Signal Processing Techniques and Neural Network Algorithms Applied to the Production Line Reality for an Automatic Identification of Mechanical Defects in Electric Motors

R.M. Rodriguez and C. Cristalli (Italy)


Applications (Manufacturing), Time frequency signal analysis, Neuronal Networks for signal processing.


In order to perform an automatic on-line test for mechanical fault detection in universal motors, signals acquired from accelerometers have been processed. The signals extracted from the transducers have been analysed and in a few seconds processed in order to find peculiar features that will allow the discrimination between a good (noiseless) and faulty (noisy) motor. A wide range of defects has been detected and analysed both from the mechanical and the electrical anomalous conditions of the motor (e.g. brush noise, rotor unbalance, defects of the bearings, defects on the commutator segments). The system replaces the subjectivity of human inspection and testing with an objective and accurate assessment of product quality. This is achieved by adopting advanced techniques of Pattern Recognition and integrating them into complete systems for data acquisition and classification.

Important Links:

Go Back