【摘要】 为了避免集装箱吊具作业过程中因结构承载性能退化而引发巨大的安全事故问题，提出以服役信息数据为驱动的吊具结构疲劳寿命预测方法及系统。从集装箱吊具服役工况、载荷分布以及结构应力薄弱位置入手，搭建吊具信息采集系统，获取实时工作状态信息和性能参数信息，经数据处理形成用于应力预测和寿命评估的驱动数据。通过随机搜索(RS)改进随机森林(RF),构建基于RS-RF的应力预测模型，利用驱动数据实现吊具结构测点应力的实时预测，并从中提取双参数应力谱，根据Pairs模型和Miner准则，确定结构测点的损伤度和疲劳寿命评估结果。然后，从工程应用的角度，以上述方法为支撑，通过封装集成的方式，围绕运行过程监控、信息可视化、应力预测、寿命评估和数据管理等5大模块开发吊具结构疲劳寿命快速预测系统。最后，以QTH45S型可移动双箱吊具为例，利用上述方法和系统得到吊具结构的疲劳寿命为15.61年，相比基于RF、XGBoost和GBDT的寿命预测模型，预测结果精度分别提高了最少2.75%、1.54%和8.46%,验证了所提方法和系统的科学性和可行性。更多还原

【Abstract】 The significant safety accidents are easy to cause by the degradation of structural load-bearing performance of container spreaders operation process.To avoid the occurrence of the above issues, a structural fatigue life prediction method and system for spreaders driven by service information data were proposed.Considering the service conditions, load distribution and weak structural stress locations of spreaders, a lifting equipment information collection system was established to obtain real-time working status information and performance parameter information.After data processing, driving data for stress prediction and life assessment was formed.The random forest model(RF) was modified by the randomly searching algorithm(RS),and a stress prediction model based on RS-RF was constructed.The real-time stress prediction of spreader structure measurement points was achieved using driving data, and dual parameter stress spectra were extracted from it.Based on the Pairs model and Miner criterion, the damage degree and fatigue life evaluation results of structural measurement points were determined.Then, from the perspective of engineering applications, supported by the above methods, a rapid fatigue life prediction system for spreader structures was developed through encapsulation and integration, which focused on five modules: operation process monitoring, information visualization, stress prediction, life assessment and data management.Finally, taking the QTH45S movable double box spreader as an example, the fatigue life of spreader structure was obtained to be 15.61 years by the above method and system.Compared with the life prediction models based on RF,XGBoost and GBDT,the accuracy of the prediction results is improved by at least 2.75%,1.54% and 8.46%,respectively, the scientific and feasibility of the proposed method and system are verified.更多还原