US 11,669,082 B2
Online fault localization in industrial processes without utilizing a dynamic system model
Xiaofan Wu, North Brunswick, NJ (US); and Ulrich Münz, Princeton, NJ (US)
Assigned to SIEMENS AKTIENGESELLSCHAFT, Munich (DE)
Appl. No. 17/51,379
Filed by Siemens Aktiengesellschaft, Munich (DE)
PCT Filed May 14, 2019, PCT No. PCT/US2019/032125
§ 371(c)(1), (2) Date Oct. 28, 2020,
PCT Pub. No. WO2019/222152, PCT Pub. Date Nov. 21, 2019.
Claims priority of provisional application 62/673,467, filed on May 18, 2018.
Prior Publication US 2021/0232132 A1, Jul. 29, 2021
Int. Cl. G05B 23/02 (2006.01)
CPC G05B 23/0254 (2013.01) [G05B 23/0218 (2013.01); G05B 23/0245 (2013.01)] 14 Claims
OG exemplary drawing
 
1. A method for localizing faults in an industrial process, the industrial process comprising an industrial plant including a plurality of components, the method comprising:
receiving structural plant data describing physical connections of the plurality of components of the industrial plant, wherein the physical connections include piping connections;
digitizing the structural plant data to identify a linearized system matrix, wherein a non-zero term in the linearized system matrix represents a physical connection between two components;
generating, in the form of a look-up table, fault patterns for specific faults of the components of the industrial plant using the structural plant data, wherein the look-up table includes a mapping between specific faults of the components and terms of the linearized system matrix that are outside a term-specific range of normal operational values;
receiving sensor data from sensors measuring characteristics of the plurality of components;
identifying a linearized system model of the industrial plant based on the sensor data and the digitized structural plant data;
continuously updating values of terms of the linearized system matrix based on the sensor data for a specified amount of time during fault free operation of the industrial plant;
determining the term-specific range of normal operational values for each term of the linearized system matrix based on the updating;
detecting and localizing faults during operation of the industrial plant based on the range of normal operational values for each term of the linearized system matrix and the look-up table; and
displaying fault information via a display to an operator.