KR100532237B1 - Model based error detection system for electric motors - Google Patents

Abstract

The present invention relates to a model based error detection system and method for monitoring an electric motor and predicting maintenance needs.

The method and system of the present invention are much less expensive than conventional methods because they are software based and use data from non-mandatory measurements.

The system includes computer means attached to sensors that provide continuous real-time information on input voltage, current and motor speed.

This system and method uses a multivariate experimental modeling algorithm to obtain a mathematical description of the motor.

The algorithm quantifies the comparison result by the residual value generated by subtracting each signal by comparing the measured result with the modeled result.

The diagnostic observer analyzes the residual value to determine if the motor is error free. When an impending error is detected, the diagnostic observer evaluates the measured parameters of the motor to determine deviations from the reference value and develops a diagnosis of the component that has failed or exhibits signs of failure.

Another embodiment of the invention is particularly advantageous for performing fractional horsepower electric motors, in particular quality control tests.

Description

Model-Based Error Detection System for Electric Motors

The present invention relates to an electric motor, and more particularly, to a method and apparatus for monitoring and predicting and repairing a state of an electric motor.

Electric motors are widely used in industrial equipment and processes, particularly in moving goods from one workshop to another, along the assembly line, or as a power source for tool use.

An example is an air compressor that provides compressed air to power screwdrivers, paint applicators and other small equipment.

Buildings and automobiles use a larger powered electric motor to cool, heat, and transport air through a heating and cooling system to maintain environmental control.

In the home and office environment, electric motors are used in various ways from computers to vacuum cleaners.

As is commonly known, such fixtures are a major source of noise and vibration. Thus, there has been an increasing desire in the industry for motors that are quieter and have less vibration.

In manufacturing conditions, unexpected failures of the motor are undesirable, as well as damaging.

In industrial installations, if a motor fails during operation, the assembly line will be shut down for the time it is to be repaired or replaced, which will result in large financial losses.

Furthermore, in some manufacturing processes, such as semiconductor manufacturing, major motor failures can have a major impact on product production.

Therefore, there is an increasing demand to improve the reliability of electric motors in general.In particular, there is a demand in the industry to detect the impending failure so that the motor can be repaired or replaced through routine maintenance, not after the failure of the motor. It is increasing.

It is also desirable to improve the reliability of the electric motor through improved quality control monitoring during the manufacture of the electric motor.

In addition, performance monitoring during operation is required to detect motor failures before major failures occur.

Recently, an error detection and diagnosis method (FDD method) has been developed for comparing output signals of complex systems with output signals obtained from mathematical models of error-free systems. The comparison of these signals is quantified as "residual", which is the difference between the two signals. The residuals are analyzed to determine the type of error.

This analysis involves a statistical method for comparing a database of residuals and residuals for systems with known errors.

Until recently, it has been difficult to obtain accurate and real-time models for multivariate systems, that is, systems with one or more inputs and / or one or more outputs.

If the model of the system is not accurate, the residuals will contain modeling errors and it is very difficult to separate them from the effects of real errors.

Another problem with this FDD method involves the difficulty of creating a database that statistically tests residuals to classify errors. Such a database development would require all conventional information relating to all possible errors and the impact each of these errors had on the residual value.

Therefore, it takes a lot of time to monitor a faulty device and a normal device, and to develop a database including an error symbol for error classification. This process is both costly and time consuming. The database must also match what the specific FDD design requires.

Since mechanical defects are the result of vibrations, detecting and analyzing vibrations is a common element of many conventional detection designs. This technique requires the development of a library that demonstrates the previously experienced motor vibration patterns associated with detected defects.

A disadvantage of mechanical defect detection is that it requires deductive information about the defect signal in order to relate the detected signal to the actual defect. This correlation requires extensive database development, analysis of motors, and a significant level of expertise.

Another drawback of mechanical defect detection is that it is difficult to recalculate the measurement. For example, measuring vibration using an accelerometer is highly dependent on how the sensor is mounted and the location of the sensor in order to ensure the repeated detection of the signal.

Even if the sensor is properly mounted and positioned, signal detection can be misleading due to changes in operating conditions such as background vibration and running speed, input voltage and motor load.

In systems that rely on mechanical fault detection, the likelihood of misrepresenting errors is very high. For example, determining the bearing state of a motor involves analyzing the mechanical vibrations of the motor and separating only certain frequencies (and / or sum and difference frequencies and associated counts) for bearing scratches.

Unfortunately, the presence of and other coincidences with other vibrations in the vibration spectrum often hinder the detection of the desired signal.

Thus, obtaining the desired information requires expensive and sophisticated means, and the success of the system for detecting and predicting errors is not as desired.

Therefore, it is desirable to eliminate the complexity caused by incorrectly displaying and not displaying modeling errors and motor errors.

It would also be desirable to eliminate the need to build extensive databases and develop expertise in analyzing the failure causes of electric motors.

Furthermore, it would be desirable to eliminate the need for expensive and complex means to obtain and process information indicating error presence.

1 is a schematic diagram showing an electric motor useful for practicing a preferred embodiment of the present invention.

2 is a plan view of a typical motor enclosure

3 and 4 are exemplary input and output waveform diagrams for practicing one embodiment of the present invention.

5 shows a system level configuration of a preferred embodiment of the present invention.

6 is a block diagram of an error detection and diagnosis system according to an embodiment of the present invention.

7A-7B and 8A-8B are flowcharts of the operation of the error detection and diagnosis system according to an embodiment of the present invention.

The present invention relates to a model based error detection system and method for monitoring electric motors, in particular fractional horsepower electric motors and predicting their maintenance needs. Using this system, it is possible to obtain information for early diagnosis of an impending mechanical fault of an electric motor in an operating environment under unknown loading conditions.

The method and system of the present invention is significantly lower in performance compared to conventional maintenance methods because it is based on software and uses data from non-mandatory methods.

The system comprises computer means coupled to the voltage, current and speed sensors by means of multifunction data acquisition.

The sensors provide continuous real-time information of the input voltage and current and provide real-time information of the output voltage signal generated by the tachometer of the motor.

The computer means utilizes this information in error detection and diagnostic algorithms in cooperation with the diagnostic observer.

The system and method of the present invention are structured in order of differential equations describing the motor mathematically, inductance, motor resistance, moment of inertia, non-physical parameters such as A, B and C mattresses of state equations describing the motor and other selected parameters. A multivariate experimental modeling algorithm is used to obtain a model of the electric motor by measuring the constant variables of the equivalent motor.

In a preferred embodiment, the model of the electric motor is developed when the motor is known to be running free of defects after its initial installation. The model output voltage signal is then calculated on the basis of the actual input voltage and current applied to the motor during operation and contrasted with the measured output voltage signal of the motor. The algorithm quantifies the comparison value by the residual value generated by subtracting each signal.

The diagnostic observer analyzes the residual value to determine whether the motor is intact. Under trouble-free operation, the residual value is ideally zero, even though the selected error threshold can be selected to compensate for modeling errors, noise, etc., which may result in non-zero residual values in operation.

When the motor components deteriorate and the motor is operating outside its intended operating range, or when an error actually occurs, the residual value will have a non-zero value outside of the allowed error threshold.

If the computer means immediately detects a non-zero residual value, a fault is imminent and an alert is sent to find the appropriate means to minimize the effect that a failed motor can cause.

Upon detecting an impending fault, the diagnostic observer evaluates the measured parameters of the motor and obtains a deviation from the reference value to perform a diagnosis of the likely or defective component.

In another embodiment of the present invention, a system for detecting and diagnosing mechanical failure of an electric motor is disclosed.

Rather than building an extensive database to correlate the error with the measured signal, the present invention incorporates a mathematical model of a faultless motor to measure motor operating parameters without being affected by environmental, operational and payload torsion.

This example is particularly advantageous for the production of fractional horsepower electric motors and in particular for carrying out quantitative control tests.

After manufacturing a plurality of motors, a multivariate system recognition algorithm is used to develop a basic model using the entire available motor population.

Since the population may include multiple error generating motors, it is necessary to select the error threshold and retest each motor against the model to refine the model.

Motors beyond the threshold are removed from the population and the remaining motors are used to build a new base model.

This newly built base model is stored in computer means for quality control testing of all continuously manufactured motors.

During the quality control test, if the motor's parameters, such as inductance, motor resistance friction coefficient or moment of inertia, deviate from the critical error established in the basic motor model, the motor under test is classified as having an error. By comparing the parameters of the motor under test with a basic motor model with different error limits, it is possible to classify motor errors and display diagnostic information.

Hereinafter, the present invention will be described in detail with reference to the accompanying drawings.

Figure 1 shows a system including an electric motor 10, such as a fractional horsepower electric motor. For illustration, the motor 10 comprises an axis 16 whose end is supported by a rotor winding 12, a stopper 14 and a bearing 18. Pulley 20 couples shaft 16 to a load not shown. The collector 22 is connected to the rotor 12 and its rotor to form a magnetic field that directs current into or out of the armature 24, which moves the motor.

Those skilled in the art will appreciate that the motor 10 may have a commutator or a rotor without windings. Motor 10 is mounted in case 26 to isolate the motor from dust, moisture and other external conditions.

FIG. 2 is a plan view of the motor enclosure with the bottom of case 26 secured to the cap in a manner well known in the art, such as screws and nuts 28.

5, a preferred embodiment of the motor condition monitoring system 30 according to the present invention is shown.

System 30 includes a power supply 32, a plurality of sensors 34, 35 and 38, a multifunction board 37 and a computer 42, which may be a motor 10, a line voltage or a power supply such as Hewlett Packard 6010A.

When voltage is applied, motor 12 is typically operated at operating speed within 25 milliseconds after power is applied and shaft 16 rotates at a speed that is dependent upon the partially applied voltage and load. The speed of the motor 12 is detected by the tachometer sensor 36 converted from the analog signal to the digital signal by the multifunction input / output board 37 and transmitted to the computer 42. The tachometer sensor 36 can be a speed encoder or a bulk-in tachometer designed in motor 10.

The multifunction board is furthermore coupled to a voltage sensor 34 and a current sensor 35, which may be a 1: 100 voltage split probe, for example, with a response time of at least 23 nanoseconds (examples of acceptable current sensors are Tektronix 6303, 100 amp ac / dc current probe, Tektronix 502a power module and Tektronix 503b ac / dc current probe amplifier).

The signals from sensors 34 and 35 are also conditioned by board 37 and input to computer 42. Computer 42 writes sensor data to memory. Computer 42 gives a fault detection and diagnostic model of an ideal motor that is also stored in memory. In a preferred embodiment, the model tone of the monitor is an experimental modeling tool (EMT) developed by a multivariate system recognition algorithm, Ahmet Duyar, and currently available from Advanced Systems, Inc., Floria 33431, Boca Raton, North Ocean Boulerard 4201 Suite 206. Is developed first.

EMT is an experimental modeling tool that creates mathematical equations describing the dynamic relationships between input and output measurements obtained from experiments designed to characterize a system under a range of possible modes of operation.

This information includes enough input signals to test the system over the system's bandwidth, optimal scan rate and durability, and the system's bandwidth. As is known in the art, experimental modeling is the selection of mathematical relationships that match the observed input and output data. Thus, equations are developed that describe the behavior of various system elements and their interconnections during the modeling process.

The experimental model of the system is described by a set differential equation in the form of a matrix.

The EMT program determines the structure of the system, that is, the order of the system, the parameters and constant coefficients of the variables of the differential equations.

In a preferred embodiment, the structure is determined by developing an information matrix using input and output data.

This matrix is row searched to determine the structure of the system.

The theoretical concept of judging row rank search is explained in detail in the next book. State Space Representation of the Open-Loop Dynamics of the Space Shuttle Main Engine, Ahmet Duyar, Vasfi Eldem, Walter C. Merrill and Ten-Huei Guo, 1991 December, Vol 113, Journal of Dynamic Systems, Measurements and Control, pp 684 ~ 690.

Once the system architecture is determined, the number of parameters contained in the differential equation set is known.

The measured data is used in conjunction with a set of differential equations containing unknown constants to produce various equations.

The number of equations generated is greater than the unknown constant.

The least squares technique is used to determine unknown constants by methods known in the art.

The model-based error detection and diagnostic scheme of the present invention describes an error-free motor with a series of equations described below.

Since the error in motor 10 changes the parameter, the equation for motor 10 will be different from the expected equation generated by the model.

The scheme of the present invention relies on the concept of analytical redundancy, in which the signal generated by the model is compared with the measured signal obtained from motor 10 to determine whether the motor is operating normally. This model replaces the need to develop information about the motor. Based on the comparison, computer 42 generates residuals and analyzes them to determine whether the motor is running without error.

The present invention develops the prediction of information that is important for early diagnosis of the impending failure of an electric motor while operating under unknown load.

To illustrate, consider an error-free system described by the following discrete state space equation.

x (k + 1) = Ax (k) + B u (k) .... (1)

y (k) = Cx (k) ..... (2)

Where x, u and y are n x 1 state vectors, p x 1 input vectors and q x 1 output vectors, respectively, and k represents discrete time increments.

A, B and C are known nominal mattresses (paramets) of systems with appropriate dimensions. For example, a fractional horsepower electric motor is used, and the experimental model uses input voltage, current and speed measurements.

3 shows a graph of the input voltage 38 used to power up the motor 10.

In a preferred embodiment, the input voltage 38 is a step input and is represented in the experimental model as a row vector containing the measured voltage.

4 shows experimentally determined current and speed output signals 39 and 40, respectively, and the measured current and speed output signals are shown in solid lines.

The resulting system description can be represented by equations (3) and (4), where A matrix representing, for example, state-space is of the form:

And the B matrix is of the form

-2.6188

0.0012

4.3719

0.0092

-3.5824

-0.0259

1.0257

0.0156

1.0915

0.0000

And the output C matrix, which combines the variable with the output, is of the form

0 0 0 0 0 0 0 0 1 0

0 0 0 0 0 0 0 0 0 1

In addition to the discrete A, B and C matrices of the system determined by the modeling program, a standard error of estimate (SEE) is also determined.

SEE compares the modeled and measured outputs and provides an estimate of the modeling error.

In this example, the SEE for the model is 2.8 for the current output and 0.67% for the speed output.

If ten motors fail, the parameters and consequently the response of system 30 will be different.

Wrong parameters and variable error in the system can be represented by equations describing the system:

xf (k + 1) = Af xf (k) + Bf uf (k) .... (3)

yf (k) = Cf xf (k) ... (4)

In the simplest case, the residual vector r (k) is defined as the difference between the output of the error-free system and the output of the error system.

r (k) = yf (k)-y (k) .... (5)

Without noise and modeling errors, the residual (residual) vector r (k) is zero under error-free conditions. Non-zero residual vectors indicate that an error exists. In the presence of noise and modeling errors, the effect should be separated from the error effect by comparing the residual size to the selected threshold. Using the residual distribution observed under error-free conditions, the threshold selects a confidence level (3). Error alarms and absence errors can be minimized.

Referring to FIG. 6, an EMT, which is a multivariable identification algorithm, is used to develop a baseline experimental model 44 of the motor 10. Model 44 includes parameters of other equations, namely A, B and C, and their order, n in equations (1) and (2). In contrast to theoretically derived model parameters, experimental model parameters do not provide a physical meaning. In other words, changes to these parameters cannot be used to understand cause-effect relationships. Although the physical meaning of these parameters is missing, the experimental model does not derive from the assumptions, so it accurately represents motor 10. However, system 30 eliminates the need to rely on information about the structure of the motor 10, aside from the assumption that the motor 10 was initially error free.

The output of model 44 is evaluated by the computer 42 with the EMT algorithm using measurements from voltage sensor 34, speed sensor 36 and current sensor 26 to obtain model output. The model output is compared to the output of the monitor displayed by summer 46 to produce a residual value r (k). Comparator 48 determines whether the residual vector r (k) is a zero vector and correspondingly whether the motor is operating without error. If comparator 48 determines that residual vector r (k) does not have a value of zero, one or more errors are displayed. However, since there are typically noise and modeling errors, the residual vector r (k) is compared with the threshold chosen first to avoid error reading. If the residual value is below the threshold, the nonzero value is considered to be due to this noise or modeling error and the motor 10 will be considered to be error free. System 30 then reports that the system is error free as indicated by box 50. However, if the residual value exceeds the threshold, the error is shown in Table 2 and system 30 begins analysis 52 of the error.

Based on the results of the analysis, the errors are categorized and reported to the user at 54 or stored on computer 42 for future reference.

By using model-based diagnostic routines, the current response of the motor can be modeled under error-free conditions and compared with the current response of the same motor during operation. In the present invention, the computer 42 includes means for repeating the error detection to predict, detect, and classify the mechanical error of the electric motor. The systems and methods of the present invention can be used in both manufacturing and operating environments.

Error classification is accomplished by measuring the change in the motor 10 parameters and correlating the change with the motor error using the physical parameters of the model theoretically obtained.

Consider the simplified theoretical ice equations (6) and (7) that represent a universal motor that can operate on direct current or alternating current for DC voltage input.

L di / dt + Ri = V + k1 wi ...... (6)

J dw / dt + fw = k2 i2 + M ... (7)

Where L, R, J and f are the inductance, resistance, moment of inertia and coefficient of friction, respectively, and k1 and k2 are motor constants. In equations (6) and (7), the output variable, current and speed are represented by j and w, respectively, while the input variable voltage is represented by V.

In the MCM algorithm, the load M is generally available or not easily measured. Therefore, it is necessary to operate in equations (6) and (7) to remove the load for use by diagnostic observers. In one embodiment, the diagnostic observer is only based on the model in equation (6) that is independent of the load. In this embodiment, although some information is provided to the diagnostic observer, motor friction and constant k2 are not available and there may be a larger% unknown error record.

Thus, if such information is needed, the observer may choose a derivative of equation (7) to assume that the load is constant and remove the load item.

As will be apparent to those of ordinary skill in the art, other possible mathematical means are available to remove the load item by describing equations (6) and (7) in matrix form and multiplying both sides with the appropriate matrix operator. Do.

Referring again to FIGS. 1 and 2, conventional mechanical defects may result from rotor 12 that is out of balance, screw 28 of uneven torque or defective bearing 18, collector 22, or pulley 20.

These mechanical defects cause vibration and noise once motor 10 is installed and operated under load M.

Recognizing that mechanical vibrations result in physical displacements, the vibrations caused by the bearing grooves will cause periodic displacements on axis 16, but in the electric motor the drive shaft is rotated by the armature assembly.

Mechanical defects will cause the rotor to mismatch, which in turn causes the air gap to be asymmetric, changing the inductance, resistance and motor constant parameters included in equation (6).

The current through the motor is partly a function of the magnetic field of the air gap between the armature and the stator (or field coil). The periodic displacement induced on the drive shaft affects the symmetry of the air gap and the magnetic field in the air gap. The magnetic field in the air gap again affects the current through the motor. The effect on the magnetic field in the air gap is periodic and its frequency is known and so it is the effect on the current.

Thus, the nominal value change of the inductance parameter, L, is associated with an unbalanced rotor error.

The observed change in resistance parameter R is believed to indicate collector error.

Bearing error is judged when the change in inductance constant indicates oscillation behavior and / or when both the inductance and friction constant change in line.

The standard deviations of error-free and error-free and error-free parameters are shown in Tables 1 and 2.

In Table 1, for a given voltage V and load M, the current and velocity outputs predicted by model 44 are shown with the selected error parameters (3 standard deviations) and examples of current and velocity measurements.

As noted, the current measurement exceeds the expected value by more than 3 standard deviations. Thus an error appeared.

Error The parameters of motor 10 are checked in table 2.

As noted, the inductance L of the error motor 10 is greater than the corresponding inductance parameter predicted by the model 44 by more than 1 standard deviation while the other parameters are less than 1 standard deviation than the predicted value.

As noted above, this kind of error is indicative of an unstable rotor error reported by error classification element 54 of system 30.

The flowcharts in 7A-7B summarize the steps for implementing System 30 once Model 44 has been developed.

More specifically, computer 42 loads model 44 into memory at selected intervals (step 62), and displays information on the display of computer 42 for the user (step 64).

Upon being instructed to initiate monitoring of motor 10, at predetermined intervals or continuously, system 30 begins to obtain data from sensors 34-38 (steps 66 and 68). Data acquisition continues at a rate that the user can determine. Computer 42 calculates the residual value r (k) and this value is then compared with the predicted residual value developed by model 44 (step 72). If the residual value is within the threshold range, the motor will run without error and this information is displayed to the user via the display of computer 42 (step 74). However, if an error appears, this information is shown on the display in step 76.

Once an error is detected, system 30 evaluates the error and provides diagnostic information to the user. By using the predictability of the present invention, it is possible to avoid unexpected disasters financially. As shown in FIG. 7B, the diagnostic observer site of model 44 evaluates physical parameters such as current i and speed w of motor 10 in step 78 and compares these parameters with corresponding parameters of model 44 (see Table 2). . Based on the comparison, the system 30 may classify and indicate a mechanical basis for degradation or failure of motor performance as shown in step 82. Model 44 excludes the need to develop information about the motor.

The algorithm performed by the computer 42 is called Motor Condition Monitor (MCM) in FIGS. 7A and 7B. The basic concept of monitoring motor status is to observe parameter deviations intermittently or continuously, referring to the same parameters evaluated when the motor is known to operate satisfactorily, for example when the motor is known to be operating without defects. The output deviation is obtained while the motor is running continuously and the deviation is then compared with a predetermined threshold. If the deviation exceeds that threshold it is detected as an error. This error is classified by evaluating the parameters of the diagnostic model and comparing the parameters back to their initial values using the appropriate thresholds for these parameters.

For electric motors, it is possible to develop models that cover a range of manufacturing process variations, rather than using parameters from a single motor as described above in describing MCM systems and methods.

This concept is used during the manufacturing process to develop methods for detecting and diagnosing mechanical faults in electric motors as part of the testing process, especially for the quality verification process performed by most manufacturers immediately before shipping the motor.

A method and algorithm called Motor Quality Monitor (MQM) will be described using the method of the present invention for the application of quality assurance.

The basic function of the MQM algorithm is to test the electric motor, display the test results, control the test results (ie, the base model development will be described later) and store the measured and digitized data in memory for record keeping. It is.

Since there is no reliable technique or measurement to identify fault-free motors, the first method of obtaining a model of a typical fault-free motor ("base model") was developed.

A more detailed description of the MQM method is shown in FIGS. 8A-8F. The MQM method includes two basic functions: (1) development of a base motor model and (2) ongoing quality assurance testing of fractional horsepower electric motors.

The user can select any function from a menu displayed on the display of the computer 42. In a preferred embodiment, a "user difined" parameter is entered, such as, for example, the threshold limit and the number of motors to be tested before the user selects one of the following three options.

① Base Motor Model Development

② Select a Base Motor Model, or

③ Quality Assurance Test

If the base motor is not available and the step 90, "Base motor model development" option (step 92) needs to be selected first, where the user is asked to supply the information shown in Table 3 if it differs from the default setting. Step 94.

The choice of the "base motor model development" option is mandatory when the MQM is first installed. The user has the option of developing a base motor for electric motors of different tolerances or even electric motors of the same kind with multipliers of different tolerances. The model of the motor, its parameters and their standard deviations are obtained and stored in the specified data file.

Base motor models are mostly developed from a family of motors known to contain error-free motors. In a preferred embodiment of the present invention, the data obtained from the electric motor group is used to develop a base motor model. As is well known to those skilled in the art, this group of motors may include error-free motors, as well as several faulty motors due to the inefficiencies inherent in manufacturing and testing.

An experimental model of the selected motor type is developed using the EMT software program, which represents the characteristics of the selected motor type (steps 98-100).

In step 102, the model is evaluated for apparent modeling and critical nozzles (steps 102-104). Using a base motor model developed from the group, each motor in each group is then tested against an experimental base motor model using tolerance values obtained from the planned standard deviation of the SEE (step 106).

If the output of any of the motors in the motor group deviates from the output of the experimental model by more than each tolerance value, the motor is removed from the motor group and the data file is trimmed to remove erroneous data (steps 108-112). ).

Furthermore, refinement of the base motor model is then performed using the test data for the motors remaining in the motor group.

After removing all motors that deviate from the tolerance values set by the experimental model, in the modeling of that motor group again, the mean and standard deviations are returned until the motor group includes only those motors whose output is within the tolerance factor selected for the experimental model. It is possible to evaluate and refine the experimental model again (step 114).

After repeating this iteration, the experimental model will show the characteristics of a faultless motor manufactured to the same specification. This experimental model is stored as a base motor model in a database stored in the memory of computer 42 for the following reference (step 116).

If the base motor model already exists, the procedure can be reduced by simply reloading the base motor model into the active memory of the computer 42. The user selects the option "Select base motor model" and then selects "Quality Assurance Test". Start execution.

For example, the base motor model may correspond to a universal, shaded pole induction motor, synchronous motor or other small electric motor.

Referring again to FIG. 8A, if the "Select Base Motor Model" option is selected or the "Quality Assurance Test" option is selected, the appropriate base motor model for the motor under test is loaded into the computer memory, The test begins (step 120).

The user can then make adjustments to the tolerance multiplier for error detection and classification (steps 122 and 124).

The MQM algorithm then calculates the appropriate error detection and error classification thresholds (steps 126-128).

Fig. 8B shows the measurement part of the MQM algorithm, where the measured value of the motor output is compared with the output obtained from the base motor model using the threshold selected during the electric motor test during the manufacturing process for quality assurance purposes.

The threshold is determined by multiplying the tolerance value used to develop the experimental base motor by tolerance multipliers.

The MQM algorithm allows a multiplier to be determined by a quality assurance engineer who takes into account the allowable output deviation of the motor due to normal manufacturing deviations.

If the deviation exceeds a preset threshold, the motor tested is defined as having an error.

In detail, once the base motor model is selected, the user inputs the parameters necessary to perform the "quality assurance test" in steps 130 to 134 as summarized in Table 4 below.

When the "quality calibration test" is performed, the algorithm calculates the error detection and classification limits according to the selected motor type and the appropriate tolerance multiplier. The algorithm initiates data acquisition to obtain real time voltage, speed and current signals from the motor under test (step 134).

These signals are counted using the scan speed and scan time values previously input (steps 130 to 132). The digitized signal is stored in memory (step 136) and preprocessed to remove noise using a Butterworth software filter or a commercially available filter (step 140).

Real-time voltage, speed and current signals are used by the base model motor to determine the modeling state of the motor under its current state (steps 142 and 144). As shown in step 146, the base model motor estimated residual value and the actual residual value of the motor under test are calculated and compared in step 148. The deviation of the calculated residual value is then compared with the error detection threshold. If the output deviation of the motor being tested is within tolerance limits, the motor is recognized as a fault-free motor and displayed in a message (step 150).

If the motor is detected to be an error, a message is displayed (step 152) and error classification is achieved using the diagnostic model in a manner similar to that described above as shown in step 154. In summary, theoretically derived equations (6) and (7) describing an electric motor are used as diagnostic models.

The physical parameters of the diagnostic model are determined experimentally from the data obtained from the motor groups mentioned above. The physical parameters of the diagnostic model and their associated standard deviations are stored in the computer 42's memory. As motor faults are detected, the physical parameters of the faulty motor are evaluated by the MQM algorithm and compared with the corresponding parameters of the base motor model (steps 156-162). This comparison is used to classify motor faults and display diagnostic information.

If the deviation of the residual value is higher than the threshold, the motor status is classified as "FAULT FOUND" or a similar phrase is displayed on the display of computer 42. Once recognized, the physical parameters of the faulty motor are evaluated. These parameters are compared with the physical parameters of the base motor model using the classification thresholds (see Table 4).

For universal electric motors, the physical parameters are expressed in inductance, resistance and friction coefficient as shown in equations (5) and (6); And the motor constant.

Each of the parameters obtained from the faulty motor is compared with the fault classification threshold mentioned above. Steps 164-170 illustrate representative examples of one possible decision tree for classifying errors.

For example, if the inductance parameter of a faulty motor is greater than the error classification threshold for inductance, the decision is marked as "CHECK BALANCE."

If the resistance parameter of the faulty motor exceeds the fault classification threshold for the resistance, the decision is marked as "CHECK COLLECTOR".

If the frictional and inductance parameters of the faulty motor exceed the fault classification threshold, the determination is marked as "CHECK BEARING."

If more than one threshold is exceeded at the same time, all results are displayed.

If all parameters are less than the corresponding threshold, the decision is shown on the display as "UNCLASSIFIED." This may occur due to the cumulative effect of the change of each parameter on the motor output.

In such situations, the model may have small errors that can accumulate multiple or cause the model output to exceed the threshold.

However, since the threshold value is selected by the user, it is possible to enhance the tolerance value for each parameter so that such a limit error can be detected.

The MQM method is particularly suitable for use for fault diagnosis and preventive maintenance in electric motor repair shops. In this application, base motor models for various electric motors of different sizes and manufacturers are stored on computer 42. Thus, when a faulty motor is connected, the repairman selects the base motor model of the motor being tested and performs error detection and diagnosis.

The method and apparatus may also be used for condition monitoring and preventive maintenance. In this embodiment, the third embodiment, the MQM algorithm is used instead of the MCM algorithm for intermittent or continuous condition monitoring application.

In another embodiment of the present invention, the MQM and MCM algorithms are used in conjunction with existing quality assurance or condition monitoring systems that already have data acquisition capabilities for measuring voltage, speed and current.

In conclusion, the MCM algorithm and the MQM algorithm are very similar or each of them is different in two respects.

First, in the MCM algorithm, the system does not develop a base motor model. This is due to the nature of condition monitoring, which involves the system monitoring only one motor. For this reason, the MCM method benefits from the customary model of the motor being monitored. The customary model is developed when it is known that the motor is running without error.

MQM, on the other hand, develops a base model that includes variables related to large populations. Thus it is possible for a marginally operating monitor to pass the test threshold set in the MQM model, but since the MCM model is effective for individual motors, subsequent degradation cannot be detected by the MCM.

Secondly, MCM is inevitably constrained by operational requirements. For example, the input signal applied to the motor depends on the requirements imposed by that application. It will be solved that the input applied to model 44 cannot be "rich" as the input signal can be applied during the MQM test. Furthermore, under the MCM test, the actual load applied to the motor is unknown and can vary during the time periods in which measurements are obtained from sensors 34-38. Under these conditions, only model parts that are not affected by the load are modeled.

For example, the current signal may be modeled using Equation (6) using the measured voltage and velocity input signals to obtain a result using the diagnostic observer. In alternative embodiments, unknown load items can be excluded using techniques such as taking the derivative of equation (7) for constant load. In such an embodiment, the derivation of equations (6) and (7) may be combined to enhance the results obtained by the diagnostic observer.

Although described by way of example embodiments, these are merely illustrative of the present invention and by no means limit the scope of the present invention. Furthermore, those skilled in the art will appreciate that modifications may be made without departing from the scope of the present invention as set forth above.

Claims (28)

In the error detection system for monitoring the operating state of the motor operating with an unknown load, The system, A sensor coupled to the motor for measuring a selected operating parameter; And Connected to the sensor, Determining an ideal residual value of 0 when the motor is operating in error free, selecting a non-zero threshold tolerance level, the ideal residual value is multiplied by the selected operating parameter multiplied by an invariant value and summed up the multiplication result Computer means for determining a plurality of residual values of the motor during operation; It is configured to include, The computer means compares each of the plurality of residual values with the ideal residual value in a memory, and if the plurality of residual values is less than a threshold tolerance, the motor operates without error or at least one of the plurality of residual values. Includes a memory and a display, where the display indicates a message indicating that the motor is operating in error if the threshold tolerance is exceeded. Motor error detection system, characterized in that. The system of claim 1, wherein the operating parameter includes an applied voltage, an output current and a speed of the motor, and the operating parameter is measured by an analog sensor. The system of claim 1, wherein the operating parameter is measured with an analog sensor. 4. The system of claim 3, wherein the system further comprises data acquiring means for coupling the sensor to the computer means and converting the analog signal into a digital representation of the analog signal. 3. The system of claim 2, wherein the motor is an electric motor. 3. The system of claim 2, wherein the motor is a fractional horsepower motor. A method of monitoring the operation of an electric motor to detect mechanical faults that cause a motor fault before the motor actually shuts down, the method comprising: Developing a model of the motor in a computer connected to the motor by a plurality of sensors; Measuring a plurality of actuation signals of the motor as the sensor; Applying the measured plurality of actuation signals to solve a linear discrete-time state equation; Comparing a solution presented by the model by calculating a solution and a residual value of the state equation; Determining whether the motor is operating without error according to the comparison result; Informing the presence of the fault to associate changes to mechanical faults and prevent unexpected motor stops when the motor is operating with faults; And At predetermined intervals during operation of the motor, Repeating the above steps except developing the model; Monitoring method comprising a. The method of claim 7, wherein the measuring of the plurality of operation signals measures a current output of the motor, a voltage applied to the motor, and a speed of the motor for a predetermined interval. 8. The method of claim 7, wherein the motor is a fractional horsepower electric motor. 9. The method of claim 8, wherein the step of developing a model of the motor includes obtaining a motor invariance of inductance and resistance (resistance) of the motor and combining the invariant and the measured signal according to the following equation. How to. L di / dt + Ri = V + k1 wi K1 is the motor constant. 9. The method of claim 8, wherein the step of associating and informing the presence of a mechanical defect in the motor further comprises: Pointing to an unbalanced rotor in response to a change in the L di / dt operating parameter; Indicating a collector error in response to a change in the Ri parameter; Indicating a bearing defect in response to the oscillation change of L di / dt; And Indicating a bearing defect in response to a change in both the L di / dt and the fw parameters; Method comprising a. 9. The method of claim 8, wherein the interval is preferably in the range of 400 milliseconds to 1000 milliseconds. 13. The method of claim 12, wherein the operating parameter is sampled at a sampling frequency in the range of 500 Hz to 24 kHz. Measuring the voltage, current and speed of the electric motor with a plurality of sensors when the electric motor is operating without error, Multiplying the measured voltage, current and speed of the electric motor by a constant variable, Calculating and holding the result of the discrete state space equation; x (k + 1) = Ax (k) + Bu (k) y (k) = C x (k) Repeating multiplying the measurement; Discrete State-Space Equation xf (k + 1) = Af xf (k) + Bf uf (k) yf (k) = Cf x (k) Calculating a result of; comparing the difference between y (k) and yf (k); And Repeating the iteration, calculation and comparison pure until the difference exceeds a selected threshold; Defect monitoring and detection method of the electric motor comprising a. 15. The method of claim 14, further comprising the step of calculating a difference value that exceeds the selected threshold: Selecting parameter thresholds for inductance, motor resistance, motor inertia and motor constants; And Comparing each enemy L di / dt, Ri, J dw / dt, fw, k1wi and i2 k2 with the corresponding selected threshold; Method comprising the 16. The method of claim 15, further comprising displaying a result of said comparing step. In the method of inspecting the manufacturing quality of the motor group and detecting mechanical defects that may cause the error of the electric motor, The method, Selecting a motor group comprising a motor operating without error and a motor operating with one or more unknown errors; Measuring a plurality of actuation signals of the motor and applying the actuation signals to solve the next discrete state space equation; x (k + 1) = Ax (k) + Bu (k) y (k) = C x (k) Developing, on computer means, an experimental model having a critical tolerance based on two standard deviations of the mean of the motor group model; Re-measuring the plurality of operating signals to test each motor from the motor group and removing the motor from the motor group if the equation of the motor under test exceeds a threshold; Repeating the measuring, developing, and testing steps until all motors in the motor group are within a threshold; The motors refine an experimental model based on the motors remaining in the motor group; And Storing the experimental model in the computer means. Method comprising a. 18. The method of claim 17, wherein the testing step further comprises: measuring voltage, current, and speed of the electric motor with a plurality of sensors; Multiplying the measured voltage, current and speed of the electric motor by a selected constant value; And Discrete Space Equation xf (k + 1) = Af xf (k) + Bf uf (k) yf (k) = Cf x (k) Calculating a result of and comparing the difference between y (k) and yf (k); Method comprising a. 18. The method of claim 17, further comprising the steps of: measuring the voltage, current, and speed for each of the plurality of motors when the plurality of motors different from the motor group are tested against the experimental model; Applying the measured plurality of actuation signals to solve the discrete state space equation; Comparing the solution proposed by the experimental model with the solution of the state equation by comparing the residual value; And Determining whether the motor is operating without a detected error based on the comparing step; Method comprising a. The method of claim 19 further comprising Informing the presence of the error in order to correlate the change with a mechanical fault and prevent unexpected motor failure when the motor is operating with the detected error; Method comprising the The method of claim 20 wherein said associating Evaluating the following equation. L di / dt + Ri = V + k1 wi J dw / dt + fw = k2 i2 + M In the model-based error detection and diagnosis system for detecting an error in the motor group and developing diagnostic information for correcting the error, Means for generating a system model representing an error-free motor according to the following formula; And Means for determining a system error by measuring a parameter of the motor while comparing with the system model, And the determining means compares an equation representing each motor with an equation representing the system model to determine an operating state of the motor. x (k + 1) = Ax (k) + Bu (k) and y (k) = Cx (k) In which A, B and C represent parameters of the system model. The method of claim 22, wherein the determining means, Compare the equations representing each of the motors with the equations representing the system model in memory, Each of the motors is operating in an error free state where the difference between the equation representing each motor and the equation representing the system model is less than the selected threshold tolerance, or To display on the display device a message indicating whether a failure of each of the motors is operating in an impending state when the difference between the equation representing each of the motors and the equation representing the system model is greater than a critical tolerance, And computer means having a memory and a display. 23. The system of claim 22, wherein the motor is a fractional horsepower electric motor. In a method of testing a plurality of motors of a conventional type that do not know the operating state, The method, Measuring the voltage, current and speed of the motor; Multiplying the measured voltage, current and speed of the electric motor by a selected constant value; Discrete Spatial Equations x (k + 1) = Ax (k) + Bu (k) calculating and holding the result of y (k) = Cx (k); Repeating the measuring and multiplying steps; Discrete Space Equation xf (k + 1) = Afxf (k) + Bf uf (k) calculating a result of yf (k) = Cfx (k); comparing between y (k) and yf (k); And Repeating the repeating, calculating and comparing steps until the difference exceeds a selected threshold; Method comprising a. 26. The method of claim 25, further comprising selecting a parameter threshold for inductance, motor resistance, when the result of the step results in a difference that exceeds the selected threshold. 27. The method of claim 26, further comprising displaying a result of said comparing step. 27. The method of claim 26, wherein said comparing step further Indicating an unbalanced rotor in response to a change in the L di / dt operating parameter; Indicating a collector error in response to the oscillation change of the R ₁ parameter; Indicating a bearing error in response to the oscillation change of the L di / dt parameter; And Indicating a bearing error in response to a change in both the L di / dt and the fw parameters; Method comprising the