Abstract:
Abstract: Working in harsh environments and bumpy roads, ground vehicles such as the agricultural vehicle and construction machinery often suffer from random load which may obey different load distributions. Aiming at the problem that the single distribution is difficult to fit the multi-peak form of the load, the traditional load spectrum compiling method is improved to obtain the load spectrum that reflects actual working conditions. During the field test of wheel loader, the semi-axle load data, the speed and the bucket cylinder displacement are collected through sensors and data acquisition system. The operation modes of wheel loader can be reflected by the above data. Then the semi-axle load data are divided into 6 sections according to the operation modes of wheel loader. The 6 sections are no load forward section, spading section, full load backward section, full load forward section, unloading section and no load backward section. The load mean, load amplitude and their corresponding frequency are obtained after conducting rain-flow counting for each section. Both the single distribution and the mixed distribution are applied to fit the load mean and load amplitude in each section. The maximum likelihood estimation method is used for single distribution estimation, and the maximum expectation algorithm is used for mixed distribution estimation. The log-likelihood function values and decision coefficients are applied to the fitting test. The larger the log-likelihood function value and the decision coefficients, the better the fitting results. The fitting test results show that the fitting effects of the mixed distribution for both load mean and load amplitude are better than those of the single distribution in no load forward section, spading section, full load backward section, unloading section and no load backward section. The decision coefficients of the mixed distribution for load mean are 32%, 2.3%, 25.1%, 40.1% and 160.8% respectively larger than the corresponding decision coefficients of the single distribution. The decision coefficients of the mixed distribution for load amplitude are 8.3%, 6.7%, 1.4%, 6.2% and 1.2% respectively larger than the corresponding decision coefficients of the single distribution. For the load mean in full load forward section, the corresponding fitting test values of the mixed distribution are larger than those of the single distribution, which shows the fitting effect of the mixed distribution is better. Different conclusions are obtained for the load amplitude in the full load forward section. The fitting test values of the load amplitude are ?1 417.10 and 0.995 9 for the single distribution, while ?143 0.50 and 0.962 2 for the corresponding mixed distribution. The above values show that the fitting effects of the single Weibull distribution with the parameters of (8.717 1, 0.887 1, 9.344 1) for the load amplitude are better. It is clear to demonstrate that the load data of different sections may present different characteristics and obey different distributions. Furthermore, the load spectrum can be affected by the fitting effect caused by the distribution type according to the compiling procedure. In this paper, the load spectrum compiling method is improved. The comparisons of fitting effects of different distributions are added before deciding the distribution function. Based on the distribution which has good fitting effects in each section, the maximum values of the load mean and load amplitude of each section are determined. The joint probability density functions are also obtained. Then the frequency extrapolation and synthesis are carried out. Next the two-dimensional load spectrum is obtained. According to the Goodman theory and the fatigue damage theory, the two-dimensional load spectrum is converted into a one-dimensional load spectrum with load mean of 0. The proposed method is helpful to solve the problem of fitting multi-peak form of the load and contributes to compile the load spectrum that reflects actual semi-axle load conditions of wheel loader.