`
`[19]
`
`[11] Patent Number:
`
`5,555,190
`
`Derby et al.
`
`
`
`[45] Date of Patent: Sep. 10, 1996
`
`U8005555190A
`
`’
`
`[54]
`
`METHOD AND APPARATUS FOR ADAPTIVE
`LINE ENHANCEMENT IN CORIOLIS MASS
`FLOW METER MEASUREMENT
`
`[75]
`
`Inventors: Howard V. Derby, Boulder; Tamal
`Bose, Denver, both of C010.; Seeraman
`Rajan, Bombay, Ind.
`
`[73] Assignee: Micro Motion, Inc., Boulder, C010.
`
`[21] Appl. No.: 501,411
`
`[22] Filed:
`
`Jul. 12, 1995
`
`Int. Cl.6 ........................................................ G01F 1/84
`[51]
`[52] US. Cl.
`....................................... 364/510; 73/861356
`[58] Field of Search ..................................... 364/509, 510,
`364/572; 73/861.38, 861.37, 861.03, 861;
`324/76, 77-85, 601
`
`[56]
`
`References Cited
`U.S. PATENT DOCUMENTS
`
`1/1983 Smith ................................... 73/861318
`Re. 31,450
`8/1978 Smith ................... 73/861.37
`4,109,524
`
`..... 73/86138
`1/1985 Smith et a1.
`4,491,025
`
`11/1989 Zolock ............. 73/861.38
`4,879,911
`
`6/1990 Romano ............................... 73/86138
`4,934,196
`4/1991 Kalotay ................................ 73/86138
`5,009,109
`
`10/1991 Christ et a1.
`..... 73/861.38
`5,052,231
`7/1994 Zolock ............. 73/861355
`5,331,859
` 5,429,002 7/1995 Colman . 73/86138
`
`
`5,469,748
`11/1995 Kalotay ................................ 73/861.38
`
`[57]
`
`ABSTRACT
`
`An apparatus and method for determining frequency and
`phase relationships of vibrating flow tubes in a Coriolis mass
`flow meter. Adaptive line enhancement (ALE) techniques
`and apparatus are used in a digital signal processing (DSP)
`device to accurately determine frequency and phase rela-
`tionships of the vibrating flow tube and to thereby more
`accurately determine mass flow rate of a material flowing
`through the mass flow meter. In a first embodiment, an
`adaptive notch filter is used to enhance the signal from each
`corresponding sensor signal on the vibrating flow tubes. In
`a second embodiment, a plurality of adaptive notch filters
`are cascaded to enhance the signal from each corresponding
`sensor signal. In both embodiments an anti—aliasing deci-
`mation filter associated with each sensdr signal reduces the
`computational complexity by reducing the number of
`samples from a fixed frequency AID sampling device asso-
`ciated with each sensor signal. Computational adjustments
`are performed to compensate for spectral leakage between
`the fixed sampling frequency and the variable fundamental
`frequency of the vibrating flow tubes. Despite this added
`computational complexity, the present invention is simpler
`than prior designs and provides better noise immunity due to
`the adaptive notch filtration. Heuristics are applied to the
`weight adaptation algorithms of the notch filters to improve
`convergence of the digital filters and to reduce the possibility
`of instability of the filters interfering with mass flow mea-
`surements.
`
`Primary Examiner—James P. Trammell
`Attorney, Agent, or Firm—Duft, Graziano & Forest, RC.
`
`35 Claims, 20 Drawing Sheets
`
`250
`
`157
`
`158
`
`
`48:1
`
`
`DECIMATION
`
`
`
`204
`
`NOTCH
`FILTER
`
`204
`
`
`
`48:1
`
`DECIMATION
`
`moron
`FILTER
`
`262
`
`258
`
`WEIGHT
`ADAPTATION
`
`
`
`K
`21o
`
`HALF
`SAMPNO
`
`wmnow
`
`
`206
`
`260
`
`PHASE
`
`205
`
`268
`/
`
`264
`
`208
`AT
`CALCULATION
`
`294
`)
`I
`
`290
`s
`MASS
`
`FLOW
`COMPU-
`TATIDN
`
`PHASE
`
`266
`
`‘55
`
`UTILIZATION
`MEANS
`
` 268
`
`292
`
`60
`
`FREQUENCY
`CALCULATION
`
`212
`
`Micro Motion 1016
`
`1
`
`Micro Motion 1016
`
`
`
`US. Patent
`
`Sep. 10, 1996
`
`Sheet 1 of 20
`
`5,555,190
`
`26.:
`
`zoEEMEDE
`
`mm<z
`
`2
`
`
`
`
`US. Patent
`
`Sep. 10, 1996
`
`Sheet 2 0f 20
`
`5,555,190
`
`2252::
`
`wzfiz
`
`mmm
`
`mmm
`
`>0ZmDOmE
`
`zo:.<..:o._<o
`
`Nrm
`
`NGE
`
`om
`
`Clam;
`
`zo_H<.E<n_<
`
`05
`
`me
`
`56.:
`
`382:5
`
`0mm
`
`m2:
`
`26:
`
`-228
`
`ZOE:
`
`Em
`
`
`
`momvmm
`
`mwN
`
`mmfii
`
`:05:
`
`ESE
`
`tmv
`
`zo:.<_2_omo
`
`mmr
`
`c.9“:
`
`mom
`
`h<
`
`295150.20
`
`mom
`
`vow
`
`mm<zm
`
`mmm
`
`1052
`
`ESE
`
`F
`
`mv
`
`zo;<_>__oma
`
`com
`
`mom
`
`vow.
`
`NON
`
`omm
`
`com
`
`:.omv5
`
`m2
`
`.:.9“:
`
`3
`
`
`
`
`
`
`
`
`
`U.S. Patent
`
`Sep. 10, 1996
`
`Sheet 3 of 20
`
`5,555,190
`
`.mom
`
` oom
`
`4
`
`
`
`US. Patent
`
`Sep. 10, 1996
`
`Sheet 4 of 20
`
`5,555,190
`
`
`
`5
`
`
`
`US. Patent
`
`Sep. 10, 1996
`
`Sheet 5 of 20
`
`5,
`
`555,190
`
`vom
`
`6
`
`
`
`US. Patent
`
`Sep. 10, 1996
`
`Sheet 6 of 20
`
`5,555,190
`
`9-:v_x_923;.
`
`
`
`“mimicousmso
`
`
`
`gm:2525
`
`“ETSEB
`
`mom
`
`x
`
`>._._.=m<5
`
`55.
`
`
`
`7
`
`
`
`US. Patent
`
`Sep. 10, 1996
`
`Sheet 7 of 20
`
`5,555,190
`
`wE
`
`53E
`
`62:2:
`
`“3%
`
`k G
`
`E
`
`BOOZE
`
`8.9“;EN
`
`8
`
`
`
`US. Patent
`
`Sep. 10, 1996
`
`Sheet 8 of 20
`
`5,555,190
`
`8.9”:
`
`3%
`
`a.or:
`
`own
`
`mom
`
`$52K8m
`:2.&mom ,mzmxvow
`
`
`5%)“wow
`
`m.UE
`
`+
`
`a5i: vwwwE.mom
`Engamom vmm
`
`00mm86...:
`
`5o
`
`
`
`Em8wmmm.H_-E«E
`
`woOhm
`
`9
`
`
`
`
`
`
`
`US. Patent
`
`Sep. 10, 1996
`
`Sheet 9 of 20
`
`5,555,190
`
`
`
`10
`
`10
`
`
`
`US. Patent
`
`Sep. 10, 1996
`
`Sheet 10 of 20
`
`5,555,190
`
`
`
`2GE
`
`11
`
`11
`
`
`
`U.S. Patent
`
`Sep. 10, 1996
`
`Sheet 11 of 20
`
`5,555,190
`
`E
`
`+25%_382.3
`
`mmwv
`
`00:.
`
`382;)
`
`825529
`
`
`
`02n=2<m
`
`NR
`
`a.9“;
`
`n__>_<w>>
`
`I:
`
`>.aliasL”V_>
`
`12
`
`:..GE
`
`12
`
`
`
`
`
`
`
`
`
`
`US. Patent
`
`Sep. 10, 1996
`
`Sheet 12 of 20
`
`5,555,190
`
`mom
`
`
`
`®
`
`00$
`
`
`
`NmNF
`
`+av+
`
`alum
`
`NFGE
`
`
`
`a6.“:fim
`
`382:5
`
`“EST.
`
`mmmw
`
`momw
`
`13
`
`8m
`
`8.9”;
`
`8m
`
`8.9”:
`
`13
`
`
`
`
`
`
`US. Patent
`
`Sep. 10, 1996
`
`Sheet 13 of 20
`
`5,555,190
`
`vow
`
`mom
`
`
`
`
`
`.2..GE
`
`0mm?
`
`oomw
`
`58.
`
`a6.”:
`
`Nmmr
`
`wmmw
`
`SN
`
`a.03
`
`14
`
`mom?
`
`14
`
`
`
`US. Patent
`
`Sep. 10,
`
`1996
`
`Sheet 14 of 20
`
`5,555,190
`
`0mm.mmm—
`
`
`
` mom;
`
`3..GE
`
`
`
`.mmm.vmm
`
`Nmmtommr
`
`.002
`
`15
`
`15
`
`
`
`US. Patent
`
`Sep. 10, 1996
`
`Sheet 15 of 20
`
`5,555,190
`
`3-3
`
`Nmmr.oomr
`
` NSF
`
`.NomF
`
`NE;
`
`m“.GE
`
`16
`
`16
`
`
`
`
`_ US. Patent
`
`Sep. 10, 1996
`
`Sheet 16 of 20
`
`5,555,190
`
`vowr
`
`wmm
`
`v30
`u
`
`30325,
`
`u_._<I
`
`mpGE
`
`vomw
`
`17
`
`17
`
`
`
`
`
`US. Patent
`
`Sep. 10, 1996
`
`Sheet 17 of 20
`
`5,555,190
`
`FIG- 1 7
`
`A / D INTERRUPT
`
`1700
`
`
`
`
`
`
`
`
`
`
`
`
`1702
`
`READ A/D
`SAMPLE INTO
`CIRCULAR
`BUFFER
`
`
`
`1 704
`
`DONE
`
`DONE
`
`
` 8 NEW
`SAMPLES READ
`
`SINCE LAST
`CONVOLUTION
`?
`
`
`
`
`
`
`
`DETERMINE THE
`CONVOLUTION OF SAMPLES
`
`
`AND STORE IN SECOND
`
`
`CIRCULAR BUFFER
`
`1706
`
`
`
`
`
`
`SINCE LAST
`CONVOLUTION
`
`
`
`
`
`DETERMINE THE
`CONVOLUTION
`VALUES IN SECOND
`BUFFER AND STORE IN
`SAMPLE BUFFER
`
`
`
`1710
`
`DONE
`
`18
`
`18
`
`
`
`US. Patent
`
`Sep. 10, 1996
`
`Sheet 18 of 20
`
`5,555,190
`
`FIG. 1a
`
`
`
`
`
`INITIALIZE CIRCULAR BUFFERS FOR
`A/ D DECIMATION AND
`ENABLE A/ D INTERRUPTS
`
`
`
`
`WAIT FOR DECIMATED
`SAMPLE AVAILABILITY
`
`1800
`
`1802
`
`ACCUMULATE SIGNAL
`AND NOISE VALUES
`
`
`
`APPLY NOTCH FILTER TO PRODUCE
`ENHANCED SAMPLE
`
`UPDATE FILTER PARAMETERS
`
`
`
`
`
`UPDATE GOERTZEL
`FILTER TO ACCUMULATE
`
`
`COMPLEX NUMBER
`
`
`FOR PHASE
`
`
`1804
`
`1806
`
`
`
`
`
`1812
` HALF
`WINDOW
`START
`
`DETERMINE SNR VALUE
`FROM ACCUMULATED
`SIGNAL AND NOISE VALUES
`
`1814
`
`
`?
`
`
`
`COMPUTATIONS
`
` RESET FILTER
`ADAPTATION
`
`
`
`DETERMINE At FOR PREVIOUS WINDOW AND
`APPLY UTILIZATION MEANS AND
`
`DETERMINE GOERTZEL FILTER WEIGHTS
`
`
`
`
`
`
`1820
`
`19
`
`19
`
`
`
`US. Patent
`
`Sep. 10, 1996
`
`Sheet 19 of 20
`
`5,555,190
`
`FIG. 19
`
`1 806
`
`DETERMINE
`UPDATED FORGETTING
`FACTOR
`
`DETERMINE
`UPDATED GAIN
`FACTOR
`
`-
`
`DETERMINE
`UPDATED DEBIASING
`PARAMETER
`
`DETERMINE
`UPDATE COVARIANCE
`MATRIX
`
`DETERMINE
`UPDATED NOTCH FILTER
`WEIGHTS
`
`
`
`UPDATED
`WEIGHTS
`STABLE
`?
`
`APPLY UPDATED
`WEIGHTS TO
`NOTCH FILTERS
`
`20
`
`20
`
`
`
`US. Patent
`
`Sep. 10, 1996
`
`Sheet 20 of 20
`
`5,555,190
`
`”mama
`
`H5015
`
`21
`
`21
`
`
`
`5,555,190
`
`1
`NIETHOD AND APPARATUS FOR ADAPTIVE
`LINE ENHANCEMENT 1N CORIOLIS MASS
`FLOW METER MEASUREMENT
`
`FIELD OF THE INVENTION
`
`The present invention relates to mass flow rate measure-
`ment and in particular to the use of digital signal processing
`adaptive filtration methods and apparatus in Coriolis mass
`flow meters.
`
`PROBLEM
`
`It is known to use Coriolis mass flowmeters to measure
`
`mass flow and other information for materials flowing
`through a conduit. Such flowmeters are disclosed in US.
`Pat. Nos. 4,109,524 of Aug. 29, 1978, U.S. Pat. No. 4,491,
`025 ofJarr. l, 1985, and Re. 31,450 of Feb. 11, 1982, all to
`J. E. Smith et a1. These flowmeters have one or more flow
`
`tubes of straight or curved configuration. Each flow tube
`configuration in a Coriolis mass flowmeter has a set of
`natural vibration modes, which may be of a simple bending,
`torsional or coupled type. Each flow tube is driven to
`oscillate at resonance in one of these natural modes. Material
`flows into the flowmeter from a connected conduit on the
`inlet side of the flowmeter, is directed through the flow tube
`or tubes, and exits the flowmeter through the outlet side. The
`natural vibration modes of the vibrating, fluid filled system
`are defined in part by the combined mass of the flow tubes
`and the material within the flow tubes.
`
`When there is no flow through the flowmeter, all points
`along the flow tube oscillate about a pivot point with
`identical phase due to an applied driver force. As material
`begins to flow, Coriolis accelerations cause each point along
`the flow tube to have a different phase. The phase on the inlet
`side of the flow tube lags the driver, while the phase on the
`outlet side leads the driver. Sensors are placed on the flow
`tube to produce sinusoidal signals representative of the
`motion of the flow tube. The phase difference between two
`sensor signals is proportional
`to the mass flow rate of
`material through the flow tube.
`A complicating factor in this measurement is that the
`density of typical process material varies. Changes in den—
`sity cause the frequencies of the natural modes to vary. Since
`the flowmeter’s control system maintains resonance,
`the
`oscillation frequency varies in response to changes in den-
`sity. Mass flow rate in this situation is proportional to the
`ratio of phase difference and oscillation frequency.
`The above—mentioned US. Pat. No. Re. 31,450 to Smith
`discloses a Coriolis flowmeter that avoids the need for
`
`measuring both phase difference and oscillation frequency.
`Phase difference is determined by measuring the time delay
`between level crossings of the two sinusoidal sensor output
`signals of the flowmeter. When this method is used, the
`variations in the oscillation frequency cancel, and mass flow
`rate is proportional to the measured time delay. This mea-
`surement method is hereinafter referred to as a time delay or
`At measurement.
`Measurements in a Coriolis mass flowmeter must be made
`with great accuracy since it is often a requirement that the
`derived flow rate information have an accuracy of at least
`0.15% of reading. The signal processing circuin which
`receives the sensor output signals measures this phase
`difference with precision and generates the desired charac-
`teristics of the flowing process material to the required
`accuracy of at least 0.15% of reading.
`
`2
`
`In order to achieve these accuracies, it is necessary that
`the signal processing circuitry operate with precision in
`measuring the phase shift of the two signals it receives from
`the flowmeter. Since the phase shift between the two output
`signals of the meter is the information used by the process-
`ing circuitry to derive the material characteristics,
`it
`is
`necessary that the processing circuitry not introduce any
`phase shift which would mask the phase shift information
`provided by the sensor output signals. In practice,
`it is
`necessary that this processing circuitry have an extremely
`low inherent phase shift so that the phase of each input
`signal is shifted by less than 0.0010 and, in some cases, less
`than a few parts per million. Phase accuracy of this magni-
`tude is required if the derived information regarding the
`process material is to have an accuracy of less than 0.15%.
`The frequencies of the Coriolis flowmeter output signals
`fall in the frequency range of many industrially generated
`noises. Also, the amplitude of the sensor output signals is
`often small and, in many cases, is not significantly above the
`amplitude of the noise signals. This limits the sensitivity of
`the flowmeter and makes the extraction of the useful infor-
`mation quite difiicult.
`There is not much a designer can do either to move the
`meter output signals frequency out of the noise band or to
`increase the amplitude of the output signals. Practical Corio«
`lis sensor and flowmeter design requires compromises that
`result in the generation of output signals having a less than
`optimum signal to noise ratio and dynamic range. This
`limitation determines the flowmeter characteristics and
`specifications including the minimum and maximum flow
`rates which may be reliably derived from the flowmeter’s
`output signals.
`The magnitude of the minimum time delay that can be
`measured between the two Coriolis flowmeter output signals
`at a given drive frequency is limited by various factors
`including the signal to noise ratio, the complexity of asso-
`ciated circuitry and hardware, and economic considerations
`that limit the cost and complexity of the associated circuitry
`and hardware. Also, in order to achieve a flowmeter that is
`economically attractive, the low limit of time delay mea-
`surement must be as low as possible. The processing cir-
`cuitry that receives the two output signals must be able to
`reliably measure the time delay between the two signals in
`order to provide a meter having the high sensitivity needed
`to measure the flowing characteristics of materials having a
`low density and mass such as, for example, gases.
`There are limitations regarding the extent to which con-
`ventional analog circuit design can, by itself, permit accurate
`time delay measurements under all possible operating con-
`ditions of a Coriolis flowmeter. These limitations are due to
`
`in any electronic equipment
`the inherent noise present
`including the imperfections of semi—conductor devices and
`noise generated by other circuit elements. These limitations
`are also due to ambient noise which similarly limits the
`measurement can be reduced to some extent by techniques
`such as shielding, guarding, grounding, etc. Another limi-
`tation is the signal to noise ratio of the sensor output signals
`themselves.
`
`Good analog circuit design can overcome some of the
`problems regarding noise in the electronic equipment as well
`as the ambient noise in the environment. However, an
`improvement in the signal to noise ratio of the output signals
`cannot be achieved without the use of analog filters. But
`analog filters alter the amplitude and phase of the signals to
`be processed. This is undesirable, since the time delay
`between the two signals is the base information used to
`
`10
`
`15
`
`20
`
`25
`
`30
`
`35
`
`4O
`
`45
`
`50
`
`55
`
`65
`
`22
`
`22
`
`
`
`3
`
`4
`
`5,555,190
`
`derive characteristics of the process fluid. The use of filters
`having unknown or varying amplitude and/or phase charac—
`teristics can unacceptably alter the phase difference between
`the two sensor output signals and preclude the derivation of
`accurate information of the flowing material.
`The flowmeter’ s drive signal is typically derived from one
`of the sensor output signals after it is conditioned, phase
`shifted and used to produce the sinusoidal drive voltage for
`the drive coil of the meter. This has the disadvantage that
`harmonics and noise components present in the sensor signal
`are amplified and applied to the drive coil to vibrate the flow
`tubes at their resonant frequency. However, an undesirable
`drive signal can also be generated by unwanted mechanical
`vibrations and electrical interferences that are fed back to the
`meter drive circuit and reinforced in a closed loop so that
`they create relatively high amplitude self-perpetuating dis-
`turbing signals which further degrade the precision and
`accuracy of the time delay measurement.
`There are several well known methods and circuit designs
`which seek to reduce the above problems. One such suc-
`cessful technique to reduce some of the above problems is
`described in U.S. Pat. No. 5,231,884 to M. Zolock and U.S.
`Pat. No. 5,228,327 to Bruck. These patents describe Coriolis
`flowmeter signal processing circuitry that uses three identi-
`cal channels having precision integrators as filters. A first
`one of these channels is permanently connected to one
`sensor signal, say, for example,
`the left. The other two
`channels (second and third) are alternately connected, one at
`a time, to the right sensor signal. When one of these, say the
`second channel, is connected to the right sensor signal, the
`third channel is connected, along with the first channel, to
`the left sensor signal. The inherent phase delay between the
`. first and third channel is measured by comparing the time
`difference between the outputs of the two channels now both
`connected to the left signal. Once this characteristic delay is
`determined, the role of this third channel and the second
`channel connected to the right sensor signal is switched. In
`this new configuration,
`the second channel undergoes a
`calibration of its delay characteristics while the third cali-
`brated channel is connected to the right sensor signal. The
`roles of second and third channels are alternately switched
`by a control circuit approximately once every minute. Dur-
`ing this one-minute interval (about 30 to 60 seconds), aging,
`temperature, and other effects have no meaningful influence
`on the phase—shift of the filters and therefore their phase
`relationship is known and considered defined.
`The precisely calibrated integrators used by Zolock pro-
`vide a signal to noise ratio improvement amounting to about
`6 db/octave roll-off in the amplitude transfer function of the
`integrator. Unfortunately, this 6 db/octave improvement is
`not enough under many circumstances in which Coriolis
`flowmeters are operated (such as light material or exces—
`sively noisy environments). The reason for this is that a
`single-pole filter, such as the Zolock integrator, has a rela-
`tively wide band width. As a result, noise signals generated
`by unwanted flow tube vibration modes, noisy environment,
`material flow noise, electromagnetic or radio frequency
`interference and disturbances are not removed to the extent
`necessary for high meter sensitivity required for precision.
`Depending on their frequency, their amplitude is reduced
`somewhat, but they can still interfere with the precision time
`delay measurement between the two sensor output signals
`when measuring low mass materials such as gases.
`There is another source for errors in the Zolock and Bruck
`systems. The integrator time delay measurements are made
`at three (3) certain well defined points of the sinusoidal
`sensor signals. The two sensor signals are ideal only when
`
`10
`
`15
`
`20
`
`25
`
`30
`
`35
`
`40
`
`45
`
`50
`
`55
`
`60
`
`65
`
`their shape is the same and is symmetrical around their peak
`values. However, when the two magnetic circuits (sensors)
`that generate the sensor signals are not identical, the result—
`ing non-ideal wave forms contain different amounts of
`harmonics with possibly undefined phase conditions which
`can alter their shape and potentially change their symmetri~
`cal character. The result of such variations is such that when,
`during normal operations, a Zolock integrator is calibrated
`with one wave form and is subsequently used to measure
`another wave form,
`the difference in wave forms may
`produce an undefined and unknown amount of error due to
`its harmonic content and its undefined and varying phase of
`its harmonics.
`
`Other analog circuit design techniques suffer from similar
`problems of complexity,
`insufiicient noise immunity, or
`insufficient harmonic rejection.
`There are techniques currently available, such as digital
`signal processing (hereinafter referred to as DSP) and asso-
`ciated digital filtering,
`to overcome the above—discussed
`problems and simultaneously improve the signal to noise
`ratio of the signals being processed. However, these alter—
`natives have been more complicated and expensive than
`conventional analog circuit designs. In addition, these prior
`DSP designs have shown only modest improvement over
`conventional analog circuit designs with respect to noise
`immunity and harmonic rejection. U.S. Pat. No. 4,934,196,
`issued Jun. 19, 1990 to Romano, teaches a DSP design for
`computing the phase difference, At, and correlated mass flow
`rate. Romano’s design alters the sampling frequency of an
`AID converter in an attempt to maintain an integral number
`of sample times within each periodic cycle of the vibrating
`flow tubes. This need for variable frequency sampling
`complicates Romano’s DSP design. Although this DSP
`design is structurally quite distinct from prior discrete ana-
`log circuit designs, it has proven to provide only modest
`improvements over analog designs in measurement accu-
`racy because it provides significant improvement in filtration
`only at integer multiples of the fundamental frequency.
`However, many signal components result from mechanical
`vibration modes of the flow tubes whose resonant frequen-
`cies are not integer multiples of the fundamental frequency
`and are therefore poorly rejected by the prior DSP designs.
`Neither prior approach (analog nor prior DSP) effectively
`rejects non-harmonic or broadband noise. From the above
`discussion, it can be seen that there is a need for an improved
`method and apparatus for measuring mass flow rate in a
`Coriolis mass flow meter.
`
`SOLUTION
`
`The present invention solves the above identified prob-
`lems and achieves an advance in the art by applying digital
`filtering and digital signal processing (DSP) methods and
`apparatus to improve the accuracy of mass flow measure-
`ments in a Coriolis mass flow meter. The present invention
`comprises a DSP design which includes adaptive notch
`filters to improve the accuracy of frequency and phase
`measurements used in the computation of mass flow rate.
`The use of adaptive notch filtration in the present invention
`is one application of the technology commonly referred to as
`Adaptive Line Enhancement (ALE).
`In the present invention, the signal from each vibrating
`flow tube sensor is sampled, digitized, and then processed by
`a digital adaptive notch filter which passes all noise signals
`outside a narrow frequency band (a notch) around the
`fundamental frequency. This digitized filtered signal is then
`
`23
`
`23
`
`
`
`5
`
`6
`
`5,555,190
`
`subtracted from the original digitized signal to produce an
`enhanced signal representing the sensor output signal wave-
`form at the fundamental frequency with virtually all noise
`signals eliminated. This method and apparatus eliminates
`harmonic as well as non-harmonic noise signals. Initially the
`width of the notch filter’s “notch” is wide and is adapted
`over time to narrow as it converges on the fundamental
`frequency. Adaptation algorithms rapidly adapt the notch
`frequency of the adaptive filter to track changes over time in
`the fundamental frequency of the vibrating flow tubes.
`The DSP design of the present invention uses a fixed
`sampling frequency as distinct from Romano’s variable
`frequency design. This fixed sampling frequency approach
`permits rapid convergence of the adaptive notch filters on
`the fundamental frequency of the vibrating flow tubes and
`simplifies the total circuit design. The fixed sampling rate
`eliminates the need exhibited in Romano to provide addi-
`tional circuitry to vary the sampling rate. The present design
`performs computational adjustments to compensate for
`spectral leakage between the fixed sampling frequency and
`the variable fundamental frequency of the vibrating flow
`tubes. Despite this added computational complexity,
`the
`present invention is simpler than prior designs exemplified
`by Romano and provides better noise immunity due to the
`use of adaptive notch filtration.
`The present invention provides superior noise immunity
`and harmonic rejection as compared to all known designs
`and simplifies aspects of the DSP design disclosed by
`Romano. This permits improved accuracy of the flow rate
`measurements even in particularly noisy environments as
`well as applications with low density flow materials (such as
`gas).
`Since the flow tubes vibrate at the same fundamental
`
`frequency, adaptation of the notch filters is determined by
`samples from only one of the two notch filters. The adap-
`tation weights so determined are applied to both notch
`filters. Heuristics applied to the computations by the present
`invention prevent the notch filters from diverging from the
`fundamental frequency due to instability in the computa—
`tions. Other heuristics restart convergence computations for
`the adaptation when the signal to noise ratio measured by the
`notch filter is too small. A small signal
`to noise ratio
`indicates that the adaptive notch filter is not converged on
`the fundamental frequency. This may be due to a shift in the
`fundamental frequency of the vibrating flow tubes.
`In a first embodiment of the present invention, the output
`signal from each vibrating flow tube sensor is sampled at a
`fixed frequency by a corresponding A/D converter. The
`sampled value generated by each A/D converter is then
`applied to a corresponding decimation filter to reduce com-
`putational complexity by reducing the number of samples
`used in subsequent computations. The decimation filters also
`provide a degree of anti-aliasing filtration to smooth the
`sampled analog signals. The decimated signals are then each
`applied to a corresponding adaptive notch filter to further
`enhance the signal from each sensor. The enhanced output
`signal from each sensor, after being filtered of most noise
`and harmonics, is then applied to a corresponding phase
`computation element
`to determine the phase difference
`between the two enhanced signals. The output of each phase
`computation element is applied to a computation element to
`determine the time difference between the enhanced sensor
`signals and hence the proportional mass flow rate.
`In a second embodiment of the methods of the present
`invention, four adaptive notch filters are utilized,
`two in
`series on each of the left and right channel signals. The two
`
`filters on each of the left and right channels are “cascaded”
`in that the first filter utilizes a low—Q (wide notch) filter to
`supply limited signal enhancement but the ability to rapidly
`converge on changes in the fundamental frequency of the
`vibrating flow tubes. The signal output from the first cas-
`caded notch filter is then applied to a second cascaded notch
`filter. The second notch filter utilizes a high~Q (narrow
`notch) filter to provide superior noise and harmonic rejection
`over that of previous solutions or over that of the first
`embodiment described above. Despite the narrow notch
`(high-Q) of the second notch filter, it can still rapidly adapt
`to changes in the fundamental frequency of the vibrating
`flow tubes due to the limited enhancement (filtration) per-
`formed by the first notch filter. The reduced noise and
`harmonic levels in the signal applied to the second notch
`filter allow it to also rapidly converge on changes in the
`fundamental frequency of the vibrating flow tubes.
`An additional notch filter (fifth filter) having a notch
`shape even wider than that of the first cascaded notch filter
`is used to provide an estimate of the fundamental frequency
`of the vibrating flow tubes. This estimate is used by weight
`adaptation computations to set the frequency parameter of
`the first cascaded notch filters for both the left and right
`channels. The output from the second cascaded notch filters
`is used by weight adaptation computations to adjust the
`frequency parameter of the second cascaded notch filters.
`This combination of two (or more) cascaded adaptive
`notch filters to enhance the output signal from each sensor
`further enhances both the rejection characteristics of the
`filtration and the speed with which the adaptive filters
`converge on changes in the fundamental frequency of the
`vibrating flow tubes.
`The term “adaptive notch filter” as used herein refers
`broadly to a filter with variable parameters. This definition
`contrasts with a more widely accepted definition which
`combines a variable parameter filter with a mechanism for
`automatically tuning the parameters of the filter based on the
`filter’s own inputs and outputs. As used herein, the adapta—
`tion of some notch filters is computed based on the operation
`of other filters rather than each filters own inputs and
`outputs. In other words, some notch filters in the present
`invention are slaved to the operation of other notch filter
`computations. For this reason, the detailed discussions of the
`filters and the adaptation mechanisms are separated. One
`adaptation computation may adjust the parameters for mul-
`tiple notch filters based on inputs from a single filter.
`The above and other aspects of the present invention will
`become apparent from the following description and the
`attached drawing.
`
`BRIEF DESCRIPTION OF THE DRAWING
`
`FIG. 1 shows a typical Coriolis mass flow meter attached
`to meter electronics which embody the apparatus and meth-
`ods of the present invention;
`FIG. 2 shows a block diagram of the computational
`elements within the meter electronics which determine mass
`flow rate through the flow meter in accordance with the
`present invention;
`FIG. 3 shows additional detail of a first embodiment of the
`present invention shown in FIG. 2 wherein a single adaptive
`notch filter is used in conjunction with each sensor signal;
`FIGS. 4—12 show additional detail of the computational
`elements of the first embodiment of the present invention
`shown in FIG. 3;
`
`10
`
`15
`
`20
`
`25
`
`30
`
`35
`
`45
`
`50
`
`55
`
`60
`
`65
`
`24
`
`24
`
`
`
`5,555,190
`
`7
`FIG. 13 shows additional detail of a second embodiment
`of the present invention shown in FIG. 2 wherein two
`cascaded adaptive notch filters are used in conjunction with
`each sensor signal;
`FIGS. 14—16 show additional detail of the computational
`elements of the second embodiment of the present invention
`shown in FIG. 13; .
`
`FIG. 17 is a flowchart of a software implementation of the
`first embodiment of the present invention and depicts inter-
`rupt processing for servicing of an AID converter and
`associated decimation of the samples;
`FIG. 18 is a flowchart of a software implementation of the
`first embodiment of the present invention and depicts pro-
`cessing of decimated samples for purposes of filtering and
`determination of At phase difference;
`FIG. 19 is a flowchart depicting additional detail of an
`element of FIG. 18 which determines updated filter param-
`eters after each decimated sample is processed; and
`FIG. 20 is a block diagram of digital signal processing
`electronics suitable to perform the software methods of the
`present invention.
`
`DETAILED DESCRIPTION OF THE
`PREFERRED EMBODIMENT
`
`A typical Coriolis mass flowmeter 10 is illustrated in FIG.
`1 as having two cantilever mounted flow tubes 12, 14 affixed
`to a manifold body 30 so as to have substantially identical
`spring constants and moments of inertia about their respec-
`tive out-of-phase bending axes W—W and W'—W'.
`A drive coil and magnet 20 are mounted at a midpoint
`region between the top portion 130 and 130' of flow tubes
`12, 14 to oscillate flow tubes 12, 14 out of phase about axes
`W—W and W’—W'. Left sensor 16 and right sensor 18 are
`mounted near the respective ends of the top portions of flow
`tubes 12, 14 to sense the relative movement of flow tubes 12,
`14. This sensing may be done in many ways including by
`measuring the movement of the top ends of the flow tubes
`12, 14 through their zero crossings or some other pre-defined
`point. Flow tubes 12 and 14 have left side legs 131 and 131'
`and right side legs 134 and 134'. The side legs converge
`downwardly toward each other and are aflixed to surfaces
`120 and 120' of manifold elements 121 and 121'. Brace bars
`140R and 140L are brazed to 'the legs of flow tubes 12, 14
`and serve to define the axes W—W and W'~—W' about which
`the flow tubes oscillate out of phase when driver 20 is
`energized over path 156. The position of axes W—W and
`W—W' is determined by the placement of brace bars 140R
`and 140L on flow tube side legs 131, 131' and 134, 134'.
`Temperature detector 22 is mounted on side leg 131 of
`flow tube 14 to measure the flow tube’s temperature and the
`approximate temperature of the material flowing therein.
`This temperature information is used to determine changes
`in the spring constant of the flow tubes. Driver 20, sensors
`16 and 18 and temperature detector 22 are connected to mass
`flow instrumentation 24 by paths 156, 157, 158 and 159,
`respectively. Mass flow instrumentation 24 includes at least
`one microprocessor which processes the signals received
`from sensors 16, 18, and 22 to determine the mass flow rate
`of the material flowing through flowmeter 10 as well as other
`measurements, such as material density and temperature.
`Mass flow instrumentation 24 also applies a drive signal
`over path 156 to driver 20 to oscillate tubes 12 and 14
`out-of-phase about axes W—W and W‘-—W'.
`Manifold body 30 is formed of casting 150, 150'. Casting
`elements 150, 150' are attachable to a supply conduit and
`
`10
`
`15
`
`20
`
`25
`
`30
`
`35
`
`45
`
`50
`
`55
`
`60
`
`65
`
`8
`
`exit conduit (not shown), by flanges 103, 103'. Manifold
`body 30 diverts the material flow from the supply cond