(19)
(11)EP 3 260 944 B1

(12)EUROPEAN PATENT SPECIFICATION

(45)Mention of the grant of the patent:
04.11.2020 Bulletin 2020/45

(21)Application number: 17176585.2

(22)Date of filing:  19.06.2017
(51)International Patent Classification (IPC): 
G05B 23/02(2006.01)

(54)

DEVICE AND METHOD FOR MACHINE MONITORING

VORRICHTUNG UND VERFAHREN ZUR MASCHINENÜBERWACHUNG

DISPOSITIF ET PROCÉDÉ DE SURVEILLANCE DE MACHINES


(84)Designated Contracting States:
AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR

(30)Priority: 21.06.2016 US 201615188137

(43)Date of publication of application:
27.12.2017 Bulletin 2017/52

(73)Proprietor: General Electric Company
Schenectady, NY 12345 (US)

(72)Inventor:
  • ANTA MARTINEZ, Adolfo
    Niskayuna, NY New York 12309 (US)

(74)Representative: BRP Renaud & Partner mbB Rechtsanwälte Patentanwälte Steuerberater et al
Königstraße 28
70173 Stuttgart
70173 Stuttgart (DE)


(56)References cited: : 
EP-A1- 2 690 513
US-A1- 2012 029 892
EP-A2- 2 149 824
  
      
    Note: Within nine months from the publication of the mention of the grant of the European patent, any person may give notice to the European Patent Office of opposition to the European patent granted. Notice of opposition shall be filed in a written reasoned statement. It shall not be deemed to have been filed until the opposition fee has been paid. (Art. 99(1) European Patent Convention).


    Description


    [0001] This invention relates to devices useful for monitoring machine performance. In particular, the present invention relates to machine monitoring devices which rely in part on prediction of machine responses to changes in machine operating conditions.

    BACKGROUND



    [0002] Equipment monitoring devices are of fundamental importance to modern engineering. Increasingly, data taken from machines in operation is compared with models generated from earlier data taken from the same or similar machines to assure compliance with design specifications and to predict future performance. In large, complex machines such predictive models tend to be based on data gathered through painstaking experimentation during which machine operating conditions are systematically varied while machine outputs are recorded. Particularly when a machine is initially deployed, there may be little or no historical data available with which to predict machine responses to changes in machine operating conditions. A significant amount of historical data may be required in order to establish models useful in machine monitoring devices.

    [0003] There is a need for machine monitoring devices which are capable of establishing useful predictive models with limited data, and which can alert operators to potential deviations from acceptable machine performance based on outputs of the predictive models.

    [0004] US 2012/0029892 A1 describes a method for monitoring a wind turbine whereby a subsystem of the wind turbine is defined and a simulation model for that subsystem is provided. During normal operation of the wind turbine an input parameter of the subsystem is determined. A behavior of the subsystem is simulated using the input parameter as an input of the simulation model. Based on the simulated behavior, it is determined if the subsystem operates within a given specification.

    [0005] EP 2 149 824 A2 describes an estimation system for estimating an operating parameter. The estimation system includes a primary model configured to receive at least a sensor input and to estimate the operating parameter using at least the sensor input, and a redundancy system configured to receive the estimated operating parameter and to determine whether an anomaly is present within the estimated operating parameter.

    [0006] EP 2 690 513 A1 describes providing a dynamic model of a machine or a portion of a machine. The dynamic model is provided with model parameters, such as mass inertia, spring stiffness, or attenuation values. The frequency analysis for the machine or the portion of the machine is performed. The new value for model parameter from the frequency analysis is determined. The new value of the model parameter is compared with an initial value of the model parameter. The state of the machine or the portion of machine is detected from the comparison result, for targeted maintenance.

    BRIEF DESCRIPTION



    [0007] The present invention is defined in the accompanying claims.

    [0008] In one embodiment, the present invention provides a machine monitoring device, the device comprising: (a) a communications link to interrogate a machine with a probe signal and receive one or more measured machine operating condition outputs; and (b) a device controller for: (i) selecting a machine operating condition input variable for which a corresponding machine operating condition output is unknown; (ii) applying a predictive model in which the machine operating condition input variable serves as an argument of a predicted machine operating condition output; (iii) updating a library of predicted machine operating condition outputs; and (iv) alerting a human operator if a measured or predicted machine operating condition output is at variance with a predetermined threshold machine operating condition output; wherein the predictive model comprises at least two independent primary models, for each of which primary models at least one correspondence between a primary model machine operating condition and a corresponding primary model machine output are known; the primary models sharing a single common basis within the predictive model.

    [0009] In an alternate embodiment, the present invention provides a system comprising a machine monitoring device operationally coupled to a machine, the machine monitoring device comprising: (a) a communications link to interrogate the machine with a probe signal and receive one or more measured machine operating condition outputs; and (b) a device controller for: (i) selecting a machine operating condition input variable for which a corresponding machine operating condition output is unknown; (ii) applying a predictive model in which the machine operating condition input variable serves as an argument of a predicted machine operating condition output; (iii) updating a library of predicted machine operating condition outputs; and (iv) alerting a human operator if a measured or predicted machine operating condition output is at variance with a predetermined threshold machine operating condition output; wherein the predictive model comprises at least two independent primary models, for each of which primary models at least one correspondence between a primary model machine operating condition and a corresponding primary model machine output are known; the primary models sharing a single common basis within the predictive model.

    [0010] In yet another embodiment, the present invention provides a method of monitoring a machine with a machine monitoring device, the method comprising: (a) interrogating a targeted machine component with a probe signal via a communications link and receiving one or more measured machine operating condition outputs; and (b) selecting a targeted machine component operating condition input variable for which a corresponding targeted machine component operating condition output is unknown; (c) applying a predictive model in which the targeted machine component operating condition input variable serves as an argument of a predicted targeted machine component operating condition output; (d) updating a library of predicted targeted machine component operating condition outputs; and (e) alerting a human operator if a measured or predicted targeted machine component operating condition output exceeds a predetermined threshold targeted machine component operating condition output; wherein the predictive model comprises at least two independent primary models, for each of which primary models at least one correspondence between a primary model targeted machine component operating condition and a corresponding primary model targeted machine component output are known; the primary models sharing a single common basis within the predictive model.

    BRIEF DESCRIPTION OF THE DRAWING FIGURES



    [0011] Various features, aspects, and advantages of the present invention will become better understood when the following detailed description is read with reference to the accompanying drawings in which like characters may represent like parts throughout the drawings. Unless otherwise indicated, the drawings provided herein are meant to illustrate key inventive features of the invention. These key inventive features are believed to be applicable in a wide variety of systems which comprising one or more embodiments of the invention. As such, the drawings are not meant to include all conventional features known by those of ordinary skill in the art to be required for the practice of the invention.

    Fig. 1 illustrates a machine monitoring device provided by the present invention.

    Fig. 2 illustrates a system provided by the present invention.

    Fig. 3 illustrates a method provided by the present invention.

    Fig. 4 illustrates frequency responses of a machine to a probe signal according to one or more embodiments of the present invention.

    Fig. 5 illustrates one or more aspects of the present invention.

    Fig. 6 illustrates one or more aspects of the present invention.

    Fig. 7 illustrates frequency responses of a machine to a probe signal under two different machine operating conditions.

    Fig. 8 illustrates a graphical representation of a predictive model according to one or more embodiments of the present invention.


    DETAILED DESCRIPTION



    [0012] In the following specification and the claims, which follow, reference will be made to a number of terms, which shall be defined to have the following meanings.

    [0013] The singular forms "a", "an", and "the" include plural referents unless the context clearly dictates otherwise.

    [0014] "Optional" or "optionally" means that the subsequently described event or circumstance may or may not occur, and that the description includes instances where the event occurs and instances where it does not.

    [0015] Approximating language, as used herein throughout the specification and claims, may be applied to modify any quantitative representation that could permissibly vary without resulting in a change in the basic function to which it is related. Accordingly, a value modified by a term or terms, such as "about" and "substantially", are not to be limited to the precise value specified. In at least some instances, the approximating language may correspond to the precision of an instrument for measuring the value. Here and throughout the specification and claims, range limitations may be combined and/or interchanged, such ranges are identified and include all the sub-ranges contained therein unless context or language indicates otherwise.

    [0016] In one or more embodiments the present invention provides a machine monitoring device useful for evaluating machine performance. The monitoring device is capable of interrogating the machine with one or more probe signals and receiving one or more measured machine operating condition outputs. In one or more embodiments, the probe signal interrogates the machine and generates a machine frequency response function (FRF) for a given machine operating condition, such as a given machine rotor speed. In addition, one or more sensors deployed in and/or around the machine may provide machine operating condition output data such as vibrational level, temperature, pressure, and displacement. Such machine operating condition output data may in turn be sent to the machine monitoring device controller via the communications link. In one or more embodiments, changes in the probe signal appear as changes in the frequency of onset of probe signal attenuation and/or amplification in response to a change in a machine operating condition. These changes are the result of machine-probe signal interactions and may be measured and correlated with one or more machine operating condition outputs. The extent to which the probe signal is altered, and the nature of such alteration may be a function of the state of the machine during interrogation by the probe signal. Thus, the machine may be in a stationary or an active operational (non-stationary) state. In either case, the probe signal may be sensitive to external parameters such as ambient temperature, ambient humidity, barometric pressure which also constitute elements of the machine state. Primarily, however, the probe signal will be sensitive to machine operational parameters such as machine component rotational speed, machine component temperature, machine component displacement, and the like.

    [0017] Those of ordinary skill in the art will appreciate that a variety of probe signals are available for use in machine interrogation. These include electric signals, optical signals, acoustic signals, radiofrequency signals, and combinations thereof. In one or more embodiments, the probe signal is a high frequency electrical signal. In one or more alternate embodiments, the probe signal is an optical signal. In one or more embodiments, the probe signal is an orthogonal random phase multi-sine electrical signal such as are known in the art.

    [0018] As noted, the machine monitoring device comprises a communications link which permits the device to interrogate one or more machine components with a probe signal, to receive measured machine operating condition outputs, and to detect changes in probe signal characteristics such as its attenuation as a function of frequency under different machine operating conditions. Machine operating condition outputs such as machine component vibrational level, machine component temperature and pressure will be dependent on the state of the machine at the time of its interrogation. It is useful to think of the machine state in terms of individual parameters such as rotational speed, or sets of parameters such as ambient temperature and rotational speed. Such individual parameters and sets of parameters are at times herein referred to as machine operating condition input variables. It follows that machine operating condition outputs reflect a machine state characterized at least in part by the individual parameter or set of parameters being considered.

    [0019] In one or more embodiments, the communications link is configured such that a probe signal having known voltage and frequency characteristics may be injected into a machine power supply at a frequency greater than the fundamental electrical frequency (and its low order harmonics) powering the machine. This creates a high frequency current signal which may be measured as part of a voltage and current sensor interfaced with the power line. Such signal injection components are described in detail in co-pending United States patent application serial number 14/663691.

    [0020] In one aspect, an advantage offered by the machine monitoring device provided by the present invention lies in its ability to predict the machine operating condition output for which no correlation yet exists with a corresponding machine operating condition input variable. For example, a machine operating condition output such as the vibrational state of a machine component of an untested machine state may be predicted in advance of testing at such machine state, or in many cases in lieu of such testing. The predictive power of the machine monitoring device thus enables the more rapid deployment and commissioning of complex machines such as compressors and turbines, and can also alert operators to potential problems when a measured or predicted machine operating condition output is at variance with a predetermined threshold machine operating condition output.

    [0021] According to the invention, a device controller is programmed to select a machine operating condition input variable, for example a rotor speed, for which a corresponding machine operating condition output is unknown. The controller applies a predictive model of the machine's performance wherein the input variable, for example rotor speed, serves as an argument of a predicted machine operating condition output, for example the expected vibrational level of a particular machine component at the given rotor speed. The controller then updates a library of predicted machine operating condition outputs. The controller is similarly configured to update a library of measured machine operating condition outputs which may be used to update the predictive model. Finally, the device controller is programmed to alert a human operator if a measured or predicted machine operating condition output is at variance with a predetermined threshold machine operating condition output.

    [0022] The predictive model, described in detail herein, relies upon at least two independent primary models. Each of the primary models includes at least one known correspondence between a primary model machine operating condition and a corresponding primary model machine output. Such correspondences are illustrated by frequency response plots 401 and 402 shown in Fig. 4, at times herein referred to as frequency response functions (FRFs). FRFs 401 and 402 were experimentally determined on a rotary test rig being interrogated by a complex electrical probe signal of the orthogonal random phase multi-sine type superimposed onto the test rig power supply. Within the predictive model, the primary models share a common basis. The primary models themselves may or may not initially share a common basis. Where the primary models do not share a common basis, conversion to a common basis as components of the predictive model is required.

    [0023] In one or more embodiments, primary model machine outputs and the predicted machine operating condition output include one or more of a machine temperature, a machine pressure, a machine vibrational characteristic, a machine rotational speed, a machine translational speed, a machine acceleration, a machine force, a machine torque, a machine power input variable, time of operation, machine age, a machine weight characteristic, a machine geometry characteristic, and a characteristic of matter being processed by the machine.

    [0024] The predictive model is created by first creating and combining at least two independent primary models. The independent primary models can be created using known modeling tools such as MATLAB using a limited number of correspondences between a machine operating condition and a corresponding machine output.

    [0025] Turning now to the figures, Fig. 1 illustrates a machine monitoring device 10 provided by the present invention. In the embodiment shown, a communications link 20 transmits a probe signal 22 to a machine 100 being monitored at a limited number of machine operating conditions and receives data 23 from the machine. In one or more embodiments, the data is detected as the attenuation and/or amplification 23a of the probe signal as a function of frequency which is used to generate a frequency response function (FRF) for a given machine operating condition. In addition, machine sensor data 23b is received through the communications link during this limited operation of the machine. Thus, during a limited set of machine operating conditions designated input variables, for example rotor speeds X1, X2, and X3, and the corresponding machine operating condition outputs Y1, Y2, and Y3 are recorded by the machine monitoring device. Again, these operating condition outputs include both the machine frequency response behavior 23a and machine sensor data 23b. The limited number of such X to Y correspondences are used in combination with the predictive model discussed above to predict machine operating condition outputs (Yn+1, Yn, Yn-1 ...) for which such correspondences with machine operating input variables (Xn+1, Xn , Xn-1 ...) are unknown. In the embodiment shown, device controller 30 has access to a range of machine operating inputs X1- Xn+1. The device controller selects a machine operating input variable Xn 32 for which a corresponding machine operating condition output Yn 34 is unknown and predicts such output. A library 50 containing predicted machine operating condition outputs is then updated. The machine monitoring device alerts a human operator via alerting unit 40 if the machine operating condition output Yn 34 is at variance with a predetermined threshold machine operating condition output. According to the invention this is carried out by comparing the machine operating condition output 34 with a predetermined acceptable range of machine operating condition outputs. In one or more embodiments, the alerting unit 40 comprises a control center computer, a portable electronic device, an optical warning device, an audible warning device or a combination of two or more of the foregoing.

    [0026] In one or more embodiments, the machine operating condition input variable Xn is extrapolative relative to at least two X to Y correspondences between the primary model machine operating conditions and the corresponding primary model machine outputs. In one or more alternate embodiments, the machine operating condition input variable Xn is interpolative relative to at least two X to Y correspondences between the primary model machine operating conditions and the corresponding primary model machine outputs.

    [0027] Referring to Fig. 2, the figure represents a system 200 provided by the present invention, the system comprising a machine 100 powered by a variable frequency drive 210, the machine being monitored by machine monitoring device 10, machine monitoring device 10 comprising as subunits, communications link 20 and device controller 30. In the embodiment shown, the machine monitoring device 10 is operationally coupled to the machine 100 via a power transmission cable 222 linking the machine to the associated variable frequency drive 210. The machine monitoring device 10 is configured to interface with the machine 100 via electrical interface 28a and the variable frequency drive 210 via electrical interface 28b. As such, the machine monitoring device 10 also serves as the electrical connection between the variable frequency drive and allied components on the one hand, and the machine driven by the variable frequency drive on the other. Machine monitoring device 10 functions essentially as described in the discussion of Fig. 1 and elsewhere herein. In the embodiment shown, a high frequency electric probe signal 22 interrogates the machine 100 via power transmission cable 222 via a capacitive or inductive coupling unit 25 attached to the power transmission cable. The probe signal 22 is generated by processor-microcontroller unit 38 working in tandem with switching network 27 to inject the probe signal directly into the power transmission cable. Blocking filter 26 inhibits the transmission of the probe signal into other components of the system, such as the variable frequency drive. Attenuation and/or amplification of probe signal 23a and machine sensor signals 23b are detected by voltage current sensor 24 and are transmitted to processor-microcontroller unit 38 via analog to digital converter 36. As discussed in reference to Fig. 1, a library, 50, containing measured and predicted machine operating condition outputs is updated during the course of operation of the machine. The machine monitoring device alerts a human operator via alerting unit 40 if one or more measured or predicted machine operating condition outputs is at variance with a predetermined threshold machine operating condition output.

    [0028] Referring to Fig. 3, the figure illustrates a method 300 of monitoring a machine according to one or more embodiments of the present invention. In a first method step 301, a targeted machine component, for example an active magnetic bearing, is interrogated with a probe signal via a communications link. Such interrogation comprises operating the machine at a first operating condition and measuring the alteration of the probe signal resulting at least in part by the interaction of the probe signal with the targeted machine component, and recording the correspondence between the alteration of the probe signal with the first operating condition. A given machine operating condition may at times herein be referred to as a machine component operating condition input variable. Similarly, the alteration of the probe signal at the given machine operating condition may at times herein be referred to as the machine component operating condition output. In addition to recording the frequency response behavior of the probe signal, one or more machine component attributes, for example rotor displacement within the targeted magnetic bearing is sensed and recorded via the communications link during the operation of the machine at the first operating condition. Such machine component attributes also constitute machine operating condition outputs. In a second method step, 302, a targeted machine component operating condition input variable, for example a rotor speed, is selected for which a corresponding targeted machine component operating condition output is unknown. In a third method step, 303, a predictive model is applied in which the targeted machine component operating condition input variable serves as an argument of a predicted targeted machine component operating condition output. The predictive model comprises at least two independent primary models, for each of which primary models at least one correspondence between a primary model targeted machine component operating condition and a corresponding primary model targeted machine component output are known; the primary models sharing or having been converted into the predictive model in which the primary models share a single common basis. In a fourth method step, 304, a library of predicted targeted machine component operating condition outputs is updated. In a fifth method step, 305, a human operator is alerted if a predicted targeted machine component operating condition output is at variance with a predetermined threshold targeted machine component operating condition output.

    [0029] Referring to Fig. 4, the figure represents a frequency response 400 of a machine to a probe signal according to one or more embodiments of the present invention and shows the magnitude 404 of the probe signal as a function of the frequency 403 of the probe signal. In the embodiment shown, a machine first frequency response function (FRF) 401 is obtained at a first rotor speed Ω1 and a second frequency response function 402 of the same machine is obtained at a second rotor speed Ω2. For example, the probe signal, at times herein referred to as an excitation signal, may be applied to one or more machine components such as a stator coil or a magnetic bearing at different rotor speeds. The illustrated frequency response functions 401 and 402 are obtained by applying the probe signal at rotor speed Ω1 and subsequently applying the same probe signal at rotor speed Ω2 and measuring the machine-induced alteration of the probe signal embodied in plots 401 and 402. In this example, rotor speeds Ω1 and Ω2 represent machine operating condition input variables and frequency responses 401 and 402 represent machine operating condition outputs and may serve as components, together with sensed machine attributes such as vibrational levels, of independent primary models with which to create a predictive model capable of predicting machine operating condition outputs in response to the probe signal under untested machine operating conditions, rotor speeds Ω3n.

    [0030] In one or more embodiments, the machine monitoring device computes the predictive model by constructing a common basis for the available primary models of machine behavior. In order to draw parallelisms and establish connections between the two (or more) available primary models, it is first necessary to ensure that the coordinates of both models coincide, that is, state vector x in the state space description corresponds to the same set of variables in both models. A common basis or coordinates set for all available models is then obtained. Since the at least two of the primary models were obtained from experimental data gathered under different machine operating conditions (e.g. different rotor speeds), there is no guarantee that the two state space representations use the same set of coordinates for each state. The inputs and outputs are clearly defined since they correspond to physical signals (e.g. vibrations and currents in the Example herein), but the same cannot be guaranteed for the states. In one or more embodiments, the machine monitoring device controller computes a common basis for the primary models by (i) choosing an input/output pair of the multiple-input multiple-output system, (ii) computing the poles and zeros of this single-input single-output (SISO) system and sorting these poles and zeros in the same order for all systems, and (iii) dividing this SISO system into a combination of first order and second order subsystems and applying a canonical transformation to each subsystem. This representation is unique provided the sorting in step is done properly. Once all available systems have been transformed, an interpolation or extrapolation can be done directly on the mathematical models.

    [0031] Additional guidance on methodologies used herein is available in Blom R. S., Delft, 2011, 244 pages, ISBN: 9789491104077 and J. De Caigny et al. Proceedings of ISMA2008.

    Example



    [0032] We consider here the deployment and commissioning of a machine whose behavior depends on the rotor speed having a nominal operating range of 0 to 1000rpm. In this example, a requirement for safe operation is that the vibrations in the machine do not surpass 3 centimeter per second squared (3cm/s2). The magnitude of machine vibrations increases as the rotor speed of the machine increases. In order to commission this machine, it is necessary to demonstrate that vibrations will not exceed the 3cm/s2 safety requirement. A comparative system 200 in normal operation is depicted in Fig. 5. wherein electric current I 102 is provided via cable 222 to power a machine 100 while sensor 104 senses and transmits the vibrational level 34 (Yi to Yn) of the machine at machine speeds X1 to Xn. Again, the vibration level of the machine depends on the rotor speed X (32) of the machine which in turn depends on the input current 102.

    [0033] The commissioning of the machine shown in Fig. 5 would be carried out as follows. First, the vibrational level of the machine at 0 rpm is measured. If the safe vibrational level is not exceeded, the rotor speed of the machine is increased to 100 rpm and the vibration level is again be measured. If the safe vibrational level is not exceeded, the rotor speed of the machine would be increased to 200 rpm and the vibration level would again be measured. This incremental approach would be repeated until the maximum nominal operating speed (1000 rpm) is reached.

    [0034] Such conventional commissioning procedures are inherently limited. First, it is time consuming to repeat measurements at incrementally increasing speeds across the desired operational range. Second, because the true vibrational behavior of the machine during operation is unknown until the machine is put into operation, critical threshold vibrational limits may be exceeded during initial testing. Under various circumstances, including unforeseen and undetected damage done to the machine during transport, there is a risk that a machine will not function properly even within its nominal range of operating conditions. Under such circumstances, as the speed is increased within the nominal operating range, the machine vibrational level could exceed a predetermined threshold operating vibrational level, 3cm/s2 in this example, and be damaged. In large high kinetic energy rotating machines such excursions can be catastrophic.

    [0035] The present invention, illustrated by the non-limiting embodiment shown in Fig. 6, provides solutions to these and other challenges. Thus, primary models of the machine in operation are obtained at 0 rpm and at 100 rpm, speeds at which the machine vibrations are expected to be very low and therefore safe operation is highly probable. The primary models either initially share a common basis or are converted into primary models sharing a common basis before being incorporated into a predictive model which can be used to obtain a prediction of the vibration levels for any speed in the range 0 rpm to 1000rpm.

    [0036] The present invention requires only measurements of two machine operating condition outputs; the frequency response function of a probe signal reference and the measured vibrational level at 0 rpm and 100 rpm. To obtain a predictive model in this example, the machine is probed with a probe signal R (22) (in this example an auxiliary current) to interrogate and study the machine response. The probe signal is injected into power supply cable 222 where it is added to machine input current 102. The present invention requires only one probe signal be used to record frequency response functions for the machine at two speeds 0 rpm and 100 rpm. In the embodiment shown, the attenuated and/or amplified probe signal 23b is detected at machine sensor 122 and relayed via a communications link 20 to a machine monitoring device controller 30. These frequency response functions Gi (701) and G2 (702) are shown in Fig. 7. Because one of the two speeds is 0 rpm, a single input current corresponding to a machine speed of 100 rpm may be used. At 0 rpm the same input current required to produce a machine speed of 100 rpm is applied, but the machine is mechanically locked in a stationary state. Machine vibration levels Y1 and Y2 (34) corresponding to machine speeds 0 rpm and 100 rpm are measured via a sensor 104 (Fig. 6). The vibration levels Y1 and Y2 are a function of the machine operating condition input variable (speed), the probe signal 22 magnitude (R) and the input current 102 (I). In this example the probe signal 22 is a sine wave and the input current 102 is held constant at two amperes (I(t) = 2 amperes). The magnitude R of probe signal 22 varies with time and can be expressed mathematically as



    [0037] The measured vibration levels also vary with time at constant speed. Vibrational data is collected from sensor 104. In stationary state (0 rpm) the vibrational level of the machine can be expressed



    [0038] Similarly, the measured vibration level when the machine is rotating at 100rpm may be expressed as



    [0039] Next, the frequency response functions Gi (701) and G2 (702) (Fig. 7) can be parametrized using art known parametrization techniques and expressed mathematically as

    having units of cm/s2/ampere and wherein s is the probe signal frequency. The predictive model for all machine speeds the can be expressed as a transfer function

    having units of cm/s2/ampere. The predictive model can be represented as well by the family of frequency response functions 800 shown in Fig. 8 wherein each curve illustrates machine vibration output at a given speed within the range 0 rpm to 100 rpm. Those of ordinary skill in the art will understand that, in the example just given Y1(t), Y2(t), G1(s) and G2(s) constitute the primary models used to generate the predictive model.


    Claims

    1. A machine monitoring device (10), the device comprising:

    (a) a communications link (20) to interrogate a machine (100) with a probe signal (22) and receive one or more measured machine operating condition outputs reflecting the machine state characterized by machine operating condition input variables representing an individual parameter or set of parameters being considered; and

    (b) a device controller (30) arranged for:

    (i) selecting (302) a machine operating condition input variable (Xn) for which a correlation with a corresponding machine operating condition output (Yn) is unknown;

    (ii) applying (303) a predictive model in which the machine operating condition input variable serves as an argument of a predicted machine operating condition output, wherein the predictive model comprises at least two independent primary models, the primary models being used to create the predictive model, for each of which primary models at least one correspondence between a primary model machine operating condition (X1, X2) and a corresponding primary model machine output (Y1, Y2) are known; the primary models either initially share a common basis or are converted into primary models sharing a common basis before being incorporated into the predictive model if the primary models did not share a common basis prior to being converted into the predictive model;

    (iii) updating (304) a library (50) of predicted machine operating condition outputs (Xn, Yn); and

    (iv) alerting (305) a human operator if one of the measured or the predicted machine operating condition output (34) is at variance with a predetermined threshold machine operating condition output by comparing the machine operating condition output (34) with a predetermined acceptable range of machine operating condition outputs.


     
    2. The machine monitoring device (10) according to claim 1, wherein the primary model machine outputs and the predicted machine operating condition output include one or more of a machine temperature, a machine pressure, a machine vibrational characteristic, a machine rotational speed, a machine translational speed, a machine acceleration, a machine force, a machine torque, a machine power input variable, time of operation, machine age, a machine weight characteristic, a machine geometry characteristic, a characteristic of matter being processed by the machine (100).
     
    3. The machine monitoring device (10) according to any preceding claim, wherein the primary model machine outputs and the predicted machine operating condition output include a machine component rotational speed.
     
    4. The machine monitoring device (10) according to any preceding claim, wherein the primary model machine outputs reflect the machine in non-stationary states.
     
    5. The machine monitoring device (10) according to any preceding claim, wherein the machine is selected from the group consisting of turbines, wind turbines, compressors, turboexpanders, motors, and pumps.
     
    6. The machine monitoring device (10) according to any preceding claim, wherein the probe signal (22) is a high frequency electrical signal.
     
    7. The machine monitoring device (10) according to any of claims 1 to 6, wherein the probe signal (22) is an optical signal.
     
    8. The machine monitoring device (10) according to any preceding claim, wherein the machine operating condition input variable is extrapolative relative to at least two correspondences between the primary model machine operating condition and the corresponding primary model machine output.
     
    9. The machine monitoring device (10) according to any preceding claim, wherein the machine operating condition input variable is interpolative relative to at least two correspondences between the primary model machine operating condition and the corresponding primary model machine output.
     
    10. A system comprising:

    a machine monitoring device (10) according to any preceding claim, operationally coupled to the machine (100).


     
    11. A method of monitoring a machine (100) with a machine monitoring device (10), the method comprising:

    (a) interrogating (301) a targeted machine component with a probe signal (22) via a communications link (20) and receiving one or more measured machine operating condition outputs reflecting the machine state characterized by machine operating condition input variables representing an individual parameter or set of parameters being considered;

    (b) selecting (302) a machine operating condition input variable (Xn) for which a correlation with a corresponding machine operating condition output (Yn) is unknown;

    (c) applying (303) a predictive model in which the machine operating condition input variable serves as an argument of a predicted machine operating condition output, wherein the predictive model comprises at least two independent primary models, the primary models being used to create the predictive model, for each of which primary models at least one correspondence between a primary model machine operating condition (X1, X2) and a corresponding primary model machine output (Y1, Y2) are known; the primary models either initially share a common basis or are converted into primary models sharing a common basis before being incorporated into the predictive model if the primary models did not share a common basis prior to being converted into the predictive model;

    (d) updating (304) a library (50) of predicted machine operating condition outputs (Xn, Yn); and

    (e) alerting (305) a human operator if one of the measured or the predicted machine operating condition output (34) is at variance with a predetermined threshold machine operating condition output by comparing the machine operating condition output (34) with a predetermined acceptable range of machine operating condition outputs.


     
    12. The method according to claim 11, wherein the targeted machine component is a rotor.
     
    13. The method according to claim 11, wherein the targeted machine component is a rotor of a permanent magnet motor.
     


    Ansprüche

    1. Vorrichtung zur Maschinenüberwachung (10), wobei die Vorrichtung umfasst:

    (a) eine Kommunikationsverbindung (20) zum Abfragen einer Maschine (100) mit einem Sondensignal (22) und Empfangen einer oder mehrerer gemessener Maschinenbetriebsbedingungsausgaben, welche den Maschinenzustand widerspiegeln, dadurch gekennzeichnet, dass Maschinenbetriebsbedingungseingabevariablen berücksichtigt werden, welche einen einzelnen Parameter oder Satz von Parametern darstellen; und

    (b) eine Vorrichtungssteuerung (30), welche eingerichtet ist zum:

    (i) Auswählen (302) einer Maschinenbetriebsbedingungseingabevariable (Xn), für welche eine Korrelation mit einer entsprechenden Maschinenbetriebsbedingungsausgabe (Yn) unbekannt ist;

    (ii) Anwenden (303) eines Vorhersagemodells, bei welchem die Maschinenbetriebsbedingungseingabevariable als ein Parameter einer vorhergesagten Maschinenbetriebsbedingungsausgabe dient, wobei das Vorhersagemodell mindestens zwei unabhängige Primärmodelle umfasst, wobei die Primärmodelle verwendet werden, um das Vorhersagemodell zu erzeugen, wobei für jedes der Primärmodelle mindestens eine Entsprechung zwischen einer Primärmodell-Maschinenbetriebsbedingung (X1, X2) und einer entsprechenden Primärmodell-Maschinenausgabe (Y1, Y2) bekannt ist; die Primärmodelle entweder anfänglich eine gemeinsame Basis aufweisen oder in Primärmodelle umgewandelt werden, die eine gemeinsame Basis aufweisen, bevor sie in das Vorhersagemodell integriert werden, wenn die Primärmodelle keine gemeinsame Basis aufweisen, bevor sie in das Vorhersagemodell umgewandelt werden;

    (iii) Aktualisieren (304) einer Bibliothek (50) vorhergesagter Maschinenbetriebsbedingungsausgaben (Xn, Yn); und

    (iv) Alarmieren (305) eines physischen Bedieners, wenn eine der gemessenen oder der vorhergesagten Maschinenbetriebsbedingungsausgaben (34) durch Vergleich der Maschinenbetriebsbedingungsausgabe (34) mit einem vorbestimmten akzeptablen Bereich von Maschinenbetriebsbedingungsausgaben einen vorbestimmten Maschinenbetriebsbedingungsausgabenschwellenwert überschreitet.


     
    2. Vorrichtung zur Maschinenüberwachung (10) nach Anspruch 1, wobei die Primärmodell-Maschinenausgaben und die vorhergesagte Maschinenbetriebsbedingungsausgabe eines oder mehrere aus einer Maschinentemperatur, einem Maschinendruck, einer Maschinenvibrationseigenschaft, einer Maschinendrehzahl, einer Maschinenlineargeschwindigkeit, einer Maschinenbeschleunigung, einer Maschinenkraft, einem Maschinendrehmoment, einer Maschinenleistungseingabevariable, einer Betriebszeit, einem Maschinenalter, einer Maschinengewichtseigenschaft, einer Maschinengeometrieeigenschaft, einer Materialeigenschaft einschließen, die von der Maschine verarbeitet werden (100).
     
    3. Vorrichtung zur Maschinenüberwachung (10) nach einem der vorhergehenden Ansprüche, wobei die Primärmodell-Maschinenausgaben und die vorhergesagte Maschinenbetriebsbedingungsausgabe eine Maschinenkomponentendrehzahl einschließen.
     
    4. Vorrichtung zur Maschinenüberwachung (10) nach einem der vorhergehenden Ansprüche, wobei die Primärmodell-Maschinenausgaben die Maschine in nicht stationären Zuständen widerspiegeln.
     
    5. Vorrichtung zur Maschinenüberwachung (10) nach einem der vorhergehenden Ansprüche, wobei die Maschine aus der Gruppe ausgewählt ist, die aus Turbinen, Windturbinen, Verdichtern, Turboexpandern, Motoren und Pumpen besteht.
     
    6. Vorrichtung zur Maschinenüberwachung (10) nach einem der vorhergehenden Ansprüche, wobei das Sondensignal (22) ein elektrisches Hochfrequenzsignal ist.
     
    7. Vorrichtung zur Maschinenüberwachung (10) nach einem der Ansprüche 1 bis 6, wobei das Sondensignal (22) ein optisches Signal ist.
     
    8. Vorrichtung zur Maschinenüberwachung (10) nach einem der vorhergehenden Ansprüche, wobei die Maschinenbetriebsbedingungseingabevariable extrapolierend in Bezug auf mindestens zwei Entsprechungen zwischen der Primärmodell-Maschinenbetriebsbedingung und der entsprechenden Primärmodell-Maschinenausgabe ist.
     
    9. Vorrichtung zur Maschinenüberwachung (10) nach einem der vorhergehenden Ansprüche, wobei die Maschinenbetriebsbedingungseingabevariable interpolierend in Bezug auf mindestens zwei Entsprechungen zwischen der Primärmodell-Maschinenbetriebsbedingung und der entsprechenden Primärmodell-Maschinenausgabe ist.
     
    10. System, umfassend:

    eine Vorrichtung zur Maschinenüberwachung (10) nach einem der vorhergehenden Ansprüche, die für den Betrieb gekoppelt ist mit: der Maschine (100).


     
    11. Verfahren zum Überwachen einer Maschine (100) mit einer Vorrichtung zur Maschinenüberwachung (10), wobei das Verfahren umfasst:

    (a) Abfragen (301) einer Zielmaschinenkomponente mit einem Sondensignal (22) über eine Kommunikationsverbindung (20) und Empfangen einer oder mehrerer gemessener Maschinenbetriebsbedingungsausgaben, welche den Maschinenzustand widerspiegeln, dadurch gekennzeichnet, dass Maschinenbetriebsbedingungseingabevariablen berücksichtigt werden, welche einen einzelnen Parameter oder einen Satz von Parametern darstellen;

    (b) Auswählen (302) einer Maschinenbetriebsbedingungseingabevariable (Xn), für welche eine Korrelation mit einer entsprechenden Maschinenbetriebsbedingungsausgabe (Yn) unbekannt ist;

    (c) Anwenden (303) eines Vorhersagemodells, bei welchem die Maschinenbetriebsbedingungseingabevariable als ein Parameter einer vorhergesagten Maschinenbetriebsbedingungsausgabe dient, wobei das Vorhersagemodell mindestens zwei unabhängige Primärmodelle umfasst, wobei die Primärmodelle verwendet werden, um das Vorhersagemodell zu erzeugen, wobei für jedes der Primärmodelle mindestens eine Entsprechung zwischen einer Primärmodell-Maschinenbetriebsbedingung (X1, X2) und einer entsprechenden Primärmodell-Maschinenausgabe (Y1, Y2) bekannt ist; die Primärmodelle entweder anfänglich eine gemeinsame Basis aufweisen oder in Primärmodelle umgewandelt werden, die eine gemeinsame Basis aufweisen, bevor sie in das Vorhersagemodell integriert werden, wenn die Primärmodelle keine gemeinsame Basis aufweisen, bevor sie in das Vorhersagemodell umgewandelt werden;

    (d) Aktualisieren (304) einer Bibliothek (50) vorhergesagter Maschinenbetriebsbedingungsausgaben (Xn, Yn); und

    (e) Alarmieren (305) eines physischen Bedieners, wenn eine der gemessenen oder der vorhergesagten Maschinenbetriebsbedingungsausgaben (34) nach Vergleich der Maschinenbetriebsbedingungsausgabe (34) mit einem vorbestimmten akzeptablen Bereich von Maschinenbetriebsbedingungsausgaben einen vorbestimmten Maschinenbetriebsbedingungsausgabenschwellenwert überschreitet.


     
    12. Verfahren nach Anspruch 11, wobei die Zielmaschinenkomponente ein Rotor ist.
     
    13. Verfahren nach Anspruch 11, wobei die Zielmaschinenkomponente ein Rotor eines Permanentmagnetmotors ist.
     


    Revendications

    1. Dispositif de surveillance de machine (10), le dispositif comprenant :

    (a) une liaison de communication (20) pour interroger une machine (100) avec un signal de sonde (22) et recevoir un ou plusieurs résultats relatifs à la condition de fonctionnement de machine mesurés reflétant l'état de la machine caractérisé par des variables d'entrée de condition de fonctionnement de machine représentant un paramètre individuel ou un ensemble de paramètres étant considérés ; et

    (b) un système de commande de dispositif (30) agencé pour :

    (i) sélectionner (302) une variable d'entrée de condition de fonctionnement de machine (Xn) pour laquelle une corrélation avec un résultat de condition de fonctionnement de machine correspondant (Yn) est inconnue ;

    (ii) appliquer (303) un modèle prédictif dans lequel la variable d'entrée de condition de fonctionnement de machine sert d'argument d'un résultat de condition de fonctionnement de machine prédit, dans lequel le modèle prédictif comprend au moins deux modèles indépendant primaires, les modèles primaires étant utilisés pour créer le modèle prédictif, pour chacun des modèles primaires au moins une correspondance entre une condition de fonctionnement de machine de modèle primaire (X1, X2) et un résultat de machine de modèle primaire correspondant (Y1, Y2) sont connus ; les modèles primaires partagent initialement une base commune ou sont convertis en modèles primaires partageant une base commune avant d'être incorporés dans le modèle prédictif si les modèles primaires ne partageaient pas de base commune avant d'être convertis en modèle prédictif ;

    (iii) mettre à jour (304) une bibliothèque (50) de résultats de conditions de fonctionnement de machine prédits (Xn, Yn) ; et

    (iv) alerter (305) un opérateur humain si l'une du résultat de la condition de fonctionnement de machine mesurée ou prédite (34) varie avec un résultat de seuil de la condition de fonctionnement de machine prédéterminée en comparant le résultat de la condition de fonctionnement de machine (34) avec une plage acceptable prédéterminée de résultats de condition de fonctionnement de machine.


     
    2. Dispositif de surveillance de machine (10) selon la revendication 1, dans lequel les résultats de la machine de modèle primaire et le résultat de condition de fonctionnement de machine prédit comprennent un ou plusieurs parmi une température de machine, une pression de machine, une caractéristique de vibration de la machine, une vitesse de rotation de la machine, une vitesse de translation de la machine, une accélération de la machine, une force de la machine, un couple de la machine, une variable d'entrée d'alimentation de la machine, un temps de fonctionnement, l'âge de la machine, une caractéristique de poids de la machine, une caractéristique de la géométrie de la machine, une caractéristique de la matière étant traitée par la machine (100).
     
    3. Dispositif de surveillance de machine (10) selon l'une quelconque des revendications précédentes, dans lequel les résultats de la machine de modèle primaire et le résultat de la condition de fonctionnement de machine prédite incluent une vitesse de rotation des composants de la machine.
     
    4. Dispositif de surveillance de machine (10) selon l'une quelconque des revendications précédentes, dans lequel les résultats de la machine de modèle primaire reflètent la machine dans des états non stationnaires.
     
    5. Dispositif de surveillance de machine (10) selon l'une quelconque des revendications précédentes, dans lequel la machine est choisie dans le groupe constitué par les turbines, les éoliennes, les compresseurs, les turbodétendeurs, les moteurs, et les pompes.
     
    6. Dispositif de surveillance de machine (10) selon l'une quelconque des revendications précédentes, dans lequel le signal de sonde (22) est un signal électrique haute fréquence.
     
    7. Dispositif de surveillance de machine (10) selon l'une quelconque des revendications 1 à 6, dans lequel le signal de sonde (22) est un signal optique.
     
    8. Dispositif de surveillance de machine (10) selon l'une quelconque des revendications précédentes, dans lequel la variable d'entrée de condition de fonctionnement de machine est extrapolative par rapport à au moins deux correspondances entre la condition de fonctionnement de machine de modèle primaire et le résultat de la machine de modèle primaire correspondante.
     
    9. Dispositif de surveillance de machine (10) selon l'une quelconque des revendications précédentes, dans lequel la variable d'entrée de condition de fonctionnement de machine est interpolative par rapport à au moins deux correspondances entre la condition de fonctionnement de machine de modèle primaire et le résultat de la machine de modèle primaire correspondante.
     
    10. Système comprenant :

    un dispositif de surveillance de la machine (10) selon l'une quelconque des revendications précédentes, couplé de manière opérationnelle à la machine (100).


     
    11. Procédé de surveillance d'une machine (100) avec un dispositif de surveillance de machine (10), le procédé comprenant les étapes consistant à :

    (a) interroger (301) un composant de machine ciblé avec un signal de sonde (22) par l'intermédiaire d'une liaison de communication (20) et recevoir un ou plusieurs résultats de condition de fonctionnement de machine mesurés reflétant l'état de la machine caractérisé par des variables d'entrée de condition de fonctionnement de machine représentant un paramètre individuel ou un ensemble de paramètres étant considérés ;

    (b) sélectionner (302) une variable d'entrée de condition de fonctionnement de machine (Xn) pour laquelle une corrélation avec un résultat de condition de fonctionnement de machine correspondant (Yn) est inconnue ;

    (c) appliquer (303) un modèle prédictif dans lequel la variable d'entrée de la condition de fonctionnement de machine sert d'argument d'un résultat de la condition de fonctionnement de machine prédit, dans lequel le modèle prédictif comprend au moins deux modèles primaires indépendants, les modèles primaires étant utilisés pour créer le modèle prédictif, pour chaque modèle primaire au moins une correspondance entre une condition de fonctionnement de machine de modèle primaire (X1, X2) et un résultat de machine de modèle primaire correspondant (Y1, Y2) sont connus ; les modèles primaires partagent initialement une base commune ou sont convertis en modèles primaires partageant une base commune avant d'être incorporés dans le modèle prédictif si les modèles primaires ne partageaient pas de base commune avant d'être convertis en modèle prédictif ;

    (d) mettre à jour (304) une bibliothèque (50) de résultats de conditions de fonctionnement de machine prédits (Xn, Yn) ; et

    (e) alerter (305) un opérateur humain si l'un des résultats de la condition de fonctionnement de machine mesuré ou prédit (34) est variable avec un résultat de la condition de fonctionnement de machine de seuil prédéterminé en comparant le résultat de condition de fonctionnement de machine (34) avec une plage acceptable prédéterminée de résultats de condition de fonctionnement de machine.


     
    12. Procédé selon la revendication 11, dans lequel le composant de machine ciblé est un rotor.
     
    13. Procédé selon la revendication 11, dans lequel le composant de machine ciblé est un rotor d'un moteur à aimant permanent.
     




    Drawing


























    Cited references

    REFERENCES CITED IN THE DESCRIPTION



    This list of references cited by the applicant is for the reader's convenience only. It does not form part of the European patent document. Even though great care has been taken in compiling the references, errors or omissions cannot be excluded and the EPO disclaims all liability in this regard.

    Patent documents cited in the description




    Non-patent literature cited in the description