US 4,509,110 AGranted

Method and apparatus for detecting failures in a control system

AssigneeUnited Technologies Corporation

InventorsAdelard Levesque, Jr., John P. Rembold

Life of this patent

Estimating the term…

Loading patent…

Abstract

A system for detecting the failure of a control system and its object includes a mathematical model of the control system and its object which continuously predicts the output of the object. An error signal is continuously calculated which is the difference between the model output and the actual output of the object. Error signals exceeding a predetermined minimum value or deadband are integrated; but the integral is reset to zero whenever the error signal value falls within a predetermined range indicative of normal steady state operating tolerances. If the value of the integral reaches a specified level before being reset to zero, a failure signal is triggered. This failure detection system can be made sensitive to relatively small error signals without triggering a false failure signal if the small error was the result of only a temporary problem or due to inaccuracies in model simulation.

Claims6

Dependent claims

A system for detecting the failure of apparatus, said apparatus including control means and an object of said control means, said system comprising:

model means simulating the input/output characteristics of said apparatus;
means for providing the same input to said apparatus and to said model means;
means for continuously comparing the output of said object to the apparatus input and for operating on said object to eliminate any difference between the compared input and output;
means for continuously comparing the output of said object to the output of said model means and for calculating error signals corresponding to the differences therebetween;
means for integrating, over time, the absolute value of said error signals which are outside a predetermined deadband having limits -a and +b;
means for generating a failure signal when said integral exceeds a predetermined value;
and
means for resetting said integral to zero whenever the value of the error signal is within predetermined limits -c and +d, where the absolute value of c is less than or equal to the absolute value of a, and the absolute value of d is less than or equal to the absolute value of b.

5 more elements

Dependencies are inferred from the claim text.

Facts

Application: 06/386,136Filed 7 Jun 1982
Granted: 2 Apr 1985
Term ends: Not estimatedThe source records don’t support an estimate.

Classification5 codes

G05B 23/0245
Model based detection method, e.g. first-principles knowledge model: based on a qualitative model, e.g. rule based; if-then decisions
G05B 13/048
Involving the use of models or simulators: using a predictor
G05B 17/02
Systems involving the use of models or simulators of said systems: electric
G05B 2219/42307
Servomotor, servo controller kind till VSS: compare actual feedback with predicted, simulated value to detect run away

Family

No related US applications are recorded in the indexed filing history.

Timeline

Citations

Open elsewhere

Building an agent or pulling many patents? Use the API or MCP server instead of scraping pages. Get an API key