Soft Measurement Of Ball Mill Load Based On Information Fusion And Transfer Learning

Ball mill is an important basic equipment in mineral processing industry.It is very important to use soft sensing technology to detect the load parameters of ball mill to realize the optimal control of grinding process.However,in the actual industrial production process,due to the complex and changeable operation environment,the soft sensing based on single signal source has its limitations.In addition,the actual working conditions usually fluctuate with time,resulting in the decline of the soft sensing model accuracy.In view of the above problems,the method of information fusion to fuse the soft sensing results of multiple sensors is used in this paper.At the same time,transfer learning strategy is introduced to realize the soft measurement of ball mill load under unknown working conditions.The main contents of this paper are described as follows:1.The related theoretical basis of soft measurement of the ball mill load is introduced.A small ball mill in the laboratory is designed and built to collect and analyze the vibration signal of the bearing seats,so as to provide data support for the research of soft measurement of ball mill load.2.In view of the shortcomings of evidence theory,the incomplete signal information obtained by single sensor and the poor performance of single classifier,an information fusion method of multi-classifier integration and multi-sensor fusion based on improved evidence combination algorithm is proposed,which can effectively improve the soft sensing accuracy of ball mill load.3.Aiming at the situation that it is difficult to obtain labeled samples in the target domain under the condition of single source domain,a joint discriminant higher-order moment alignment network is proposed.Compared with other deep transfer learning methods,it can effectively improve the soft sensing accuracy of ball mill load under unknown working conditions.4.In view of the difficulty of obtaining labeled samples in the target domain under the condition of multi-source domain,the two improved methods proposed in this paper are combined,and a method based on information fusion and transfer learning is proposed.Compared with the single source domain method,it can effectively improve the soft sensing accuracy of ball mill load under unknown conditions.