Flocculation-dewatering prediction of fine mineral tailings using a hybrid machine learning approach.

Polymer-assisted flocculation-dewatering of mineral processing tailings (MPT) is crucial for its environmental disposal. To reduce the number of laboratory experiments, this study proposes a novel and hybrid machine learning (ML) method for the prediction of the flocculation-dewatering performance. The proposed ML method utilizes principle component analysis (PCA) for the dimension-reduction of the input space. Then, ML prediction is performed using the combination of particle swarm optimisation (PSO) and adaptive neuro-fuzzy inference system (ANFIS). Monte Carlo simulations are used for the converged results. An experimental dataset of 102 data instances is prepared. 17 variables are chosen as inputs and the initial settling rate (ISR) is chosen as the output. Along with the raw dataset, two new datasets are prepared based on the cumulative sum of variance, namely PCA99 with 9 variables and PCA95 with 7 variables. The results show that Monte Carlo simulations need to be performed for over 100 times to reach the converged results. Based on the statistic indicators, it is found that the ML prediction on PCA99 and PCA95 is better than that on the raw dataset (average correlation coefficient is 0.85 for the raw dataset, 0.89 for the PCA99 dataset and 0.88 for the PCA95 dataset). Overall speaking, ML prediction has good prediction performance and it can be employed by the mine site to improve the efficiency and cost-effectiveness. This study presents a benchmark study for the prediction of ISR, which, with better consolidation and development, can become important tools for analysing and modelling flocculate-settling experiments.

Hybrid Artificial Intelligence Approaches for Predicting Buckling Damage of Steel Columns Under Axial Compression.

This study aims to investigate the prediction of critical buckling load of steel columns using two hybrid Artificial Intelligence (AI) models such as Adaptive Neuro-Fuzzy Inference System optimized by Genetic Algorithm (ANFIS-GA) and Adaptive Neuro-Fuzzy Inference System optimized by Particle Swarm Optimization (ANFIS-PSO). For this purpose, a total number of 57 experimental buckling tests of novel high strength steel Y-section columns were collected from the available literature to generate the dataset for training and validating the two proposed AI models. Quality assessment criteria such as coefficient of determination (R2), Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) were used to validate and evaluate the performance of the prediction models. Results showed that both ANFIS-GA and ANFIS-PSO had a strong ability in predicting the buckling load of steel columns, but ANFIS-PSO (R2 = 0.929, RMSE = 60.522 and MAE = 44.044) was slightly better than ANFIS-GA (R2 = 0.916, RMSE = 65.371 and MAE = 48.588). The two models were also robust even with the presence of input variability, as investigated via Monte Carlo simulations. This study showed that the hybrid AI techniques could help constructing an efficient numerical tool for buckling analysis.

Prediction and Sensitivity Analysis of Bubble Dissolution Time in 3D Selective Laser Sintering Using Ensemble Decision Trees.

The presence of defects like gas bubble in fabricated parts is inherent in the selective laser sintering process and the prediction of bubble shrinkage dynamics is crucial. In this paper, two artificial intelligence (AI) models based on Decision Trees algorithm were constructed in order to predict bubble dissolution time, namely the Ensemble Bagged Trees (EDT Bagged) and Ensemble Boosted Trees (EDT Boosted). A metadata including 68644 data were generated with the help of our previously developed numerical tool. The AI models used the initial bubble size, external domain size, diffusion coefficient, surface tension, viscosity, initial concentration, and chamber pressure as input parameters, whereas bubble dissolution time was considered as output variable. Evaluation of the models' performance was achieved by criteria such as Mean Absolute Error (MAE), Root Mean Squared Error (RMSE) and coefficient of determination (R2). The results showed that EDT Bagged outperformed EDT Boosted. Sensitivity analysis was then conducted thanks to the Monte Carlo approach and it was found that three most important inputs for the problem were the diffusion coefficient, initial concentration, and bubble initial size. This study might help in quick prediction of bubble dissolution time to improve the production quality from industry.