It is important to maintain every machine affecting the process of making sugar to ensure excellent product quality with minimal losses and to accelerate productivity and profitability targets. The centrifuges are widely used in industry today with some being very difficult and critical for surgery, and the collapse of the engine has the ability to cause expensive damage. One of these is the centrifugal machines, and they are expected to be efficient to produce high-quality sugar. Meanwhile, an efficient diagnostic tool to predict the correct time for centrifugal repair is vibration signal analysis namely by attaching the accelerometer sensor to the location of the centrifugal bearing to produce vibration data that is ready to be analyzed. Still, the process requires sufficient insight and experience. The manual method usually used is complicated and requires a lot of time to obtain results of a centrifugal diagnosis. Therefore, this study was conducted to design an intelligent system to diagnose centrifugal vibrations using Artificial Neural Networks (ANN). The situation is involved in applying and training the concept of vibration analysis from spectrum data to ANN to produce diagnostic results according to the spectrum diagnosis reference. The results obtained were quite good with the largest cross-entropy value of 10.67 having 0% error value with the largest Mean Square Error value being 0.0023 while the smallest regression was 0.993. The test conducted on nine new spectrums produced eight true predictions and one false. The system can provide fairly accurate results in a short time. Classification quality improvement can be made by adding training data.