Method for the monitoring of a plant

• US 6,587,737 B2
• Filed: 08/06/2001
• Issued: 07/01/2003
• Est. Priority Date: 09/14/2000
• Status: Expired due to Fees

• Current Assignee: Sulzer Markets and Technology AG
• Original Assignee: Sulzer Markets and Technology AG
• Inventors: Güttinger, Heinz, Voser, Alexandre
• Primary Examiner(s): Picard, Leo
• Assistant Examiner(s): Garland, Steven R.
• Application Number:
  US09/923,865
• Publication Number:
  US 20020052712A1
• Time in Patent Office: 694 Days
• Field of Search: 700/30, 700/29, 700/31, 700/33, 700/47, 700/52, 700/73, 700/80, 703/2, 703/7, 701/14
• US Class Current: 700/30
• CPC Class Codes: G05B 17/02 electric
• 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

A method is proposed for the monitoring of a plant having a plurality of sub-systems which is operable at variable working points. Respective measured values are detected for a fixed set of process parameters (B, U, S1, . . . , SN) at pre-settable time intervals during the operation of the plant (1). The measured values detected in a learning phase for different working points are used to prepare models (M1, M2, . . . , MN) for the operating characteristics of the sub-systems (11, 12, . . . , 1N). At least one monitoring parameter (Ri, Ri_k), which is independent of the respective current working point, is determined at pre-settable time intervals in an operating phase using the models (M1, M2, . . . , MN) and the time-dependent course of the monitoring parameter (Ri, Ri_k) is used for the monitoring of the plant (1). A pre-check (2) is carried out prior to the determination of the monitoring parameter (Ri, Ri_k), in which a check is made whether at least the measured values for those process parameters which are operating parameters (B) are within a pre-determined range.

Claims

1. A method for the monitoring of a plant having a plurality of sub-systems which is operable at variable working points and comprising the following steps:

a) respective measured values are detected for a fixed set of process parameters (B, U, S1, . . . , SN) at pre-settable time intervals during the operation of the plant (1);

b) the measured values detected in a learning phase for different working points are used to prepare models (M1, M2, . . . , MN) for the operating characteristics of the sub-systems (11, 12, . . . , 1N), with the input values of each model (M1, M2, . . . , MN) being at least a part of the process parameters (B, U, S1, . . . , SN) and the output value of each model comprising a model value (Yi, Yi_k) for at least one of the process parameters, and with the models (M1, M2, . . . , MN) being optimised by comparing the model values with the measured values;

c) at least one monitoring parameter (Ri, Ri_k), which is independent of the respective current working point, is determined at pre-settable time intervals in an operating phase using the models (M1, M2, . . . , MN);

d) the time-dependent course of the monitoring parameter (Ri, Ri_k) is used for the monitoring of the plant (1);

characterised in that a pre-check (2) is carried out prior to the determination of the monitoring parameter (Ri, Ri_k), in which a check is made whether at least the measured values for those process parameters which are operating parameters (B) are within a pre-determined range.

2. A method in accordance with claim 1 wherein a machine model (MA) is determined for the pre-check (2), the input values of said machine model (MA) being operating parameters (B) and its output values comprising a machine model value (YA) for at least one of the operating parameters (B).

3. A method in accordance with claim 2 wherein the machine model (MA) comprises at least one model (ME) with which a machine model value (YA) is determined for the deviation (DP;

DQ) of the measured value (PI;

PQ) for an operating parameter (P;

Q) from a theoretically determined desired value (PS;

QS) for said operating parameter (P;

Q).

4. A method in accordance with claim 3 wherein the model (ME) for the deviation (DP;

DQ) of the measured value (PI;

PQ) from the theoretically determined desired value (PS;

QS) is an experimental model, which is prepared and optimised on the basis of the measured values detected in the learning phase.

5. A method in accordance with claim 2 wherein the machine model (MA) is prepared using the measured values detected in the learning phase for different working points, with a deviation value being determined by comparing the machine model value (YA) with the actual measured value corresponding thereto, and model parameters of the machine model (MA) being optimised such that a model error determinable from the deviation values becomes minimal.

6. A method in accordance with claim 2 wherein a residual value (RA) is determined for the plant during the operating phase by a comparison of the machine model value (YA) with the corresponding current measured value of said operating parameter (B).

7. A method in accordance with claim 1 wherein the environmental parameters are examined for freedom from malfunction during the pre-check (2).

8. A method in accordance with claim 1 wherein the pre-check (2) comprises a range test (21) in which an examination is made as to whether individual or combinations of operating parameters (B) are within the range of working points learnt in the learning phase.

9. A method in accordance with claim 8 wherein the range test (21) is carried out by means of a histogram (5) which is prepared using the measured values detected in the learning phase.

10. A method in accordance with claim 1 wherein a plurality of monitoring parameters is determined, each of which is a residual value (Ri, Rik, RA) which identifies the deviation of a model value (Yi, Yik) or a machine model value (YA) from the current measured value corresponding thereto.

11. A method in accordance with claim 1 wherein a plurality of categories is used for the monitoring, the process parameters being capable of being classified therein in each case in dependence on their current value, with there at least being provided a category for the normal operating state, a category for an early warning, a category for a wear warning, a category for unknown states and a category for a signal warning.

12. A method in accordance with claim 1 wherein a confidence range is fixed at least for one monitoring parameter (Ri, Rik, RA), said confidence range being able to be extended if the monitoring parameter leaves the original confidence range, and the extension of the confidence range is registered.