Method for monitoring plants with mechanical components

• US 6,208,953 B1
• Filed: 01/14/1998
• Issued: 03/27/2001
• Est. Priority Date: 07/31/1997
• Status: Expired due to Fees

• Current Assignee: Sulzer Innotec AG (Sulzer AG)
• Original Assignee: Sulzer Innotec AG (Sulzer AG)
• Inventors: Milek, Janusz, Guettinger, Heinz
• Primary Examiner(s): Teska, Kevin J.
• Assistant Examiner(s): Sergent, Douglas W.
• Application Number:
  US09/006,925
• Time in Patent Office: 1,168 Days
• Field of Search: 395/500.23, 395/500.28, 702/33-35, 702/183, 706/911-912, 703/183, 703/2, 703/7, 970/691.2, 700/20, 700/2
• US Class Current: 703/7
• 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
• CPC Class Codes: G07C 3/00 Registering or indicating the condition or the working of machines or other apparatus, other than vehicles

Abstract

In a method for monitoring plants with mechanical components, measured values (x_iT; x_iu) are determined at predeterminable time intervals for a fixed set of parameters (x_i). The measured values (x_iT) which are determined for various working points during a modelling phase are used for the generation of a model for the operating behavior of the components. With the help of the model for the operating behavior at least one monitoring value (r; s_nu) is derived at predeterminable time intervals which is independent of the respective current working point. The temporal behavior of the monitoring value (r; s_nu) is used for estimating the wear in the components and/or for the detection of operating disturbances.

Claims

1. A method for monitoring plants with mechanical components that can be operated at variable working points, the method comprising:

determining measured values (x_iT;

x_iU) during the operation of the components, the measured values being determined in each case at predeterminable time intervals for a fixed set of parameters (x_i);

using the measured values (x_iT) that are determined during a modeling phase for various working points for the generation of a model (A) for the operating behavior of the components without determination of the physical relationships between the parameters, wherein the input variables of the model are at least a portion of the set of the parameters (x_i) and the output values comprise a model value (y_nT) for at least one of the parameters (x_n), wherein a residue (r_T) is formed in each case through comparison of the respective model value (y_nT) with the corresponding actual measured value (x_nT) of the modeling phase, and wherein the model is optimized by termination of model parameters in such a manner that a modeling error (E) that is derivable from the residues (r_T) is a minimum;

deriving at least one monitoring value (r;

s_nu) at predeterminable time intervals, the at least one monitoring value being derived with the help of the model of the operating behavior and being independent of the respective current working point;

determining a threshold value for the monitoring value on the basis of the residues determined for the modeling phase; and

using the temporal behavior of the monitoring value (r;

s_nu) for estimating the wear in the components and/or the detection of operating disturbances.

2. A method in accordance with claim 1 further comprising storing the determined measured values (x_iT;

x_iu) in an evaluation and storage apparatus, the generation and evaluation of the model taking place in the evaluation and storage apparatus, and wherein a warning apparatus is activated for the triggering of an alarm in the event that the monitoring value (r;

s_nu) departs from a predeterminable normal region.

3. A method in accordance with claim 1 wherein the monitoring value (r) is the difference between the respective current measured value (x_n) and the model value (y_n) corresponding to it.

4. A method in accordance with claim 2 wherein an updating takes place at predeterminable time intervals by means of the evaluation and storage apparatus for the determination of a current model (M_u) in which the measured values (x_iu) that were determined since the last updating of the model are taken into consideration.

5. A method in accordance with claim 4 wherein certain models (M_u, M_u−

1, M<sub>u−

2</sub>) that were determined at different times were stored in the evaluation and storage apparatus in each case together with characteristic values that are representative of the quality of the respective model.

6. A method in accordance with claim 5 wherein older models (M_u, M_u−

1*, M<sub>u−

2</sub>*) are taken into consideration in the updating for the determination of the current model (M_u).

7. A method in accordance with claim 6 wherein the older models are taken into consideration in a weighted manner.

8. A method in accordance with claim 4 wherein the respective current model (M_u) is optimized by means of the evaluation and storage apparatus and for the measured values (x_iu) that were determined since the last updating of the model, wherein a residue is calculated for the associated parameter (x_i) through comparison with the model values corresponding to them in each case, and wherein the model parameters are predetermined in such a manner that the modeling error that can be determined from the residues is a minimum.

9. A method in accordance with claim 4 wherein at least one reference working point is fixed, wherein after each updating of the model, the associated model value is determined for the fixed reference working point by means of the respective current model (M_u) of the associate model value as a nominal value (s_nu) for the corresponding parameter (x_n) by the evaluation and storage apparatus, and wherein the temporal behavior of the nominal value (s_nu) is used for the estimation of the wear in the components and/or for the detection of disturbances in the operation.

10. A method in accordance with claim 9 further comprising determining in each case, a plurality of nominal values for different parameters (x_i) for the fixed reference working point.

11. A method in accordance with claim 1 further comprising carrying out a fault isolation in the event of a deviation from normal operating behavior, wherein a plurality of partial models, (m₁, m₂, m_{3, m4}) are generated on the basis of the measured values determined for the parameters and/or of a total model, each of which has a model dimension (d) smaller that n−

1, with n being the number of the parameters (x_i) determined;

wherein reconstructed values for at least a portion of the parameters (x_i) are determined by means of the partial models (m₁, m₂, m₃, m₄);

wherein a reconstruction error is determined through respective comparison of the reconstructed values with measured values for the parameters (x_i) corresponding to them for each partial model (m₁, m₂, m₃, m₄);

and wherein at least one faulty parameter is isolated through comparison of the reconstruction errors with predeterminable threshold values.

12. A method in accordance with claim 11, wherein the partial models (m₁, m₂, m₃, m₄), that have reconstruction errors greater than the predeterminable threshold value are rated as faulty;

and

wherein the faulty parameter or the faulty parameters are isolated on the basis of the combinations of parameters (x_i) that lead to faulty partial models and the combinations of parameters (x_i) that lead to non-faulty models.