Neural network based refrigerant charge detection algorithm for vapor compression systems

• US 7,680,751 B2
• Filed: 05/31/2006
• Issued: 03/16/2010
• Est. Priority Date: 05/31/2006
• Status: Expired due to Fees

• Current Assignee: Honeywell International Inc.
• Original Assignee: Honeywell International Inc.
• Inventors: Lem, Lawrence A., Hosler, Brian G., Parthasarathy, Girija, Menon, Sunil K.
• Primary Examiner(s): Vincent; David R
• Assistant Examiner(s): Fernandez Rivas; Omar F
• Application Number:
  US11/444,092
• Publication Number:
  US 20070277537A1
• Time in Patent Office: 1,385 Days
• Field of Search: 706 14- 23, 62/13, 621/29, 621/32, 622/31, 621/72, 624/01, 700/275, 700/276, 700/281, 700/282, 700/299, 700/300, 703/2, 703/6, 703/9, 701/100, 701/106
• US Class Current: 706/23
• CPC Class Codes: B64D 13/06 the air being conditioned pressurising B64D13/02
• CPC Class Codes: B64D 2013/0614 with subsystems for cooling avionics
• CPC Class Codes: B64D 2013/0644 including electric motors or generators
• CPC Class Codes: B64D 2013/0674 comprising liquid subsystems
• CPC Class Codes: F25B 2500/19 Calculation of parameters
• CPC Class Codes: F25B 2600/05 Refrigerant levels
• CPC Class Codes: F25B 49/005 of safety devices F25B49/02 and F25B49/04 take precedence
• CPC Class Codes: Y02T 50/50 On board measures aiming to increase energy efficiency

Abstract

Methods and apparatus are provided for determining refrigerant charge in a vapor compressor system (VCS) of an aircraft. The methods and apparatus comprise the following steps of, and/or means for, generating a data set from historical data representative of a plurality of VCS operating conditions over time, identifying one or more steady-state data points in the generated data set, forming a revised data set that includes at least the steady-state data points, using principal components analysis (PCA) to derive values for a plurality of minimally correlated input variables, supplying the derived values for the plurality of minimally correlated input variables and the corresponding values for the VCS refrigerant charge in the revised data set to a nonlinear neural network model, and deriving a simulator model characterizing a relationship between the plurality of minimally correlated input variables and the VCS refrigerant charge.

Claims

1. A method of developing a model for determining refrigerant charge in a vapor compressor system (VCS) of an aircraft, the method comprising the steps of:

(a) receiving, in a processor, a data set generated from historical data representative of a plurality of VCS operating conditions over time, the generated data set comprising a plurality of data points, each data point comprising;

(i) one or more values for a plurality of VCS operating variables reflecting operation of the VCS over a specific time period and corresponding to a specific set of operating conditions; and

(ii) corresponding values for VCS refrigerant charge over the same time period; and

(b) the processor processing the data to;

(i) identify one or more steady-state data points in the generated data set, each steady-state data point corresponding to steady-state operation of the VCS;

(ii) form a revised data set that includes at least the steady-state data points;

(iii) use principal components analysis (PCA) to derive values for a plurality of minimally correlated input variables from the values for the plurality of VCS operating variables in the revised data set;

(iv) supply the derived values for the plurality of minimally correlated input variables, and the corresponding values for the VCS refrigerant charge in the revised data set, to a nonlinear neural network model; and

(v) derive a simulator model characterizing a relationship between the plurality of minimally correlated input variables and the VCS refrigerant charge using the nonlinear neural network model.

2. The method of claim 1, further comprising the steps of the processor processing the data to:

dividing the steady-state data points according to particular ranges of VCS operating conditions, thereby creating a revised data subset for each range of VCS operating conditions; and

calculating average values for the plurality of VCS operating variables for each data subset, for use in deriving the values for the plurality of minimally correlated input variables.

3. The method of claim 1, further comprising the step of the processor processing the data to:

filtering the values for the VCS refrigerant charge in the revised data set.

4. The method of claim 1, further comprising the steps of the processor processing the data to:

divide the steady-state data points into training data and validation data;

remove the validation data from the revised data set, so that the revised data set comprises only the training data for use in step (d); and

test the derived simulator model derived in step (f), utilizing the validation data.

5. The method of claim 1, further comprising the step of the processor processing the data to:

determine the VCS refrigerant charge, utilizing the derived simulator model.

6. The method of claim 1, wherein the plurality of VCS operating variables are selected from the group consisting of motor speed, motor power, compressor inlet temperature (TCPI), condenser outlet temperature (TCDO), evaporator outlet temperature (TEFO), condenser ram inlet temperature (TCDRI), compressor inlet pressure (PCPI), compressor outlet pressure (PCPO), subcooling temperature, and evaporator inlet temperature (TEVI).

7. The method of claim 1, wherein the plurality of VCS operating variables are selected from the group consisting of motor power, TCDO, TEFO, TCDRI, PCPI, PCPO, and subcooling temperature.

8. The method of claim 1, wherein the nonlinear neural network model is a feed forward neural network model with one hidden layer.

9. A method for developing a model for determining refrigerant charge in a vapor compressor system (VCS) of an aircraft, the method comprising the steps of:

(a) receiving, in a processor, a data set generated from historical data representative of a plurality of VCS operating conditions over time, the generated data set comprising a plurality of data points, each data point comprising;

(i) one or more values for a plurality of VCS operating variables reflecting operation of the VCS over a specific time period and corresponding to a specific set of operating conditions; and

(ii) corresponding values for VCS refrigerant charge over the same time period; and

(b) the processor processing the data to;

(i) identify one or more steady-state data points in the generated data set, each steady-state data point corresponding to steady-state operation of the VCS;

(ii) form a revised data set that includes at least the steady-state data points;

(iii) divide the steady-state data points according to particular ranges of VCS operating conditions, thereby creating a revised data subset for each range of VCS operating conditions;

(iv) calculate average values for the plurality of VCS operating variables for each data subset;

(v) supply the average values for the plurality of VCS operating variables to a nonlinear neural network model;

(vi) derive a simulator model characterizing a relationship between the plurality of VCS operating variables and the VCS refrigerant charge using the nonlinear neural network model;

(vii) use principal components analysis (PCA) to derive values for a plurality of minimally correlated input variables from the average values for the plurality of VCS operating variables in the revised data set; and

(viii) supply the derived values for the plurality of minimally correlated input variables, and corresponding values for the VCS refrigerant charge in the revised data set, to a nonlinear neural network model.

10. The method of claim 9, further comprising the step of the processor processing the data to:

filtering the values for the VCS refrigerant charge in the revised data set.

11. The method of claim 9, further comprising the steps of the processor processing the data to:

dividing the steady-state data points into training data and validation data;

removing the validation data from the revised data set, so that the revised data set comprises only the training data for use in deriving the simulator model; and

testing the derived simulator model derived in step (g), utilizing the validation data.

12. The method of claim 9, further comprising the step of the processor processing the data to:

determining the VCS refrigerant charge, utilizing the simulator model.

13. The method of claim 9, wherein the plurality of VCS operating variables are selected from the group consisting of motor speed, motor power, compressor inlet temperature (TCPI), condenser outlet temperature (TCDO), evaporator outlet temperature (TEFO), condenser ram inlet temperature (TCDRI), compressor inlet pressure (PCPI), compressor outlet pressure (PCPO), subcooling temperature, and evaporator inlet temperature (TEVI).

14. The method of claim 9, wherein the plurality of VCS operating variables are selected from the group consisting of motor power, TCDO, TEFO, TCDRI, PCPI, PCPO, and subcooling temperature.

15. The method of claim 9, wherein the nonlinear neural network model is a feed forward neural network model with one hidden layer.

16. An apparatus comprising:

(a) a processor;

(b) a memory coupled to the processor; and

(c) a program residing in memory and being executed by the processor, the program configured to provide a model for determining a refrigerant charge in a vapor compressor system (VCS) of an aircraft via a model through at least the following steps;

(i) generating a data set from historical data representative of a plurality of VCS operating conditions over time, the generated data set comprising a plurality of data points, each data point comprising one or more values for a plurality of VCS operating variables reflecting operation of the VCS over a specific time period, and corresponding values for VCS refrigerant charge over the same time period;

(ii) identifying one or more steady-state data points in the generated data set, each steady-state data point corresponding to steady-state operation of the VCS;

(iii) forming a revised data set that includes at least the steady-state data points;

(iv) using principal components analysis (PCA) to derive values for a plurality of minimally correlated input variables from the values for the plurality of VCS operating variables in the revised data set;

(v) supplying the derived values for the plurality of minimally correlated input variables, and the corresponding values for the VCS refrigerant charge in the revised data set, to a nonlinear neural network model; and

(vi) deriving a simulator model characterizing a relationship between the plurality of minimally correlated input variables and the VCS refrigerant charge using the simulator model.

17. The apparatus of claim 16, wherein the processor is operable to implement the derived simulator model.