Xu Degang

Constrained generalized predictive control strategy for three-level air tank pressure system

The three-level air tank system is widely used in the industrial manufacture as gas storage device. Because of its features like time-delay and great inertia, usually it is very difficult for the traditional PID controller to achieve ideal control performance. Aiming at solving practical problems, in this paper, we have proposed a constrained generalized predictive control (GPC) algorithm that the input amplitude, input-change-rate and output are constrained. In this algorithm, the system parameters are identified by recursive least squares (RLS) algorithm, and the prediction model is obtained by solving the Diophantine equations. The active set method is firstly applied to solve the quadratic programming, and then this rolling optimization strategy is used to calculate and obtain the optimal control law. Finally, compared with the other methods in physical experiments, our approach is more effective and can remarkably improve the control performance.

Research on dosage matching for gold and antimony froth flotation based on rough sets and case-based reasoning

Currently controlling dosage in mineral flotation process is based on the subject experiences of operators, which are influenced by different operating technique. According to different shifts, process performance of forth flotation is different so that the flotation process is often running in unstable states. Lots of production data of gold and antimony forth flotation process have been acquired, which can be used to build case rules of adding dosage for the process, build the flotation process of flotation reagents adding quantity is putted forward based on a large amount of data acquried by flotation process, combining with the characteristics of rough set to dealing uncertain knowledge, the introduction of the Cauchy inequality. A new method of flotation dose matching is proposed based on rough sets and case-based reasoning, realized the control of reagent adding in floatation process. This method has been applied in a real industrial site. The application results show that the proposed method can effectively reduce the contents of tailings and improve the recovery rate.

A novel texture extraction and classification method for mineral froth images based on complex networks

The flotation froth surface texture can be used as an indication to illustrate the production states. A novel froth image texture extraction and classification method based on complex network is presented to obtain the accurate texture features descriptors and facilitate the mineral flotation process monitoring. Firstly, froth images are pre-classified by defining a similarity coefficient. Then, designing the optimum value for the parameter p of Minkowski distance is discussed according to the pre-classification result. A network model of froth image is built utilizing complex network theory. The energy and entropy of the complex network model as texture descriptors is given in terms of the Minkowski distance. Finally, copper froth images captured are used in experiments, and texture feature are extracted. Experiment results show that the presented method can automatically select the optimum values of froth image texture extraction according to their characteristics. It can accurately describe the texture difference of different mineral flotation states, and accurately identify the floatation states.

Research on spectral clustering algorithms based on building different affinity matrix

As one of the most popular researches in the field of machine learning, spectral clustering algorithms have made great process in many different applications such as image processing. However, there are still some unsolved problems about spectral clustering algorithms, which should be immediately dealt with .These problems include how to build the affinity matrix, and how to deal with the eigenvectors. This paper mainly focuses on building the affinity matrix, which is the most important problem of spectral clustering algorithms. We propose four different methods to build the affinity matrix including the Gaussian kernel function, the Minkowski function, the nearest-correlation function and the local scale function. Then, we develop four new algorithms to contrast the clustering results. Finally, we find that building appropriate local scale function is the most available method to formulate the affinity matrix for spectral clustering algorithm.