Optimized Distribution Of Dampers Based On Machine Learning

Since the 21 st century,artificial intelligence technology has developed vigorously and has been widely and successfully applied in various fields.As an important application of artificial intelligence,machine learning algorithms have become a key research direction in many industries and fields.As a large-scale traditional industry,the civil construction industry plays an extremely important role in the development of the world economy.The combination of machine learning methods and traditional technologies in the civil construction industry will inevitably promote technological progress and promote the optimization and upgrading of traditional industrial structures.On the other hand,the earthquake disaster is one of the most serious natural disasters in the world.Buildings may be damaged during earthquakes,which will cause great damages to human beings.Therefore,it is very important to reduce the vibration response of structures caused by earthquakes so as to reduce the losses caused by earthquakes.As a kind of energy dissipation and shock absorption device,passive dampers can effectively reduce the dynamic response of structures and reduce structural damages.At present,how to arrange dampers in the structure to achieve the best control effect is a research hotspot,and there are many mature theories and methods.This research mainly applies machine learning to the optimal arrangement of dampers,and uses machine learning methods to give reasonable optimal arrangements of dampers.The main research contents are as follows:(1)Based on the supervised learning classification theory,the machine learning classification model is used to predict the damper arrangement schemes.In all cases,the arrangement of dampers is often repeated,and there are few different arrangements of dampers.Therefore,these problems can be regarded as multi-classification learning problems,and each arrangement can be labeled as a type.Using support vector machine theory and multi-layer perceptron theory as the classifier principle,the damper optimal arrangement samples obtained from time-domain analysis optimization are trained as training sets to obtain the classification model.Then the obtained machine learning classification model is used to predict the type of damper arrangement scheme of the structure.(2)Based on the supervised learning regression theory,a machine learning regression model is used to predict the damper arrangement.As the structural parameters and the number of dampers are considered as the optimal arrangement conditions of dampers,the types of damper arrangement schemes increase rapidly.Therefore,the problems can no longer be regarded as classification problems.In this case,the number of dampers in each layer can be directly predicted,which can be regarded as multi-output regression problems.Each output corresponds to the number of dampers in the corresponding structural layer.Using the multilayer perceptron theory as the regression principle,the regression model is obtained by training the damper layout optimization samples obtained from the time-domain analysis optimization,and the damper layout scheme of the structure is predicted by using the obtained machine learning regression model.(3)Based on the supervised learning regression theory,considering the difference of damper arrangement caused by site differences,the regression learning model has the ability to predict the different damper arrangement under different site conditions.Using the multi-layer perceptron theory as the regression principle,the optimal damper layout samples obtained from the frequency-domain analysis optimization with site parameters are trained to obtain the regression model.The obtained machine learning regression model is used to predict the damper arrangement scheme of the structure.At the same time,the ranges of structural parameters as optimization conditions are extended to improve the application ability of machine learning model.It is proved by calculation and analysis that the application methods proposed in this thesis can achieve better results,which provides a theoretical possibility for applying machine learning theory to the optimal distribution of dampers in the field of structural vibration control.