Soft Sensor Of Ball Mill Fill Level Based On Cloud Model Uncertainty Reasoning And Acoustic Signal

As a traditional industrial grinding device, ball mill is widely used in the fields of architecture, mining, metallurgy, power plant and chemical industry and so forth. However, ball mill often consumes high energy and runs with low efficiency. It is mainly due to the absence of effective measurement and control of fill level. The operators usually keep ball mill running with low fill level for the purpose of safety working condition. The efficiency is therefore turned down.Along with the development of the measurement technology of ball mill, the complex characteristics of ball mill system are gradually discovered. Direct measurement has already been discarded by researchers, and the indirect measurement methods assisted by second variables are more and more popular recently. It is too difficult to model via analytical soft sensor modeling for the complexity of physical process of ball mill system. Thus more and more data-driven soft sensor modelling approaches, such as PCR, PLSR, BPNN, SVM, ELM, are introduced into the research of soft sensor of ball mill fill level.Cloud model uncertainty reasoning system imitates the way of humans thinking. It represents data with cloud concepts through extracting the fuzziness and randomness of data. In addition, the conceptual way of cloud reasoning is easily understood from the perspective of the human. It is proved that cloud reasoning system is a tool for arbitrary non-linear function approximation. The cloud reasoning system therefore provides a new choice for soft sensor of non-linear system. A novel soft sensor model based on cloud model uncertainty reasoning is presented and applied to the fill level measurement of ball mill. The main research work is as follows.(1) Through extracting the cloud conceptual parameters of vibration acoustic signals of ball mill with X information based backward cloud generator, the vibration acoustic signal features and fill level concepts are represented by cloud models.(2) Soft sensor model is built by one dimension normal cloud model uncertainty reasoning.(3) The definition of sparse and dense rule of cloud model reasoning is proposed to determine the sparsity of uncertainty rule base of fill level measurement.(4) Synthesized cloud model algorithm is employed to solve the reasoning of sparse cloud reasoning rule base. With the proposed method, the ball mill fill level measurement in the case of lacking training data becomes possible.Compared with the current popular soft sensor models of ball mill fill level on the same dataset, the proposed soft sensor model achieves a relatively high precision in the case of sufficient training set and keeps a better result in the case of incomplete training set. Thus the proposed approach is feasible and robust.