Method and apparatus for predicting a fault condition using non-linear curve fitting techniques

• US 6,363,332 B1
• Filed: 12/22/1998
• Issued: 03/26/2002
• Est. Priority Date: 12/22/1998
• Status: Expired due to Fees

• Current Assignee: Caterpillar Incorporated
• Original Assignee: Caterpillar Incorporated
• Inventors: Sarangapani, Jagannathan, Rangarajan, Sri K.
• Primary Examiner(s): Assouad, Patrick
• Assistant Examiner(s): Barbee, Manuel L.
• Application Number:
  US09/218,923
• Time in Patent Office: 1,190 Days
• Field of Search: 702/34-40, 702/58, 702/59, 702/113-115, 702/182-185, 702/188, 701/29, 701/34, 361/115, 361/25, 361/27
• US Class Current: 702/185
• CPC Class Codes: G05B 19/4063 Monitoring general control system G05B19/4062 takes precedence
• CPC Class Codes: G05B 2219/34465 Safety, control of correct operation, abnormal states
• CPC Class Codes: G05B 2219/34477 Fault prediction, analyzing signal trends

Abstract

An apparatus for predicting a fault condition for a machine is disclosed. The machine has a plurality of parameters being dependent upon machine performance. A sensor is connected to the machine and is adapted to produce an electrical signal in response to one of the plurality of machine parameters. The apparatus further includes means for determining a non-linear data trend of the parameter in response to the electrical signal, and means for predicting the fault condition as a function of the trend and a threshold value indicative of the fault condition.

In a second aspect of the invention, a method for predicting a fault condition is provided. The method includes the steps of sensing a parameter having a level being dependent upon machine performance and responsively producing an electrical signal, determining a non-linear data trend of the parameter in response to the electrical signal, and predicting the fault condition as a function of the trend and threshold value indicative of the fault condition.

Claims

1. An apparatus for predicting a fault condition, comprising:

a machine having a plurality of parameters being dependent upon machine performance;

a sensor connected to the machine and adapted to produce an electrical signal in response to one of the plurality of machine parameters;

means for determining a non-linear data trend of the parameter in response to the electrical signal where said means for determining further comprises a means for fitting an exponential function to the trend and if a predetermined fit tolerance is not achieved, then fitting a polynomial function to the trend;

means for predicting the fault condition as a function of the trend and a threshold value indication of the fault condition.

2. The apparatus, as set forth in claim 1, further comprising means for computing a value as a function of the electrical signal for determining the non-linear data trend.

3. The apparatus, as set forth in claim 1, further comprising an output device for displaying a fault signal when the fault condition has been predicted.

4. The apparatus, as set forth in claim 1, wherein the means for predicting the fault condition are located on the machine.

5. The apparatus, as set forth in claim 1, wherein the means for predicting the fault condition comprise a computer.

6. A method for predicting a fault condition for a machine, the machine having a plurality of parameters being dependent upon machine performance, the method comprising the steps of:

sensing one of the parameters dependent upon machine performance and responsively producing an electrical signal;

determining a non-linear data trend of the parameter in response to the electrical signal where determining a non-linear data trend further comprises the steps of attempting to fit an exponential function to the trend, and if the exponential function does not fit the trend within a predetermined tolerance limit, attempting to fit a polynomial function to the trend; and

predicting the fault condition as a function of the trend and a threshold value indicative of the fault condition.

7. A method of predicting a fault condition comprising the steps of:

sensing with a sensor at least one of a plurality of parameters of a machine where said parameters are dependent upon machine performance and responsively producing trend data;

determining a polynomial function representative of the trend data, where the step of determining includes determining an order of the polynomial by iteratively incrementing an estimated order and computing a derivative until the resultant slope is constant, and then computing the parameters of the polynomial of the determined order; and

predicting the fault condition as a function of the polynomial function and a threshold value indicative of the fault condition.