This Chapter aims to propose two efficient strategies to predict the states of a running ball mill in different load varying conditions. The first method is to establish a mathematical model to obtain the acoustic signatures or patterns of different states of transitions of the running ball mill by analyzing the sound of the ball mill in its load varying conditions. In this approach, using Kernel Density Estimator (KDE) and curve fitting algorithm simulated patterns of the acoustics are obtained. On the other hand, in experimentation, the ball mill is stopped manually at the fixed time interval representing the states and the particles are collected and corresponding particle size distribution (psd) is calculated. In order to develop the mathematical model which predicts the acoustic signature corresponding to the states, errors between experimental and simulated patterns at different instant of time are calculated and minimized by learning the parameters of the model. The second method predicts the terminating condition of the ball mill by extracting statistical features using time domain information of the acoustics like amplitude, frequency and phase. Using principal component analysis (PCA) method, important features are selected and the data points are clustered into two groups using K-means clustering algorithm representing the End state and Start state of the ball mill. Finally, a Naive Bayesian classifier is trained to predict the termination condition of the running ball mill and when stopped the desired psd of the crushed material as targeted by the user has been obtained.