Fault detection using averaging filter with variable response time
AssigneeFord Motor Company
InventorsJohn V. James, Timothy M. Feldkamp
Life of this patent
Estimating the term…
Abstract
The time needed to verify proper operation of a fault detection system or to verify the impact of the replacement of a component in the system being monitored is reduced without substantial increase in the possibility of false indication of failure or false indication of proper operation by employing a geometric moving average with a variable time constant reset to a minimum value upon initialization and gradually increasing to a maximum value. Motor vehicle systems and components are monitored and a diagnostic indicator annunciates faults to alert the vehicle operator to malfunctions.
Claims7
A fault detection system for a vehicle comprising:
- a performance monitor coupled to said vehicle, said performance monitor producing a data stream quantifying a performance parameter of said vehicle, said data stream including random error fluctuations;
- a geometric filter filtering said data stream to produce a plurality of successive average values, said geometric filter characterized by a time constant to produce said average values, wherein said geometric filter is responsive to an initialization signal to obtain a temporary fast response of said geometric filter, wherein said time constant is initialized to a minimum value in response to said initialization signal and is incremented during successive average values to a maximum value;
- a comparator comparing said average values with a predetermined threshold and generating a fault signal when one of said average values exceeds said predetermined threshold, wherein said comparator utilizes the same predetermined threshold both prior to and following said initialization;
- and
- a diagnostic indicator annunciating a fault in response to said fault signal.
A method of fault detection for a vehicle comprising the steps of:
- monitoring a performance parameter of said vehicle to generate a data stream quantifying said performance parameter;
- initializing a time constant to a minimum value;
- filtering said data stream with a geometric filter to produce an average value in accordance with said time constant;
- comparing said average value with a predetermined threshold;
- generating a fault signal if said average value exceeds said predetermined threshold;
- incrementing said time constant by a predetermined step size and then repeating said filtering, comparing, and generating steps;
- repeating said incrementing, filtering, comparing, and generating steps until said time constant reaches a maximum value;
- and
- repeating said filtering, comparing, and generating steps using said maximum value for said time constant until said initializing step is repeated.
Dependencies are inferred from the claim text.