The system proposed in this paper aims at monitoring the torque and efficiency in induction motors in real time by employing wireless sensor networks (WSNs). An embedded system is employed for acquiring electrical signals from the motor in a noninvasive manner, and then performing local processing for torque and efficiency estimation. The values calculated by the embedded system are transmitted to a monitoring unit through an IEEE 802.15.4-based WSN. At the base unit, various motors can be monitored in real time. An experimental study was conducted for observing the relationship between the WSN performance and the spectral occupancy at the operating environment. This study demonstrated that the use of intelligent nodes, with local processing capability, is essential for this type of application. The embedded system was deployed on a workbench, and studies were conducted to analyze torque and system efficiency.

The induction motors are the main equipment used for electromechanical conversion in the world, and are present in almost all production processes, accounting for about two thirds of the industrial electrical consumption. Faults in induction motors can result in operational disasters, and stop entire sectors of a plant, causing economic and human losses. Therefore, the development of appropriate techniques for fault diagnosis in induction motors is critical. Classical methods for induction motors fault diagnosis do not always provide satisfactory results. This paper proposes a hybrid system that uses data obtained from vibration, and current sensors to detect failures at an early stage, since each technique has limitations and disadvantages when used individually. The signals are processed in the frequency and time domain through short time Fourier transform and wavelet multi-resolution analysis, which provides inputs to an intelligent system based on fuzzy logic. The failures due to unbalanced load in the motor shaft and in the motor helix were correctly detected. Experiments with broken bars were also performed, and the system was also validated for this type of failure. The signal processing, and the classification of the failure severity, through fuzzy logic, were developed using Matlab.