Data processing: measuring – calibrating – or testing – Testing system – Of mechanical system
Reexamination Certificate
1999-02-11
2002-05-21
Wachsman, Hal (Department: 2857)
Data processing: measuring, calibrating, or testing
Testing system
Of mechanical system
C702S058000, C702S185000, C318S565000, C318S568240, C318S490000, C703S004000
Reexamination Certificate
active
06393373
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates to electric motors. More particularly, the present invention relates to a method and apparatus for condition monitoring and predictive maintenance of electric motors.
Electric motors are widely used in industrial equipment and processes where such motors are used to move goods along an assembly line from one work station to another or as a power source for power tools used by assemblers. Examples include air compressor that provide compressed air to power screw drivers, paint sprayers and other small hand-held appliances. Larger horsepower electrical motors maintain environmental control by cooling, heating and transporting air through the heating and cooling system in buildings and vehicles. In the home and office environment, electric motors are used in appliances ranging from computers to vacuum cleaners. As is generally known, such appliances constitute a major source of noise and vibration. Therefore, ever increasing demand from the market for quieter and vibration free motors can only be fulfilled by the design and production of fault free and quieter motors.
In the manufacturing environment, unexpected failure of the motor is both undesirable and costly. In the industrial setting, motor failure could have significant financial impact if an assembly line is shut down during the time it takes to repair or replace the motor. Further, in some manufacturing processes, such as in a semiconductor fabrication facility, failure of a critical motor could result in damage to the product if control over the environment is compromised.
Accordingly, there is a growing demand to improve the reliability of electric motors in general and, especially in industrial applications, detect impending faults so the motors can be repaired or replaced during routine maintenance rather than after failure has occurred. It is also desirable to improve reliability of electric motors through improved quality control monitoring during manufacture of the electric motors. It is further desirable to detect motor faults prior to catastrophic failure through performance monitoring during operation.
Recently, fault detection and diagnosis (FDD) methods have been developed that compare the output signals of complex systems with the output signal obtained from a mathematical model of the fault free system. The comparison of these signals is quantified in terms of a “residual” which is the difference between the two signals. Analysis of the residuals is carried out to determine the type of the fault. This analysis includes statistical methods to compare the residuals with a database of residuals for systems with known faults.
Until recently it has been difficult to obtain accurate, real-time models for multivariable systems, that is, systems with more than one inputs and/or one outputs. If the model of the system is not accurate, the residuals will contain modeling errors that are very difficult to separate from the effect of actual faults.
Another shortcoming of such FDD methods relates to the difficulty in generating a data base for statistical testing of residuals to classify faults. Developing such a database requires a priori information about all possible faults and the effect each such fault has on the residuals. Accordingly, a period of time is required to monitor defective and normal equipment and to develop a data base which contains fault signatures for fault classification purposes. This process is both costly and time consuming. Also, the data base must also meet the specific requirements of a particular FDD scheme.
Since, mechanical faults are the result of vibration, detection and analysis of vibration is a common element of many prior art detection schemes. Such techniques require development of a library showing previously experienced motor vibration patterns which are correlated with the detected fault.
A common disadvantage of mechanical fault detection is that the scheme requires a-priori information about the fault signature in order to correlate the actual fault with the detected signature. Such correlation requires development of an extensive database and a laborious analysis and a level of expertise about the motor.
Another drawback of mechanical fault detection arises from the difficulty associated with reproducing the measurements. For example, vibration measurements using an accelerometer are highly dependent on mounting method and positioning of the sensor to ensure repeatable detection of the signature. Even with proper sensor mounting and positioning, signature detection may be corrupted by background vibration and variation in operating conditions such as running speed, input voltage and motor loading. It will be appreciated that the likelihood of erroneous indication of failure in a system relying on mechanical fault detection is high. As an example, the assessment of the condition of the motor's bearings involves analyzing the mechanical vibration of the motor and separating out the specific frequencies related solely to bearing flaws ( and/or any sum and difference frequencies and related harmonics). Unfortunately, the presence of, and possible coincidence with, other vibrations in the vibration spectrum often interfere with detection of the desired signal. Expensive and sophisticated means are necessary to gain the desired information and the success of such a system in detecting or predicting a fault is less than desirable.
Accordingly, it is desirable to eliminate the complications caused by modeling errors and both false indications and missed indication of motor faults. It is also desirable to avoid having to develop an extensive database and laboriously developed expertise in analysis of the cause of faults in electric motors. It is further desirable to eliminate the need for expensive and sophisticated means for obtaining and processing information that may indicate a fault exists.
SUMMARY OF THE INVENTION
The present invention relates to a model based fault detection system and method for monitoring and predicting maintenance requirements of electric motors and more particularly fractional horsepower electric motors. Using the system, it is possible to obtain information for early diagnosis of impending mechanical failure of the electric motor in the operational environment under unknown loading conditions. Since the method and system of the present invention is software based and utilizes data obtained from non-intrusive measurements, implementation costs are significantly less than prior art maintenance methods.
The system comprises computer means coupled to voltage, current and speed sensors by a multifunction data acquisition means. The sensors provide continuous real- time information of the input voltage and current and of the output voltage signal developed by the motor's tachometer. The computer means uses such information in continuously running a fault detection and diagnostic algorithm in conjunction with a diagnostic observer.
The system and method utilize a multivariable experimental modeling algorithm to obtain a model of the electric motor by determining the structure, that is the order of the differential equations mathematically describing the motor, and the motor's invariants, that is, parameters such as inductance, motor resistance, moment of inertia, non-physical parameters such as A, B and C matrices of state equations describing the motor and other selected parameters. In the preferred embodiment, the model of the electric motor is developed when the motor is known to be running free of faults, usually after the motor is initially installed. Later, during operation, the model output voltage signal is calculated based on the actual input voltage and current applied to the motor and continuously compared to the measured output voltage signal of the motor. The algorithm quantifies the comparison in terms of a residual which is generated by subtracting the respective signals.
The diagnostic observer analyzes the residual and determines if the motor is fault free or operating in a
Albas Evren
Durakbasa Osman Tugrul
Duyar Ahmet
Serafettinoglu A. Hakan
Arcelik A.S.
Shinners Craig E.
Wachsman Hal
LandOfFree
Model-based fault detection system for electric motors does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Model-based fault detection system for electric motors, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Model-based fault detection system for electric motors will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2848783