Application of abnormal event detection technology to olefins recovery trains

• US 7,424,395 B2
• Filed: 08/26/2005
• Issued: 09/09/2008
• Est. Priority Date: 09/10/2004
• Status: Active Grant

• Current Assignee: ExxonMobil Research & Engineering Company (Exxon Mobil Corporation)
• Original Assignee: ExxonMobil Research & Engineering Company (Exxon Mobil Corporation)
• Inventors: Wang, Robert K., Woo, Stephen S., Emigholz, Kenneth F., McLain, Richard B.
• Primary Examiner(s): Barlow, Jr.; John E
• Assistant Examiner(s): Washburn; Douglas N
• Application Number:
  US11/212,434
• Publication Number:
  US 20060074599A1
• Time in Patent Office: 1,110 Days
• Field of Search: 700/269, 702/183, 702/182
• US Class Current: 702/182
• CPC Class Codes: C10G 11/187 Controlling or regulating
• CPC Class Codes: G05B 2219/31477 Display correlated data so as to represent the degree of correlation
• CPC Class Codes: G05B 2219/31478 Display all processes together or select only one
• 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

Abstract

The present invention is a method for detecting an abnormal event for process units of an ethylene processing system. The method compares the operation of the process units to a model developed by principal components analysis of normal operation for these units. If the difference between the operation of a process unit and the normal operation indicates an abnormal condition, then the cause of the abnormal condition is determined and corrected.

Claims

1. A method for abnormal event detection (AED) for current operation for some process units of an ethylene processing system comprising:

(a) comparing online measurements from the process units to a set of models including principal component analysis models and/or engineering models for normal operation of the corresponding process units, wherein said ethylene processing system has been divided into equipment groups and process operating modes wherein there is minimal interaction between said equipment groups and principal component analysis models correspond to equipment groups and process operating modes,

(b) determining if the current operation differs from expected normal operations so as to indicate the presence of an abnormal condition in a process unit, and

(c) determining the underlying cause of an abnormal condition in the ethylene processing system.

2. The method of claim 1 wherein said set of models corresponds to equipment groups and process operating modes, one model for each group which may include one or more operating modes.

3. The method of claim 1 wherein said set of models correspond to equipment groups and process operating modes, one model for each group and each mode.

4. The method of claim 1 wherein said set of models of normal operations further includes engineering models.

5. The method of claim 1 wherein said ethylene processing system is divided into operational areas with a principal component analysis model for each area.

6. The method of claim 1 further comprising the steps of

(a) dividing the ethylene process into equipment groups with one or more operating modes,

(b) determining input variables and their operating range for said equipment groups and/or operating modes,

(c) determining historical training and testing data for said input variables,

(d) creating a set of principle component analysis models and/or engineering models representing normal operations, where each model corresponds to one of said equipment groups including one or more operating modes,

(e) dividing the ethylene process into operational areas where a set of one or more models corresponds an operational area, and

(f) comparing online measurements from the ethylene process units to a set of models for normal operation of the corresponding operational area.

7. The method of claim 1 wherein said equipment groups have minimal interactions among each other.

8. The method of claim 1 wherein said equipment groups include feedforward information from upstream equipment groups and feedback information from downstream equipment groups.

9. The method of claim 1 wherein inputs to the models are measurements of the performance and/or physical state of the ethylene process, and/or control signals sent to field actuators from the control system, and/or external factors affecting the ethylene process.

10. The method of claim 1 wherein the inputs to the models determine the physical state of the ethylene process.

11. The method of claim 1 wherein said indication of an abnormal event further comprises of suppressing model calculations to eliminate false positives on special cause operations.

12. The method of claim 1 wherein unwanted abnormal event indications are suppressed using override suppressors.

13. The method of claim 2 wherein said equipment groups include all major material and energy interactions in the same group.

14. The method of claim 2 wherein each said model further comprises suppressing model calculations to eliminate false positives on special cause operations.

15. The method of claim 2 further comprising the steps of(a)-determining said each model begins with a rough model based on questionable data, (b) using said rough model to gather high quality training data, and improve the model, and (e) repeating step (b) to further improve the model.

16. The method of claim 2 wherein said model includes transformed variables.

17. The method of claim 2 wherein principal components include process variables provided by online measurements.

18. The method of claim 4 wherein said engineering model further comprises determining consistency between some control valves and flow meters.

19. The method of claim 4 wherein the principle components for the principle component analysis models are selected based on named degrees of freedom in the equipment group.

20. The method of claim 4 wherein the number of principal components is chosen such that coefficients of the non-chosen principal components become about equal in size.

21. The method of claim 5 wherein there are eleven operational areas.

22. The method of claim 5 wherein said principal components analysis models include process variable values measured by sensors.

23. The method of claim 5 wherein said principal components analysis models for different process units include some process variable values measured by the same sensor.

24. The method of claim 5 wherein a number of principal components is selected by a magnitude of total process variation represented by successive components.

25. The method of claim 6 wherein operating modes are specific time periods wherein the ethylene process behavior is significantly different.

26. The method of claim 6 wherein historical data (raining sets arc segregated according to operating modes.

27. The method of claim 6 wherein said models includes transformed variables.

28. The method of claim 6 further comprising the steps

(a) determining said model begins with a rough model based on questionable data,

(b) using said rough model to gather high quality training data, and improve the model, and

(c) repeating step (b) to further improve the model.

29. The method of claim 6 wherein a set of models of normal operations includes one or more Principle Component Analysis Models and/or one or more engineering models.

30. The method of claim 6 wherein the number of principle components is selected by the magnitude of total process variation represented by successive principle components.

31. The method of claim 6 wherein said principal component analysis models for different equipment groups include as inputs some process variable values measured by the same sensor.

32. The method of claim 10 wherein said inputs to the models include temperatures, and/or pressures, and/or flows, and/or analyzer readings, and/or vibrations, and/or electrical current, and/or signals to field actuators.

33. The method of claim 12 wherein override suppressors act automatically whenever a defined condition exists.

34. The method of claim 12 wherein the operator initiates the action of an override suppressor.

35. The method of claim 13 wherein said equipment groups include quick recycles in the same group.

36. The method of claim 15 wherein said training data includes historical data for the model of the processing unit.

37. The method of claim 16 wherein said transformed variables include reflux to feed ratio in distillation columns, log of composition in high purity distillation, pressure compensated temperature measurement, sidestream yield, flow to valve position, and reaction rate to exponential temperature change.

38. The method of claim 16 wherein some pairs of measurements for two variables are brought into time synchronization by one of the variables using a dynamic transfer function.

39. The method of claim 16 wherein variables of process measurements that are affected by operating point changes in process operations are converted to deviation variables.

40. The method of claim 16 wherein said model inputs are corrected for noise.

41. The method of claim 16 wherein the measurements of a variable are scaled.

42. The method of claim 16 wherein measurements are transformed to bring tern into time synchronization, with one of the other variables using a dynamic transfer function.

43. The method of claim 17 wherein principal components include some measurement pairs that are time synchronized to one of the variables using a dynamic filter.

44. The method of claim 17 wherein process variables affected by operating point changes in the process operations are convened to deviation variables.

45. The method of claim 21 wherein the eleven operational areas of the ethylene processing system include;

charge gas compressor;

drier, cold box, and methane booster;

ethylene and propylene refrigeration;

demethanizer;

deethanizer;

acetylene converter;

C2 splitter;

methanator and H2 drier;

depropanizer;

MAPD hydrofiner; and

propylene rerun and fractionator.

46. The method of claim 27 wherein said transformed variables include reflux to feed ratio in distillation columns, log of composition in high purity distillation, pressure compensated temperature measurement, sidestream yield, flow to valve position, and reaction rate to exponential temperature change.

47. The method of claim 28 wherein said training data includes historical data for the model of the processing unit.

48. The method of claim 36 wherein the number of principal components is chosen such that coefficients of each principal component become about equal m size.

49. The method of claim 36 wherein the historical data periods selected for training and testing data sets are selected to increase cross correlation among inputs and/or increase the signal to noise ratio of the inputs in the data.

50. The method of claim 39 wherein a rolling average of inputs from recent operations is subtracted from the current value of the measurements.

51. The method of claim 40 wherein said model inputs are corrected by filtering or eliminating noisy measurements of variables.

52. The method of claim 41 wherein the measurements are scaled to the expected normal range of that variable.

53. The method of claim 49 wherein data from time periods of steady operations are eliminated from the training data.

54. The method of claim 49 wherein data from severe abnormal operations are eliminated from the training data.

55. The method of claim 49 wherein potential model inputs are deleted from the inputs used in the model.

56. A system for abnormal event detection (AED) for some process units of an ethylene processing system comprising:

(a) on-line data acquisition of measurements and control signals,

(b) a set of models including principal component models and/or engineering models included in the set describing operations of the process units of said ethylene processing system, wherein said ethylene process system has been divided into equipment groups, wherein said set of models includes at least a principal component model corresponding to an equipment group,

(c) a set of displays which indicate if the current operation differs from expected normal operations so as to indicate the presence of an abnormal condition in the process unit, and

(d) a set of displays which indicates the underlying cause of an abnormal condition in the ethylene processing system.

57. The system of claim 56 wherein said model for each process unit is either a principal components model or an engineering model.

58. The system of claim 56 wherein inputs to the system are measurements of the performance and/or physical state of the ethylene process, and/or control signals sent to process field devices from the control system, and/or external factors affecting the ethylene process.

59. The system of claim 56 wherein said models further comprises suppression calculations to eliminate unwanted notifications of known conditions.

60. The system of claim 56 wherein an ethylene processing system is partitioned into operational sections and the models are grouped into sets according to their associated operational section.

61. The system of claim 56 wherein the operator interface includes a continuous indication of the likelihood of an abnormality within each operational section of the ethylene processing system.

62. The system of claim 56 wherein the operator interface allows the operator to navigate to a display associated with the highest likelihood of an abnormal condition with a single mouse click.

63. The system of claim 57 wherein an ethylene processing system is partitioned into operational sections with a principal components model for each section.

64. The system of claim 57 wherein said principal components include process variables provided by online measurements.

65. The system of claim 57 wherein said model further comprises calculations to eliminate operator induced notifications and false positives.

66. The system of claim 57 wherein said model includes transformed variables.

67. The system of claim 58 wherein said inputs to the system include temperatures, and/or pressures, and/or flows, and/or analyzer readings, and/or vibrations, and/or electrical current, and/or signals to field actuators.

68. The system of claim 58 wherein said inputs to the system become transformed inputs to the models.

69. A system of claim 59 wherein override suppressors act automatically whenever a defined condition exists.

70. A system of claim 59 wherein the operator initiates the action of an override suppressor.

71. The system of claim 61 wherein the likelihood of an abnormality within an operational section is indicated to the operator with a trend of the highest likelihood of an abnormality.

72. The system of claim 61 wherein the operator is alerted whenever the likelihood of an abnormality exceeds a defined threshold.

73. The system of claim 62 wherein said displays include Pareto charts.

74. The system of claim 62 wherein, for a set of models, the Pareto charts list models in the order of magnitude of the likelihood of an abnormal condition as indicated by each model.

75. The system of claim 62 wherein said displays include X-Y plots.

76. The system of claim 63 wherein there are eleven operational sections.

77. The system of claim 63 wherein fuzzy Petri nets are used to convert outputs from models to the likelihood of an abnormal process operation and/or equipment problem.

78. The system of claim 66 wherein said transformed variables include reflux to total product flow in distillation columns, log of composition and overhead pressure in distillation columns, pressure compensated temperature measurements, flow to valve position and bed differential temperature and pressure.

79. The system of claim 66 wherein some measurement pairs are time synchronized to one of the variables using a dynamic filter.

80. The system of claim 66 wherein process measurement variables affected by operating point changes in the process operations are converted to deviation variables.

81. The system of claim 66 wherein the measurements and transforms of a variable are scaled prior to model identification.

82. The system of claim 66 wherein a number of principal components is selected by a magnitude of total process variation represented by successive components.

83. The system of claim 66 wherein noise is removed from the system inputs and/or the transformed variables.

84. The system of claim 66 wherein the measurements of a variable are scaled prior to model calculation.

85. The system of claim 73 wherein, for principle component analysis models, the Pareto charts show in order of magnitude the specific contribution of each input to the estimate of the likelihood of an abnormal condition.

86. The system of claim 75 wherein, two dimensional models are shown as X-Y plots.

87. The system of claim 75 wherein, the X-Y plot shows the estimate of the normal operating relationship as well as the confidence region around that estimate.

88. The system of claim 75 wherein the operator can recalibrate the X-Y plot on demand.

89. The system of claim 75 wherein time adjacent data sample points are connected.

90. The system of claim 75 wherein the normal process operating range of the X axis variable is indicated.

91. The system of claim 77 wherein the likelihood of abnormality indicated by models within a set are combined using fuzzy Petri nets to estimate the likelihood of an abnormality within an operational section.

92. The system of claim 80 wherein the measurements are scaled by the expected normal range of that transformed and/or deviation variable.

93. The system of claim 83 wherein noise is removed using digital filters.

94. The system of claim 83 wherein noise is removed using logic.

95. The system of claim 85 wherein the operator can navigate to time plots of inputs whose order is determined by each input'"'"'s position on the Pareto chart.