Abstract
Industrial monoblock centrifugal pumps are critical pieces of rotational machinery that play an important role in manufacturing operations. The critical components must be in proper working order for the industry to continue operating. State monitoring is essential for monitoring and analysing the condition of equipment. Bearing failure, cavitation, a broken impeller, and other issues are common in monoblock centrifugal pumps. Traditional procedures for calculating outcomes have been proven to be time-consuming and difficult. At regular intervals, time domain vibrational signals are collected for the defective pump. These vibrational indicators are evaluated to the healthy, defect-free pump. To acquire the accuracy, these images are fed into an efficient deep convolutional neural network (DCNN). This research examines two types of failures outer race bearing seal failure and cavitation. The visuals are trained and assessed in proportions of 70:30. Finally, the DCNN architecture's fault diagnosis accuracy is 99.07%.