Load Spectrum Compiling Method Based On Modified Extrapolation In Construction Machinery

Characteristics of construction machinery are various working materials,complicated working modes and harsh working environment,which affect the reliability and safety of the whole machine due to the high frequency of the components fatigue and failures.In order to realize the design of fatigue durability and controllable life in the initial stage of product design and ensure the service reliability of the whole machine and key parts,using scientific and reasonable load spectrum is the foundation and objective basis to realize the above design requirements.Therefore,research on the load spectrum compiling method of construction machinery has important theoretical significance and engineering application value.During the actual operation of the construction machinery,it is difficult to obtain the full life cycle load spectrum of the components,and only limited time periods of load under certain working condition can be performed for time consuming and costly.The reasonable extrapolation is involved to transfer limited periods load to longer periods load,especially to a design life,which provides reasonable data information for the fatigue analysis and life prediction.At the same time,it can also guarantee that the existing construction machinery products and key components have the basis of load in the process of optimization design.This paper studies the process of load spectrum compiling systematically.The load acquisition,characteristics and extrapolation method are studied,combined with the extreme value theory,decision making technology and optimization algorithm.Firstly,load acquisition schemes for the axle-shaft torque of wheel loader and the hydraulic pump pressure of the excavator are developed.The composition of the measured system and the typical working conditions have been illustrated.And the working mode has been divided in phases to ensure the authenticity of the load data source and the stability requirements of the data analysis.Based on the load data of each working phase of the excavator,the energy consumption is used to evaluate the drivers' operation performance to study influence of the load characteristics caused by different drivers.According to the analysis of the load characteristics of the excavating-unloading working cycle,the average time difference of each driver is within 7%,the average power value difference is within 4%,and the energy consumption difference is within 10%.The result verifies the existence of the differences between the drivers and shows the degree of the influence.And the above values can provide reference for determine the necessity of drivers' individual differences when developing an acquisition scheme.Aiming at the high sensitivity problem of threshold selection of load extrapolation in time domain(LETD)based on POT,a threshold selection method based on multi-criteria decision making(MCDM)is proposed.The entropy,VIKOR and POT requirements are included.From the perspective of fitting test results,the entropy is used to determine the weight of the test criterion.The VIKOR is used to obtain the compromise solutions.According to the POT requirements,the minimum value of the compromise solutions is selected as the optimal threshold.The feasibility of the proposed method is verified by using 500 groups of simulation data.Finally,the method is applied to the axle-shaft spading phase load.Compared with the classical method,the results show that the extreme load 9161 Nm based on the MCDM threshold selection method is suitable,and the established POT model not only retains more valid load information,but also shows a better comprehensive fitting effect.Aiming at the problem of load multi-peak form,the mixture distribution is suggested as the load distribution model firstly.To avoid the analysis of the correlation between load mean and load range,a load extrapolation model based on rainflow matrix is proposed,which uses the three-parameter Weibull-normal distribution(3W-n)as the mixture distribution basic function type,and components number of the mixture is determined by the minimum AIC value.Then,the genetic algorithm(GA)is proposed when the EM does not converge.Through maximizing the value of log likelihood function,the GA method can realize the parameter estimation of the mixture distribution based on the rainflow matrix.To verify the feasibility of the GA method based on 3W-n,the EM method based on 3W-n is used as a comparison.The result shows,the two AIC values are close to each other with a difference of 26,which is 0.13% of the optimal value 20778(EM).Therefore,the GA method can be taken as an alternative especially when the EM does not converge.In addition,the GA method model is simple and suitable for function distribution of various complex types.Discrete operation of the traditional parameter extrapolation method(PE)leads to loss of measured load information,and an improved PE method is proposed.This method puts forward the idea of “extrapolation after equivalent”.Based on the 3W-n mixture distribution model of rainflow matrix,the edge one-dimensional load spectrum of mean and range are established.To achieve continuous load spectrum dimension reduction,the equivalent stress and range are established according to the stress correction method,then a continuous one-dimensional equivalent stress spectrum has obtained according to the functional relationship transformation.Finally,the superiority of the improved PE method is verified from the perspective of the load information utilization,the edge distribution function of the average range load and the equivalent stress acquisition.The analysis results show that the improved PE method can make the best of the load information.The load spectrum obtained by the improved PE method are closer to the actual working condition,because the relative integrated absolute error(RIAE)values of the mean and range are smaller than the corresponding values of the traditional PE method,and the difference is 1.26 times or more.While the one-dimensional equivalent stress obtained by the traditional PE method has the problems of large dispersion and wide distribution range.In order to obtain the possible load during the full life cycle of the components accurately,the differences of the selected extrapolation methods of the load in different domains are studied based on the pseudo-damage theory.The LETD method based on MCDM and the improved PE method are used as examples,the pseudo-damage values obtained by each method under different extrapolation factors are calculated,together with the damage ratio.According to the coefficient of variation of pseudo-damage,the value 0.0102 shows the randomness of the extrapolated extreme value has no effect on the damage of LETD.Compared with the damage results obtained by different methods in different domains under different extrapolation factors,it shows that the reliability of the specific damage value of LETD method depends on sample size.Finally,the application scope of each method is summarized with the comparison results.The axle-shaft torque of wheel loader and the hydraulic pump outlet pressure of excavator are used to compile load spectrum,and the corresponding extrapolation methods are determined by the requirement of application scope.The load extrapolation is performed according to the working phase and the synthesizing is used to obtain the load spectrum of the full working phases.In this paper,a set of load extrapolation methods are proposed in time domain and rainflow domain respectively.These methods provide a reference for the load spectrum compiling of construction machinery components,and provide theoretical foundation for the subsequent reliability test and accelerated life testing.