Research On Time-variant Reliability Analysis And Optimization Method For Key Components Of Industrial Robot

As one of the core equipment of intelligent manufacturing,industrial robot has become an important symbol to measure the manufacturing level,and scientific and technological level of a country.However,due to the blockade of foreign core technologies and the complex structural composition,the reliability of industrial robot produced by our country is obviously lower than that of foreign countries' advanced level.In particular,the key components of industrial robot are very prone to failure including reducer,controller,driver,servo motor,etc.Thus,it is very valuable to improve the quality of domenstic robots by studying on the method of the reliability analysis and optimization for key components of industrial robot.At present,with the continuous development of reliability technology,time-variant reliability has gained a wide attention in academia and industry,which aims to analyze the reliability of a product under the comprehensive influence of random process load and material performance degradation.In actual operation of industrial robot,due to the repetitive back and forth motion,the key components are usually subjected to stationary random process load and their structural performance is also gradually degraded.Under these cases,the conventional static reliability methods cannot accurately depict the time correlation of random process load and the cumulative effect of material performance degradation.For these reasons,the time-variant reliability analysis and optimization for key components of industrial robot are studied in this dissertation.The main research contents and contributions are summarized as follows:(1)Time-variant reliability analysis under a single failure mode based on Kriging surrogate modelDue to the extreme-value-based double-loop time-variant reliability method has a huge computational cost,this dissertation proposes a novel single-loop time-variant reliability analysis strategy.The best values are in current samples are used to approximate the extreme values and thus the inner extremal optimization in double-loop procedure for time-variant reliability analysis can be avoided,which can effectively decouple the double-loop procedure into single-loop procedure.Based on the approximated values,the surrogate model of extreme value response surface can be constructed by using Kriging.The relevant strategy and learning function can be further adopted to update the surrogate model iteratively until the convergence criterion is satisfied.Therefore,the time-variant reliability analysis can be performed based on the well-updated surrogate model.Meanwhile,the proposed single-loop strategy is compared with the existing single-loop procedure for time-variant reliability to illustrate a higher efficiency and it is also applied into time-variant reliability analysis of industrial robot controller to further demonstrate the effectiveness.(2)Time-variant reliability analysis under multiple failure modes based on multiple response Gaussian processUnder the comprehensive influence of random process load and structural material performance degradation,the failure mode of key components of industrial robot behaves not only time-variant characteristics but also various,namely it belongs to time-variant reliability problem under multiple failure modes.However,the classical Kriging method cannot directly construct the surrogate model of multi-output variables and the correlation between output variables cannot even be depicted in process of constructing model.In term of this situation,this dissertation introduces the multiple response Gaussian process(MRGP)to construct surrogate model of extreme value response surface based on the presented single-loop strategy.Thus,the time-variant reliability analysis can be performed based the surrogate model which is updated well by the relevant strategy and learning function.Simultaneously,the proposed method is applied into time-variant reliability analysis of industrial robot driver and the Monte Carlo simulation(MCS)is also used to demonstrate the effectiveness.(3)Time-variant reliability analysis for a small failure probability problem by combining MRGP and subset simulationWith the improvement of domestic manufacturing technology,key components of industrial robot involve not only the multiple failure modes but also a small failure probability.In terms of these cases,it is a major difficulty in actual engineering that how to accurately and efficiently assess the failure probability.Therefore,based on the proposed time-variant reliability method under multiple failure modes,this dissertation introduces a MRGP-SS approach for time-variant reliability problem under a small failure probability by combining the MRGP and subset simulation(SS).The MRGP model is herein used to depict the correlation between multiple failure modes and the SS method is adopted to assess the small failure probability.Meanwhile,the presented method is compared with the exsiting time-variant reliability method under multiple failure modes to illustrate a higher efficiency.The proposed method is also applied into the time-variant reliability analysis of industrial robot RV reducer and the MCS method is adopted to futher demonstrate the effectiveness.(4)Time-variant sequential optimization and reliability assessment based on the surrogate model of extreme value response surfaceConsidering the actual operating space of industrial robot,the volume of key components is usually not too large and tends to be miniaturized.Thus,under ensuring the reliability and safety of various performance indictors,it is very valuable for industrial robot to reduce the volume of key components as much as possible.In view of this,based on the proposed time-variant reliability methods,this dissertation presents a time-variant sequential optimization and reliability assessment(SORA)method based on the surrogate model of extreme value response surface.For the case that time-variant reliability constraints cannot be usually expressed analytically,the proposed single-loop procedure for time-variant reliability analysis is adopted to build the surrogate model of extreme value response surface and thus the time-variant reliability constraints can be equivalently converted into the static reliability constraints based on the surrogate model.Further,the SORA method can be used for optimization.Meanwhile,the presented method is applied into the time-variant reliability optimization of industrial robot harmonic reducer to demonstrate its effectiveness.