Vibrating machine automated diagnosis with supervised learning
Supervised learning is implemented to improve the accuracy of automated diagnoses performed by monitoring units installed at a machine. The monitoring units perform indicator acquisition and automated diagnoses based on a Bayesian model derived in accordance with the machine's known configuration. Raw data is collected, including machine vibration data and other diagnostic data. The data is analyzed to diagnose for specific fault defect assumptions so as to generate the automated diagnoses results and a rating for overall health of the machine. The results are uploaded to an external environment that can be accessed by an expert for review and correction. Based upon the expert's corrections, the Bayesian model is adjusted using supervised learning to improve the automated diagnoses performed by the monitoring units.
1. A method with learning for generating an overall diagnostic health rating for a machine having moving parts, comprising the steps:
sensing machine diagnostic data, including vibration data, using one or more sensors installed at the machine;
deriving with one or more first processors a prescribed set of indicators from the machine diagnostic data;
for a first set of said machine diagnostic data, automatically diagnosing by a monitoring unit processor a first overall diagnostic health rating of the machine from the first set of said machine diagnostic data and a subset of the prescribed set of indicators, wherein said automatic diagnosing comprises said monitoring unit processor implementing a neural network classification model configured with a set of parameters defined for a prescribed set of symptoms for a prescribed set of diagnostic conditions monitored at the machine;
in response to a review of the prescribed set of indicators and the first set of said machine diagnostic data, storing in memory a corrected overall diagnostic health rating of the machine which differs from said first overall diagnostic health rating of the machine;
modifying with a third processor the neural network classification model so as to generate said corrected overall diagnostic health rating in response to said first set of machine diagnostic data, wherein said learning comprises said modifying step;
downloading the modified neural network classification model to a monitoring unit comprising said monitoring unit processor;
repeating said steps of sensing and deriving to derive a newly-derived set of indicators from newly-collected machine diagnostic data; and
based on the newly-derived set of indicators, automatically diagnosing by said monitoring unit processor a second overall diagnostic health rating of the machine using the modified neural network classification model.
2. The method of claim 1 , wherein the prescribed set of diagnostic conditions monitored at the machine is a first set of diagnostic conditions, and further comprising the steps:
monitoring a second prescribed set of diagnostic conditions of the machine encompassing more conditions than said first set of diagnostic conditions;
automatically diagnosing a fault severity probability for each one of said second prescribed set of diagnostic conditions; and
storing in said memory a corrected fault severity probability for a first diagnostic condition among said second prescribed set of diagnostic conditions; and
wherein said modifying with the third processor the neural network classification model comprises further configuring the modified neural network classification model with another parameter so as to be configured by a second set of parameters, said another parameter being defined for a first symptom for said first diagnostic condition.
3. The method of claim 2 , wherein said modifying with the third processor the neural network classification model further comprises modifying a value of at least one parameter among said second set of parameters.
4. The method of claim 1 , wherein said modifying with the third processor the neural network classification model comprises modifying a value of at least one parameter among said set of parameters.