Automatic fault detection and diagnosis in complex physical systems

• US 20150235141A1
• Filed: 02/18/2015
• Published: 08/20/2015
• Est. Priority Date: 02/18/2014
• Status: Active Grant

• Current Assignee: Cimetrics Inc.
• Original Assignee: Cimetrics Inc.
• Inventors: Rensing, Paul, Corbett, Keith, Frankston, Charles
• Granted Patent: US 10,372,567 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G05B 23/0224 Process history based detection method, e.g. whereby history implies the availability of large amounts of data
• CPC Class Codes: G05B 23/0278 Qualitative, e.g. if-then rules; Fuzzy logic; Lookup tables; Symptomatic search; FMEA
• CPC Class Codes: G06F 11/22 Detection or location of defective computer hardware by testing during standby operation or during idle time, e.g. start-up testing
• CPC Class Codes: G06N 5/025 Extracting rules from data

Abstract

A fault detection and diagnosis (FDD) system designed to automatically and efficiently determine the true fault underlying a set of symptoms, presenting the operator with a clear cause and effect diagram for all faults and symptoms. An Event Cluster is used to group related symptoms and causes, potentially across multiple pieces of equipment, and establish a cause and effect relationship chain.

Claims

1. A method for automated detection of faults in a complex engineered system, comprising the steps of:

a) collecting data from a plurality of sensors associated with the complex engineered system;

b) performing a plurality of time-series calculations on the data to identify a plurality of symptom flags;

c) converting the plurality of symptom flags to a plurality of events;

d) grouping the plurality of events into a plurality of event ranges;

e) using cluster rules to group the plurality of event ranges into a plurality of event clusters; and

f) assigning a likelihood score to each of the plurality of event clusters.

2. The method of claim 1, further comprising the step of performing an adjustment to the complex engineered system, based on the likelihood scores of the plurality of event clusters.

3. The method of claim 1, in which step (a) comprises the steps of:

polling the plurality of sensors at periodic intervals to retrieve a plurality of sensor data points; and

storing the plurality of sensor data points in a database, each sensor data point being marked with a time stamp.

4. The method of claim 1, in which step (c) further comprises the step of de-bouncing comprising combining pairs of symptom flags which occur within a selected time period.

5. The method of claim 1, in which step (c) further comprises assigning an event type to each of the plurality of events.

6. The method of claim 5, in which step (d) comprises grouping events which occur related to a single piece of equipment and are of the same event type into an event range.

7. The method of claim 1, in which the event ranges of step (d) are created with a start time set to a start time of the time stamp of a first data point of the first event in the event range.

8. The method of claim 7, further comprising the step of setting an end time when a fault has been fixed or closed.

9. The method of claim 1, in which step (e) comprises the steps of:

i) looking for matches between an applicable-to pattern for a cluster rule and the plurality of event ranges;

ii) if a match is found in step (i), check for matches on all event types that do not have an allowCreate flag set;

iii) if the event types all match and a time range function is present for the cluster rule, then run the time range function on all matched event ranges, and if a result of the time range function is equal to or greater than a specified threshold, and the allowCreate flag is on, create a diagnosis event range;

iv) for all the rules in an event cluster, link all event ranges that have an isCause flag set to all event ranges that do not have the isCause flag set; and

v) repeat the step from step (e)(i) until all cluster rules have been examined.

10. The method of claim 1, in which step (f) comprises the steps of:

i) assigning a score to each event range, representing a likelihood that a symptom underlying the event range represents a fault;

ii) assigning a weight to each cluster rule, representing a probability that the cluster rule was valid; and

iii) computing a total likelihood score for each of the cluster rules.

11. The method of claim 10, in which the likelihood in step (i) is a cross product of values derived from a plurality of intrinsic weighting factors assigned to each symptom calculation, a total amount of time the symptom has been present, a weighting factor based on a weighted average of lengths of each event in the event range.

12. The method of claim 11 in which each of the values is a fraction between 0.0 and 1.0 inclusive.

13. The method of claim 10, in which the weights assigned in step (ii) are between 0.0 and 1.0 inclusive.

14. The method of claim 10, in which the total likelihood score is computed by summing all even cluster links multiplied by the score of each event cluster.

15. A fault detection system for a complex engineered system having a plurality of components, comprising:

a) a plurality of sensors, each sensor being associated with at least one of the plurality of components;

b) a system controller coupled to each of the plurality of sensors, the system controller comprising one or more processors, one or more computer-readable memories, and one or more computer-readable, tangible storage devices;

c) program instructions, stored on at least one of the one or more storage devices for execution by at least one of the one or more processors via at least one of the one or more memories, to collect data from the plurality of sensors associated with the components of the complex engineered system;

d) program instructions, stored on at least one of the one or more storage devices for execution by at least one of the one or more processors via at least one of the one or more memories, to perform a plurality of time-series calculations on the data to identify a plurality of symptom flags;

e) program instructions, stored on at least one of the one or more storage devices for execution by at least one of the one or more processors via at least one of the one or more memories, to program instructions, stored on at least one of the one or more storage devices for execution by at least one of the one or more processors via at least one of the one or more memories, to convert the plurality of symptom flags to a plurality of events;

f) program instructions, stored on at least one of the one or more storage devices for execution by at least one of the one or more processors via at least one of the one or more memories, to group the plurality of events into a plurality of event ranges;

g) program instructions, stored on at least one of the one or more storage devices for execution by at least one of the one or more processors via at least one of the one or more memories, to use cluster rules to group the plurality of event ranges into a plurality of event clusters; and

h) program instructions, stored on at least one of the one or more storage devices for execution by at least one of the one or more processors via at least one of the one or more memories, to assign a likelihood score to each of the plurality of event clusters.

16. The system of claim 15, in which the complex engineered system further comprises a building control system, and the fault detection system further comprises program instructions, stored on at least one of the one or more storage devices for execution by at least one of the one or more processors via at least one of the one or more memories, to send commands to the building control system to adjust parameters in the complex engineered system.