Plant diagnostic system

• US 4,755,925 A
• Filed: 09/23/1986
• Issued: 07/05/1988
• Est. Priority Date: 09/24/1985
• Status: Expired due to Fees

• Current Assignee: Toshiba Corporation
• Original Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Inventors: Tsuchiya, Minoru, Ohmori, Kazunori
• Primary Examiner(s): Smith, Jerry
• Assistant Examiner(s): NOT, DEFINED
• Application Number:
  US06/910,687
• Time in Patent Office: 651 Days
• Field of Search: 364/150, 364/151, 364/156, 364/157, 364/551, 364/184, 364/185
• US Class Current: 700/30
• 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 plant diagnostic system is disclosed wherein a plant to be diagnosed is expressed as an auto regressive model. A correlation function with respect to a difference obtained on the basis of an estimated value based on the auto regressive model and a plant value of the plant to be diagnosed is calculated. There is additionally provided a time constant determination unit for determining a time constant at which a change of one plant value has the greatest influence on that plant value or other plant values by using a coefficient matrix of the auto regressive model. This enables calculation of the correlation function in consideration of the time constant of the plant to be diagnosed. In this manner, an identification of an abnormality or failure source which is in conformity with the characteristic of the plant to be diagnosed can be conducted.

Claims

1. A plant diagnosis system comprising:

a plant diagnostic model for representing a plant system to be diagnosed by a mathematical model;

first means for determining a time constant at which a change of one plant value has the greatest influence on the plant value or other plant values by using a coefficient matrix of an auto-regressive model;

second means for calculating a time series of residual difference from an estimated value calculated by using said plant diagnostic model and an actual plant value obtained from said plant system so as to obtain the time series of residual differences;

third means for qualifying whiteness of the time series of residual difference from said second means and the time constant from said first means so as to produce an index;

fourth means for determining an abnormality from the index produced by said third means;

fifth means for determining a failure source from the index and the result; and

sixth means for outputting the abnormality and the failure source according to the results obtained by said fourth means and said fifth means.

2. A plant diagnostic system as set forth in claim 1, wherein said first means calculates impulse response a_ij (τ

) indicating to what degree a change of a certain plant value (item number i) affects a change of another plant value (item number j) from said auto regressive model when auto regressive coefficient matrices ##EQU16## are given, thus to calculate said time constant τ

_ij which has the greatest influence.

3. An apparatus, comprising:

a plant system having a plurality of plant variables with actual values;

a plant diagnostic model comprising a mathematical model for calculating estimated values of said plant variables;

time constant means for determining time constants at which a change of each of said plant variables has the greatest influence on the other of said plant variables;

difference determining means for determining differences between said estimated values of said plant variables and said actual values of said plant variables; and

means for determining an abnormality and a cause of said abnormality based on said differences determined by said difference determining means and said time constants determined by said time constant determining means.