Enhancing the productivity of cement grinding system by observing the vibration response

Abstract

At a cement plant, the grinding process is the last phase of production. the process of turning kiln-ground cement clinker into final cement by mixing it with 4-5% gypsum, limestone, and potential additives. The grinding of cement must be fine enough to meet the strength properties requirements. For a particular cement type, the productivity of grinding process and the loading of ball mill has a proportional relationship, but it's important to note that the relationship is not always straightforward and can be influenced by various factors such as the characteristics of the raw material (mainly clinker), the design and condition of the mill, the speed of the mill, and the size and shape of the grinding media. As of now, there are no reliable ways for identifying the mill blockage condition of a cement ball mill, which occurs when the mill is suddenly overloaded with material to the point of obstruction and rapid drop in grinding productivity. Mill operators intentionally reduce the grinding output by feeding the mill with less material in order to prevent overloading and subsequent mill failure. This results in a less efficient and more power-intensive grinding process. There are very few external controls that can be used to create better conditions. Only the data extracted from sensors fixed at mill motor bearings do not provide accurate readings for mill fill level. Additional vibration responses and torque responses need to be considered for better fill level predictions. Time domain vibration signals are those that are obtained through the use of an accelerometer. Sensor array design and development has been done according to capture features of vibration signals of mill at various feed rates. Proper filtering has been used to remove noises of vibration signals. Fast Fourier Transform (FFT), with the use of DALOG BusyBee software has been used to extract features from time constrained vibration information. The features extracted were utilized as an ANN’s input parameters. The material feed rate to the ball mill is estimated using the ANN's output. Regression-based Deep Learning neural network fit for the cement mill operation automation and cement mill feed rate can be predicted without forcing mill blockages by analysis of vibration responses of mill motor and mill gearbox and torque response of mill shaft.