Dynamic prediction of risk levels for manufacturing operations through leading risk indicators: alarm-based intelligence and insights

• US 9,495,863 B2
• Filed: 10/10/2014
• Issued: 11/15/2016
• Est. Priority Date: 09/12/2011
• Status: Active Grant

• Current Assignee: Near-Miss Management, LLC
• Original Assignee: Near-Miss Management, LLC
• Inventors: Pariyani, Ankur, Oktem, Ulku G.
• Primary Examiner(s): HOFSASS, JEFFERY A
• Application Number:
  US14/511,729
• Publication Number:
  US 20150061860A1
• Time in Patent Office: 767 Days
• Field of Search: 340/540, 340/999, 702/179, 702/182, 702/183
• US Class Current: 1/1
• CPC Class Codes: G05B 23/024 Quantitative history assessment, e.g. mathematical relationships between available data; Functions therefor; Principal component analysis [PCA]; Partial least square [PLS]; Statistical classifiers, e.g. Bayesian networks, linear regression or correlation analysis; Neural networks
• CPC Class Codes: G05B 23/0278 Qualitative, e.g. if-then rules; Fuzzy logic; Lookup tables; Symptomatic search; FMEA
• CPC Class Codes: G06F 16/122 using management policies backup systems G06F11/1446; file migration policies for HSM systems G06F16/185
• CPC Class Codes: G08B 29/02 Monitoring continuously signalling or alarm systems

Abstract

Provided are methodologies to properly assess and manage operational risks at operations sites, e.g., a manufacturing, production or processing facility, such as a refinery, chemical plant, fluid-catalytic-cracking units, or nuclear energy plant, or a biological or waste management facility, airport or even financial institutions, or at any facility in which operations are often accompanied by risk associated with many high-probability, low-consequence events, often resulting in near-misses. In some operations, processes are monitored by alarms, but the invention operates on either process data or alarm data. The methods are based upon measurement of one or more variables, and/or utilization and management of the concept of “hidden process near-miss(es)” to identify a change or escalation, if any, in probability of occurrence of an adverse incident. The methodologies combine a plurality of subsets (also useful independently) of dynamically calculated leading risk indicators for dynamic risk management.

Claims

1. A method for managing operational risks at an operations site comprising:

collecting measured process data for an alarm-monitored process variable monitored by an alarm in an operation occurring within the operations site or for measured alarm data associated with the alarm-monitored process variable;

determining ranges for the alarm-monitored process variable or frequency values for the measured alarm data during a measurement period of the operation;

identifying a change or escalation in a probability of an occurrence of at least one adverse incident recorded in the measured process data during the measurement period by;

comparing the ranges of the alarm-monitored process variable or the frequency values for the measured alarm data against normal operating conditions, whereby the normal operating conditions are derived from long term measured process data comprising at least one of an alarm duration, an alarm frequency, a number of significant alarms, or an alarm flood; and

determining criticality level of the at least one adverse incident by assessing deviation between the measured process data relative to the normal operation conditions;

identifying at least one hidden process near miss based on the criticality level of the at least one adverse incident; and

initiating corrective action to reduce or avert the at least one adverse incident or a catastrophic failure of the operation.

2. The method of claim 1, wherein the measurement period may be one minute, 3 minutes, 10 minutes, hourly, daily, weekly, biweekly, monthly, quarterly, or yearly.

3. The method of claim 1, wherein the alarm indicates an activation or deactivation of an event in the operation when a value of the alarm-monitored process variable crosses one or more thresholds that can be statically or dynamically adjusted.

4. The method of claim 1, wherein the change or escalation in the probability of the occurrence of the at least one adverse incident comprises one or more alarm floods identified by X₁ alarm activations occurring in less than or equal to Y₁ time or ending when X₂ alarms occur in more than Y₂ time, wherein starting points of Y₁ and Y₂ are incremented within the measurement period.

5. The method of claim 4, wherein X₁ and X₂ are integers and wherein X₁ is greater than X₂.

6. The method of claim 1, wherein the change or escalation in the probability of the occurrence of the adverse incident comprises one or more alarm attacks identified by X₅ fresh standing alarms or ending with X₆ fresh standing alarms, wherein fresh standing alarms comprise alarms active for less than a predetermined time period.

7. The method of claim 6, wherein X₅ and X₆ are integers, wherein X₅ is greater than X₆, and wherein the criticality level of the at least one adverse incident is derived from the long term measured process data for one or more of the alarm duration, the alarm frequency, the number of significant alarms, or the intensity of the alarm flood.

8. The method of claim 6, wherein an escalation rate of an alarm attack is determined by identifying each of the X₅ fresh standing alarms that triggered the alarm attack by calculating time distance from the starting point of the alarm attack for each of the X₅ fresh standing alarms, and by calculating a median or mean value of all X₅ time distance values to obtain the escalation rate.

9. The method of claim 1, wherein fractional intensity of significant alarms is determined from an amount of the significant alarms related to the alarm attack, wherein the significant alarms comprise one or more of high priority alarms or medium priority alarms.

10. The method of claim 1, wherein the long term measured process data is associated with at least one alarm-monitored process variable for a time period preceding the adverse incident.

11. The method of claim 1, wherein the criticality level of the at least one adverse incident is derived from the long term measured process data for one or more of the alarm duration, the alarm frequency, the number of significant alarms, or the intensity of the alarm flood.

12. The method of claim 1, wherein determining the criticality level of the at least one adverse incident further comprises:

grouping the alarm floods based on priority of the measured alarm data;

calculating weighted alarm frequency for each of the alarm flood based on priority of the measured alarm data;

calculating the significant alarms for each alarm flood, wherein significant alarms comprising at least one of high priority alarms or medium priority alarms; and

calculating a number of the fresh standing alarms, wherein the fresh standing alarms are alarms that are active for less than a predetermined time period.

13. The method of claim 1, wherein the criticality level of the at least one adverse incident is determined by ranking the at least one adverse incident against an abnormal event matrix that provides stages and criticality levels of adverse events, and then prioritizing the abnormal incidents for the corrective action.

14. The method of claim 1, wherein the criticality level of the at least one adverse event is determined by predetermined notable alarms that are based on at least one or more of total recovery time, average recovery time, frequency of alarms, average inter-arrival time, average neighborhood time, or average kick-off time.

15. The method of claim 1, wherein the criticality level of the at least one adverse event is determined by rare alarms that are defined by a maximum number of alarms N₁ and the minimum time period T₁ for the measurement period.

16. The method of claim 1 further comprising:

determining the criticality levels of the at least one adverse event from the measure process data by conducting time segment analysis of at least one or more of kickoff time analysis, acceleration time analysis, deceleration time analysis, or neighborhood time analysis,

wherein the kickoff time analysis comprises evaluating time elapsed from a last inactivation of an alarm to a next activation,

wherein acceleration time analysis comprises evaluating time difference between alarm activations of two consecutive alarm levels that belong to a same variable,

wherein deceleration time analysis comprises evaluating time difference between inactivations or deactivations of two consecutive alarm levels that belong to a same variable, and

wherein the neighborhood time analysis comprises evaluating total time that a variable is in close proximity to a referenced value.

17. The method of claim 1 wherein identifying the at least one hidden process near miss comprises sending alert signals in real time about approaching or instantly occurring adverse incidents.

18. The method of claim 1 wherein identifying the at least one hidden process near miss comprises providing on-demand alerts of adverse incidents.

19. A method for identifying risk and impact of a hidden process near-miss as a measure of a deviation of a process condition from normal conditions of an operations site, the method comprising:

collecting measured alarm data associated with at least one alarm-monitored process variable of the process conditions of the operations site during a measurement period;

determining alarm frequency values or value ranges for the at least one alarm monitored process variable for the measurement period;

collecting long-term alarm data for a time period preceding the measured alarm data;

determining normal alarm frequency values or value ranges for the at least one alarm monitored process variable, or combination thereof, based on the long term data;

comparing the alarm frequency data of the at least one alarm-monitored process variable for the measurement period with the collected long-term alarm frequency data to determine deviation between the collected alarm frequency data relative to normal frequency values, and

initiating corrective action to reduce or avert the at least one adverse incident or a catastrophic failure of the operation based on results of comparing the alarm frequency data to the collected long-term alarm frequency data.

20. A system for identifying risk and impact of a hidden process near-miss as a measure of deviation of process condition of an operations site from normal conditions, the system comprising:

an alarm fitness analyzer, wherein the alarm fitness analyzer is configured to analyze process data and alarm data of the operations site in order to identify at least one hidden process near miss based on the criticality level of at least one adverse incident, whereby the alarm fitness analyzer;

collects measured process data for an alarm-monitored process variable monitored by an alarm in the operation or a measured alarm data associated with at least one alarm-monitored process variable;

determines ranges for the alarm-monitored process variable or frequency values for the measured alarm data during a measurement period of the operation;

identifies a change or escalation in a probability of an occurrence of at least one adverse incident recorded in the measured process data during the measurement period by;

comparing the ranges of the alarm-monitored process variable or the frequency values for the measured alarm data against normal operating conditions, whereby the normal operating conditions are derived from long term measured process data comprising at least one of alarm duration, alarm frequency, number of significant alarms, or intensity of alarm flood; and

determining criticality level of the at least one adverse incident by assessing deviation between the measured process data relative to normal operation conditions; and

initiates corrective action needed to reduce or avert the at least one adverse incident or a catastrophic failure of the operation.