Method and system for synthesizing vibration data

• US 5,995,910 A
• Filed: 08/29/1997
• Issued: 11/30/1999
• Est. Priority Date: 08/29/1997
• Status: Expired due to Term

• Current Assignee: Reliance Electric Industrial Company
• Original Assignee: Reliance Electric Industrial Company
• Inventors: Discenzo, Frederick M.
• Primary Examiner(s): Hoff, Marc S.
• Assistant Examiner(s): Bui, Bryan
• Application Number:
  US08/921,409
• Time in Patent Office: 823 Days
• Field of Search: 702/34, 702/56, 702/141, 702/127, 73/577, 73/660, 73/659, 73/650
• US Class Current: 702/56
• CPC Class Codes: G01H 1/003 of rotating machines G01H1/10 takes precedence
• CPC Class Codes: G01M 13/045 Acoustic or vibration analysis

Abstract

A system for training a neural network to synthesize vibration data relating to the operation of a machine. The system includes a first sensor operatively coupleable to the machine, the first sensor adapted to obtain at least one vibration signal relating to the operation of the machine. The system further includes a second sensor operatively coupleable to a power lead of the machine, the second sensor adapted to obtain at least one current signal relating to the operation of a machine. Additionally, the system includes a neural network operatively coupleable to the second sensor, the neural network being trainable to generate at least one synthesized vibration signal from the current signal, wherein the synthesized vibration signal is substantially equivalent to the vibration signal obtained from the first sensor.

Claims

1. A system for training a neural network to synthesize vibration data relating to the operation of a machine, comprising:

a first sensor operatively coupleable to the machine, the first sensor adapted to obtain at least one vibration signal relating to the operation of the machine;

a second sensor operatively coupleable to a power lead of the machine, the second sensor adapted to obtain at least one current signal relating to the operation of a machine; and

a neural network operatively coupleable to the second sensor, the neural network being trainable to generate at least one synthesized vibration signal from the current signal, wherein the synthesized vibration signal is substantially equivalent to the vibration signal obtained from the first sensor.

2. The system of claim 1, wherein the first sensor is an accelerometer.

3. The system of claim 2 wherein the second sensor is a current transformer.

4. The system of claim 1, further including a processor operatively coupleable to the first and second sensors, the processor adapted to generate fast fourier transforms of the at least one vibration signal and the at least one current signal.

5. The system of claim 4, wherein the processor inputs the fast fourier transform of the at least one current signal to the neural network, and inputs the fast fourier transform of the at least one vibration signal to a comparator, the comparator being operatively coupleable to the neural network.

6. The system of claim 5, wherein the comparator compares a synthesized vibration signal generated by the neural network against the fast fourier transform of the at least one vibration signal.

7. The system of claim 1, further including a processor operatively coupleable to the first and second sensors, the processor adapted to generate wavelet transforms of the at least one vibration signal and the at least one current signal.

8. The system of claim 1, wherein the current signal corresponds to instantaneous values of current.

9. The system of claim 1, wherein the neural network is an autoassociator trained using current signal values and vibration signal values obtained from the machine operating in various known states.

10. The system of claim 1, wherein the neural network includes:

a plurality of neurons adapted to receive signals and adapted to present an output;

a plurality of connective synapses providing a weighted coupling between the neurons;

wherein the neural network is capable of adapting itself to produce a predetermined output in response to an input by changing the value of the weighted coupling between the neurons.

11. The system of claim 1, wherein the machine is an AC induction motor.

12. The system of claim 1, wherein the machine is a pump.

13. The system of claim 1, wherein the machine is a generator.

14. A method for synthesizing vibration data, comprising the steps of:

collecting at least one sample of vibration data signal relating to the operation of a machine;

collecting at least one sample of current data signal relating to the operation of the machine;

inputting the at least one sample of current data signal to a neural network; and

using the neural network to synthesize at least one vibration data signal from the at least one sampled current data, the at least one synthesized vibration signal being substantially equivalent to the at least one sample of vibration data signal.

15. The method of claim 14, wherein the step of collecting at least one sample of vibration data signal relating to the operation of the machine is accomplished by employing a vibration sensor.

16. The method of claim 15, further including the step of inputting the at least one sample of vibration data signal to a comparator operatively coupleable to the neural network, wherein the comparator compares the output of the neural network against the at least one sample of vibration data signal out from the accelerometer.

17. The method of claim 14, wherein the step of collecting at least one sample of current data signal relating to the operation of the machine is accomplished by employing a current transformer.

18. The method of claim 14, further including the step of training the neural network to use current signal values and vibration signal values obtained from the machine operating in various known states, wherein the neural network includes:

a plurality of neurons adapted to receive signals and adapted to present an output;

a plurality of connective synapses providing a weighted coupling between the neurons;

wherein the neural network is capable of adapting itself to produce a predetermined output in response to an input by changing the value of the weighted coupling between the neurons.

19. A system for diagnosing a machine using synthesized vibration data, comprising:

a sensor for obtaining current data relating to the operation of a machine;

a neural network operatively coupleable to the sensor, the neural network adapted to synthesize a vibration signal from the sampled current data, and

a processor operatively coupleable to the neural network, the processor adapted to receive the synthesized vibration signal and employ the synthesized vibration signal to facilitate diagnosing the state of the machine, the synthesized vibration signal being substantially equivalent to a vibration signal that would be obtained from a vibration sensor if Attached to the machine.

20. The system of claim 19, wherein the sensor is a current transformer.

21. The system of claim 19, wherein the processor performs classical vibration analysis on the machine using the synthesized vibration signal.

22. The system of claim 21, further including a memory operatively coupleable to the processor, the memory adapted to store historical vibration analysis data which the processor can access when performing the classical vibration analysis.

23. The system of claim 19, wherein the machine is an AC induction motor.

24. The system of claim 19, wherein the machine is a pump.

25. The system of claim 19, wherein the machine is a generator.

26. A method for diagnosing a machine, comprising the steps of:

collecting at least one sample of current data relating to the operation of the machine;

inputting the at least one sample of current data to a neural network;

using the neural network to synthesize a vibration data signal from the at least sample of current data, the synthesize vibration data signal being substantially equivalent to a vibration signal that would have been obtained from a vibration sensor if attached to the machine;

determining the diagnostic state of the machine from the synthesized vibration data signal.

27. The method of claim 26, wherein the step of collecting at least one sample of current data signal relating to the operation of the machine is accomplished by employing a current transformer.

28. The method of claim 26, wherein the step of determining the diagnostic state of the machine from the at least one synthesized vibration data signal is accomplished via a processor employing classical vibration analysis.

29. The method of claim 28, further including the step of using the processor to access historical vibration analysis data stored in a memory operatively coupleable to the processor.

30. A system for diagnosing a plurality of machines using synthesized vibration data, comprising:

a plurality of sensors for obtaining current data relating to the operation of the plurality of machines;

a channel interface operatively coupleable to the plurality of sensors, the channel interface designating a separate channel for the current data obtained from each of the plurality of sensors, respectively;

a host computer operatively coupleable to the channel interface, the host computer including a neural network operatively coupleable to each channel of the channel interface, the neural network adapted to synthesize vibration signals corresponding to each of the plurality of machines from the sampled current data;

wherein a processor of the host computer cycles through each of the channels, the processor performing classical vibration analysis on each of the plurality of machines using the synthesized vibration signal for each respective machine.

31. A system of claim 30, further including a memory operatively coupleable to the processor, the memory adapted to store historical vibration analysis data which the processor can access when performing the classical vibration analysis.

32. The system of claim 30, wherein the plurality of machines includes at least one of either:

an AC induction machine, a pump, and a generator.

33. A system for training a neural network to synthesize data relating to the operation of a machine, comprising:

a first sensor operatively coupleable to the machine, the first sensor adapted to obtain a first signal relating to the operation of the machine;

a second sensor operatively coupleable to the machine, the second sensor adapted to obtain a second signal relating to the operation of a machine, the second signal having data relating to the first signal encoded therein; and

a neural network operatively coupleable to the second sensor, the neural network being trainable to generate a synthesized first signal from the second signal, wherein the synthesized first signal is substantially equivalent to the first signal obtained from the first sensor.