TBM Working State Prediction And Parameter Allocation Research

Controlling of the advance rate of the tunneling boring machine(TBM)and the speed of the cutterhead accurately and efficiently is the key technology to guarantee the safe and fast tunneling of the TBM.There are no uniform standard and better method in engineering for how to optimize the parameters of the TBM to ensure that it works at the desired propulsion speed and cutterhead speed.Currently,the construction personnel completely rely on their experience in the parameter settings,and therefore the desired state still can not be efficiently controlled.For To solve this engineering problem,the data collected in the construction site of the TBM are analyzed and modeled to predict the advance rate and the speed of the cutterhead and optimize the parameters of TBM.The main work is as follows:1.The training samples with good ergodicity and the number as few as possible are screened based on the clustering idea for the massive data.A method in which the SSE and its slopes are used to adaptively determine the number of clusters of each type of samples is proposed.In order to eliminate isolated samples and speed up the modeling,the samples were cleaned and the training samples were further screened.First,the non-working state and the abnormal samples were removed according to the advancement speed.Then,the samples are divided into four categories according to the surrounding rock.Both the SSE and its slope are calculated and according to them the number of clusters in each type samples is determined.Finally,K-means clustering is performed to screen the final training samples.2.The models of BP and RBF neural network are constructed to predict the propulsion speed and cutter speed of TBM.The prediction absolute average error percentages of the BP and RBF models is 9.23% and 9.6%,respectively for the propulsion speed.The prediction absolute average error percentages of the cutterhead speed is 0.13% and 0.37% for the BP and RBF model respectively.3.A method of reversing adjustment the parameters of the equipment based on the BP prediction model of the propulsion speed and cutter head speed is proposed.A reasonable setting of the equipment parameters which make the TBM work under the expected state are obtained.Firstly,the BP prediction model of propulsion speed and cutter speed was established.Then,in the case of fixed weight and activation function,the equipment parameters were randomly input and they were adjusted based on the gradient descent until the propulsion speed and the cutter speed reach the desired working state.4.A TBM parameter estimation model based on BP neural network was established.A two-layer BP estimation model is constructed based on propulsion speed,cutter speed and soil type to evaluate the suitable vales of TBM parameters.After the net has trained,reasonable TBM parameters can be obtained by inputing the expectation propulsion speed and cutter speed and soil quality.The experimental results show that the BP model is superior to RBF for TBM,regardless of propulsion speed,cutter speed or equipment parameter estimation.In addition,all the prediction accuracy of the working state and the estimation accuracy of the equipment parameters meet the project requirements.This means that the methods proposed in this thesis are effective.