Measurement Automation Monitoring, Dec 2016, no 12, vol 62, ISSN 2450-2855 411 Iryna GRYSHANOVA, Ivan KOROBKO, Pavlo POGREBNIY NATIONAL TECHNICAL UNIVERSITY OF UKRAINE «IGOR SIKORSKY KYIV POLITECHNIK INSTITUTE», 37 Peremogy Ave, Kyiv, 03056, Ukraine Increasing of accuracy of multipath ultrasonic flow meters by intelligent correction Abstract In the paper intelligent compensation of flow errors, caused by profile distortion has been presented This increases the accuracy of ultrasonic multipath meters used in such conditions The goal of this intelligent correction consists in search of optimal layout and minimal sufficient quantity of chords in measurement transducer under different installation effects Keywords: chord layout, multi-path ultrasonic meter, flow profile distortion, intelligent correction Fig 1 5S layout Fig 2 5C layout Fig 3 5P layout 1 Introduction Ultrasonic flow measurement is a well-known and widely applied technology Today there are a lot of ultrasonic flow meters with different number of acoustic paths Single path meters are less expensive and commonly used but having not-enough straight pipe sections they can t provide highly accurate measurements They are very sensitive to non-ideal flow profiles caused by flow distortions It is known that measurement errors of ultrasonic flowmeters with single diametric path are larger than for the meter with dual paths and more To eliminate this problem multipath meters with different quantity of chords and their arrangement are applied But this approach may be very cost-based That is why optimization of topology of acoustic paths with the view of minimal sufficient quantity of chords is very important question So, an aim is to further improve the accuracy of ultrasonic multipath flow meters without complication of internal geometry of measurement transducer and without increasing the number of measuring channels just by optimizing and improving intelligent component of measuring device Intelligent correction is a new direction of research that became available only in the past decade with increasing of computing capacity 2 Selection of chord layouts for research Fig 4 4S layout arrangement Fig 7 3S layout arrangement Fig 5 4C layout Fig 8 3C layout Fig 6 4P layout arrangement Fig 9 3P layout Practically all methods realized in multipath meters are based on velocity distribution There are two methods of calculation, which are used to derive the mean velocity of a multipath ultrasonic meter, namely, averaged and integrated methods The last ones are based on weighted sum of the velocities along the individual paths while the first ones involve an equally weighted average of the path velocities For multipath meter numerical integration method gives position of each acoustic path together with weighting factor Quadrature methods were introduced and compared by many researchers These methods are pretty good discussed in literature [1-3] but we would like to propose another approach Research objectives: - Discovering of the impact of flow under different installation effects on meter s performance basing on mathematical model for multipath ultrasonic flow meter, - Optimization of chord layout, - Development of intelligent system of correction of readings of multipath ultrasonic meters, For these purposes we propose the set of following research configurations (Figs 1 12) Fig 10 2S layout Fig 11 2C layout 3 Error caused by flow distortion Fig 12 1P layout As far as flow distortion leads to profiles without rotational symmetry about the pipe axis the performance of chords layouts we will assess at different orientation angles relatively to the flow profile (Re=const) (1) where is set average velocity on the inlet of the pipeline without impact of flow distortion;
412 Measurement Automation Monitoring, Dec 2016, no 12, vol 62, ISSN 2450-2855 (2) where is average velocity in consideration of orientation angle of the meter relatively to the flow profile; is number of chords; h i is a distance from a chord to the center of flow metering section; k H is a weighted factor Values of this factor were founded during simulation of flow in the straight pipeline without impact of flow distortion on the distance of 30 DN from the inlet where the flow becomes fully developed (see Fig 13) At this we use the same turbulence model as while simulating real flow under different installation effects Fig 16 Pattern of spatial Fig 13 Reference flow profile on the distance of 30 DN from the inlet of the pipeline As a result we represent a table with values of for different Tab 1 Values of weighted factor for flow profile on the distance of 30 DN from the inlet of the pipeline Fig 17 Flow profiles after spatial 90 0 - bend on the distances of: a) 5DN, b)10dn, and c)-20dn 0 01 02 03 04 09722 09726 09746 09795 09867 05 06 07 08 09 10001 10277 10718 11302 12814 We will assess errors caused by flow distortion after (Fig 14), spatial bend (Fig 16), orifice plate (Fig18), contraction (Fig 20) and expansion (Fig 22) on distances 5DN, 10DN, 20DN (Tab 3, 4, 5) Fig 18 Pattern of orifice plate Fig 19 Flow profiles after orifice plate on the distances of: a) 5DN, b) 10DN, and c)-20dn Fig 14 Pattern of Flow profiles in cross-sections after in Fig 15 are presented Fig 15 Flow profiles after 90 0 - bend on the distances of: a)- 5DN, b)-10dn and c)-20dn Fig 20 Pattern of contraction with angle
Chord layout i (designation) Chord quantity Quantity of generated models Boundary value of correction error Reliability Number of type II errors Probability If significance level meets the condition Measurement Automation Monitoring, Dec 2016, no 12, vol 62, ISSN 2450-2855 413 Fig 21 Flow profiles after contraction with angle on the distances of: a) 5DN, b) 10DN, and c) 20DN Fig 22 Pattern of expansion with angle Unfortunately theoretical determination of this criterion causes difficulties, that is why we will apply one of Monte Carlo approaches [4] For each of layouts there is a set of flow profiles For every case of layout we generate N times flow model considering rotation angle of measurement transducer relatively to flow distortion, where d{ } (number of flow distortion), d{ } (number of distance between measurement transducer and flow distortion), (considering rotation angle of measurement transducer relatively to flow distortion) Further, setting boundary value of error of intelligent correction, we will find the most optimal correction model for every case and check its adequacy If this model is incorrect, then we get the type II error Having number of type II errors, number of generated models and set power of test we keep this chord layout if significance level meets the condition In other case we make a conclusion that for specified layout quantity of chord is insufficient Let s evaluate all chords layouts, specifying number of generated models, boundary value of correction error with the difference depending on chord quantity, reliability (see Table 5) This table gives us a possibility to formulate the criterion of chords sufficiency at set limitations But it will not be accurate enough because it demands more flow profiles and appropriately generated models It can be seen that at set limitations the criterion of chords sufficiency for all layouts is So if chords quantity an intelligent correction of results can be applied At this boundary value of correction error will decrease while chords quantity increasing Fig 23 Flow profiles after expansion with angle b)10dn, and c) 20DN on the distances of: a) 5DN, Tab 5 Determination of criterion of sufficient chords quantity at set limitations As far as these investigations are very volumetric we will demonstrate them only by example of But looking at this huge data array there is a reasonable question how to evaluate all obtained information to make correct conclusions We propose to create a system of intelligent correction 4 Intelligent compensation of errors caused by flow distortion To create intelligent compensation system we should determine: - Chord layout; - Quantity of chords for given layout; - Flow profile caused by certain flow distortion Let s designate chord layout considering number of chords as, where is layout number ( ), is quantity of chords in layout ( ) Flow profile we designate as, where is number of flow distortion ( ), is number for distance from flow meter to flow distortion (after (Fig 14), spatial bend (Fig 16), orifice plate (Fig18), contraction (Fig 20) and expansion (Fig 22)) ( ) Mentioned values are equal to:,,, 1(S) 2(C) 3(H) 5 1000 0001 099 0 1000 + 4 1000 0002 099 1 0999 + 3 1000 0003 099 3 0997 + 2 1000 0005 099 11 0989-1 1000 0010 099 23 0977 5 1000 0001 099 1 0999 + 4 1000 0002 099 1 0999 + 3 1000 0003 099 5 0995 + 2 1000 0005 099 16 0984 1 1000 0010 099 23 0977 5 1000 0001 099 3 0997 + 4 1000 0002 099 3 0997 + 3 1000 0003 099 5 0995 + 2 1000 0005 099 16 0984 1 1000 0010 099 23 0977 41 Criterion of sufficient quantity of chords For correct realizability of intelligent compensation for ultrasonic flow meter the criterion of sufficient quantity of chords should be represented It means to include minimal chord quantity for layout ( ), at which with preset error that could be brought by intelligent correction right compensation model will be selected with specified power of test (reliability) So, type II error ( ) will not be made 42 Compensation based on direct search Assessment of selection criterion for compensation model To create an intelligent model of error compensation for multipath ultrasonic flow meter under impact of installation effects it is necessary to have a base of compensation models
414 Measurement Automation Monitoring, Dec 2016, no 12, vol 62, ISSN 2450-2855 Compensation model implicates relative data obtained along each chord For each chords layout, where is layout number, is chords quantity in layout, sufficient number of compensation models should be generated Tab 2 Errors caused by flow distortion after 90 0 -bend on the distance of 5 DN Cord q-th For S layout For C layout For P layout 5 4 3 2/1 Tab 3 Errors caused by flow distortion after 90 0 -bend on the distance of 10 DN Cord q-th For S layout For C layout For P layout 5 4 3 2/1
Measurement Automation Monitoring, Dec 2016, no 12, vol 62, ISSN 2450-2855 415 Tab 4 Errors caused by flow distortion after 90 0 -bend on the distance of 20 DN Cord q-th For S layout For C layout For P layout 5 4 3 2/1 Denote the set of compensation models for one of chords layouts as, where is number of flow disturbance ( ), is number of upstream straight pipe section ( ), is number of rotation angle of flow disturbance regarding ultrasonic meter Sufficiency criterion of compensation models should be built at sampling frequencies of distance interval between ultrasonic meter and flow disturbance, the full angle of rotation relative to flow disturbance ( - quantization), and also number of commonly encountered flow disturbances For each model we set,, and get the capacity for compensation model set, which obviously does not satisfies the criterion of sufficiency, but is sufficient for further research Each compensation model is set of elements with number, which is equivalent to chord quantity in certain layout where is compensation factor along th chord for number of distance from flow meter to th flow disturbance, number of selected flow disturbance and number of revolution angle of flow disturbance relatively flow meter for selected chord layout is the velocity along the th chord; is the average velocity along all chords adjusted for ; is the hydrodynamic factor associated with chord ; In ideal case, when there is no influence of flow disturbance Intelligent compensation will be done by searching the database corresponding compensation models The criterion for selection of compensation model will be acceptable deviation of sum of squared errors of compensation factor of meter s chords layout (3) (4) regarding compensation models from the database The relative error of chords layout is: where is weight factor along of meter s chord j It is necessary to create a function that describes the dependence of error of choosing the compensation factor of chord from the error submitted into meter readings because of intelligent correction : Knowing this dependence for each chord layout we can set boundary value of error from intelligent correction and find permissible error of choosing the compensation factor of chord: If after search in database we found set of permissible compensation models which elements meet the condition: Then at next step we choose the best acceptable model under criterion of minimum error of chord compensation factor: Dependence can t be created analytically, and also can t be plotted once as a dependency But we can embrace by experiment to find error for pointwise values of deviation And after that approximation of those values will give us dependency Taking one of approximated profiles developed after flow disturbance with number, located on the distance with number from flow disturbance for selected layout it is possible programmatically by changing revolution angle of the meter arrangement relatively to flow disturbance create two chord (5) (6) (7) (8)
416 Measurement Automation Monitoring, Dec 2016, no 12, vol 62, ISSN 2450-2855 layouts For these layouts relative to each other will be equal to given with so high accuracy that is possible to consider is not a random quantity ( ) If you generate a finite number of such variants for each flow profile, then you get sampling of errors from intelligent correction of meter readings value Let s conduct appropriate research for layout for set of points CF ={00002, 00005, 0001, 0002, 0005} (see Table 6) Under research for all samplings it has been put forward a null hypothesis about normal law of error, which was confirmed with reliability of 095 according to 2 Pearson criterion Tab 6 Statistical observations for samplings generated for given (reliability 095) Null hypothesis about normal law of error (reliability 095) 0001 000002 000021 0000412 confirmed 0002 000001 000038 0000745 confirmed 0005 000001 000050 000098 confirmed 001 000005 000130 0002548 confirmed 002 000012 000590 0011564 confirmed Here E is mathematical expectation, is root-mean-square error Considering normality of distribution the mathematical expectation and root-mean-square error have been obtained Findings lead us to error value which is brought into meter readings because of intelligent correction with 095 reliability Basing on data of Table 6 we will plot the dependency (Fig 24) Fig 24 δ cor 002 001 dependency 5 Discussion of results It can be seen that ratio So, for the error of chord compensation factor the error which is brought by correction will be approximately twice less This dependency gives us a possibility to specify a criterion of choosing of compensation model For example, if we are satisfied with error of intelligent compensation (but at this we can compensate the error from impact of flow disturbance on the order of few percent), then we should choose the criterion 6 Conclusions 0 0 0005 001 0015 002 0025 δ CF For appropriate intelligent correction of ultrasonic multipath meter the criterion of sufficient chords quantity has been applied This criterion consists in minimal chords quantity for layout ( ), at which with set error, that can brought by intelligent correction the right compensation model will be chosen with set power of test (reliability), so the type II error ( ) will not be committed Then specified criterion has been formulated numerically for each layout with reliability Further based on previous criterion the system of intelligent correction have been developed Compensation applies direct search of compensation model, for which evaluation of selection criterion of such model is provided So, there is a clear dependence between error from compensation model selection and error of determination of this model To create mentioned system the set of chord layouts was prepared and CFD simulations in different conditions, considering impact of flow disturbance, straight pipe distances, rotation angle of measurement transducer have been conducted 7 References [1] Moore Pamela I, Brown Gregor J, and Stimpson Brian P: Ultrasonic transit-time flowmeters modelled with theoretical velocity profiles: methodology Meas Sci Technol, Vol 11 pp 1802-1811, October 2000 [2] Zheng D, Zhang P, Zhang T, Zhao D:A method based on a novel flow pattern model for the flow adaptability study of ultrasonic flowmeter, Flow measurement and Instrumentation, Vol 29, pp 25-31, 2013 [3] Zanker KJ: The effects of Reynolds number, wall roughness, and profile asymmetry on single- and multi-path ultrasonic meters In Proc of North Sea Flow Measurement Workshop, 1999, pp117-129, Oslo, Norway [4] Wojtishek AB: Osnovy metoda Monte-Carlo w algoritmach i zadaczach Nowosibirsk, izd NGY, 1999 Received: 21092016 Paper reviewed Accepted: 02112016 Iryna GRYSHANOVA, Associate Professor Iryna Gryshanova is an associate professor in National Technical University of Ukraine Igor Sikorsky Kyiv Polytechnic Institute She received MSc in Mechanical Engineering, 1998, PhD in Mechanical Engineering, 2002, in National Technical University of Ukraine Kyiv Polytechnic Institute, Kyiv, Ukraine Researcher with over 19 years experience, specialized in fluid flow measurement (water & wastewater), heat measurement, having both a solid academic background and technical experience e-mail: irgryshanova@gmailcom Ivan KOROBKO, PhD, eng He graduated Kyiv Polytechnic Institute on specialty Precision mechanics instruments in 1980 In 1993 received scientific degree of candidate of engineering sciences In 2015 defended his doctoral thesis on specialty Devices and methods of measurement of mechanical quantities His research interests include: measurement of liquid flow rate and volume of fuel and energy resources and water, energy saving and efficient energy systems e-mail: ikorobko@kpiua Pavlo POHREBNYI, MSc, eng He specializes in theoretical and applied computer science, programming, big data analysis, information systems, mathematical modeling systems and computer simulation He works as a senior engineer at SoftServe inc in Kyiv, Ukraine e-mail: pogrebnij@gmailcom