System, method and algorithm for data-driven equipment performance monitoring

• US 7,966,152 B2
• Filed: 04/23/2008
• Issued: 06/21/2011
• Est. Priority Date: 04/23/2008
• Status: Active Grant

• Current Assignee: Honeywell International Inc.
• Original Assignee: Honeywell International Inc.
• Inventors: Stluka, Petr, Rojicek, Jiri
• Primary Examiner(s): Nghiem; Michael P
• Assistant Examiner(s): Ngon; Ricky
• Application Number:
  US12/108,076
• Publication Number:
  US 20090271150A1
• Time in Patent Office: 1,154 Days
• Field of Search: 702 57- 59, 702 81- 84, 702/105, 702182-184, 702190-195, 700/108, 700/109, 700/110, 700/275, 700276-282, 700299-306, 703/2, 703/6, 703/13
• US Class Current: 702/182
• CPC Class Codes: G05B 23/0232 based on qualitative trend analysis, e.g. system evolution

Abstract

A method for automatically monitoring the performance of equipment includes compiling current operating conditions associated with current conditions (PMc). A historical database including a plurality of stored operating conditions and associated stored performance measure (PM*) is searched, each stored operating condition including at least one stored sensor reading, wherein at least one similar operating condition is identified in the search using distances between the current operating conditions and the stored operating conditions. The performance measure (PM*) associated with the similar operating condition is fit to generate a regression model. The regression model is applied to the current operating condition to generate an estimate for the performance measure for the current operating condition (PMe). A difference between PMc and PMe is computed. The first difference is compared to a predetermined threshold, wherein a warning is automatically generated if the first difference has a value greater than the predetermined threshold.

Claims

1. A method for automatically monitoring the performance of equipment, comprising the steps of:

compiling current operating conditions of said equipment comprising at least one current sensor reading and at least one current performance measurement (PMc) associated with said current operating conditions;

searching, using a computing system including non-transitory machine readable storage, a historical database comprising a plurality of stored operating conditions and at least one associated stored performance measure (PM*) that are stored in the non-transitory machine readable storage, each said stored operating conditions comprising at least one stored sensor reading wherein at least one similar operating condition is identified using distances between said current operating conditions and said plurality of stored operating conditions;

fitting said stored performance measure (PM*) associated with said similar operating condition to generate a regression model;

applying said regression model to said current operating conditions to generate an estimate for said performance measure for said current operating conditions (PMe);

computing a first difference between said PMc and said PMe, and

comparing said first difference to a predetermined threshold, wherein a warning is automatically generated if said first difference has a value greater than said predetermined threshold.

2. The method of claim 1, wherein said stored operating conditions comprise respective run times, further wherein said run times are used as an independent variable in said fitting step, and said applying said regression model further includes a time associated with said current operating conditions.

3. The method of claim 1, further comprising the step of applying said regression model to said similar operating condition evaluated at an initial period of operation for said equipment to generate an ideal performance for said performance measure (PMi);

said computing further comprises a second difference between said PMc and said PMi, and said comparing comprises comparing at least one of said first and second difference to respective predetermined thresholds, wherein a first warning is generated if said first difference is above said first predetermined threshold and a second warning different from said first warning is generated if said second difference is above said second predetermined threshold.

4. The method of claim 3, wherein both said first difference and said second difference are calculated in said calculating and compared in said comparing step.

5. The method of claim 1, wherein said current operating conditions further comprises at least one current control signal and said stored operating conditions each further comprise at least one stored control signal.

6. The method of claim 1, wherein said equipment comprises industrial processing equipment.

7. The method of claim 1, wherein said equipment comprises building HVAC equipment.

8. The method of claim 1, wherein said distances are based on Euclidean distances.

9. The method of claim 1, wherein said searching comprises searching over a selected subset of said operating conditions.

10. The method of claim 1, wherein said sensor reading comprises at least one of temperature, pressure and flow rate.

11. The method of claim 1, wherein said PMc and said PMs are calculated from said operating conditions using a formula.

12. The method of claim 1, wherein said PMc and said PMs are obtained from one of said operating conditions.

13. The method of claim 1, further comprising the step of using said warning to determined when to perform maintenance or cleaning on said equipment.

14. A method for automatically monitoring the performance of equipment, comprising the steps of:

compiling current operating conditions of said equipment comprising at least one current sensor reading and at least one current performance measurement (PMc) associated with said current operating conditions;

searching, using a computing system including non-transitory machine readable storage, a historical database comprising a plurality of stored operating conditions and at least one stored associated performance measurement (PM*) associated therewith that are stored in the non-transitory machine readable storage, each of said stored operating conditions comprising at least one stored sensor reading and a run time, wherein at least one similar operating condition is identified using distances between said current operating conditions and said plurality of stored operating conditions;

fitting said stored performance measurement (PM*) associated with said similar operating condition to generate a regression model, wherein said run time is an independent variable;

applying said regression model to said current operating conditions and a time associated with said current operating conditions to generate an estimate for said stored performance measurement for said current operating conditions (PMe);

applying said regression model to said similar operating condition evaluated at an initial period of operation for said equipment to generate an ideal performance for said stored performance measurement (PMi);

computing at least one of a first difference between said PMc and said PMi and a second difference between said PMc and said PMe, and

comparing at least one of said first and second difference to respective first and second predetermined thresholds, wherein a first warning is generated if said first differences is above said first predetermined threshold and a second warning different from said first warning is generated if said second difference is above said second predetermined threshold.

15. The method of claim 14, wherein said current operating conditions further comprises at least one current control signal and said stored operating conditions each further comprise at least one stored control signal.

16. A non-transitory machine readable storage for automatically monitoring and assessing the performance of equipment having an associated computing system including a historical database comprising a plurality of stored operating conditions each comprising at least one associated stored performance measure (PM*), said non-transitory machine readable storage having stored thereon a computer program, the non-transitory machine readable storage comprising:

code for searching said historical database, wherein at least one similar operating condition is identified using distances between a current operating condition comprising at least one current sensor reading and at least one current performance measurement associated with said current conditions (PMc) and said plurality of stored operating conditions;

code for fitting said stored performance measure (PM*) associated with said similar operating condition to generate a regression model;

code for applying said regression model to said current operating conditions to generate an estimate for said performance measure for said current operating conditions (PMe);

code for computing a first difference between said PMc and said PMe, and

code for comparing said first difference to a predetermined threshold, wherein a warning is generated if said first difference is above said respective predetermined threshold.

17. A system comprising equipment having automatic monitoring for assessing the performance of said equipment, comprising:

a plurality of sensors for reporting sensor readings and a plurality of actuators for controlling operational aspects of said equipment;

a computing system coupled to said equipment comprising a controller, processor and software module, and memory including a stored historical database comprising a plurality of stored operating conditions and associated stored performance measures (PM*);

wherein said software module includes;

code for compiling current operating conditions comprising at least one current sensor reading and at least one current control signal, and at least one current performance measurement associated with said current operating conditions (PMc);

code for searching said historical database comprising said plurality of stored operating conditions and identifying at least one similar operating condition using distances between said current operating conditions and said plurality of stored operating conditions;

code for fitting said stored performance measure (PM*) associated with said similar operating condition to generate a regression model;

code for applying said regression model to said current operating conditions to generate an estimate for said performance measure for said current operating conditions (PMe);

code for computing a first difference said PMc and said PMe, and

code for comparing said first difference to a predetermined threshold, wherein warnings are generated if said first difference is above said predetermined threshold.

18. The system of claim 17, wherein said equipment comprises industrial processing equipment.

19. The system of claim 17, wherein said equipment comprises building HVAC equipment.