Virtual sensor system and method

• US 20080312756A1
• Filed: 06/15/2007
• Published: 12/18/2008
• Est. Priority Date: 06/15/2007
• Status: Active Grant

• Current Assignee: Caterpillar Incorporated
• Original Assignee: Caterpillar Incorporated
• Inventors: Jayachandran, Amit, Grichnik, Anthony J., Kesse, Mary L., Seskin, Michael
• Granted Patent: US 7,787,969 B2
• Time in Patent Office: Days
• US Class Current: 700/29
• CPC Class Codes: F02D 41/1405 Neural network control
• CPC Class Codes: G05B 19/0423 Input/output
• CPC Class Codes: G05B 2219/37537 Virtual sensor

Abstract

A method is provide for providing sensors for a machine. The method may include obtaining data records including data from a plurality of sensors for the machine and determining a virtual sensor corresponding to one of the plurality of sensors. The method may also include establishing a virtual sensor process model of the virtual sensor indicative of interrelationships between at least one sensing parameters and a plurality of measured parameters based on the data records and obtaining a set of values corresponding to the plurality of measured parameters. Further, the method may include calculating the values of the at least one sensing parameters substantially simultaneously based upon the set of values corresponding to the plurality of measured parameters and the virtual sensor process model and providing the values of the at least one sensing parameters to a control system.

Claims

1. A method providing sensors for a machine, comprising:

obtaining data records including data from a plurality of sensors for the machine;

determining a virtual sensor corresponding to one of the plurality of sensors;

establishing a virtual sensor process model of the virtual sensor indicative of interrelationships between at least one sensing parameters and a plurality of measured parameters based on the data records;

obtaining a set of values corresponding to the plurality of measured parameters;

calculating the values of the at least one sensing parameters substantially simultaneously based upon the set of values corresponding to the plurality of measured parameters and the virtual sensor process model; and

providing the values of the at least one sensing parameters to a control system.

2. The method according to claim 1, wherein determining includes:

calculating correlation values between any two sensors of the plurality of sensors;

separating the plurality of sensors into a plurality of sensor groups based on the correlation values; and

determining the virtual sensor corresponding to the one of the plurality of sensors from a desired sensor group.

3. The method according to claim 2, wherein separating includes:

creating a correlation matrix with rows representing the respective plurality of sensors, columns representing the respective plurality of sensors, and each element representing a correlation value between the corresponding two sensors from a row and a column, respectively;

determining a score for each of the plurality of sensors based on the correlation matrix; and

separating the plurality of sensors into the plurality of sensor groups based on the score of each of the plurality of sensors.

4. The method according to claim 3, wherein the plurality of sensor groups include:

a first sensor group containing desired physical sensors;

a second sensor group containing sensors with combinations of physical sensors and virtual sensors; and

a third sensor group containing virtual sensors.

5. The method according to claim 1, wherein the establishing includes:

obtaining data records associated with one or more input variables and the at least one sensing parameters;

selecting the plurality of measured parameters from the one or more input variables;

generating a computational model indicative of the interrelationships between the plurality of measured parameters and the at least one sensing parameters;

determining desired statistical distributions of the plurality of measured parameters of the computational model; and

recalibrating the plurality of measured parameters based on the desired statistical distributions to define a desired input space.

6. The method according to claim 5, wherein selecting further includes:

pre-processing the data records; and

using a genetic algorithm to select the plurality of measured parameters from the one or more input variables based on a mahalanobis distance between a normal data set and an abnormal data set of the data records.

7. The method according to claim 5, wherein generating further includes:

creating a neural network computational model;

training the neural network computational model using the data records; and

validating the neural network computation model using the data records.

8. The method according to claim 5, wherein determining further includes:

determining a candidate set of the measured parameters with a maximum zeta statistic using a genetic algorithm; and

determining the desired distributions of the measured parameters based on the candidate set,

wherein the zeta statistic ζ

is represented by;

$\mathrm{\zeta <br><br>}=\underset{1}{\overset{j}{\sum <br><br>}}<br><br>\underset{1}{\overset{i}{\sum <br><br>}}<br><br><br><br>S_{\mathrm{ij}}<br><br><br><br>\left(\frac{{\mathrm{\sigma <br><br>}}_i}{{\stackrel{\_}{x}}_i}\right)<br><br>\left(\frac{{\stackrel{\_}{x}}_j}{{\mathrm{\sigma <br><br>}}_j}\right),$

provided that x_i represents a mean of an ith input;

x_j represents a mean of a jth output;

σ

_i represents a standard deviation of the ith input;

σ

_j represents a standard deviation of the jth output; and

|S_ij| represents sensitivity of the jth output to the ith input of the computational model.

9. The method according to claim 1, wherein the providing includes:

separately obtaining values of the at least one sensing parameters from a physical sensor;

determining that the physical sensor has failed; and

providing the values of the at least one sensing parameters from the virtual sensor process model to the control system.

10. A method for retrofitting a first machine lacking a supporting physical sensor with a virtual sensor created based on a second machine with a supporting physical sensor, comprising:

obtaining data records including data from a plurality of sensors that are available on both of the first machine and the second machine, and from the supporting physical sensor on the second machine;

calculating correlation values between the supporting physical sensor and each of the plurality of sensors based on the data records;

selecting correlated sensors from the plurality of sensors based on the correlation values;

creating the virtual sensor for the supporting physical sensor based on the correlated sensors; and

using the virtual sensor in the first machine to provide functionalities that were provided by the supporting physical sensor on the second machine.

11. The method according to claim 10, wherein creating includes:

establishing the virtual sensor process model indicative of interrelationships between at least one sensing parameters provided by the plurality of sensors and a plurality of measured parameters of the supporting physical sensor based on the data records.

12. The method according to claim 10, wherein using includes:

obtaining a set of values corresponding to the plurality of measured parameters;

calculating the values of the at least one sensing parameters simultaneously based upon the set of values corresponding to the plurality of measured parameters and the virtual sensor process model; and

providing the values of the at least one sensing parameters to a control system of the first machine.

13. The method according to claim 10, wherein selecting includes:

selecting the correlated sensors each with a correlation value beyond a predetermined threshold.

14. The method according to claim 11, wherein the establishing includes:

obtaining data records associated with one or more input variables and the at least one sensing parameters;

selecting the plurality of measured parameters from the one or more input variables;

generating a computational model indicative of the interrelationships between the plurality of measured parameters and the at least one sensing parameters;

determining desired statistical distributions of the plurality of measured parameters of the computational model; and

recalibrating the plurality of measured parameters based on the desired statistical distributions to define a desired input space.

15. The method according to claim 14, wherein selecting further includes:

pre-processing the data records; and

using a genetic algorithm to select the plurality of measured parameters from the one or more input variables based on a mahalanobis distance between a normal data set and an abnormal data set of the data records.

16. The method according to claim 14, wherein generating further includes:

creating a neural network computational model;

training the neural network computational model using the data records; and

validating the neural network computation model using the data records.

17. The method according to claim 14, wherein determining further includes:

determining a candidate set of the measured parameters with a maximum zeta statistic using a genetic algorithm; and

determining the desired distributions of the measured parameters based on the candidate set,

wherein the zeta statistic ζ

is represented by;

$\mathrm{\zeta <br><br>}=\underset{1}{\overset{j}{\sum <br><br>}}<br><br>\underset{1}{\overset{i}{\sum <br><br>}}<br><br><br><br>S_{\mathrm{ij}}<br><br><br><br>\left(\frac{{\mathrm{\sigma <br><br>}}_i}{{\stackrel{\_}{x}}_i}\right)<br><br>\left(\frac{{\stackrel{\_}{x}}_j}{{\mathrm{\sigma <br><br>}}_j}\right),$

provided that x_i represents a mean of an ith input;

x_j represents a mean of a jth output;

σ

_i represents a standard deviation of the ith input;

σ

_j represents a standard deviation of the jth output; and

|S_ij| represents sensitivity of the jth output to the ith input of the computational model.

18. A computer system, comprising:

a database configured to store information relevant to a virtual sensor process model; and

a processor configured to;

obtain data records including data from a plurality of sensors for the machine;

determine a virtual sensor corresponding to one of the plurality of sensors;

establish the virtual sensor process model of the virtual sensor indicative of interrelationships between at least one sensing parameters and a plurality of measured parameters based on the data records;

obtain a set of values corresponding to the plurality of measured parameters;

calculate the values of the at least one sensing parameters substantially simultaneously based upon the set of values corresponding to the plurality of measured parameters and the virtual sensor process model; and

provide the values of the at least one sensing parameters to a control system.

19. The computer system according to claim 18, wherein, to determine the virtual sensor, the processor is further configured to:

calculate correlation values between any two sensors of the plurality of sensors;

separate the plurality of sensors into a plurality of sensor groups based on the correlation values; and

determine the virtual sensor corresponding to the one of the plurality of sensors from a desired sensor group.

20. The computer system according to claim 19, wherein, to separate the plurality sensors, the processor is further configured to:

create a correlation matrix with rows representing the respective plurality of sensors, columns representing the respective plurality of sensors, and each element representing a correlation value between the corresponding two sensors from a row and a column, respectively;

determine a score for each of the plurality of sensors based on the correlation matrix; and

separate the plurality of sensors into the plurality of sensor groups based on the score of each of the plurality of sensors.

21. A machine having a retrofitted virtual sensor to provide functionalities of a corresponding physical sensor without the supporting physical sensor being installed on the machine, comprising:

a power source configured to provide power to the machine;

a control system configured to control the power source; and

a virtual sensor system, corresponding to the supporting physical sensor, including a virtual sensor process model indicative of interrelationships between at least one sensing parameters provided by a plurality of sensors and a plurality of measured parameters of the supporting physical sensor, the virtual sensor system being configured to;

obtain a set of values corresponding to the plurality of measured parameters;

calculate the values of the at least one sensing parameters simultaneously based upon the set of values corresponding to the plurality of measured parameters and the virtual sensor process model; and

provide the values of the at least one sensing parameters to the control system to provide functionalities corresponding to the supporting physical sensor,

wherein the virtual sensor is created by;

obtaining data records including data from the plurality of sensors and the supporting physical sensor;

calculating correlation values between the supporting physical sensor and the plurality of sensors based on the data records;

selecting correlated sensors from the plurality of sensors based on the correlation values; and

creating the virtual sensor of the supporting physical sensor based on the correlated sensors.

22. The method according to claim 21, wherein selecting includes:

selecting the correlated sensors each with a correlation value beyond a predetermined threshold.

23. The method according to claim 21, wherein the creating includes:

obtaining data records associated with one or more input variables and the at least one sensing parameters;

selecting the plurality of measured parameters from the one or more input variables;

generating a computational model indicative of the interrelationships between the plurality of measured parameters and the at least one sensing parameters;

determining desired statistical distributions of the plurality of measured parameters of the computational model; and

recalibrating the plurality of measured parameters based on the desired statistical distributions to define a desired input space.