimization process is put forward associated with the model validation and the time-variant sensitivity analysis.The timevariant sensitivity analysis method based on the ANOVA helps selecting the most sensitive parameters for checking and verification,increasing the computation efficiency.The actual experimental data rather than the simulated data are used to further confirm the model by the MLE method.At the same time the time-variant reliability constraints are added into the deterministic optimization.The out-crossing rate method based on PHI2+ is applied to the time-variant RBDO with the reliability index approach,which takes full consideration of the uncertainty and the time-variant feature of the design variables and design parameters.(2)The new time-variant reliability sequential optimization method is proposed based on the B-distance model validation index which expands the sequential optimization strategy again.In contrast with the maximum likelihood estimation,the Bdistance numerical index is recommended to quantify the accuracy of the model validation.When the B-distance numerical index is computed,the kernel function method is used to approximate the extreme probability density curve of the model,and the HLCDF method improves the B-distance computation efficiency significantly.(3)The time-variant harmonic transmission efficiency simulation model of high precision is constructed.The Wiener process is selected to characterize the degradation change amount of the harmonic reducer in the name of modeling statistically.Besides,according to the actual movement principle,the physics-based efficiency model is built on the harmonic drive.Then both advantages is considered and the collective simulation model is constructed integrating many environmental factors and individual differences.Finally,the basic analysis is carried out about the simulation model.(4)The two kinds of improved reliability-based sequential optimization method is applied to the harmonic flex spline design.From three aspects,the screening of the design variables and the validation parameters,the selection of objective function and the formulation of constraint conditions,the construction process of harmonic reducer optimization design model is elaborated in detail.About the lightweight design of harmonic reducer flex spline,the MLE-based method and the B-distance index-based method are compared and analyzed and it proves that the two methods are at the nearly same level in the computation efficiency,but the new sequential optimization method based on the B-distance model validation index has the characteristics of high transparency of the model validation.