ORIGINAL ARTICLE Monitoring the fill level of a ball mill using vibration sensing and artificial neural network Dilip Kumar Nayak 1 • Debi Prasad Das 2 • Santosh Kumar Behera 2 • Sarada Prasad Das 2 Received: 21 November 2018 / Accepted: 5 October 2019 Ó Springer-Verlag London Ltd., part of Springer Nature 2019 Abstract Ball mills are extensively used in the size reduction process of different ores and minerals. The fill level inside a ball mill is a crucial parameter which needs to be monitored regularly for optimal operation of the ball mill. In this paper, a vibration monitoring-based method is proposed and tested for estimating the fill level inside a laboratory-scale ball mill. A vibration signal is captured from the base of a laboratory-scale ball mill by using a ± 5 g accelerometer. Features are extracted from the vibration signal by using different transforms such as fast Fourier transform, discrete wavelet transform, wavelet packet decomposition, and empirical mode decomposition. These features are given as input to an artificial neural network which is used to predict the percentage fill level inside the ball mill. In this paper, the predicted fill level obtained by using different features are compared. It is found that the predicted fill level due to features obtained after fast Fourier transform outperforms other transforms. Keywords Artificial neural network  Ball mill  Vibration analysis  Feature extraction  Mill fill level 1 Introduction Ball mill/tumbling mill is used to grind ores and minerals during beneficiation processes. It is used for size reduction in minerals in a grinding circuit. Ball mills are huge drum- like structure inside which ores and steel balls are present. With the rotation of ball mill, ores collide with each other, with balls and with the inside surface of the ball mill so that bigger ores break into a smaller size. Ball mills have been used both for dry grinding and wet grinding. The perfor- mance of the ball mill is greatly affected by parameters such as the percentage of mill filling, the percentage of ball quantity, the percentage of water quantity, mill speed, type of ore, slurry viscosity, and shape of lifters. Therefore, it is important to monitor these parameters to optimize the performance of the ball mill. Since a ball mill is opaque and its inside has dynamic media, it is impossible to monitor the inside condition of a mill. Neither it is possible to visually inspect from outside, nor any sensor probe such as a camera can be inserted inside a ball mill. For a long time, many researchers have been trying to diagnose the ball mill to improve the performance. A simulation study using the discrete element method (DEM) was used to compute the charge motion in the semi-auto- genous (SAG) mill [1]. It was shown that by reducing the number of lifters and increasing the face angle, the desired charge motion for better grinding and peak power could be obtained. The DEM was also used to calculate the force distribution inside the ball mill [2]. The correlation between the spectral peaks of the acoustic signal and rotation speed of the ball mill and the number of lifters used inside the mill was established in [2]. A 2D DEM code was experimentally validated by using digital image analysis of velocity profile of the balls, toe and shoulder angle, and consumed power [3]. The trajectory and velocity of the balls in the system were also determined in [3]. & Debi Prasad Das dpdas@immt.res.in Dilip Kumar Nayak nayakdilipk@gmail.com Santosh Kumar Behera skbehera@immt.res.in Sarada Prasad Das spdas@immt.res.in 1 ITER, Siksha ‘O’ Anusandhan, Bhubaneswar, India 2 CSIR-Institute of Minerals and Materials Technology, Bhubaneswar, India 123 Neural Computing and Applications https://doi.org/10.1007/s00521-019-04555-5