Method and system of monitoring, sensor validation and predictive fault analysis

• US 20050210337A1
• Filed: 03/04/2004
• Published: 09/22/2005
• Est. Priority Date: 03/04/2004
• Status: Active Grant

• Current Assignee: Falconeer Technologies LLC (S&S Engineering & Construction, Inc.)
• Original Assignee: Falconeer Technologies LLC (S&S Engineering & Construction, Inc.)
• Inventors: Chester, Daniel L., Daniel, Stephen L., Fickelscherer, Richard J., Lenz, Douglas H.
• Granted Patent: US 7,451,003 B2
• Time in Patent Office: Days
• US Class Current: 714/47
• CPC Class Codes: G05B 23/0254 based on a quantitative model, e.g. mathematical relationships between inputs and outputs; functions: observer, Kalman filter, residual calculation, Neural Networks
• CPC Class Codes: G05B 23/0262 Confirmation of fault detection, e.g. extra checks to confirm that a failure has indeed occurred

Abstract

The present invention provides an improved method and system for real-time monitoring, validation, optimization and predictive fault analysis in a process control system. The invention monitors process operations by continuously analyzing sensor measurements and providing predictive alarms using models of normal process operation and statistical parameters corresponding to normal process data, and generating secondary residual process models. The invention allows for the creation of a fault analyzer directly from linearly independent models of normal process operation, and provides for automatic generation from such process models of linearly dependent process models. Fuzzy logic is used in various fault situations to compute certainty factors to identify faults and/or validate underlying assumptions. In one aspect, the invention includes a real-time sensor data communications bridge module; a state transition logic module; a sensor validation and predictive fault analysis module; and a statistical process control module; wherein each of the modules operates simultaneously.

Claims

1. A method of monitoring, validation and analysis for a process control system having normal process data and sensors, and represented by a plurality of primary process models and a plurality of process variables, comprising:

generating a plurality of primary residual process models derived from said primary process models, said normal process data, and one or more statistical parameters corresponding to said normal process data;

measuring real-time sensor data;

computing primary residual values of said primary residual process models corresponding to said real-time sensor data;

comparing said primary residual values to expected values;

computing a certainty factor for a possible fault using fuzzy logic.

2. The method as set forth in claim 1, further comprising:

measuring said normal process data.

3. The method as set forth in claim 1, further comprising:

calculating said statistical parameters.

4. The method as set forth in claim 1, further comprising:

computing a certainty factor corresponding to each of said primary residual values.

5. The method as set forth in claim 1 wherein said fuzzy logic is defined in a diagnostic rule that is regenerated each time said diagnostic rule is used.

6. The method as set forth in claim 1 wherein said fuzzy logic is defined in a diagnostic rule as follows:

FAULT-IS-PRESENT=SOME (evidence-for-fault) AND
ALL (neutral-evidence) AND
NOT (SOME (evidence-against-fault)).

7. The method as set forth in claim 6, further comprising:

determining a direction of deviation of said primary residual values from said expected values; and

determining evidence-for-fault and evidence-against-fault by comparing said direction of deviation to an expected direction of deviation consistent with a fault.

8. The method as set forth in claim 6, further comprising:

determining neutral-evidence by comparing the magnitude of a deviation of said primary residual values to zero.

9. The method as set forth in claim 1, further comprising:

generating one or more secondary residual process models, wherein each of said secondary residual process models is derived from two primary residual process models having at least one common process variable.

10. The method as set forth in claim 9, further comprising:

computing secondary residual values of said secondary residual process models corresponding to said real-time sensor data; and

comparing said secondary residual values to expected values.

11. The method as set forth in claim 10, further comprising:

determining a direction of deviation of said secondary residual values from said expected values;

comparing said direction of deviation to an expected direction of deviation consistent with a fault.

12. The method as set forth in claim 11 wherein said determination is made by calculating a first partial derivative of said secondary residual process model.

13. The method as set forth in claim 10, further comprising:

computing a certainty factor for a possible fault as a function of one or more of said secondary residual values using fuzzy logic.

14. The method as set forth in claim 13, further comprising:

computing a certainty factor corresponding to each of said secondary residual values.

15. The method as set forth in claim 13 wherein said fuzzy logic is defined in a diagnostic rule as follows:

FAULT-IS-PRESENT=SOME (evidence-for-fault) AND
ALL (neutral-evidence) AND
NOT (SOME (evidence-against-fault)).

16. The method as set forth in claim 15, further comprising:

determining neutral-evidence from one or more of said secondary residual process models which do not relate to relevant process variables.

17. The method as set forth in claim 9 wherein said secondary residual process models are functions of both measured process variables and unmeasured process variables.

18. The method as set forth in claim 9, further comprising:

regenerating said secondary residual process models.

19. The method as set forth in claim 1 wherein said statistical parameters comprise a mean and a standard deviation.

20. The method as set forth in claim 1 wherein said primary residual process models are functions of both a measured process variable and an unmeasured process variable.

21. The method as set forth in claim 1, further comprising:

regenerating said primary residual process models.

22. The method as set forth in claim 1, further comprising:

computing a first partial derivative and a second partial derivative of said primary residual process models.

23. The method as set forth in claim 1, further comprising:

reporting said certainty factor if it exceeds a predetermined threshhold.

24. The method as set forth in claim 1, further comprising:

displaying said certainty factor on a monitor.

25. The method as set forth in claim 1 wherein said certainty factor is between 0.0 and 1.0.

26. The method as set forth in claim 1, further comprising:

evaluating said real-time sensor data using statistical process control charting techniques to determine if a corresponding sensor is in control.

27. The method as set forth in claim 1, further comprising:

calculating the exponentially weighted moving average of said real-time sensor data to determine if a corresponding sensor is in control.

28. The method as set forth in claim 1, further comprising:

computing a compressed value.

29. The method as set forth in claim 1, further comprising:

computing a plurality of certainty factors for a corresponding plurality of possible faults.

30. A method of monitoring, validation and analysis for a process control system having normal process data and sensors, and represented by a plurality of primary residual process models and a plurality of process variables, comprising:

measuring real-time sensor data;

computing primary residual values of said primary residual process models corresponding to said real-time sensor data;

comparing said primary residual values to expected values;

computing a certainty factor for a possible fault using fuzzy logic.

31. The method as set forth in claim 30, further comprising:

computing a certainty factor corresponding to each of said primary residual values.

32. The method as set forth in claim 30 wherein said fuzzy logic is defined in a diagnostic rule that is regenerated each time said diagnostic rule is used.

33. The method as set forth in claim 30 wherein said fuzzy logic is defined in a diagnostic rule as follows:

FAULT-IS-PRESENT=SOME (evidence-for-fault) AND
ALL (neutral-evidence) AND
NOT (SOME (evidence-against-fault)).

34. The method as set forth in claim 33, further comprising:

determining a direction of deviation of said primary residual values from said expected values; and

determining evidence-for-fault and evidence-against-fault by comparing said direction of deviation to an expected direction of deviation consistent with a fault.

35. The method as set forth in claim 33, further comprising:

determining neutral-evidence by comparing the magnitude of a deviation of said primary residual values to zero.

36. The method as set forth in claim 30, further comprising:

generating one or more secondary residual process models, wherein each of said secondary residual process models is derived from two primary residual process models having at least one common process variable.

37. The method as set forth in claim 36, further comprising:

computing secondary residual values of said secondary residual process models corresponding to said real-time sensor data; and

comparing said secondary residual values to expected values.

38. The method as set forth in claim 37, further comprising:

computing a certainty factor for a possible fault as a function of one or more of said secondary residual values using fuzzy logic.

39. The method as set forth in claim 38, further comprising:

computing a certainty factor corresponding to each of said secondary residual values.

40. The method as set forth in claim 38 wherein said fuzzy logic is defined in a diagnostic rule as follows:

FAULT-IS-PRESENT=SOME (evidence-for-fault) AND
ALL (neutral-evidence) AND
NOT (SOME (evidence-against-fault)).

41. The method as set forth in claim 40, further comprising:

determining neutral-evidence from one or more of said secondary residual process models which do not relate to relevant process variables.

42. The method as set forth in claim 37, further comprising:

determining a direction of deviation of said secondary residual values from said expected values;

comparing said direction of deviation to an expected direction of deviation consistent with a fault.

43. The method as set forth in claim 42 wherein said determination is made by calculating a first partial derivative of said secondary residual process model.

44. The method as set forth in claim 36, further comprising:

regenerating said secondary residual process models.

45. The method as set forth in claim 30, further comprising:

regenerating said primary residual process models.

46. The method as set forth in claim 30, further comprising:

computing a first partial derivative and a second partial derivative of said primary residual process models.

47. The method as set forth in claim 30, further comprising:

evaluating said real-time sensor data using statistical process control charting techniques to determine if a corresponding sensor is in control.

48. The method as set forth in claim 30, further comprising:

computing a compressed value.

49. The method as set forth in claim 30, further comprising:

computing a plurality of certainty factors for a corresponding plurality of possible faults.

50. A method of monitoring, validation and analysis for a process control system having normal process data and sensors, and represented by a plurality of primary process models and a plurality of process variables, comprising:

generating a plurality of primary residual process models derived from said primary process models, said normal process data, and one or more statistical parameters corresponding to said normal process data;

generating one or more secondary residual process models, wherein each of said secondary residual process models is derived from two primary residual process models having at least one common variable; and

using one or more of said primary residual process models and one or more of said secondary residual process models to predict a possible fault.

51. The method as set forth in claim 50, further comprising:

measuring said normal process data.

52. The method as set forth in claim 50, further comprising:

calculating said statistical parameters.

53. The method as set forth in claim 50, further comprising:

measuring real-time sensor data.

54. The method as set forth in claim 53, further comprising:

computing primary residual values and secondary residual values of said primary residual process models and said secondary residual process models corresponding to said real-time sensor data;

comparing said primary residual values and said secondary residual values to expected values; and

computing a certainty factor for a possible fault using fuzzy logic.

55. The method as set forth in claim 54 wherein said fuzzy logic is defined in a diagnostic rule that is regenerated each time said diagnostic rule is used.

56. The method as set forth in claim 54, further comprising:

computing a certainty factor corresponding to each of said primary residual values.

57. The method as set forth in claim 54, further comprising:

computing a certainty factor corresponding to each of said secondary residual values.

58. The method as set forth in claim 50, further comprising:

regenerating said secondary residual process models.

59. The method as set forth in claim 50, further comprising:

regenerating said primary residual process models.

60. The method as set forth in claim 50, further comprising:

using one or more of said primary residual process models and one or more of said secondary residual process models to predict a plurality of possible faults.

61. A method of monitoring, validation and analysis for a process control system having normal process data and sensors, and represented by a plurality of primary process models and a plurality of process variables, comprising:

generating a plurality of primary residual process models derived from said primary process models, said normal process data, and one or more statistical parameters corresponding to said normal process data;

translating said primary residual process models into pseudo-code;

measuring real-time sensor data;

executing said pseudo-code to compute primary residual values corresponding to said real-time sensor data;

comparing said primary residual values to expected values;

computing a certainty factor for a possible fault by executing said pseudo-code.

62. The method as set forth in claim 61 wherein said pseudo-code uses fuzzy logic.

63. The method as set forth in claim 62 wherein said fuzzy logic is defined in a diagnostic rule that is regenerated each time said diagnostic rule is used.

64. The method as set forth in claim 62 wherein said fuzzy logic is defined in a diagnostic rule as follows:

FAULT-IS-PRESENT=SOME (evidence-for-fault) AND
ALL (neutral-evidence) AND
NOT (SOME (evidence-against-fault)).

65. The method as set forth in claim 64, further comprising:

determining a direction of deviation of said primary residual values from said expected values; and

determining evidence-for-fault and evidence-against-fault by comparing said direction of deviation to an expected direction of deviation consistent with a fault.

66. The method as set forth in claim 64, further comprising:

determining neutral-evidence by comparing the magnitude of a deviation of said primary residual values to zero.

67. The method as set forth in claim 61, further comprising:

generating pseudo-code for computing a certainty factor;

computing a certainty factor corresponding to each of said primary residual values.

68. The method as set forth in claim 61, further comprising:

generating one or more secondary residual process models, wherein each of said secondary residual process models is derived from two primary residual process models having at least one common process variable.

69. The method as set forth in claim 68, further comprising:

translating said secondary residual process models into pseudo-code;

measuring real-time sensor data;

executing said pseudo-code to compute secondary residual values corresponding to said real-time sensor data; and

comparing said secondary residual values to expected values.

70. The method as set forth in claim 69, further comprising:

generating pseudo-code for computing a certainty factor;

computing a certainty factor corresponding to each of said secondary residual values.

71. The method as set forth in claim 70, wherein said certainty factor is computed using fuzzy logic.

72. The method as set forth in claim 68, further comprising:

determining neutral-evidence from one or more of said secondary residual process models which do not relate to relevant process variables.

73. The method as set forth in claim 61, further comprising:

computing a plurality of certainty factors for a corresponding plurality of possible faults.

74. A computer-implemented monitoring, validation and analysis system for a process control system, comprising:

a real-time sensor data communication bridge module;

a sensor validation and predictive fault analysis module; and

a state transition logic module;

wherein each of said modules operates simultaneously.

75. The system as set forth in claim 74, further comprising:

a statistical process control module.

76. The system as set forth in claim 75, further comprising:

a master module which coordinates the operation of said real-time sensor data communication bridge, said sensor validation and predictive fault analysis module, said statistical process control module and said state transition logic module.

77. The system as set forth in claim 74, further comprising:

an operator display module.

78. The system as set forth in claim 74, further comprising:

a database module.

79. The system as set forth in claim 74, further comprising:

a master module which coordinates the operation of said real-time sensor data communication bridge, said sensor validation and predictive fault analysis module, and said state transition logic module.

80. The system as set forth in claim 74 wherein said real-time sensor data communication bridge module interfaces with a distributed control system, a programmable logic controller or a data historian associated with said process control system.

81. The system as set forth in claim 74 wherein said sensor validation and predictive fault analysis module comprises a model compiler.

82. The system as set forth in claim 74 wherein said sensor validation and predictive fault analysis module comprises a fault analyzer.

83. The system as set forth in claim 74 wherein said sensor validation and predictive fault analysis module computes a certainty factor for a possible fault using fuzzy logic.

84. A process control system, comprising:

a monitoring, validation and analysis system for a process control system having a real-time sensor data communication bridge module;

a sensor validation and predictive fault analysis module; and

a state transition logic module;

wherein each of said modules operates simultaneously.

85. The process control system as set forth in claim 84, further comprising a statistical process control module.

86. The system as set forth in claim 85, further comprising:

a master module which coordinates the operation of said real-time sensor data communication bridge module, said sensor validation and predictive fault analysis module, said statistical process control module and said state transition logic module.

87. The system as set forth in claim 84, further comprising an operator display module.

88. The system as set forth in claim 84, further comprising a database module.

89. The system as set forth in claim 84, further comprising:

a master module which coordinates the operation of said real-time sensor data communication bridge module, said sensor validation and predictive fault analysis module, and said state transition logic module.

90. A computer-readable medium having computer-executable instructions for performing a method for a process control system having normal process data and sensors, and represented by a plurality of primary process models and a plurality of process variables, comprising:

generating a plurality of primary residual process models derived from said primary process models, said normal process data, and one or more statistical parameters corresponding to said normal process data;

measuring real-time sensor data;

computing primary residual values of said primary residual process models corresponding to said real-time sensor data;

comparing said primary residual values to expected values;

computing a certainty factor for a possible fault using fuzzy logic.

91. The method as set forth in claim 90, further comprising:

measuring said normal process data.

92. The method as set forth in claim 90, further comprising:

calculating said statistical parameters.

93. The method as set forth in claim 90, further comprising:

computing a certainty factor corresponding to each of said primary residual values.

94. The method as set forth in claim 90 wherein said fuzzy logic is defined in a diagnostic rule that is regenerated each time said diagnostic rule is used.

95. The method as set forth in claim 90 wherein said fuzzy logic is defined in a diagnostic rule as follows:

FAULT-IS-PRESENT=SOME (evidence-for-fault) AND
ALL (neutral-evidence) AND
NOT (SOME (evidence-against-fault)).

96. The method as set forth in claim 95, further comprising:

determining a direction of deviation of said primary residual values from said expected values; and

determining evidence-for-fault and evidence-against-fault by comparing said direction of deviation to an expected direction of deviation consistent with a fault.

97. The method as set forth in claim 95, further comprising:

determining neutral-evidence by comparing the magnitude of a deviation of said primary residual values to zero.

98. The method as set forth in claim 90, further comprising:

generating one or more secondary residual process models, wherein each of said secondary residual process models is derived from two primary residual process models having at least one common process variable.

99. The method as set forth in claim 98, further comprising:

computing secondary residual values of said secondary residual process models corresponding to said real-time sensor data; and

comparing said secondary residual values to expected values.

100. The method as set forth in claim 99, further comprising:

determining a direction of deviation of said secondary residual values from said expected values;

comparing said direction of deviation to an expected direction of deviation consistent with a fault.

101. The method as set forth in claim 100 wherein said determination is made by calculating a first partial derivative of said secondary residual process model.

102. The method as set forth in claim 99, further comprising:

computing a certainty factor for a possible fault as a function of one or more of said secondary residual values using fuzzy logic.

103. The method as set forth in claim 102, further comprising:

computing a certainty factor corresponding to each of said secondary residual values.

104. The method as set forth in claim 102 wherein said fuzzy logic is defined in a diagnostic rule as follows:

FAULT-IS-PRESENT=SOME (evidence-for-fault) AND
ALL (neutral-evidence) AND
NOT (SOME (evidence-against-fault)).

105. The method as set forth in claim 104, further comprising:

determining neutral-evidence from one or more of said secondary residual process models which do not relate to relevant process variables.

106. The method as set forth in claim 98 wherein said secondary residual process models are functions of both measured process variables and unmeasured process variables.

107. The method as set forth in claim 98, further comprising:

regenerating said secondary residual process models.

108. The method as set forth in claim 90 wherein said statistical parameters comprise a mean and a standard deviation.

109. The method as set forth in claim 90 wherein said primary residual process models are functions of both a measured process variable and an unmeasured process variable.

110. The method as set forth in claim 90, further comprising:

regenerating said primary residual process models.

111. The method as set forth in claim 90, further comprising:

computing a first partial derivative and a second partial derivative of said primary residual process models.

112. The method as set forth in claim 90, further comprising:

reporting said certainty factor if it exceeds a predetermined threshhold.

113. The method as set forth in claim 90, further comprising:

displaying said certainty factor on a monitor.

114. The method as set forth in claim 90 wherein said certainty factor is between 0.0 and 1.0.

115. The method as set forth in claim 90, further comprising:

evaluating said real-time sensor data using statistical process control charting techniques to determine if a corresponding sensor is in control.

116. The method as set forth in claim 90, further comprising:

calculating the exponentially weighted moving average of said real-time sensor data to determine if a corresponding sensor is in control.

117. The method as set forth in claim 90, further comprising:

computing a compressed value.

118. The method as set forth in claim 90, further comprising:

computing a plurality of certainty factors for a corresponding plurality of possible faults.

119. A computer-readable medium having computer-executable instructions for performing a method for a process control system having normal process data and sensors, and represented by a plurality of primary residual process models and a plurality of process variables, comprising:

measuring real-time sensor data;

computing primary residual values of said primary residual process models corresponding to said real-time sensor data;

comparing said primary residual values to expected values;

computing a certainty factor for a possible fault using fuzzy logic.

120. The method as set forth in claim 119, further comprising:

computing a certainty factor corresponding to each of said primary residual values.

121. The method as set forth in claim 119 wherein said fuzzy logic is defined in a diagnostic rule that is regenerated each time said diagnostic rule is used.

122. The method as set forth in claim 119 wherein said fuzzy logic is defined in a diagnostic rule as follows:

FAULT-IS-PRESENT=SOME (evidence-for-fault) AND
ALL (neutral-evidence) AND
NOT (SOME (evidence-against-fault)).

123. The method as set forth in claim 122, further comprising:

determining a direction of deviation of said primary residual values from said expected values; and

determining evidence-for-fault and evidence-against-fault by comparing said direction of deviation to an expected direction of deviation consistent with a fault.

124. The method as set forth in claim 122, further comprising:

determining neutral-evidence by comparing the magnitude of a deviation of said primary residual values to zero.

125. The method as set forth in claim 119, further comprising:

generating one or more secondary residual process models, wherein each of said secondary residual process models is derived from two primary residual process models having at least one common process variable.

126. The method as set forth in claim 125, further comprising:

computing secondary residual values of said secondary residual process models corresponding to said real-time sensor data; and

comparing said secondary residual values to expected values.

127. The method as set forth in claim 126, further comprising:

computing a certainty factor for a possible fault as a function of one or more of said secondary residual values using fuzzy logic.

128. The method as set forth in claim 127, further comprising:

computing a certainty factor corresponding to each of said secondary residual values.

129. The method as set forth in claim 127 wherein said fuzzy logic is defined in a diagnostic rule as follows:

FAULT-IS-PRESENT=SOME (evidence-for-fault) AND
ALL (neutral-evidence) AND
NOT (SOME (evidence-against-fault)).

130. The method as set forth in claim 126, further comprising:

determining a direction of deviation of said secondary residual values from said expected values;

comparing said direction of deviation to an expected direction of deviation consistent with a fault.

131. The method as set forth in claim 130 wherein said determination is made by calculating a first partial derivative of said secondary residual process model.

132. The method as set forth in claim 129, further comprising:

determining neutral-evidence from one or more of said secondary residual process models which do not relate to relevant process variables.

133. The method as set forth in claim 125, further comprising:

regenerating said secondary residual process models.

134. The method as set forth in claim 119, further comprising:

regenerating said primary residual process models.

135. The method as set forth in claim 119, further comprising:

computing a first partial derivative and a second partial derivative of said primary residual process models.

136. The method as set forth in claim 119, further comprising:

evaluating said real-time sensor data using statistical process control charting techniques to determine if a corresponding sensor is in control.

137. The method as set forth in claim 119, further comprising:

computing a compressed value.

138. The method as set forth in claim 119, further comprising:

computing a plurality of certainty factors for a corresponding plurality of possible faults.

139. A method of monitoring, validation and analysis for a process control system having normal process data and sensors, and represented by a plurality of primary process models and a plurality of process variables, comprising:

generating a plurality of primary residual process models derived from said primary process models, said normal process data, and one or more statistical parameters corresponding to said normal process data;

generating one or more secondary residual process models, wherein each of said secondary residual process models is derived from two primary residual process models having at least one common variable; and

using one or more of said primary residual process models and one or more of said secondary residual process models to predict a possible fault.

140. The method as set forth in claim 139, further comprising:

measuring said normal process data.

141. The method as set forth in claim 139, further comprising:

calculating said statistical parameters.

142. The method as set forth in claim 139, further comprising:

measuring real-time sensor data.

143. The method as set forth in claim 142, further comprising:

computing primary residual values and secondary residual values of said primary residual process models and said secondary residual process models corresponding to said real-time sensor data;

comparing said primary residual values and said secondary residual values to expected values; and

computing a certainty factor for a possible fault using fuzzy logic.

144. The method as set forth in claim 143 wherein said fuzzy logic is defined in a diagnostic rule that is regenerated each time said diagnostic rule is used.

145. The method as set forth in claim 143, further comprising:

computing a certainty factor corresponding to each of said primary residual values.

146. The method as set forth in claim 143, further comprising:

computing a certainty factor corresponding to each of said secondary residual values.

147. The method as set forth in claim 139, further comprising:

regenerating said secondary residual process models.

148. The method as set forth in claim 139, further comprising:

regenerating said primary residual process models.

149. The method as set forth in claim 139, further comprising:

using one or more of said primary residual process models and one or more of said secondary residual process models to predict a plurality of possible faults.