Sensor validation apparatus and method
First Claim
1. An apparatus for detecting one or more sensor faults in a measured process comprising:
- A pre-processing unit for receiving a working vector of signals including measured sensor values, the pre-processing unit normalizing the measured sensor values, resulting in pre-processed sensor values;
A model unit coupled to the pre-processing unit, which converts pre-processed sensor values to equation error values that contain mainly measurement noise;
A structured residual unit coupled to the model unit, which contains a plurality of transforms, referred to as structured residual transforms, each such transform converting equation error values to a structured residual value, and each such transform designed to be insensitive to faults in a subset of sensors;
A detection unit coupled to the model unit, the detection unit monitoring the relationship among the equation error values, occurrence of a significant deviation of said relationship from expected relationship resulting in a detection event;
An identification unit coupled to the structured residual unit and the detection unit, the identification unit being activated by a detection event, and using the structured residual values to determine if one or more sensors are faulty, said determination resulting in an identification event;
An estimation unit coupled to the pre-processing unit, the structured residual unit and the identification unit, the estimation unit being activated by an identification event, and estimating fault sizes for each of the identified faulty sensors;
A replacement unit coupled to the estimation unit, the replacement unit calculating replacement values for the faulty measured sensor values in the working signal values by subtracting the estimated fault size from the corresponding measured sensor value for all identified faults;
A classification unit coupled to the estimation unit, the classification unit being active when the estimation unit is active, and classifying the identified sensor faults into a fixed set of fault types.
4 Assignments
0 Petitions
Accused Products
Abstract
An apparatus and method is disclosed for detecting, identifying, and classifying faults occurring in sensors measuring a process. If faults are identified in one or more sensors, the apparatus and method provide replacement values for the faulty sensors so that any process controllers and process monitoring systems that use these sensors can remain in operation during the fault period. The identification of faulty sensors is achieved through the use of a set of structured residual transforms that are uniquely designed to be insensitive to specific subsets of sensors, while being maximally sensitive to sensors not in the subset. Identified faults are classified into one of the types Complete Failure, Bias, Drift, Precision Loss, or Unknown.
-
Citations
68 Claims
-
1. An apparatus for detecting one or more sensor faults in a measured process comprising:
-
A pre-processing unit for receiving a working vector of signals including measured sensor values, the pre-processing unit normalizing the measured sensor values, resulting in pre-processed sensor values;
A model unit coupled to the pre-processing unit, which converts pre-processed sensor values to equation error values that contain mainly measurement noise;
A structured residual unit coupled to the model unit, which contains a plurality of transforms, referred to as structured residual transforms, each such transform converting equation error values to a structured residual value, and each such transform designed to be insensitive to faults in a subset of sensors;
A detection unit coupled to the model unit, the detection unit monitoring the relationship among the equation error values, occurrence of a significant deviation of said relationship from expected relationship resulting in a detection event;
An identification unit coupled to the structured residual unit and the detection unit, the identification unit being activated by a detection event, and using the structured residual values to determine if one or more sensors are faulty, said determination resulting in an identification event;
An estimation unit coupled to the pre-processing unit, the structured residual unit and the identification unit, the estimation unit being activated by an identification event, and estimating fault sizes for each of the identified faulty sensors;
A replacement unit coupled to the estimation unit, the replacement unit calculating replacement values for the faulty measured sensor values in the working signal values by subtracting the estimated fault size from the corresponding measured sensor value for all identified faults;
A classification unit coupled to the estimation unit, the classification unit being active when the estimation unit is active, and classifying the identified sensor faults into a fixed set of fault types. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34)
The normalizing in the pre-processing units is achieved by scaling and offsetting the measured sensor values.
-
-
3. The apparatus of claim 1 wherein:
The working signal includes known status information about the measured sensor values.
-
4. The apparatus of claim 1 wherein:
The converting in the model unit is achieved by multiplying the pre-processed sensor values by a matrix to produce the equation error.
-
5. The apparatus of claim 4 wherein:
The matrix is derived from the residual part of a principal component analysis.
-
6. The apparatus of claim 4 wherein:
The matrix is derived from the residual part of a set of partial least squares models, one for each sensor value.
-
7. The apparatus of claim 4 wherein:
The matrix is derived from mass balance or energy balance of the measured process.
-
8. The apparatus of claim 1 wherein:
Each transform of the structured residual unit is in the form of a vector dot product with the equation error.
-
9. The apparatus of claim 1 wherein:
Each transform of the structured residual unit is designed to be insensitive to faults in a subset of sensors, but maximally sensitive to faults in all other sensors not in the subset.
-
10. The apparatus of claim 1 wherein:
Each transform of the structured residual unit is designed to be insensitive to a fault in single sensor, but maximally sensitive to faults in all other sensors.
-
11. The apparatus of claim 1 wherein:
The subset of sensors defining each transform includes at least all sensors of known bad status.
-
12. The apparatus of claim 1 wherein the detection unit monitors the relationship among the equation error values by:
-
calculating a detection index which is a function of the equation errors; and
comparing said detection index to a threshold in order to detect occurrence of a significant deviation of said relationship from expected relationship.
-
-
13. The apparatus of claim 12 wherein:
The detection index is obtained by summing the squared values of the equation error values.
-
14. The apparatus of claim 12 wherein:
-
The equation error is multiplied by a matrix to decorrelate the equation error values, resulting in decorrelated equation error values; and
The detection index is obtained by summing the squared values of the decorrelated equation error values.
-
-
15. The apparatus of claim 12 wherein:
The threshold is determined using statistical techniques.
-
16. The apparatus of claim 12 wherein:
-
The detection index is filtered, at least according to time, in order to smooth out the effects of transients and noise, resulting in a filtered detection index; and
said filtered detection index is used in place of the detection index to monitor the relationship among the equation error values; and
said filtered detection is compared to a threshold in order to detect occurrence of a significant deviation of said relationship from expected relationship.
-
-
17. The apparatus of claim 16 wherein:
the detection index is filtered by applying an exponentially weighted moving average filter.
-
18. The apparatus of claim 17 wherein:
The threshold for the filtered detection index is calculated from the detection index threshold using an auto-correlation function.
-
19. The apparatus of claim 1 wherein:
-
The identification unit compares each structured residual value to a corresponding threshold; and
An identification event occurs if the value of all but one structured residual value, referred to as the identified structured residual value, exceeds its corresponding threshold; and
The determined faulty sensors are the subset of sensors for which, by design, the structured residual transform corresponding to the identified structured residual value is insensitive to the presence of faults in said subset.
-
-
20. The apparatus of claim 1 wherein the identification unit applies conversion operations to the structured residual values resulting in converted structured residual indices and
The identification unit compares each converted structured residual index to a corresponding threshold; - and
An identification event occurs if the converted structured residual index of all but one structured residual, referred to as the identified converted structured residual index, exceeds its corresponding threshold; and
The determined faulty sensors are the subset of sensors for which, by design, the structured residual transform corresponding the converted structured residual index is insensitive to the presence of faults in said subset.
- and
-
21. The apparatus of claim 20 wherein:
The conversion operation includes the operations of squaring and scaling.
-
22. The apparatus of claim 20 wherein:
The conversion operation includes applying an exponentially weighted moving average filter in time.
-
23. The apparatus of claim 22 wherein:
The threshold for each converted structured residual index is calculated from the corresponding structured residual threshold using an auto-correlation function.
-
24. The apparatus of claim 1 wherein the identification unit applies accumulating operations to the structured residual values resulting in accumulated structured residual indices, said operations accumulating since the time of the fault detection event, and
The identification unit compares each accumulated structured residual index to a corresponding threshold; - and
An identification event occurs if the accumulated structured residual index of all but one structured residual, referred to as the identified accumulated structured residual index, exceeds its corresponding threshold; and
The determined faulty sensors are the subset of sensors for which, by design, the structured residual transform corresponding to the identified accumulated structured residual index is insensitive to the presence of faults in said subset.
- and
-
25. The apparatus of claim 24 wherein:
The accumulating operation calculates a generalized likelihood ratio index that is designed to detect significant changes in the statistical mean of the structured residuals since the time of the fault detection event.
-
26. The apparatus of claim 24 wherein:
The accumulating operation calculates the statistical mean of the structured residuals since the time of the fault detection event, resulting in an accumulated mean index.
-
27. The apparatus of claim 24 wherein:
The accumulating operation calculates the statistical variance of the structured residuals since the time of the fault detection event, resulting in an accumulated variance index.
-
28. The apparatus of claim 1 wherein:
The estimated fault size for each of the identified sensor faults is calculated by minimizing the effect of said identified sensor faults on the equation error values.
-
29. The apparatus of claim 1 wherein:
-
The classification unit calculates a plurality of regression lines, one for each identified faulty sensor, each such regression line calculated using the measured sensor values for the corresponding faulty sensor since the time of identification, resulting in a plurality of sensor value regression lines; and
The classification unit calculates a plurality of regression lines, one for each identified faulty sensor, each such regression line calculating using estimated fault sizes for the corresponding faulty sensor since the time of identification, result in a plurality of fault size regression lines; and
The classification unit uses slope, bias, and residual information from sensor value regression lines and fault size regression lines to classify the identified sensor faults into a fixed set of fault types.
-
-
30. The apparatus of claim 29 wherein:
A faulty sensor is classified as type “
complete failure”
if the residual of the corresponding sensor value regression line is below a threshold, and if the slope of the corresponding sensor value regression line is statistically equal to zero.
-
31. The apparatus of claim 29 wherein:
A faulty sensor is classified as type “
bias”
if the residual of the corresponding fault size regression line is below a threshold, and if the slope of the corresponding fault size regression line is statistically equal to zero, and if the bias of the corresponding fault size regression line is statistically unequal to zero.
-
32. The apparatus of claim 29 wherein:
A faulty sensor is classified as type “
drift”
if the residual of the corresponding fault size regression line is below a threshold, and if the slope of the corresponding fault size regression line is statistically unequal to zero.
-
33. The apparatus of claim 29 wherein:
A faulty sensor is classified as type “
precision loss”
if the residual of the corresponding fault size regression line is above a threshold, and if the slope and bias of the corresponding fault size regression line are both statistically equal to zero.
-
34. The apparatus of claim 26 wherein:
A faulty sensor is classified as type “
precision loss”
if the identification event is triggered by an accumulated variance index exceeding its threshold.
-
35. A computer method for detecting one or more sensor faults in a measured process which comprises the steps:
-
Receiving a working vector of signals including measured sensor values, and preprocessing the measured sensor values, resulting in pre-processed sensor values;
Converting the pre-processed sensor values to equation error values that contain mainly measurement noise;
Applying a plurality of transforms to the equation error values resulting in a plurality of structured residual values, said transforms referred to as structured residual transforms and designed to be insensitive to faults in a subset of sensors;
Monitoring the relationship among the equation error values, occurrence of a significant deviation of said relationship from expected relationship resulting in a detection event;
In the case that a detection event occurs, using the structured residual values to determine if one or more sensors are faulty, said determination resulting in an identification event;
In the case that an identification event occurs, estimating fault sizes for each of the identified faulty sensors;
In the case that an identification event occurs, replacing faulty measured sensor values in the working signal with corrected values by subtracting the estimated fault size from the corresponding measured sensor value for all identified faults;
In the case that an identification event occurs, classifying the identified sensor faults into a fixed set of fault types. - View Dependent Claims (36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68)
The step of pre-processing the measured sensor values is achieved by scaling and offsetting.
-
-
37. The method of claim 35 wherein:
The working signal includes known status information about the measured sensor values.
-
38. The method of claim 35 wherein:
The step of converting the pre-processed sensor values to equation error values is achieved by multiplying the preprocessed sensor values by a matrix.
-
39. The method of claim 38 wherein:
The matrix is derived from the residual part of a principal component analysis.
-
40. The method of claim 38 wherein:
The matrix is derived from the residual part of a set of partial least squares models, one for each sensor value.
-
41. The method of claim 38 wherein:
The matrix is derived from mass balance or energy balance of the measured process.
-
42. The method of claim 35 wherein:
Each structured residual transform consists of a vector of coefficients that are applied as a dot product to the equation error.
-
43. The method of claim 35 wherein:
Each structured residual transform is designed to be insensitive to faults in a subset of sensors, but maximally sensitive to faults in all other sensors not in the subset.
-
44. The method of claim 35 wherein:
Each structured residual transform is designed to be insensitive to a fault in single sensor, but maximally sensitive to faults in all other sensors.
-
45. The method of claim 35 wherein:
The subset of sensors defining each transform includes at least all sensors of known bad status.
-
46. The method of claim 35 wherein the step of monitoring the relationship among the equation error values consists of:
-
calculating a detection index which is a function of the equation errors; and
comparing said detection index to a threshold in order to detect occurrence of a significant deviation of said relationship from expected relationship.
-
-
47. The method of claim 46 wherein:
The detection index is obtained by summing the squared values of the equation error values.
-
48. The method of claim 46 wherein:
-
The equation error is multiplied by a matrix to decorrelate the equation error values, resulting in decorrelated equation error values; and
The detection index is obtained by summing the squared values of the decorrelated equation error values.
-
-
49. The method of claim 47 or claim 48 wherein:
The threshold is determined using statistical techniques.
-
50. The method of claim 47 or claim 48 wherein:
-
The detection index is filtered, at least according to time, in order to smooth out the effects of transients and noise, resulting in a filtered detection index; and
said filtered detection index is used in place of the detection index to monitor the relationship among the equation error values; and
said filtered detection is compared to a threshold in order to detect occurrence of a significant deviation of said relationship from expected relationship.
-
-
51. The method of claim 50 wherein:
the detection index is filtered by applying an exponentially weighted moving average filter.
-
52. The method of claim 51 wherein:
The threshold for the filtered detection index is calculated from the detection index threshold using an auto-correlation function.
-
53. The method of claim 35 wherein the step of determining if one or more sensors are faulty consists of:
-
Comparing each structured residual value to a corresponding threshold; and
Generating an identification event if the value of all but one structured residual value, referred to as the identified structured residual value, exceeds its corresponding threshold; and
Determining that the faulty sensors are the subset of sensors for which, by design, the structured residual transform corresponding to the identified structured residual value is insensitive to the presence of faults in said subset.
-
-
54. The method of claim 35 wherein conversion operations are applied to the structured residual values resulting in converted structured residual indices and the step of determining if one or more sensors are faulty consists of:
-
Comparing each converted structured residual index to a corresponding threshold; and
Generating an identification event if the converted structured residual index of all but one structured residual, referred to as the identified converted structured residual index, exceeds its corresponding threshold; and
Determining that the faulty sensors are the subset of sensors for which, by design, the structured residual transform corresponding the converted structured residual index is insensitive to the presence of faults in said subset.
-
-
55. The method of claim 54 wherein:
The conversion operation includes the operations of squaring and scaling.
-
56. The method of claim 54 wherein:
The conversion operation includes applying an exponentially weighted moving average filter in time.
-
57. The method of claim 56 wherein:
The threshold for each converted structured residual index is calculated from the corresponding structured residual threshold using an auto-correlation function.
-
58. The method of claim 35 wherein accumulating operations are applied to the structured residual values resulting in accumulated structured residual indices, said operations accumulating since the time of the fault detection event, and the step of determining if one or more sensors are faulty consists of:
-
Comparing each accumulated structured residual index to a corresponding threshold; and
Generating an identification event if the accumulated structured residual index of all but one structured residual, referred to as the identified accumulated structured residual index, exceeds its corresponding threshold; and
Determining that the faulty sensors are the subset of sensors for which, by design, the structured residual transform corresponding to the identified accumulated structured residual index is insensitive to the presence of faults in said subset.
-
-
59. The method of claim 58 wherein:
The accumulating operation calculates a generalized likelihood ratio index that is designed to detect significant changes in the statistical mean of the structured residuals since the time of the fault detection event.
-
60. The method of claim 58 wherein:
The accumulating operation calculates the statistical mean of the structured residuals since the time of the fault detection event, resulting in an accumulated mean index.
-
61. The method of claim 58 wherein:
The accumulating operation calculates the statistical variance of the structured residuals since the time of the fault detection event, resulting in an accumulated variance index.
-
62. The method of claim 35 wherein:
The estimated fault size for each of the identified sensor faults is calculated by minimizing the effect of said identified sensor faults on the equation error values.
-
63. The method of claim 35 wherein the step classifying the identified sensor faults into a fixed set of fault types is achieved by:
-
Calculating a plurality of regression lines, one for each identified faulty sensor, each such regression line calculated using the measured sensor values for the corresponding faulty sensor from the time of identification, resulting in a plurality of sensor value regression lines; and
Calculating a plurality of regression lines, one for each identified faulty sensor, each such regression line calculated using estimated fault sizes for the corresponding faulty sensor from the time of identification, result in a plurality of fault size regression lines; and
Using slope, bias, and residual information from sensor value regression lines and fault size regression lines to classify the identified sensor faults into a fixed set of fault types.
-
-
64. The method of claim 63 wherein:
A faulty sensor is classified as type “
complete failure”
if the residual of the corresponding sensor value regression line is below a threshold, and if the slope of the corresponding sensor value regression line is statistically equal to zero.
-
65. The method of claim 63 wherein:
A faulty sensor is classified as type “
bias”
if the residual of the corresponding fault size regression line is below a threshold, and if the slope of the corresponding fault size regression line is statistically equal to zero, and if the bias of the corresponding fault size regression line is statistically unequal to zero.
-
66. The method of claim 63 wherein:
A faulty sensor is classified as type “
drift”
if the residual of the corresponding fault size regression line is below a threshold, and if the slope of the corresponding fault size regression line is statistically unequal to zero.
-
67. The method of claim 63 wherein:
A faulty sensor is classified as type “
precision loss”
if the residual of the corresponding fault size regression line is above a threshold, and if the slope and bias of the corresponding fault size regression line are both statistically equal to zero.
-
68. The method of claim 60 wherein:
A faulty sensor is classified as type “
precision loss”
if the identification event is triggered by an accumulated variance index exceeding its threshold.
Specification