Methods and systems for energy prognosis

• US 8,527,223 B2
• Filed: 09/10/2010
• Issued: 09/03/2013
• Est. Priority Date: 09/11/2009
• Status: Active Grant

• Current Assignee: University of Cincinnati
• Original Assignee: University of Cincinnati
• Inventors: AbuAli, Mohamed, Lee, Jay, Zhao, Wenyu
• Primary Examiner(s): Nghiem, Michael
• Assistant Examiner(s): SUAREZ, FELIX E
• Application Number:
  US12/879,439
• Publication Number:
  US 20110066391A1
• Time in Patent Office: 1,089 Days
• Field of Search: 702/60, 702/61, 702/62, 702/64, 702/179, 702/183, 702/186, 702/188, 340/657, 705/10
• US Class Current: 702/61
• CPC Class Codes: G06Q 10/06 Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
• CPC Class Codes: Y02P 80/10 Efficient use of energy, e.g. using compressed air or pressurized fluid as energy carrier
• CPC Class Codes: Y02P 90/80 Management or planning

Abstract

In one embodiment, a method for managing an industrial process wherein a processor transforms electronic data into an energy prognosis is provided. An energy related test signal including effective power test data indicative of electrical energy consumed by an energy consuming machine may be received. The energy related test signal may be partitioned into an energy related test sub-signal. Mathematical functions may be applied to the energy related test sub-signal to extract a feature set of data from the energy related test sub-signal. The feature set of data may be transformed with a transformation matrix into a reduced test feature set. The reduced test feature set may be input into a performance assessment algorithm based at least in part upon a reduced training feature set derived from data indicative of electrical energy consumed during healthy operation. A power-based performance index may be generated with the performance assessment algorithm.

Claims

1. A method for managing an industrial process wherein a processor transforms an energy related test signal into an energy prognosis, the method comprising:

receiving automatically with the processor the energy related test signal comprising effective power test data indicative of electrical energy consumed by an energy consuming machine;

partitioning automatically with the processor the energy related test signal into an energy related test sub-signal;

applying automatically with the processor mathematical functions to the energy related test sub-signal to extract a feature set of data from the energy related test sub-signal;

transforming automatically with the processor the feature set of data with a transformation matrix into a reduced test feature set;

inputting automatically with the processor the reduced test feature set into a performance assessment algorithm, wherein the performance assessment algorithm is based at least in part upon a reduced training feature set derived from data indicative of electrical energy consumed during healthy operation; and

generating automatically with the processor the energy prognosis, wherein the energy prognosis is based at least in part upon a power-based performance index calculated by the performance assessment algorithm.

2. The method for managing an industrial process of claim 1 further comprising calculating automatically with the processor a difference between the power-based performance index and a prior-consecutive power-based performance index, and generating automatically with the processor a fault indication when the difference exceeds a threshold.

3. The method for managing an industrial process of claim 1 further comprising:

inputting automatically with the processor the power-based performance index into a performance prediction algorithm, wherein the performance prediction algorithm is trained with a set of historic power-based performance indexes;

predicting automatically with the processor a future performance index with the performance prediction algorithm, wherein the future performance index is based at least in part upon the power-based performance index; and

generating automatically with the processor a fault indication when the future performance index exceeds a threshold.

4. The method for managing an industrial process of claim 1 further comprising providing a user interface automatically with the processor to receive a time block wherein the energy related test sub-signal is partitioned according to the time block.

5. The method for managing an industrial process of claim 1 further comprising:

providing reactive power training data indicative of electrical energy consumed during healthy operation;

locating automatically with the processor a first peak within the reactive power training data;

identifying automatically with the processor a first subsequent near-zero within the reactive power training data that is nearest to the first peak wherein, the first subsequent near-zero occurs after the first peak;

locating automatically with the processor a second peak within the reactive power training data wherein, the second peak occurs after the first peak;

identifying automatically with the processor a second subsequent near-zero within the reactive power training data that is nearest to the second peak wherein, the first subsequent near-zero occurs after the first peak; and

calculating automatically with the processor a cycle length based at least in part upon the first subsequent near-zero and the second subsequent near-zero.

6. The method for managing an industrial process of claim 5 further comprising calculating automatically with the processor an average cycle length of the reactive power training data, wherein the energy related test sub-signal is partitioned according to the average cycle length.

7. The method for managing an industrial process of claim 5 wherein the energy related test signal comprises reactive power test data, further comprising:

associating automatically with the processor the cycle length with the reactive power test data indicative of electrical energy consumed by the energy consuming machine, wherein the cycle length is associated with a start point corresponding to a beginning of an industrial-machine cycle;

extracting automatically with the processor a cycle step from the reactive power test data, wherein the cycle step is based at least in part upon the start point and a subsequent minimum within the reactive power test data that occurs after the start point; and

associating the cycle step with the effective power test data.

8. The method for managing an industrial process of claim 1 further comprising:

providing real power training data indicative of electrical energy consumed during healthy operation;

applying automatically with the processor the mathematical functions to the real power training data to extract a training feature set of data from the real power training data; and

transforming automatically with the processor the training feature set of data with a dimensionality reduction algorithm into the reduced training feature set, wherein the transformation matrix is defined by the dimensionality reduction algorithm.

9. The method for managing an industrial process of claim 1 further comprising calculating automatically with the processor a statistical overlap of the reduced training feature set and the reduced test feature set with a statistical pattern recognition algorithm, wherein the performance assessment algorithm comprises the statistical pattern recognition algorithm and the power-based performance index is indicative of the statistical overlap of the reduced training feature set and the reduced test feature set.

10. The method for managing an industrial process of claim 1 further comprising:

training automatically with the processor a self-organizing map with the reduced training feature set;

inputting automatically with the processor the reduced test feature set into the self-organizing map;

mapping automatically with the processor the reduced test feature set to a location in the self-organizing map that best correlates to the reduced test feature set.

11. The method for managing an industrial process of claim 10 further comprising calculating automatically with the processor a distance measure indicative of a comparison of the location in the self-organizing map and the reduced test feature set.

12. The method for managing an industrial process of claim 1 wherein the energy prognosis is indicative of an amount of energy consumed to per product.

13. The method for managing an industrial process of claim 1 wherein the energy prognosis is an indication of a statistically significant delta between an immediately prior power-based performance index and the power-based performance index, an alarm when the power-based performance index exceeds a first threshold or a prediction that a future power-based performance index will exceed a second threshold based at least in part upon the power-based performance index.

14. The method for managing an industrial process of claim 1 wherein the energy related test signal further comprises data indicative of temperature, humidity, pressure, voltage, or current.

15. The method for managing an industrial process of claim 1 wherein the mathematical functions comprise a time domain function applied to a time block, a cycle or a cycle step.

16. An energy management system comprising:

an energy consuming machine for producing a product;

an energy sensor communicably coupled to the energy consuming machine wherein the energy sensor generates an energy related test signal comprising real power test data and reactive power test data indicative of energy consumed by the energy consuming machine;

a processor communicably coupled to the energy sensor, wherein the processor executes machine readable instructions to;

receive the energy related test signal;

partition the energy related test signal into an energy related test sub-signal;

apply time domain functions and/or frequency domain functions to the energy related test sub-signal to extract a feature set of data from the energy related test sub-signal;

transform the feature set of data with a principle component analysis into a reduced test feature set;

compare the reduced test feature set to training data indicative of a normal operating condition of the energy consuming machine;

generate a power-based performance index with a performance assessment algorithm, wherein the power-based performance index is based at least in part upon a difference between the reduced test feature set and the training data; and

generate a fault indication based at least in part upon the power-based performance index.

17. The energy management system of claim 16 wherein the processor executes the machine readable instructions to:

extract a cycle step from the reactive power test data;

associate the cycle step with the real power test data; and

generate a component level performance prediction related to health of a component of the energy consuming machine.

18. The energy management system of claim 16 wherein the energy consuming machine is a milling machine, a pressing machine, an injection molding machine or a grinding machine.

19. The energy management system of claim 16 wherein the energy sensor is a temperature sensor, a humidity sensor, a pressure sensor, a current sensor, a voltage sensor, a power sensor, or an output from a controller.

20. A toolkit for managing an industrial process, wherein the toolkit is stored on a machine readable medium and comprises machine readable instructions to transform an energy related test signal into an energy prognosis, the machine readable instructions comprising:

a signal loading algorithm to receive the energy related test signal comprising real power test data indicative of electrical energy consumed by an energy consuming machine;

a partitioning algorithm to transform the energy related test signal into an energy related test sub-signal;

a feature extraction algorithm to transform the energy related test sub-signal into a feature set of data from the energy related test sub-signal;

a dimensionality reduction algorithm to transform the feature set of data into a reduced test feature set;

a performance assessment algorithm for comparing the reduced test feature set to a baseline indicative of a normal operating condition of the energy consuming machine;

a performance prediction algorithm for generating the energy prognosis that forecasts a future health condition of the energy consuming machine based at least in part upon the reduced test feature set; and

a visualization algorithm for displaying the energy prognosis.