Chongchong Qi

Comparative Study of Hybrid Artificial Intelligence Approaches for Predicting Hangingwall Stability

AbstractFive hybrid artificial intelligence (AI) approaches based on machine learning (ML) and metaheuristic algorithms were proposed to predict open stope hangingwall (HW) stability. The ML algori...

Prediction of open stope hangingwall stability using random forests

The prediction of open stope hangingwall (HW) stability is a crucial task for underground mines. In this paper, a relatively novel technique, the random forest (RF) algorithm, is introduced for the prediction of HW stability. The objective of this study is to verify the feasibility of the RF algorithm on HW stability prediction and investigate the relative importance of influencing variables. The training and verification of RF models were conducted on a dataset from the literature and a total of 115xa0HW cases were analysed. Thirteen influencing variables were selected as the inputs with the HW stability being selected as the output. The dataset was randomly divided into the training set and the testing set. Fivefold cross-validation was used as the validation method, and the grid search method was used for the hyper-parameters tuning. Performance measures were chosen to be the confusion matrix, the receiver operating characteristic (ROC) curve and the area under the ROC curve (AUC). The results show that the RF algorithm had great potential for the prediction of HW stability. AUC values achieved by the optimum RF model on the training set and the testing set were 0.884 and 0.873, respectively, indicating that the optimum RF model was excellent at predicting HW stability. The stope design method was found to be the most sensitive variable among all variables evaluated, with an importance score of 0.168 out of 1. The RQD and HW height also had a strong influence on the stability of an open stope HW.

A Real-Time Back-Analysis Technique to Infer Rheological Parameters from Field Monitoring

The long-term stress analysis of engineering projects can be significantly expedited if we can determine an appropriate rheological model and its corresponding parameters. In the present contribution, we show that an accurate and real-time estimation of rheological parameters is possible by employing deep learning and metaheuristic algorithms. A real-time back-analysis technique was proposed using a deep long short-term memory neural network (DeepLSTM) as a substitute for numerical modelling and firefly algorithm (FA) to search for the optimum parameter. The performance of the proposed technique, the DeepLSTM-FA, was verified using a tunnel response with the FLAC 2D finite difference program. Furthermore, the application of the DeepLSTM-FA to an engineering instance, namely, the Adriatic Motorway near Draga Valley, was discussed in detail, revealing that the DeepLSTM-FA can provide practitioners with an accurate and real-time estimation of rheological parameters, thereby allowing for timely stress and stability analyses. We found that an accurate estimation of rheological parameters can be made using the first few points of displacement data instead of the whole displacement profile. This technique extends recent efforts to determine rheological parameters in real time and significantly accelerates the application of stress and stability analyses in the future.

A hybrid method for improved stability prediction in construction projects: A case study of stope hangingwall stability

Abstract Artificial intelligence (AI) approaches have proliferated in stability prediction of construction projects in the past decade. However, the application of AI approaches did not reach the peak of its potential due to the inappropriate handling of missing data and the omission of state-of-the-art techniques. In the present contribution, we proposed a hybrid method for the improved stability prediction of construction projects based on individual machine learning (ML) algorithms, input missing data imputation, semi-supervised learning and the classifier ensemble. Seven ML algorithms were selected to build individual classifiers for the classifier ensemble. 5-fold cross validation was used as the validation method and the performance measures were chosen to be the confusion matrix, the receiver operating characteristic (ROC) curve and the area under ROC curve (AUC). Exhaustive grid search and firefly algorithm were used for hyper-parameters and weights tuning respectively. The capability of the proposed method was verified using an underground construction dataset, the stope hangingwall (HW) dataset. The case study shows that the input missing data imputation and semi-supervised learning improved the predictive performance of ML algorithms on HW stability prediction. The highest and average AUC values on the testing set were increased to 0.954 and 0.923 respectively on the expanded dataset, compared with 0.879 and 0.860 on the original complete dataset. Further improvement was obtained through the classifier ensemble, with the AUC value being increased to 0.976. Harnessing such method extends recent efforts for stability prediction in construction projects, and can significantly accelerate the project design and stability management.