The load spectrum simulation of actual working status is the key factor to solve the problem of crane endurance failure . T he precision and robustness of load spectrum predicting have great significance for reliability analysis of crane fatigue fracture and evaluation of its safety life .However ,the predicting performance of classic linear regression model is weaker .Compared with other data analysis algorithms ,support vector regression machine (SVRM ) has excellent performance for small sample and nonlinear properties ,including higher prediction accuracy and nice robustness .It can also overcome the difficulty of the curse of dimensionality ,local minima and over‐fitting and under‐fitting for traditional pattern recognition methods .So ,accuracy pre‐diction precision and reliability can be obtained by using SVRM .Furthermore ,an improved v‐SVRM prediction model was established with constructing new kernel function and decision func‐tion .T he results of engineering application show ed that the values of Er and of RMSRE all the three models (the BP neural network model and the SVRM model and the modified model of v‐SVRM ) gradually decreased while the fitting degrees R2 gradually increased .It proves that the modified method has higher prediction precision and nicer robustness and it also provides a new way for ob‐taining and predicting crane load spectrum .%载荷谱预测精度和鲁棒性直接影响起重机械的疲劳可靠性计算以及安全寿命评估。因此，绘制模拟实际工作状态的载荷谱是解决起重机械疲劳断裂问题的重要环节。然而传统的回归模拟算法对其预测精度较低。支持向量回归机（SVRM ）同其他数据分析算法相比，具有出色的小样本和非线性特性，预测精度高、稳健性好，可较好地解决欠学习、过学习以及局部最小值等传统回归算法的难题。因此，选用支持向量回归机预测起重机载荷谱，提高了模型的预测精度和鲁棒性。在此基础上，从核函数的构造和决策函数的建立两方面的改进，建立了改进的 v‐SVRM预测模型。工程实例分析结果表明：从BP神经网络模型、v‐SVRM模型到改进的v‐SVRM模型，Er和RMSRE逐渐减小，R2逐渐增大，验证了所提出的改进方法具有良好的实用性、鲁棒性以及较高的预测精度，为起重机载荷谱的获取与预测提供了新方法。
展开▼
