The overflow slurry concentration (OSC) of a hydrocyclone is a key performance indicator (KPI) of a semi-autogenous ball mill crusher (SABC) system. Accurate modeling and prediction of the indicator can improve the grinding efficiency and product quality of the process. However, the mechanism of this process is complex, and the gradual wear of key equipment leads to data drifts in the measured results. To address these problems, an online updating soft sensor that combines a stochastic configuration network (SCN) with a dynamic forgetting factor sliding window (SW) technique was proposed. First, an SCN was used as the basic learner to build an offline model with the initial dataset. Second, a new data stream was continuously obtained and divided into different SWs according to the time sequence. Third, a dynamic forgetting factor approach that assigned different factors to diverse SWs was designed to reduce the redundancy in the historical data and fully utilize data information at different times. Finally, the effectiveness of our approach was verified through a public case and an actual industrial SABC case. In the SABC case, our approach improved the prediction mean square error (PMSE) by 25% and 37%, prediction mean absolute error (PMAE) by 17% and 23%, and $R^{2}$ by 6% and 10%, compared with the state-of-the-art just-in-time learning (JITL) and the time difference (TD) methods, respectively. Comparative results demonstrated the effectiveness and superiority of the proposed approach.

Industrial processes are often nonlinear and multivariate and suffer from non-Gaussian noise and outliers in the process data, which cause significant challenges in data-driven modelling. To address these issues, a robust soft-sensing algorithm that integrates Huber’s M-estimation and adaptive regularisations with multilayer perceptron (MLP) is proposed in this paper. The proposed algorithm, called RAdLASSO-MLP, starts with an initially well-trained MLP for nonlinear data-driven modelling. Subsequently, the residuals of the proposed model are robustified with Huber’s M-estimation to improve the resistance to non-Gaussian noise and outliers. Moreover, a double L1-regularisation mechanism is introduced to minimise redundancies in the input and hidden layers of MLP. In addition, the maximal information coefficient (MIC) index is investigated and used to design the adaptive operator for the L1-regularisation of the input neurons to improve biased estimations with L1-regularisation. Including shrinkage parameters and Huber’s M-estimation parameter, the hyperparameters are determined via grid search and cross-validation. To evaluate the proposed algorithm, simulations were conducted with both an artificial dataset and an industrial dataset from a practical gasoline treatment process. The results indicate that the proposed algorithm is superior in terms of predictive accuracy and robustness to the classic MLP and the regularised soft-sensing approaches LASSO-MLP and dLASSO-MLP.

Zhao, J.; Zhou, J.; Li, J.; Song, T., and Zhou, C., 2019. Process sampling technique for a kind of suspension liquid and application for online grade analysis in mineral processing. In: Guido-Aldana, P.A. and Mulahasan, S. (eds.), Advances in Water Resources and Exploration. Journal of Coastal Research, Special Issue No. 93, pp. 102–112. Coconut Creek (Florida), ISSN 0749-0208.Suspensions, which usually contain slurries of various sizes of particles, which can settle quickly in a short period of time and contain a large amount of unavoidable interfering bubbles inside, so automatic unattended sampling and measurement of such suspensions is very difficult. However, sampling and measurement of suspensions plays an important role in online grade analysis in mineral processing control. Based on the above problems, this paper proposes an on-line analysis and prediction algorithm for suspension based on X-ray fluorescence analysis. Based on this, a large-scale online analysis system for slurry suspension is developed, which is successfully applied to the analysis of slurry grade in mineral processing control process. Unlike other liquids, mud does not allow for complex and slow process sampling and inspection processes. Since the mud detection accuracy is easily affected by the physical form of the slurry, the detection process must ensure transientity. It can be seen that the process sampling and mixing of the slurry is another key technology. In this paper, BOXA team finished simulation and experiment for primary sampling, secondary division and slurry carrying in cell based on CFD (Computational Fluid Dynamics) fluid simulation technology and analysis it, the laboratory and industrial data shows that sampling technique of BOXA can handle the slurry into standard state rapidly and steadily to ensure accuracy and reliability for the accuracy of assay and measurement.