Data processing: measuring – calibrating – or testing – Measurement system in a specific environment – Mechanical measurement system
Reexamination Certificate
2000-08-02
2002-06-25
Hilten, John S. (Department: 2863)
Data processing: measuring, calibrating, or testing
Measurement system in a specific environment
Mechanical measurement system
C702S035000, C702S081000, C702S181000, C702S182000, C700S021000, C700S079000
Reexamination Certificate
active
06411908
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to a method for estimating the remaining life with confidence bounds in an operating machine and deciding when to replace/repair an operating machine based on the cost of its estimated performance over its predicted remaining life.
2. Description of Related Art
Probabilistic modeling of machine life and other non-parametric reliability methods developed over the past five decades consider only age, and not condition, as a predictor of remaining life [see, for example, Barlow & Proschan (1975) or Ohtsuka et al., U.S. Pat. No. 5,608,845]. Now that new sensor technologies offer a means to track condition as well as age, better estimates of residual life can result. Recent work by Jardine et al. (1999) using Cox's proportional hazard modeling concept integrates both age and condition information into an Age Replacement decision model without differentiating between various failure modes.
Neural network models for estimating residual machine life have been proposed both as a “virtual” sensing technology [e.g., Roemer & Rieger (1995) and Upadhyaya et al. (1994)] and as a means to estimate remaining life [e.g., Talbott (1999)]. Husseiny, U.S. Pat. No. 5,210,704, describes “an expert system, rule-based failure data bank, a predictor, a performance evaluator and a system identifier” for prognosis of helicopter gearboxes and other rotating equipment.
Damage accumulation models of residual machine life describe remaining life as a function of material creep, fatigue, embrittlement, or corrosion damage propagation until some failure limit is reached. These models presuppose knowledge of a “damage limit” which is analogous to current practice of establishing “alarm” limits currently in vogue within the predictive maintenance community [see Talbott (1997)]. For example, Hardy et al., U.S. Pat. No. 4,525,763, estimate remaining useful life of an electric motor based on thermal damage and its relationship to a user-specified “desired life” limit.
Koyama et al., U.S. Pat. No. 4,768,383, predict remaining lifetime in metal materials, such as used in boilers and other equipment subject to high pressure and temperature, that are subject to creep damage. Their method depends upon a priori knowledge of a “creep rupture time” as a function of the degree of grain elongation and deformation over time.
Schricker, U.S. Pat. No. 5,750,887, approximates the remaining life of engine oil by trending certain oil quality estimators such as soot, oxidation, viscosity, and total base number and extrapolating a function of these estimators in time to a “predetermined threshold” value.
Binieda et al., U.S. Pat. No. 5,777,211, determine remaining useful life of automatic transmission fluid based on an empirically derived function of multiple factors affecting a “maximum life index . . . constant for each fluid/vehicle combination . . .” without specifying as to how such a constant is derived.
Patino et al., U.S. Pat. No. 6,023,150, claim a method to estimate remaining life cycles in a rechargeable battery pack by means of a look-up table relating battery cell impedance after recharge, adjusted for temperature, to remaining life cycles; however, the means for deriving such a look-up table are not disclosed.
Variations on the theme of damage accumulation modeling estimate residual life as a direct function of various environmental parameters by assuming that these parameters are indeed causing material creep, fatigue, embrittlement, or corrosion damage propagation. For example, Soga et al., U.S. Pat. No. 5,867,809, claim a remaining life estimation system for integrated circuit components on a printed circuit board based on a “predetermined life equation” such as the Coffin Manson Equation. Their system approximates remaining life in an electronic component as a function of the maximum temperature, largest temperature difference, number of temperature cycles, and other such environmental parameters collected over the operating lifetime of the component. Their system also claims a replacement decision process that compares estimated remaining life against a certain “guaranteed life” criterion.
Current industry practices for machinery condition monitoring are focused on diagnostic matters and, in terms of predicting remaining life, rely on end-users to establish their own “alarm” or “failure” levels for each deployed condition monitoring technology based on the end-user's experience or engineering judgment. Available software programs then employ simple regression models for time-to-alarm forecasting without benefit of any knowledge regarding the value of remaining life about to be discarded.
The invention described below offers machinery end-users an estimate of remaining life with a statistical confidence bounds in a machine diagnosed with a specific failure mode and presenting a set of multivariate symptoms associated with this particular failure mode. This invention, though it does not claim any diagnostic content, requires a moderate quantity of historical failure data coupled with an accurate diagnostic methodology.
SUMMARY OF THE INVENTION
According to a preferred embodiment of this invention, the described method requires condition-based data histories of same-type machines that have all lived in same-type operational environments and that have all failed according to the same failure mode. These data are obtained from end-users who monitor their machines with appropriate sensor technologies and have autopsies performed on failed machines in order to confirm failure mode. This approach offers experiential validity over prognostic algorithms based on physics-of-failure logic.
The invention aggregates condition-based data histories into “prognostic knowledge-bases” associated with a given failure mode or category of similar failure modes. These prognostic knowledge bases and the prognostic methodology described herein serve as an analysis “audit trail” for later review by authorized end users. This feature is an advantage over “black box” prognostic methods that have no visibility of method or associated data.
The invention's prognostic methodology yields, for an operating machine, a point estimate of remaining life conditioned on failure mode specific symptom values and a confidence bounds around this point estimate that measures statistical uncertainty. This confidence bounds aspect is an advantage over point-only estimates because end-users can incorporate this statistical uncertainty measure into their machine replacement decisions. For the purposes of this specification and claims, an “operating machine” is a machine that is operating in a given environment, diagnosed as having a primary failure mode underway, and presenting a set of symptom and condition variable measurements made at points in time.
With sufficient prior history data, the invention can estimate both remaining machine life and other performance output metrics (such as energy consumption cost) for an operating machine presenting a set of multivariate symptoms and environmental conditions. With this data in-hand, the invention provides a decision methodology as to when to replace an operating machine. This replacement decision methodology is based on the expected performance output over an estimated remaining life of the operating machine versus that for a healthy machine (i.e., new, rebuilt to good-as-new condition, or used but in good-as-new condition). For the purposes of this specification and claims, “expected performance output over an estimated remaining life” is defined as a cumulative value of some performance or cost metric associated with an operating machine from the point of its latest operating time stamp value to the point of its expected failure. Furthermore, an “operating time stamp” is defined as a value, in units of calendar time, age, operating cycles, cumulative distance traveled, or other similar life-related metric, recorded when a set of symptom and condition measurements are made on an o
Hilten John S.
Machinery Prognosis, Inc.
Pauley Petersen Kinne & Erickson
Vo Hien
LandOfFree
Condition-based prognosis for machinery does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Condition-based prognosis for machinery, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Condition-based prognosis for machinery will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2902777