This study focuses on real-Time condition monitoring of the compressed ignition (CI) engine using the various vibration signatures acquired for the different combinations of fuels. The vibration signals acquired through shear accelerometer by maintaining a constant speed of 1800 rpm with varying load conditions from no load (0kg) to full load (20kg) with equal intervals of 5kg. Statistical features are extracted from each particular signal and determination of features is carried out. Classification is performed using a random tree and naive Bayes algorithms which identifies the minimal vibrations and best performance among the selected blends. The best classification accuracy is found for naive Bayes algorithm with 96.4%. The selected best combination blend data is used further for training the developed neural network model with neurons in hidden layers and a trained model was tested to predict the vibration signatures with the average 0.97 correlation coefficient. © 2020 IEEE.