Wear Prediction And Layout Optimization Design Of Shield Disc Cutter

With the continuous increase in urbanization,more and more attentions are paid to the development of underground space in our country.As the main shield construction tool,the shield machine is more and more invested in the construction of underground space.During the process of tunneling,the disc cutter on the cutter disk is at the front end of the tunneling direction,which is very easy to be worn,therefore,the effective prediction and control of disc cutter wear is of great significance for reducing construction costs and improving tunneling efficiency.The thesis focuses on two aspects of shield disc cutter wear prediction and optimization of disc cutter layout.By combining Gramian Angular Difference Field(GADF)and Convolutional Neural Network(CNN),a disc cutter wear prediction model is established,and a disc cutter wear prediction system is designed to realize the effective prediction of disc cutter wear quantity.Based on the prediction system,the multi-objective optimization model of the center disc cutter layout is constructed and solved,and the optimized center disc cutter layout scheme is obtained.The main research contents are as following:(1)The wear mechanism and wear failure modes of shield disc cutter are analyzed,and the research object is determined to be disc cutter with normal and uniform wear.Relying on the shield interval data of a certain project,combined with the empirical formula of wear prediction,a set of disc cutter wear data set is established.The data set is preprocessed by GADF and converted into a set of 2D images.(2)An improved CNN is constructed based on the Py Torch deep learning framework.The preprocessed disc cutter wear data set is used as the input layer,and the output layer is the disc cutter wear quantity.Through the training comparison,the prediction performances of the model under different hyper-parameters are analyzed,and the optimal settings of hyper-parameters of the network model are determined.On this basis,the influences of different convolutional layers and different convolutional kernel sizes on the prediction performance of the model are analyzed.The optimal structure of the model is determined,and a disc cutter wear prediction model is established based on GADF-CNN.(3)By the prediction result comparisons of the GADF-CNN model with the BP neural network model and the unimproved GADF-CNN model,it shows that the prediction deviation of the GADF-CNN model is significantly smaller,and its comprehensive prediction effect is better.On this basis,shield disc cutter wear prediction system is designed with Py Qt5,including front-end human-computer interaction interface and back-end data conversion and data prediction,which can easily and quickly obtain the prediction results of disc cutter wear quantity.(4)Combined with the shield disc cutter wear prediction system,the differences of wear quantity of center disc cutter,face disc cutter and edge disc cutter on the cutter disk are analyzed,and the layout optimization design of the center disc cutter is determined.Taking the radial load of the cutter disk,the overturning moment of the cutter disk and the standard deviation of the rock breaking quantity of the disc cutters as the optimization goals,a layout optimization model of the shield center disc cutter is constructed.The multi-objective optimization solution is carried out by using Nondominated Sorting Genetic Algorithm Ⅲ(NSGA-III).And the result comparison shows that the optimized disc cutter wear condition and the force status of the cutter disk are both improved to a certain extent.