Faulty variable identification technique for data-driven fault detection within a process plant
AssigneeEmerson Process Management Power & Water Solutions, Inc.
InventorXu Cheng
Life of this patent
Estimating the term…
Abstract
A real-time control system includes a faulty variable identification technique to implement a data-driven fault detection function that provides an operator with information that enables a higher level of situational awareness of the current and likely future operating conditions of the process plant. The faulty variable identification technique enables an operator to recognize when a process plant component is behaving abnormally to potentially take action, in a current time step, to alleviate the underlying cause of the problem, thus reducing the likelihood of or preventing a stall of the process control system or a failure of the process plant component.
Claims20
A system for identifying faulty variables in on-line data indicative of an on-line operation of a process within a process plant, comprising:
- a process controller coupled to the process for controlling the process;
- and
- a fault detection system for identifying faulty variables in on-line data received from the process controller, wherein the fault detection system is configured to:
- detect an abnormal operating condition of the process by applying a multivariate state estimation technique (MSET) using a system matrix to the on-line data;
- build a respective new matrix for each unique pair of process variables represented by data included in the system matrix, the data representing measurements of the process variables at distinct moments in time;
- train a respective miniature system matrix for each unique pair of process variables using the respective new matrix;
- define a respective fault index for each process variable included in the system matrix;
- determine estimated data for each unique pair of process variables by applying the MSET using the respective miniature system matrix to a respective portion of the on-line data;
- increment the respective fault index of both process variables included in a unique pair of process variables when an estimation error corresponding to the estimated data for one of the process variables included in the unique pair exceeds a corresponding estimation error threshold;
- and
- identify a faulty variable by determining a respective process variable with a largest fault index.
A method of identifying faulty variables in on-line data indicative of an on-line operation of a process within a process plant, comprising:
- receiving, from a process controller, the on-line data;
- detecting an abnormal operating condition of the process by applying a multivariate state estimation technique (MSET) using a system matrix to the on-line data;
- building a respective new matrix for each unique pair of process variables represented by data included in the system matrix, the data representing measurements of the process variables at distinct moments in time;
- training a respective miniature system matrix for each unique pair of process variables using the respective new matrix;
- defining a respective fault index for each process variable included in the system matrix;
- determining estimated data for each unique pair of process variables by applying the MSET using the respective miniature system matrix to a respective portion of the on-line data;
- incrementing the respective fault index of both process variables included in a unique pair of process variables when an estimation error corresponding to the estimated data for one of the process variables included in the unique pair exceeds a corresponding estimation error threshold;
- and
- identifying a faulty variable by determining a respective process variable with a largest fault index.
A method of locating faulty variables in on-line data indicative of an on-line operation of a process within a process plant, comprising:
- receiving, from a process controller, the on-line data representing measurements of process variables at distinct moments in time;
- detecting an abnormal operating condition of the process by applying a multivariate state estimation technique (MSET) using a current system matrix to the on-line data;
- (a) dividing the current system matrix into a first new matrix and a second new matrix;
- (b) training a first system matrix using the first new matrix;
- (c) training a second system matrix using the second new matrix;
- (d) determining a first estimated data set by applying the MSET using the first system matrix to a respective portion of the on-line data;
- (e) determining whether abnormal operating conditions exist within the respective portion of the on-line data by comparing the first estimated data set to the respective portion of the on-line data;
- (f) responsive to determining that abnormal operating conditions exist within the respective portion of the on-line data:
- determining whether a size of the first system matrix satisfies a size threshold, and
- responsive to determining that the size of the first system matrix does not satisfy the size threshold, designating the first system matrix as the current system matrix;
- (g) responsive to determining that normal operating conditions exist within the respective portion of the on-line data:
- determining whether a size of the second system matrix satisfies the size threshold, and
- responsive to determining that the size of the second system matrix does not satisfy the size threshold, designating the second system matrix as the current system matrix;
- and
- (h) iteratively performing steps (a)-(h) until the size of the first system matrix or the size of the second system matrix satisfies the size threshold.
Dependencies are inferred from the claim text.