Method and apparatus to detect a stator turn fault in an AC motor

A system to detect a stator turn fault in an AC motor comprising:

• at least one input to input operating data from an AC motor to a microprocessor, wherein the operating data comprises positive and negative voltage sequence components of a line voltage, and a positive current sequence component of a line current;
• a computer readable storage medium connected to the microprocessor having therein a computer program that, when executed, performs a synchronous reference frame transformation on the operating data and initiates a feed forward neural network to calculate estimated negative current sequence components based on the transformed data;
• and
• at least one output to output the estimated negative sequence component.

An apparatus for detecting a stator turn fault in an AC motor, comprising:

• data acquisition means for receiving data from at least one AC motor wherein the data includes line voltage signals and line current signals from the at least one AC motor;
• a synchronous reference frame transformation means communicating with the data acquisition means for determining sequence components of at least a portion of the data;
• an integrating means, having a feed forward neural network, communicating with the transformation means for receiving the sequence components of at least a portion of the data and communicating with an outputting means for outputting estimated data of the at least one AC motor;
• and
• a comparing means for comparing the estimated data to instantaneous data acquired from the at least one AC motor during real-time.

A computer-readable storage medium having stored thereon a computer program which, when executed by a computer, will cause the computer to:

• acquire operating data from an AC motor during a known normal operation;
• input at least a portion of the operating data to a feed forward neural network having a number of weights;
• train the feed forward network to determine converged values for each of the number of weights;
• store the converged values in memory;
• obtain an instantaneous operating parameter from the AC motor during instantaneous operation of the AC motor;
• perform a synchronous reference frame transformation on the instantaneous operating parameter to determine sequence components of the instantaneous operating parameter;
• input the sequence components into the feed forward neural network;
• obtain an estimated operating parameter of instantaneous operation of the AC motor;
• and
• compare the estimated operating parameter to the instantaneous operating parameter to determine a stator turn fault in the AC motor.

An apparatus to detect a stator turn fault of an AC motor comprising:

• at least two current sensors for obtaining at least two AC motor current signals;
• at least two voltage sensors for obtaining at least two AC motor voltage signals;
• an analog to digital converter for converting the at least two AC motor current signals to digitized current signals and the at least two AC motor voltage signal to digitized voltage signals;
• a microprocessor to receive the digitized signals and perform a synchronous reference frame transformation to determine sequence parameters of the AC motor;
• and
• a feed forward neural network to determine estimated fundamental frequency sequence parameters of the AC induction motor.

A method of detecting a stator turn fault in a motor, comprising the steps of:

• acquiring training parameters from the motor during a known good condition of the motor;
• determining sequence phasors from the training parameters using a synchronous reference frame transformation;
• determining estimated values indicative of a stator turn fault from the motor using a feed forward neural network;
• acquiring instantaneous values from the motor while the motor is in service irrespective of whether the motor is running;
• comparing the instantaneous values to the estimated values to determine a stator turn fault;
• and
• indicating a presence of a stator turn fault based on the comparison.

A method of stator turn fault detection for an AC induction motor comprising the steps of:

• (A) selecting the sequence phasor parameters for a feed forward neural network;
• (B) training the feed forward neural network to learn a model of the AC induction motor under healthy operating conditions;
• (C) acquiring a measured value of the AC induction motor while the AC induction motor is in service;
• (D) performing a synchronous reference frame transformation on the measured value to determine sequence parameters of the AC induction motor;
• (E) inputting the sequence parameters to the feed forward neural network to determine an estimated value of AC induction motor operation during healthy operating conditions;
• (F) comparing the estimated value of AC induction motor operation with a measured value of the AC induction motor operation;
• (G) periodically repeating steps (C)-(F);
• and
• (H) indicating a stator turn fault of the AC induction motor if the step of comparing results in a fault value that exceeds a vigilance threshold.