Neural network state diagnostic system for equipment

• US 5,333,240 A
• Filed: 04/13/1990
• Issued: 07/26/1994
• Est. Priority Date: 04/14/1989
• Status: Expired due to Fees

• Current Assignee: Hitachi, Ltd.
• Original Assignee: Hitachi, Ltd.
• Inventors: Saito, Tadayoshi, Yokoyama, Hiroshi, Baba, Kenji, Matsumoto, Hiroshi, Kawakami, Junzo, Nomura, Masahide, Kaji, Akira, Shimoda, Makoto, Katayama, Yasunori, Nigawara, Seiitsu
• Primary Examiner(s): MacDonald, Allen R.
• Assistant Examiner(s): DAVIS, GEORGE B
• Application Number:
  US07/508,664
• Time in Patent Office: 1,565 Days
• Field of Search: 364/513, 364/200, 364/148, 364/590, 364/508, 395/23, 395/906, 395/914, 395/24, 395/11, 395/904, 395/22
• US Class Current: 706/20
• CPC Class Codes: F01D 21/003 Arrangements for testing or measuring for measuring vibrations G01H
• CPC Class Codes: F05D 2220/31 in steam turbines
• CPC Class Codes: F05D 2260/80 Diagnostics
• CPC Class Codes: F05D 2270/334 Vibration measurements
• CPC Class Codes: F05D 2270/709 with neural networks
• CPC Class Codes: G01H 1/003 of rotating machines G01H1/10 takes precedence
• CPC Class Codes: G06F 11/2257 using expert systems
• CPC Class Codes: G07C 3/00 Registering or indicating the condition or the working of machines or other apparatus, other than vehicles
• CPC Class Codes: Y10S 706/904 Manufacturing or machine, e.g. agricultural machinery, machine tool

Abstract

A state diagnostic system for equipment is disclosed. The system is constructed of a neural network model for learning in advance one or more samples of information on vibrations, which are produced in a specific operation state of the equipment, in association with the corresponding operation state and obtaining an output signal corresponding to results of the learning when information on vibrations produced upon operation of the equipment is inputted; an input unit for inputting to the neural network model the information on the vibrations produced upon operation of the equipment; and an output unit for outputting the output signal from the neural network model as diagnostic results to a user. State diagnostic methods, learning system, preview/predict system, diagnosis training system, service life estimation assisting system, service life estimation assisting system, and maintenance assisting system are also disclosed.

Claims

1. A state diagnostic system for a rotating machine, comprising:

a neural network model for learning in advance one or more samples of time dependent waveforms of vibrations, which are produced in abnormal and normal operation states of the rotating machine, in association with the corresponding operation state, and producing an output signal corresponding to diagnostic results when data related to time dependent waveforms of vibrations produced upon operation of the rotating machine are inputted;

means for inputting to the neural network model the data related to time dependent waveforms of the vibrations produced upon operation of the rotating machine;

means for converting the output signal from the neural network mode into a message to a user and outputting the message as at least a part of diagnostic results; and

a man-machine unit having a screen for displaying said message.

2. A state diagnostic system according to claim 1, and further including:

learning means for successively feeding the one or more samples of time dependent waveforms of vibrations, which are produced in abnormal and normal operation states of the rotating machine, to the neural network model and causing the neural network model to learn to produce a different output signal for each operation state.

3. A state diagnostic system for a rotating machine, comprising:

a neural network model for learning in advance at least one sample of time dependent waveforms of vibrations, which are produced in abnormal and normal operation states of the rotating machine, in association with the corresponding operation state, and producing an output signal corresponding to diagnostic results when data related to time dependent waveforms of vibrations produced upon-operation of the rotating machine are inputted;

means for inputting to the neural network model the data related to time dependent waveforms of vibrations produced upon operation of the rotating machine;

learning means for successively feeding the one or more samples of time dependent waveforms of vibrations, which are produced in abnormal and normal operation states of the rotating machine, to the neural network model and causing the neural network model to learn to obtain a different output signal for each operation state; and

output means for indicating the state of learning of the neural network model.

4. A method for diagnosing a state of operation of a rotating machine using a state diagnostic system, said system having a neural network model capable of diagnosing a state of operation of the rotating machine, comprising the steps of:

learning one or more samples of time dependent waveforms of vibrations, which are produced in abnormal and normal operation states of the rotating machine, in association with the corresponding operation state, to the neural network model;

inputting data related to time dependent waveforms of vibrations, produced in an operation state of the rotating machine, to the neural network model to obtain an output signal corresponding to diagnostic results; and

converting the output signal from the neural network model into a message to a user and outputting the message as at least a part of diagnostic results.

5. A state diagnostic system for a rotating machine, said system being adapted to detect a state of vibrations produced upon operation of the rotating machine and to diagnose a state of the operation of the rotating machine on the basis of detected time dependent waveform information, said system comprising:

a data conversion unit for converting the detected time dependent waveform information, from the rotating machine, to time dependent abnormality monitoring data;

a neural network model for learning from samples of the detected time dependent waveform information, said neural network model determining the presence or absence of an abnormality and details of the abnormality by a position of an output signal which appears at an output unit upon input of the time dependent abnormality monitoring data;

means for converting said output signal from said neural network model into a message and for outputting the message as at least a part of diagnostic results; and

a man-machine unit having a screen for displaying said message.

6. A state diagnostic system for a rotating machine, said system being adapted to detect a state of vibrations produced upon operation of the rotating machine and to diagnose a state of the operation of the rotating machine on the basis of detected time dependent waveform information, said system comprising:

a data conversion unit for converting detected time dependent waveform information, from the rotating machine, to time dependent abnormality monitoring data;

a neural network model for learning from samples of the detected time dependent waveform information, said neural network model being capable of determining the presence or absence of an abnormality and details of the abnormality by a position of an output signal which appears at an output unit upon input of the time dependent abnormality monitoring data;

a learning sample storage unit for storing, for learning purpose, the samples of the detected time dependent waveform information;

a learning control unit for causing the neural network model to learn the samples stored in the learning sample storage unit; and

a learning man-machine unit for indicating progress of results of the learning of the samples to a user.

7. A system for previewing/predicting an abnormality in the operation of a rotating machine, said system being adapted to detect vibrations produced upon operation of the rotating machine and to preview/predict an occurrence of abnormality in the rotating machine on the basis of detected time dependent waveform information, said system comprising:

a data conversion unit for converting detected time dependent waveform information, from the rotating machine, to time dependent abnormality monitoring data;

a neural network model for learning samples of the time dependent waveform information, said neural network model previewing/predicting an abnormality in accordance with a position of a signal which appears at an output unit upon input of the time dependent abnormality monitoring data; and

a diagnostic man-machine unit for providing a user with results of the preview/predict as output information from the neural network model the man-machine unit having a screen for displaying diagnostic results, and the results of the preview/predict to be displayed on the screen including at least one of information describing phenomenon, information describing the cause, or information on a counter-measure guidance.

8. A system for previewing/predicting an abnormality according to claim 7, further comprising:

means for determining certainty factors by levels of the corresponding outputs from the neural network model and displaying the certainty factors on the screen along with respective possible causes.

9. A diagnosis training system for training the diagnosis of a state of operation of a rotating machine, said system comprising:

a neural network model for determining the presence or absence of an abnormality and details thereof on the basis of a position of a signal at an output unit upon input of time dependent abnormality monitoring data on a vibration phenomenon of the rotating machine as a target of the training of the diagnosis;

an assumed abnormality producing unit for setting, as the time dependent abnormality monitoring data, time dependent waveform information from the rotating machine and corresponding to various assumed abnormality causes of the rotating machine at an input portion of the neural network model; and

a man-machine unit for indicating, to a user, diagnosis results which appear at the output unit when the data generated from the assumed abnormality producing unit are inputted in the neural network model.

10. A service life estimation assisting system for detecting time dependent waveforms of vibrations produced upon operation of a rotating machine and assisting in estimation of the service life of the rotating machine on the basis of detected time dependent waveform information, which comprises:

a data conversion unit for converting detected time dependent waveform information, from the rotating machine, to service life estimating data;

a neural network model for learning samples of various time dependent waveform information, said neural network model estimating the service life of the rotating machine on the basis of the magnitude of a signal which appears at an output unit upon input of the service life estimating data; and

a service life estimating man-machine unit for providing a user with results of the service life estimation as output information from the neural network model.

11. A state diagnostic system for equipment having elements constructed in a hierarchial structure, said system being adapted to detect time dependent information indicative of a state of operation of the equipment, which occurs upon operation of the equipment, and to diagnose the state of operation of the equipment on the basis of the time dependent information thus detected, said system comprising:

neural network models for learning samples of time dependent information indicative of various states of the equipment and performing a determination of the presence or absence of an abnormality and details thereof on the basis of a signal which appears at an output unit when time dependent monitoring information from the equipment is inputted, one or more of said neural network models being provided corresponding to each of a hierarchy of stages, and said neural network models being connected in such a way that each neural network model in a relatively upper stage coordinates one or more neural network models in a relatively lower stage and the neural network model in the relatively upper stage uses, as time dependent monitoring information, diagnostic results by the neural network models in the relatively lower stage.

12. A state diagnostic system for equipment, said system being adapted to detect time dependent vibration information produced upon operation of the equipment and to diagnose a state of operation of the equipment on the basis of time dependent information thus detected, said system comprising:

neural network models for learning samples of time dependent information indicative of various states of the equipment and performing a determination of the presence or absence of an abnormality and details thereof on the basis of a signal which appears at an output unit when time dependent monitoring information from the equipment is inputted, said neural network models being provided corresponding to time dependent information to be detected by respective detectors arranged at a like plural number of locations on the equipment so that a diagnosis is independently feasible at each of the locations.

13. A state diagnostic system for equipment, said system being adapted to detect time dependent vibration information produced upon operation of the equipment and to diagnose a state of operation of the equipment on the basis of the detected time dependent information, said system comprising:

a data conversion unit for converting in a time sharing manner time dependent information, from the equipment, which is detected respectively by plural detectors arranged at a like number of plural locations on the equipment, to time dependent monitoring information; and

a like plural number of neural network models for learning samples of time dependent information indicative of various states of the equipment and performing a determination of the presence or absence of an abnormality and details thereof on the basis of a signal which appears at an output unit when the time dependent monitoring information is inputted, and said neural network models being provided corresponding to the detectors to receive corresponding time dependent monitoring information which have been converted in the time sharing manner, whereby a diagnosis is independently feasible at each of the locations.

14. A state diagnostic system for equipment, said system being adapted to detect time dependent information produced upon operation of the equipment and to diagnose a state of operation of the equipment on the basis of the detected time dependent information, said system comprising:

a data conversion unit for converting in a time sharing manner time dependent information, from the equipment, which are detected respectively by plural detectors arranged at a like plural locations on the equipment, to time dependent monitoring information; and

at least one neural network model for learning samples of time dependent information indicative of various states of the equipment and performing in a time sharing manner a determination of the presence or absence of an abnormality and details thereof on the basis of signals which appear at an output unit when the time dependent monitoring information are inputted in a time sharing manner, whereby a diagnosis is independently feasible with respect to each of the detectors.

15. A state diagnostic system for equipment, comprising:

a neural network model for learning in advance one or more samples of time dependent waveforms of vibrations, which are produced in abnormal and normal operation states of the equipment, in association with the corresponding operation state, and producing an output signal corresponding to diagnostic results when data related to time dependent waveforms of vibrations produced upon operation of the equipment is inputted;

means for inputting to the neural network model the data related to time dependent waveforms of vibrations produced upon operation of the equipment;

means for converting the output signal from the neural network model into a message to a user and outputting the message as at least a part of diagnostic results; and

`a man-machine unit having a screen for displaying said message.

16. A state diagnostic system for according to claim 15, and further including:

learning means for successively feeding the one or more samples of time dependent waveforms of vibrations, which are produced in abnormal and normal operation states of the equipment, to the neural network model and causing the neural network model to learn to produce a different output signal for each operation state.

17. A state diagnostic system for equipment, comprising:

a neural network model for learning in advance at least one sample of time dependent waveforms of vibrations, which are produced in abnormal and normal operation states of the equipment, in association with the corresponding operation state, and producing an output signal corresponding to diagnostic results when data related to time dependent waveforms of vibrations produced upon-operation of the equipment is inputted;

means for inputting to the neural network model the data related to time dependent waveforms of vibrations produced upon operation of the equipment;

learning means for successively feeding the at least one sample of the time dependent waveforms of vibrations, which are produced in abnormal and normal operation states of the equipment, to the neural network model and causing the neural network model to learn to obtain a different output signal for each operation state; and

output means for indicating the state of learning of the neural network model.

18. A method for diagnosing a state of operation of equipment using a state diagnostic system, said system having a neural network model for diagnosing a state of operation of the equipment, comprising the steps of:

learning one or more samples of time dependent waveforms of vibrations, which are produced in abnormal and normal operation states of the equipment, in association with the corresponding operation state, to the neural network model;

inputting data related to the time dependent waveforms of vibrations, produced in an operation state of the equipment, to the neural network model to obtain an output signal corresponding to diagnostic results; and

converting the output signal from the neural network model into a message to a user and outputting the message as at least a part of diagnostic results.

19. A state diagnostic system for equipment, said system being adapted to detect time dependent waveform vibration information produced upon operation of the equipment and to diagnose a state of the operation of the equipment on the basis of detected time dependent waveform vibration information, said system comprising:

a data conversion unit for converting the detected time dependent waveform vibration information, from the equipment, to time dependent abnormality monitoring data;

a neural network model for learning samples of the detected time dependent waveform information, said neural network model determining the presence or absence of an abnormality and details of the abnormality by a position of an output signal which appears at an output unit upon input of the time dependent abnormality monitoring data;

means for converting said output signal from said neural network model into a message and for outputting the message as at least a part of diagnostic results; and

a man-machine unit having a screen for displaying said message.

20. A state diagnostic system for equipment, said system being adapted to detect time dependent waveform vibration information produced upon operation of the equipment to diagnose a state of the operation of the equipment on the basis of detected time dependent waveform vibration information, said system comprising:

a data conversion unit for converting detected time dependent waveform vibration information, from the equipment, to time dependent abnormality monitoring data;

a neural network model for learning samples of the detected time dependent waveform vibration information, said neural network model determining the presence or absence of an abnormality and details of the abnormality by a position of an output signal which appears at an output unit upon input of the time dependent abnormality monitoring data;

a learning sample storage unit for storing, for learning purpose, samples of the detected time dependent waveform vibration information;

a learning control unit for causing the neural network model to learn the samples stored in the learning sample storage unit; and

a learning man-machine unit for indicating progress of results of the learning of the samples to a user.

21. A system for previewing/predicting an abnormality in the operation of equipment, said system being adapted to detect time dependent waveform vibration information produced upon operation of the equipment and to preview/predict an occurrence of abnormality in the equipment on the basis of detected time dependent waveform vibration information, said system comprising:

a data conversion unit for converting detected time dependent waveform vibration information, from the equipment, to time dependent abnormality monitoring data;

a neural network model for learning samples of various abnormal time dependent waveform vibration information, said neural network model previewing/predicting an abnormality in accordance with a position of a signal which appears at an output unit upon input of the time dependent abnormality monitoring data; and

a diagnostic man-machine unit for providing a user with results of the preview/predict as output information from the neural network model the man-machine unit having a screen for displaying diagnostic results, and the results of the preview/predict to be displayed on the screen including at least one of information describing phenomenon, information describing the cause, or information on a counter-measure guidance.

22. A diagnosis training system for training the diagnosis of a state of operation of equipment, said system comprising:

a neural network model for determining the presence or absence of an abnormality and details thereof on the basis of a position of a signal at an output unit upon input of time dependent abnormality monitoring data of a vibration phenomenon of the equipment as a target of the training of the diagnosis;

an assumed abnormality producing unit for setting, as the time dependent abnormality monitoring data, time dependent waveform information from the equipment and corresponding to various assumed abnormality causes of the equipment, at an input portion of the neural network model; and

a man-machine unit for indicating, to a user, diagnosis results which appear at the output unit when the data generated from the assumed abnormality producing unit are inputted in the neural network model.

23. A service life estimation assisting system for detecting time dependent waveform vibration information produced upon operation of equipment and assisting in estimation of the service life of the equipment on the basis of detected time dependent waveform vibration information, said system comprising:

a data conversion unit for converting detected time dependent waveform vibration information, related to data from the equipment, to service life estimating data;

a neural network model for learning samples of time dependent waveform vibration information, said neural network model estimating the service life of the equipment on the basis of the magnitude of a signal which appears at an output unit upon input of the service life estimating data; and

a service life estimating man-machine unit for providing a user with results of the service life estimation as output information from the neural network model.