Life Prediction Method Of Crane Metal Structure Driven By High Fidelity Model

Cranes,the most common type of special equipment in China,are widely utilized in the equipment manufacturing,aerospace,water conservancy and hydropower,and nuclear power building industries.The security of their product service procedure is critical to the successful implementation of projects.Metal structure,as a typical welded structure,is an essential part of cranes and serves a crucial role in transporting and transmitting loads.It is vulnerable to damage,challenging to replace,and its fatigue life directly affects the machine’s service life.Blindly discarding materials based only on the existence or absence of evident fatigue cracks would either waste resources or raise customer use prices,or it will present substantial safety risks that might result in catastrophic catastrophes.Therefore,determining the fatigue life of crane metal structures scientifically,thoroughly,and effectively,taking reinforcement measures before fatigue failure occurs,switching from passive to active maintenance,and minimizing the likelihood of fatigue fracture accidents are of utmost importance for increasing production safety,preventing accidents,cutting down on financial losses,and creating a circular economy.This article primarily undertakes the study described below in response to the aforementioned problems:1)Dynamic prediction of load spectrum.An enhanced CNN-Bi LSTM hybrid neural network load spectrum optimal prediction model under attention mechanism is established in response to the uncertainty,diversity,and randomness of the load of in-service general bridge cranes as well as the challenge in obtaining load spectrum data caused by the complexity of on-site measurement environment.The acquired small sample data is processed using the 5-fold cross validation method,and the split data set is expanded using Latin hypercube sampling while retaining the original data features.The construction of an enhanced CNN-Bi LSTM model(ICNN-Bi LSTM-AM)based on the attention mechanism.The attention mechanism assigns various weights to each input piece of information according to its importance through automatic learning to achieve effective resource allocation for information processing after the convolutional layer extracts the spatial features of the data and the two-way short-term memory network extracts the temporal features.To complete the dynamic prediction of the load spectrum and provide accurate and reliable load information for the subsequent fatigue life analysis,the Bayesian optimization algorithm is used to determine the optimal hyperparameter of the model,and the sliding window method is used to realize the dynamic update of the input data.2)High fidelity model construction.A safe and reliable universal high-fidelity modeling technique for general bridge cranes is explored in order to provide real-time monitoring of the full lifetime of in-service cranes,accomplish safe production,high quality and efficiency,rapid reaction,and prompt decision-making.Define the meaning and typical features of high-fidelity models,and explain the modeling method for general bridge crane high-fidelity models from three perspectives: geometry,mechanism,and data.Detail the process of acquiring,transferring,evolving,and updating high-fidelity model knowledge during crane service to achieve dynamic virtual-real mapping of high-fidelity models.Ultimately,provide a convenient and practical high-fidelity virtual model of cranes for subsequent fatigue life analysis.3)Multiaxial fatigue life assessment.To achieve fast and precise evaluation of the fatigue life of metal structures in in-service cranes,and reach accurate judgment of whether cranes can be utilized safely,we construct a multi-criteria critical surface method for assessing multi-axis fatigue life of metal structures.Establish a real-time acquisition system for crane service information,complete the real-time collection of indicator information,statistics,induction,and conversion into load characteristic data,and use the ICNN-Bi LSTM-AM model to depict the load spectrum during the fixed inspection period,using the general bridge crane as the research object.By using joint modeling and virtual transformation,a virtual prototype corresponding to the physical entity was created,the stress information at fatigue testing points was extracted using rigidly coupled dynamics simulation and stress analysis,and the accuracy of the theoretical simulation model was verified in conjunction with the testing scheme,so that the expansion of the stress information at nodes was completed through simulation compensation.The node stress information is converted into an equivalent stress history using three critical surface methods under four different types of multi axis high cycle fatigue criteria,and the damage parameters on the critical interface are extracted using a combination of multi axis rain flow counting.The coefficient correction technique is presented to compute the stress life curve of the structure based on the life curve of the parent material.The damage parameters are utilized to compute the degree of fatigue damage at the hazardous places of the metal structure during each cycle,and the cumulative total damage and fatigue life are derived by combining the linear damage accumulation theory.