System and method for estimating remaining useful life

• US 8,781,982 B1
• Filed: 09/23/2011
• Issued: 07/15/2014
• Est. Priority Date: 09/23/2011
• Status: Active Grant

• Current Assignee: Lockheed Martin Corporation (Martin Marietta Corporation)
• Original Assignee: Lockheed Martin Corporation (Martin Marietta Corporation)
• Inventors: Das, Sreerupa, Harrison, Gregory A., Hall, Richard, Bodkin, Michael A.
• Primary Examiner(s): Gaffin, Jeffrey A
• Assistant Examiner(s): Traktovenko, Ilya
• Application Number:
  US13/241,674
• Time in Patent Office: 1,026 Days
• Field of Search: 706/21, 706/16, 706/15, 706/62
• US Class Current: 706/21
• CPC Class Codes: G05B 13/04 involving the use of models or simulators
• CPC Class Codes: G05B 23/0283 Predictive maintenance, e.g. involving the monitoring of a system and, based on the monitoring results, taking decisions on the maintenance schedule of the monitored system; Estimating remaining useful life [RUL] preventive maintenance, i.e. planning maintenance according to the available resources without monitoring the system G06Q10/06
• CPC Class Codes: G06E 1/00 Devices for processing exclusively digital data
• CPC Class Codes: G06E 3/00 Devices not provided for in group G06E1/00, e.g. for processing analogue or hybrid data
• CPC Class Codes: G06G 7/00 Devices in which the computing operation is performed by varying electric or magnetic quantities
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: G06N 3/048 Activation functions
• CPC Class Codes: G06N 3/049 Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
• CPC Class Codes: G06N 3/08 Learning methods
• CPC Class Codes: G06N 3/084 Backpropagation, e.g. using gradient descent

Abstract

Mechanisms for predicting a remaining useful life of a cutter head of a milling machine that includes a plurality of flutes are disclosed. Features are extracted from reference data associated with a plurality of reference cutter heads. The reference data includes reference vibration data and reference wear data. At least two Neural Network predictive models are trained in parallel for predicting the remaining life of a new cutter head based upon the extracted features. Operational data associated with the new cutter head is obtained. The operational data includes operational vibration data and operational wear data. Features extracted from the operational data are input into an optimal predictive model of the at least two Neural Network predictive models. A remaining useful life of the new cutter head is estimated by the optimal predictive model.

Claims

1. A method to predict a remaining useful life of a milling machine cutter head comprising a plurality of flutes, the method comprising the steps of:

extracting features from reference data from a plurality of reference cutter heads, the reference data including reference vibration data that quantifies vibrations of the plurality of reference cutter heads in three dimensions in operation and reference wear data that quantifies a wearing of the flutes of the plurality of reference cutter heads after each of a plurality of respective cuts by the plurality of reference cutter heads;

training at least two Neural Network predictive models in parallel for predicting the remaining useful life of a target cutter head based upon the extracted features from the reference data;

evaluating the at least two Neural Network predictive models and selecting an optimal predictive model;

obtaining operational data associated with the target cutter head comprising a plurality of flutes, the operational data including operational vibration data that quantifies vibrations of the target cutter head in three dimensions in operation and operational wear data that quantifies a wearing of the plurality of flutes after each of a plurality of cuts by the target cutter head;

inputting features extracted from the operational data into the optimal predictive model; and

estimating, by the optimal predictive model, the remaining useful life of the target cutter head.

2. The method of claim 1, further comprising:

conditioning the reference data and the operational data prior to extracting the features; and

applying a resilient back propagation learning algorithm having at least one weighting function at the evaluating step for optimizing the at least two Neural Network predictive models.

3. The method of claim 2, wherein the conditioning of the reference data and the operational data is performed by applying at least one noise reduction filter to the reference data and the operational data.

4. The method of claim 1, wherein estimating the remaining useful life of the target cutter head comprises estimating a number of cuts remaining for the target cutter head.

5. The method of claim 1, wherein the operational data is sensed real time from at least one sensor in signal communication with the target cutter head.

6. The method of claim 1, wherein the reference data further comprises reference acoustic emission data that quantifies acoustic emissions of the plurality of reference cutter heads in operation, and the operational data further comprises operational acoustic emission data that quantifies acoustic emissions of the target cutter head in operation.

7. The method of claim 1, wherein the reference wear data quantifies the wearing of each flute of the plurality of reference cutter heads in terms of a reduction of the respective flute in micrometers, and the operational wear data quantifies the wearing of each flute of the target cutter head in micrometers.

8. The method of claim 1, wherein the extracted features comprise time domain features that comprise time domain data statistics selected from a group consisting of root mean square (RMS), crest factor, variance, skewness, and kurtosis.

9. The method of claim 1, wherein the reference data includes reference force data that quantifies a force of the plurality of reference cutter heads in three dimensions on respective work pieces during operation, and the operational data includes operational force data that quantifies a force of the target cutter head in three dimensions on a work piece during operation.

10. The method of claim 1, wherein the evaluating step is performed by applying an evaluating function operable for comparing the at least two Neural Network predictive models against weighting functions to determine which Neural Network predictive model is optimal and based upon the results of the evaluating step, performing at least one of:

generating new Neural Network predictive models and performing the evaluating step; and

selecting the optimal Neural Network predictive model and proceeding to the next step.

11. The method of claim 1, wherein the estimating comprises assessing a plurality of operational data prior to an end of equipment useful life.

12. A non-transitory program storage device for predicting a remaining useful life of a milling machine cutter head comprising a plurality of flutes, the program storage device embodying instructions executable by a computer to carry out the steps of:

extracting features from reference data from a plurality of reference cutter heads, the reference data including reference vibration data that quantifies vibrations of the plurality of reference cutter heads in three dimensions in operation and reference wear data that quantifies a wearing of the flutes of the plurality of reference cutter heads after each of a plurality of respective cuts by the plurality of reference cutter heads;

training at least two Neural Network predictive models in parallel for predicting the remaining useful life of a target cutter head based upon the extracted features from the reference data;

evaluating the at least two Neural Network predictive models and selecting an optimal predictive model;

obtaining operational data associated with the target cutter head comprising a plurality of flutes, the operational data including operational vibration data that quantifies vibrations of the target cutter head in three dimensions in operation and operational wear data that quantifies a wearing of the plurality of flutes after each of a plurality of cuts by the target cutter head;

inputting features extracted from the operational data into the optimal predictive model; and

estimating, by the optimal predictive model, the remaining useful life of the target cutter head.

13. A system for estimating a remaining useful life of a milling machine cutter head comprising a plurality of flutes, the system comprising:

a database comprising reference data from a plurality of reference cutter heads, the reference data including reference vibration data that quantifies vibrations of the plurality of reference cutter heads in three dimensions in operation and reference wear data that quantifies a wearing of the flutes of the plurality of reference cutter heads after each of a plurality of respective cuts by the plurality of reference cutter head;

a processor in signal communication with the database, the processor configured to;

extract features from the reference data;

train a Neural Network predictive model based on the features for predicting a remaining useful life of a target cutter head comprising a plurality of flutes;

validate the Neural Network predictive model;

obtain operational data associated with the target cutter head, the operational data including operational vibration data that quantifies vibrations of the target cutter head in three dimensions in operation and operational wear data that quantifies a wearing of the plurality of flutes after each of a plurality of cuts by the target cutter head;

input extracted features from the operational data into the Neural Network predictive model;

estimate, by the Neural Network predictive model, a remaining useful life of the target cutter head; and

output results that quantify the remaining useful life.

14. The system of claim 13, wherein the Neural Network predictive model is configured to estimate the remaining useful life of the target cutter head in terms of a number of cuts remaining for the target cutter head.

15. The system of claim 13, wherein the validating step is performed by applying an evaluating function operable for comparing the Neural Network predictive model against weighting functions to determine if the Neural Network predictive model is acceptable and based upon the results of the validating step, performing at least one of:

generating a new Neural Network predictive model and performing the validating step; and

accepting the Neural Network predictive model and proceeding to the next step.

16. The system of claim 13, wherein the operational data is sensed real time from at least one sensor in signal communication with the target cutter head.

17. The system of claim 13, wherein the reference wear data quantifies the wearing of each flute of the plurality of reference cutter heads in terms of a reduction of the respective flute in micrometers, and the operational wear data quantifies the wearing of the plurality of flutes of the target cutter head in micrometers.

18. The system of claim 13, wherein the reference data includes reference force data that quantifies a force of the plurality of reference cutter heads in three dimensions on respective work pieces during operation, and the operational data includes operational force data that quantifies a force of the target cutter head in three dimensions on a work piece during operation.

19. The system of claim 13, wherein the Neural Network predictive model comprises three layers.

20. The system of claim 13, further comprising:

determining if the output results are within a defined range and based upon the results of the determining step, generating a signal.