Classification Of Anchor Grouting Quality Based On Random Forest Algorithm

Bolt support is now widely used in various geotechnical projects as a commonly used reinforcement method.Affected by construction factors,there are not a few safety problems caused by unqualified grouting quality in the project,so the quality inspection of anchor bolt grouting should be studied.In this paper,the random forest algorithm is used to carry out signal analysis to detect the quality of bolt grouting.Since the sample data of grouting quality belongs to unbalanced data,the random forest is improved from feature selection and the algorithm itself,and verified on experimental and simulation data.The test proves that the improved random forest algorithm performs better in the classification of anchor grouting quality.The main work of this article is as follows:(1)Introduced the principle of stress wave detection of the quality of anchor bolt grouting,and pre-processed the data set.In the data preprocessing work,the theoretical basis of wavelet transform and wavelet packet is introduced.First,the db10 wavelet fixed soft threshold denoising method is used to denoise the anchor simulation signal,and then the 16-dimensional energy value is extracted by 4-layer wavelet packet decomposition Represented signal characteristics.(2)Based on the simulation data,a random forest algorithm anchor bolt grouting quality classification was made.First,the principle of random forest classification algorithm and classification evaluation method are described,and then the principle of calculation of feature importance of random forest algorithm is introduced;based on the unbalanced anchor data,balanced random forest and weighted random forest algorithm are used to complete the classification of anchor data.(3)An improved random forest algorithm is proposed,which uses this method to complete the classification of the quality of anchor grouting.On the one hand,the random forest algorithm is optimized from feature selection,and then combined with the advantages of the balanced random forest algorithm and the weighted random forest algorithm,the balanced weighted random forest algorithm is used to complete the classification of anchor data.On the other hand,the feature importance discrimination method of random forest is improved,and the method is verified using the line graph method,and then the feature extraction is completed from the strong and weak feature areas obtained by using the probabilistic feature extraction method.Combining balanced random forest and weighted random forest to form a balanced weighted random forest algorithm,this method is used to complete the classification of anchor data.(4)Use the experimental data to complete the verification of the quality of anchor grouting.Firstly,the acquisition of the anchorage grouting quality detection signal was introduced,and then the anchor data collected by experiments were used to verify the improved random forest algorithm for feature selection and the improved balanced weighted random forest algorithm.