Mechanical structure abnormal state rapid identification method and storage medium

Abstract

The invention discloses a mechanical structure abnormal state rapid identification method. The method includes: deploying a sensor for real-time monitoring on a mechanical structure; Constructing a finite element model, arranging a virtual sensor corresponding to a sensor arranged in the mechanical structure in the finite element model; randomly obtaining a load sample according to a mechanical structure load condition for calculation; judging whether the mechanical structure is abnormal or not according to calculation; if not, collecting virtual sensor data; if yes, abandoning the load sample; and inputting the virtual sensor strain data into a machine learning model as a learning sample, carrying out model training, inputting the sensor data into the machine learning model, carrying outconfidence judgment without abnormal structure, and combining a numerical simulation method with a machine learning method, thereby solving the problems of training data source and rapid identification and prediction.

Description

technical field [0001] The invention belongs to the technical field of structural health state monitoring and evaluation, and in particular relates to a method for quickly identifying abnormal states of mechanical structures. Background technique [0002] Civil engineering, mechanical engineering, bridge engineering, hydraulic engineering and other fields often involve mechanical structures such as trusses and dams. Factors such as extreme working loads and material fatigue may lead to abnormal states such as plastic deformation, yielding, or even fracture of the structure, which in turn affects the work of the entire system and even causes major losses. Therefore, how to monitor the structural health status and predict the occurrence of abnormal status in time is an important problem to be solved in this field. In order to realize the health monitoring of the structure, the common method is to deploy a large number of sensors on the structure to form a sensor network, and ...

AI Technical Summary

Problems solved by technology

[0003] In recent years, the application of machine learning methods has provided new research methods for data-driven analysis methods. The basic idea is to use historical monitoring data to train machine learning models, and then use real-time monitoring data to predict structural states. However, a large amount of effective historical data in specific applications are often difficult to obtain, limiting the application of the method

Numerical simulation method can simulate various abnormal state parameters in the design stage of the structure, but the problem we have to solve is when the structure is abnormal during work. If the calculation is based on the actual load and boundary conditions in the work, it will take a lot of money Computational time, resulting in the inability to predict the state of the structure in time

Images