Jtam-A4.dvi JOURNAL OF THEORETICAL AND APPLIED MECHANICS 50, 4, pp. 889-902, Warsaw 2012 50th Anniversary of JTAM UNIVERSAL WEIGHTING FUNCTION IN MODELING TRANSIENT CAVITATING PIPE FLOW Kamil Urbanowicz, Zbigniew Zarzycki, Sylwester Kudźma West Pomeranian University of Technology, Department of Mechanics and Machine Elements, Szczecin, Poland e-mail: kamil.urbanowicz@zut.edu.pl; zbigniew.zarzycki@zut.edu.pl; sylwester.kudzma@zut.edu.pl This work concerns modeling of time-varying resistance during transient cavitating liquid pipe flow. The wall shear stress is presented in the way introduced by Zielke as the integral convolution of liquid local velocity changes and a weighting function. A newprocedure for determination, so-called, a universal laminar-turbulentweighting func- tion, which combines functions of Zielke and Vardy and Brown or Zielke and Zarzycki, is presented. Based on these weighting functions, the method of simulation transients in the pressure lines in awide range ofReynolds numbers is presented. It eliminates numerical pro- blems associatedwith the change of laminar flow into turbulent and vice versa.An example application of this method for simulation of water hammer is presented. The calculation results are compared against experiments presented by Bergant and Simpson, and good agreement is found. Key words: transient pipe flow, unsteady wall shear stress, weighting function, cavitation 1. Introduction One of the most common methods used for unsteady pipe flow simulation is the method of characteristics (MOC). To use this method, a model of the wall shear stress (and thus the hydraulic resistance) has to be assumed. The most popular and easiest, in terms of numerical calculations, is theassumptionof a quasi steadymodel. Its determination is basedon the formula of Darcy-Weisbach. This assumption, however, is correct only for slow changes in the liquid velocity field in the pipe cross-section, i.e. for low frequencies in the case of pulsating flow, and in the case of waterhammer for the first-wave cycle. This model does not take into account a time-varying velocity gradient in the radial direction, and thus the variable energy dissipation. The assumption of quasi-steady flow significantly deviates from the reality, especially for fast transients. In this case, it is appropriate to present the instantaneous wall shear stress τ as a sum of the quasi-steady τq and unsteady τu components, i.e. τ = τq+τu (1.1) There are two main groups of the unsteady wall shear stress τu models. In the first group, the wall shear stress is proportional to the acceleration of the liquid, both the local acceleration (Daily et al., 1956; Cartens and Roller, 1959; Sawfat and Polder, 1973) and convective accele- ration (Brunone et al., 1991). Vitkovsky et al. (2000) improved the model of Brunone et al. by introducing a “sgn |v|” to the component of the convective acceleration. The second group of models of the unsteady wall shear stress τu is based on the history of the instantaneous local velocity changes of flow. These models, unlike the first group, better reflect the experiment not only in the degree of attenuation of pressurewaves, but also its shape. They are based on the two-dimensional (2D) equation ofmotion, so they take a time-dependent distribution of the liquid velocity field in cross-section of the pipe. 890 K. Urbanowicz et al. Zielke (1968) first introduced the unsteady wall shear stress τu as a integral convolution of a weighting function and mean local acceleration of the liquid. His model concerned the unste- ady laminar flow. Using this model to simulate transients requires a large computer memory. The computation time grows exponentially in this case as the number of time steps increases. Therefore, the model underwent further improvements by Trikha (1975), Kagawa et al. (1983), Suzuki et al. (1991), Schohl (1993) making it much more efficient. In the case of unsteady turbulent flow, the models of unsteady shear stress τu are based mainly on the distribution of eddy viscosity coefficient in the pipe cross-section which was determined for a steady flow. The authors of these solutions use different numbers of flow layers in which the distribution of eddy viscosity is described. There are two-layer models (Vardy et al., 1991, 1993; Vardy and Brown, 1995, 2003), three-layer (Brown et al., 1969) and four-layer (Ohmi, 1976, Zarzycki 1994, 2000; Zarzyski and Kudźma, 2004). It is worth to mention that the above cited models are related to smooth pipes, however, Vardy and Brown in 2004 applied the idealized viscosity distribution model for the flow in a rough pipe. Rahl and Barlamond (1996) and then Ghidaoui andMansour (2002) reported that the use of Vardy and Brown two-layer model to simulate waterhammer gives good consistency with the experiment. The main problem of this work is focused on simulations of unsteady cavitating pipe flows taking into account the time-dependent hydraulic resistance. The majority of authors in their publications assume that unsteady friction, in case of turbulent flow, is dependent on a fixed Reynolds number (representing the flow before the transient) which influences the shape of the weighting function. A novelty of this work is an attempt to take into account, during the numerical calculation of transient, a variable Reynolds number which constantly modifies the shape of the weighting function. 2. Mathematical models of transient cavitating pipe flow In order to simulate transient cavitating pipe, flow two common models of cavitation – CSM (column separationmodel) andBCM(bubbly cavitationmodel) are used.Thefirstmodel,CSM, is a discrete model. It assumes that cavitation takes place at a given place of liquid flow and that leads to a disturbance of this flow continuity. In this case, the liquid flow in all flow areas is described by equations [30]: — continuity equation ∂p ∂t +ρlc 2∂v ∂x =0 (2.1) —momentum equation ∂p ∂x +ρl ∂v ∂t +ρlg sinγ+ 2 R τ =0 (2.2) where: v = v(x,t) is the mean velocity of the liquid in the pipe cross-section, p = p(x,t) – mean pressure in the pipe cross-section, R – inner radius of the pipe, τ – wall shear stress, ρl – liquid density, g – acceleration due to gravity, γ – the angle of inclination of the hydraulic line, c – velocity of pressure wave propagation, t – time, x – distance along pipe. The next model, BCM, is a continuous, homogeneous cavitation model, where cavitation is assumed to take place along the axis of a pipeline. The fundamental flow equations are given below (Shu, 2003): — continuity equation 1 c2 ∂p ∂t +(ρl−ρv) ∂α ∂t +ρm ∂ ∂x (v α ) =0 (2.3) Universal weighting function in modeling... 891 —momentum equation ρm ∂ ∂t (v α ) + ∂p ∂x + 2 R τw+ρmgsinγ =0 (2.4) where: ρm is the density of mixture, α – coefficient of liquid phase concentration, ρv – vapour density. Themixture density ρm can be expressed ρm =αρl+(1−α)ρv (2.5) It should be noted that the system of equations (2.1), (2.2) and (2.3), (2.4) is not closed, because it contains three variables: v, p and τ. Thus the main emphasis is placed on the work onmodeling the unsteady wall shear stress τ. Among all methods enabling the solution of equations of this type, particularly noteworthy is the method of characteristics (MOC), which perfectly interprets the essence of the natural phenomenon of unsteady flow, and at the same time, characterized by rapid convergence as well as ease of taking into account different boundary conditions and high accuracy of calculation results. Method of characteristics enables one to solve partial differential equations of quasi-linear hyperbolic type (2.1), (2.2) and (2.3), (2.4) in a simple way (Shu, 2003; Wylie and Streeter, 1993). The solution is to find the equivalent to the four ordinary differential equations which is then solved using finite differences schemes. An approximation of the finite difference method of the first order gives satisfactory results (Streeter and Lai, 1962). Zielke in hiswork (Zielke, 1966, 1968) presented the inverse Laplace transformof a hydraulic line impedance, shown by Brown (Brown, 1962; Brown and Nelson, 1965), and received the following relation describing the instantaneous wall shear stress τ(t)= λρv|v| 8 ︸ ︷︷ ︸ τq + 2µ R t∫ 0 w(t−u) ∂v ∂t (u) du ︸ ︷︷ ︸ τu (2.6) where λ is the coefficient of friction. The first component τq in equation (2.6) represents the quasi-steady wall shear stress while the second τu (which is the integral convolution of local fluid acceleration and a weighting function) takes into account the impact of unsteady flow on the wall shear stress. It can be easily seen that for a fixed flow, in which there is no local acceleration of the liquid, τu =0. Zielke (1968) presented also the proper weighting function for laminar flow. For turbulent flow there are two well known weighting functions: Vardy’s and Brown’s (Vardy and Brown, 2003, 2004) and Zarzycki’s (Zarzycki, 2000; Zarzycki and Kudźma, 2004). For effective solution of the second component of thewall shear stress τu (2.6), theweighting functionmust be a finite expression of exponential terms (preferably nomore than 30 phrases). In our latest article (Urbanowicz andZarzycki, 2012) was presented the newweighting functions of effective type, which are characterized by an increased range of applicability, and a very good degree of fit to their prototypes, i.e. the classical weighting function according to Zielke for laminar flow (Zielke, 1968) and according to the Vardy and Brown (2003, 2004) (alternatively according to Zarzycki (2000) and Zarzycki and Kudźma (2004)) for turbulent flow. In this work, a novelty is the inclusion in the model describing unsteady flow in pipes a time-varying hydraulic resistance which is calculated in an effective manner for both laminar and turbulent flow. 892 K. Urbanowicz et al. 3. The universal weighting function For some time, one can note two types of approaches formodeling unsteadyhydraulic resistance: • The older approaches (Bergant et al., 2006; Shu, 2003; Vitkovsky et al., 2004) assume the shape ofweighting function selection before the simulation on the basis of knownReynolds number Re=Reo (constant) – for the steadyflow shortly before the onset of the transient. When the number Reo is less than or equal to the critical value = 2320, the shape of the weighting function is consistent with the shape of the classic weighting function for the laminar flow by Zielke (1968) but when there is a turbulent flow, the shape is determined using a well-known classical turbulent weighting function by Vardy and Brown (2003) (alternatively by Zarzycki (2000)). This approach assumes constancy of the chosen shape for the entire course of the transient simulation. • In later approaches (Kudźma, 2005; Kudźma and Zarzycki, 2005; Zarzycki and Kudźma, 2004), changes in the shape of a weighting function due to Reynolds number variation are taken into account. There is no assumption of a single weighting function based on initial conditions, but to date, together with any change in the current local Reynolds number Rec (in the range of turbulent flow) its new shape is determined. To the best knowledge of authors, there are no papers on the simulation and experimental verification of the transient turbulent cavitating flow in a long liquid line associated with the starting of turbo-machinery with focus on unsteady friction problems. Hence, it can only be considered as authors using the first approach described above could determine a weighting function – probably from the known Reynolds number after reaching a steady state Res. Such an approach is rather simplistic to the problem of modeling unsteady hydraulic resistance. New efficient weighting functions described in detail in our preview article (Urbanowicz and Zarzycki, 2012) are suitable for simulation of transient flows in the case of the first of those approaches. Fig. 1. Courses of weighting functions (determined at Recr =2320): (a) log-linear scale, (b) log-log scale Detailed analysis of the second approach shows that transition from the laminar weighting function to the turbulent one and vice versa raises the problem of numerical nature. This issue is related to inconsistency of the laminar and turbulentweighting functions for a critical number of Recr =2320 (see Fig. 1). It is reflected in the variation of pressure and velocity flow simulated runs in the form of abrupt faults. In connection with a different number of components that make up the effective weighting function for laminar flow (26 terms) and the effective weigh- ting function for turbulent flow (16 terms of Vardy-Brown function – alternatively 24 terms of Universal weighting function in modeling... 893 Zarzycki’s function) (Urbanowicz and Zarzycki, 2012) the current numerical procedure for the transition from one type of flow into another assumes parameters from the equation: τu = 2µ R k ∑ i=1 ( yi(t)e −ni∆t̂+mie −ni ∆t̂ 2 [v(t+∆t)−v(t)] ︸ ︷︷ ︸ yi(t+∆t) ) (3.1) which takes into account the history in changes of velocity to be equal to zero yi(tt)=0 (where tt – time of transition). This assumption is valid as an initial condition (for a steady flow), but it certainly is a source of errorwhen there is transition from laminar to turbulent unsteady flow. To avoid the above mentioned problems by using the second approach, in the present work an effective universal laminar-turbulent weighting function model is proposed. Its shape in the case of laminar flow will coincide with Zielke’s function (unchanging when Rec ¬ 2320) while for turbulent flow with Vardy’s and Brown’s turbulent weighting function (alternatively with Zarzycki’s turbulent weighting function). Detailed analysis of the courses of classical weighting functions according to Zielke (1968), Vardy and Brown (2003) and Zarzycki (2000) shows that for the critical Reynolds number (Recr =2320) these functions have different shapes (Fig. 1). The function according to Zarzycki in the range of dimensionless time (10−10 < t̂ < ∞) takes values greater than the weighting function according to Zielke. In practice, it means that the simulation results with its usage will becharacterized by increaseddamping.The functionaccording toVardyandBrown for the range of dimensionless time 7.244 ·10−4 < t̂<∞) takes instead values smaller than Zielke’s function. Thatmeans that simulations with its use will be characterized by lower attenuation than when Zielke’s or Zarzycki’s functions are used. It was also noted that the shape of this function is greatly changed with the increase of Re. The larger the Reynolds number characterizing the transient flow the faster values of this function tend to zero (see Fig. 2). Fig. 2. Runs of the Vardy and Brownweighting function: (a) log-linear scale, (b) log-log scale Figure 3 presents a schematic procedure for determination of the universal laminar-turbulent weighting function. In the case of laminar flow, the universal weighting function is characterized by a shape consistent with Zielke’s weighting function (Zielke, 1968). As soon as there is a turbulent flow, the coefficients representing the effective function must be appropriately scaled. This process is modeled with the use of the procedure presented by Vitkovsky et al. (2004). Assuming that the shape of the universal weighting function in the case of turbulent flow should be convergent to the original model of Vardy and Brown (2003), the process of finding the current coefficients should be carried out according to the scheme shown in Fig. 3 (an alternative procedure is presented in Appendix – there the function for the case of turbulent flow are convergent to the original model of Zarzycki (2000)). 894 K. Urbanowicz et al. Fig. 3. Schematic determination of the current form of the universal weighting function wc in the numerical process. It retains the shape consistent with the classical weighting function according to Vardy and Brown (2003) Universal coefficients, needed for determination of the current formof theuniversalweighting function are calculated (for Recr = 2320) before the start of the simulation using following formulas m1u = m1 A∗ m2u = m2 A∗ · · · m26u = m26 A∗ n1u =n1−B ∗ n2u =n2−B ∗ · · · n26u =n26−B ∗ (3.2) where A∗ = √ 1 4π B∗ = Reκ 12.86 = 2320κ 12.86 κ= log 15.29 Re0.0567 = log 15.29 23200.0567 and mi, ni are the coefficients of the laminar effective weighting function presented in our previous paper (Urbanowicz and Zarzycki, 2012). Degrees of match of the universal weighting function with the classical counterparts are shown by the graphs presented in Fig. 4. Qualitative analysis of the new universal functionmatching shows that it is difficult to note any significant deviations, for Reynolds numbers Re ­ 104 (Figs. 4b-4e), from its classical counterpart. The quantitative analysis for the dimensionless time range 10−9 ¬ t̂ ¬ 10−3, and Reynolds numbers 2320 ¬ Re ¬ 107 shows that the absolute percentage error does not exceed 11% (for the case presented in Appendix, the percentage absolute error in the same ranges did not exceed 14%). The relative percentage error, as shown inFig. 4f, is large but only for values of dimensionless time t̂ > 10−3. The values of weighting function for the dimensionless time t̂ > 10−3 are very small – as shown in Fig. 5. This means that if the dimensionless time step is significantly small during numerical calculations, these errors will not affect significantly the result of simulation of the wall shear stress. Universal weighting function in modeling... 895 Fig. 4. Comparison of the universal weighting function (maintaining the shape consistent with the model of Vardy and Brown) with the classical models Fig. 5. The values of turbulent weighting function for the dimensionless time 10−3 896 K. Urbanowicz et al. 4. Numerical example In order to compare the accuracy of unsteady (with the use of a universal weighting function) and quasi-steady models of friction in relation to experimental data, simulations of a simple waterhammer case (tank – long liquid line and cut-off valve) were conducted. The computed results (using transient cavitating pipe flowmodels – CSM (Wylie and Stre- eter, 1993) and BCM (Shu, 2003)) were compared with the experimental data reported by Bergant and Simpson (1996, 1999). Bergant and Simpson conducted an experiment on a test rig installed at the University of Adelaide (Fig. 6). The rig comprises of copper pipe with internal diameter D = 0.0221m and length L=37.2m connecting two pressurized tanks. The liquid used in the experimentwas water havingkinematic viscosity ν =1·10−6m2/s.Themeasured sound speedwas c=1319m/s and the initial flow velocity v0 =1.40m/s (Re=30940) and v0 =1.50m/s (Re=33150). Fig. 6. Test rig The downstream valve was rapidly closed in the pipe line during flow. Pressure fluctuation wasmeasuredat the endpoint of the line (near the valve). Fromthe aboveparameters, it followed that it was a case of turbulent flow. The results of simulations compared to experimental data are shown in Figs. 7 and 8. Fig. 7. Pressure variations at the valve: v0 =1.4m/s (Re=30940), pt =2.158 ·10 5Pa (a) quasi-steady frictionmodel, (b) unsteady frictionmodel From the above graphs showing the runs of pressure changes (at the valve), it is clear that the use of unsteady hydraulic resistance in simulation brings the results much closer to the experimental observations. In the case of cavitating flow, the model of friction used affects Universal weighting function in modeling... 897 Fig. 8. Pressure variations at the valve: v0 =1.5m/s (Re=33150), pt =3.139 ·10 5Pa (a) quasi-steady frictionmodel, (b) unsteady frictionmodel not only the maximum pressure values for consecutive amplitudes of pressure, but also the duration of the cavitation. Therefore, the role of the frictionmodel is essential for themodeling of transients in the pressure lines with cavitation. 4.1. Quantitative analysis A quantitative analysis of transient flow, especially valuation of maximum pressures values and times of its occurrences, is very important from the point of view of automatic control engineering. In many available articles concerning transient flow with cavitation, there are no shown mathematical quantitative methods useful for comparison of simulated runs in relation to the experimental ones. All authors concentrated only on qualitative estimation. The quantitative analysis in this research work is estimated by two parameters (pp and tp) characterizing the degree of matching with the experimental course. Fig. 9. An example of pressure course during transient pipe flowwith cavitation Parameter pp is determined on the basis of maximum pressures (Fig. 9 – p1 to pn) of successive amplitudes of the analyzed course in the following way pp = n∑ i=1 |ppi| n (4.1) where ppi = pis−pie pie ·100% (4.2) 898 K. Urbanowicz et al. and pie is the value ofmaximumpressure on the analyzed i-th amplitude, based on the analysis of experimental results, pis – maximum pressure on the analyzed i-th amplitude, based on the simulated results. Similar analysis can be carried out for the time of occurrence of the pressure amplitudes (their maxima) (Fig. 9 – t1 to tn) tp = n ∑ i=1 |tpi| n (4.3) where tpi = tis− tie tie ·100% (4.4) and tie is the time of occurrence of the i-th amplitude of pressure, based on the analysis of experimental results, tis – time of occurrence of the i-th amplitude of pressure, based on the analysis of simulated results. When the values of pp and tp get smaller, then the discrepancy between the simulated and experimental results is smaller. The results of quantitative analyses of the pressure runs, presented in Section 4, are shown in Tables 1 and 2. Table 1.Quantitative analysis – time-dependent hydraulic resistance Parameter CSM BCM model model pp [%] (Run I – Re=30940) 0.7 1.2 pp [%] (Run II – Re=33150) 2.3 3.1 tp [%] (Run I – Re=30940) 2.7 2.7 tp [%] (Run II – Re=33150) 2.8 2.8 Table 2.Quantitative analysis – quasi-steady hydraulic resistance Parameter CSM BCM model model pp [%] (Run I – Re=30940) 7.5 8.0 pp [%] (Run II – Re=33150) 3.7 3.8 tp [%] (Run I – Re=30940) 7.0 6.4 tp [%] (Run II – Re=33150) 8.5 8.0 The quantitative analysis shows that the results of simulations with usage of the CSMmo- del (with taking into account also time-dependent hydraulic losses) have the best aggrement in simulating maximum pressures. The presented analysis confirms also the conclusion from qualitative analysis (Figs. 7 and 8) that the results originated from models in which the time- dependent hydraulic resistance were includedmatch definitely better with the experiment than models using of quasi-steady ones (see percentage results in Tables 1 and 2). 5. Summary Thepresentedmethod enables the determination of variable hydraulic resistance for laminar and turbulent flow. A new function includes a range of laminar and turbulent flow. The coefficients of the new function are updated every step of the calculation along with the change of Re. Universal weighting function in modeling... 899 Using the proposed approach eliminates numerical problems associated with the transition from laminar flow into turbulent one and vice versa. In view of the fact that in literature there are two weighting functions for the turbulent flow, two approaches are presented at the time of occurrence of turbulent flow. From the numerical examples showing the runs of pressure changes (at the valve), it is clear that the use of time-dependent hydraulic resistance in simulation brings the results much closer to the experimental observations. Important is also the fact that in the case of cavitating flow, the used model of friction affects not only the maximum pressure values for consecutive amplitudes of pressure but also the duration of cavitation. Therefore, the role of the friction model is essential for modeling transients in the pressure lines with cavitation. A. Appendix Assuming that the shape of the universal weighting function remained convergent to the model by Zarzycki in the case of turbulent flow, the process of searching for current coefficients of the function can be donewith the use of the scheme shown in Fig. 10. Universal coefficients, needed further for determination of the current form of the universal weighting function, are computed before starting the simulation process from the following relations m1u =Dm1 m2u =Dm2 · · · m26u =Dm26 n1u =n1 n2u =n2 · · · n26u =n26 (A.1) where D= Re0.005535 0.299635 = 23200.005535 0.299635 and mi, ni are the coefficients of the laminar effective weighting function presented in our previous paper Urbanowicz and Zarzycki (2012). Fig. 10. Schematic determination of the current form of the universal weighting function wc in the numerical process. It retains the shape consistent with the classical weighting function according to Zarzycki for turbulent flow (Zarzycki, 2000) Degrees of match of the universal weighting function with the classical counterparts one shown by the graphs presented in Fig. 11. 900 K. Urbanowicz et al. Fig. 11. Comparison of the universal weighting function (maintaining the shape consistent with the model of Zarzycki) with the classical models References 1. Bergant A., SimpsonA.R., 1996, Interesting lessons from column separation experiments,Pro- ceedings of the 7th International Conference on Pressure Surges and Fluid Transients in Pipelines and Open Channels, 19, BHRGroup, 83-97 2. BergantA., SimpsonA.R., 1999,Pipeline column separationflow regimes,Journal of Hydraulic Engineering, August, 835-848 3. Bergant A., Vitkovsky J., Simpson A.R., Lambert M., Tijsseling A.S., 2006, Discrete vapour cavity model with efficient and accurate convolution type unsteady friction term, Int. J. Kurokawa, Proceedings 23rd IAHR Symposium on Hydraulic Machinery and Systems, Yokohama, Japan, Paper 109 IAHR 4. Brown F.T., 1962, The transient response of fluid lines,Trans. ASME, Journal of Basic Engine- ering, 84, 547-553 Universal weighting function in modeling... 901 5. Brown F.T., Margolis D.L., Shah R.P., 1969, Small amplitude frequency behaviour of fluid lines with turbulent flow,Trans. ASME, Journal of Basic Engineering, 91, 4, 678-692 6. Brown F.T., Nelson S.E., 1965, Step responses of liquid lines with frequency-dependent effects of viscosity,Trans. ASME, Journal of Basic Engineering, 87, 504-510 7. Brunone B., Golia U.M., Greco M., 1991, Some remarks on the momentum equations for fast transients, Proceedings of the International Meeting on Hydraulic Transients with Column Separation, 9th Round Table, IAHR, Valencia, Spain, 201-209 8. Carstens M.R., Roller J.E., 1959, Boundary-shear stress in unsteady turbulent pipe flow, ASCE Journal of the Hydraulics Division, 85 (HY2), 67-81 9. 9] Dailey J.W., Hankey W.L., Olive R.W., Jordaan J.M., 1956, Resistance coefficient for accelerated and decelerated flows through smooth tubes and orifices, Trans. ASME, Journal of Basic Engineering, 78, July, 1071-1077 10. Ghidaoui M.S., Mansour S., 2002, Efficient treatment of the Vardy-Brown unsteady shear in pipe transients, Journal of Hydraulic Engineering, ASCE January, 102-112 11. Kagawa T., Lee I., Kitagawa A., Takenaka T., 1983, High speed and accurate computing method of frequency-dependent friction in laminar pipe flow for characteristics method, Trans. Jpn. Soc. Mech. Eng., Ser. A, 49, 447, 2638–2644 [in Japanese] 12. KudźmaS., 2005,Modelingandsimulationdynamical runs in closedconduits ofhydraulics systems using unsteady frictionmodel, PhDThesis, Szczecin University of Technology [in Polish] 13. KudźmaS., ZarzyckiZ., 2005,Application of the efficient numericalmethod of unsteady friction simulation for investigationof transientflow inhydraulic systemswith long line,ScientificPapers of theDepartment ofAppliedMechanics, no 29,XLIVSympozjonPTMTSModelowanie wMechanice, PolandGliwice, 29, 279-284 [in Polish] 14. Ohmi M, Tateo U., 1976, Pressure and velocity distributions in pulsating turbulent pipe flow (Part I – Theoretical treatments),Bulletin of the JSME, 19, 129, 307-313 15. Rahl F. M., Barlamont J., 1996, Unsteady pipe flow simulations using unsteady friction flow formula, 7th International Conference on Pressure Surges and Transients in Pipeline and Open Channels, BHRGroup, Harrogate, UK, 313-322 16. Safwat H.H., Polder J.V.D., 1973, Experimental and analytic data correlation study of water column separation, Journal of Fluids Engineering, March, 91-97 17. Schohl G.A., 1993, Improved approximate method for simulating frequency-dependent friction in transient laminar flow, Journal of Fluids Engineering, Trans. ASME, 115, 420-424 18. Shu J.J., 2003, Modelling vaporous cavitation on fluid transients, Intern. Journal of Pressure Vessels and Piping, 80, 187-195 19. Streeter V.L., Lai C., 1962, Water-hammer analysis including fluid friction, Journal of the Hydraulics Division, 88, HY 3, 79-112 20. Suzuki K., Taketomi T., Sato S., 1991, Improving Zielke’s method of simulating frequency- dependent friction in laminar liquid pipe flow, Journal of Fluids Engineering, Trans. ASME, 113, 569-573 21. Trikha A.K., 1975, An efficient method for simulating frequency-dependent friction in transient liquid flow, Journal of Fluids Engineering, Trans. ASME, March, 97-105 22. UrbanowiczK., ZarzyckiZ., 2012,Newefficient approximationofweighting functions for simu- lations of unsteady friction losses in liquid pipe flow,Journal of Theoretical andAppliedMechanics, 50, 2, 487-508 23. VardyA.E.,BrownJ.M.B., 1995,Transient turbulent smoothpipe friction,Journal ofHydraulic Research, 33, 435-456 24. Vardy A.E., Brown J.M.B., 2003, Transient turbulent friction in smooth pipe flows, Journal of Sound and Vibration, 259, 5, 1011-1036 902 K. Urbanowicz et al. 25. VardyA.E., Brown J.M.B., 2004,Transient turbulent friction in fully roughpipe flows, Journal of Sound and Vibration, 270, 233-257 26. Vardy A.E., Hwang K.L., 1991, A characteristics model of transient friction in pipes, Journal of Hydraulic Research, 29, 5, 669-684 27. Vardy A.E., Hwang K.L., Brown J.M.B., 1993, A weighting function model of transient turbulent pipe friction, Journal of Hydraulic Research, 31, 4, 533-548 28. Vitkovsky J.P., LambertM.F, SimpsonA.R., BergantA., 2000,Advances in unsteady fric- tion modelling in transient pipe flow,Proceedings of the 8th International Conference on Pressure Surges, BHRGroup Conf. Series Publ., The Hague, The Netherlands, 471-482 29. Vitkovsky J.P., Stephens M.L., Bergant A., Simpson A.R., Lambert M.F., 2004, Effi- cient and accurate calculation of Zilke and Vardy-Brown unsteady friction in pipe transients, 9th International Conference on Pressure Surges, Chester, United Kingdom, 405-419 30. Wylie E.B., Streeter V.L., 1993,Fluid Transients in Systems, Prentice-Hall Inc., Englewood Cliffs, New Jersey, USA 31. Zarzycki Z., 1994,A Hydraulic Resistance of Unsteady Fluid Flow in Pipes, Published by Tech- nical University of Szczecin, 516, Szczecin [in Polish] 32. Zarzycki Z., 2000, On weighting function for wall shear stress during unsteady turbulent pipe flow, 8th International Conference on Pressure Surges, BHRGroup, The Hague, 529-543 33. Zarzycki Z., Kudźma S., 2004, Simulations of transient turbulent flow in liquid lines using time – dependent frictional losses, The 9th International Conference on Pressure Surges, BHR Group, Chester, UK, 439-455 34. ZielkeW., 1966,Frequency dependent friction in transient pipe flow,TheUniversity ofMichigan, PhDThesis 35. Zielke W., 1968, Frequency-dependent friction in transient pipe flow, Journal of ASME, 90, March, 109-115 Uniwersalna funkcja wagi w modelowaniu niestacjonarnych przepływów z kawitacją występujących w przewodach ciśnieniowych Streszczenie Praca dotyczy modelowania zmiennych w czasie oporów podczas nieustalonego przepływu cieczy wprzewodach ciśnieniowych.Naprężenie styczne na ściance przewoduprzedstawione zostałow klasyczny sposób (wg. Zielke) jako całka splotowa z przyśpieszenia cieczy i pewnej funkcji wagowej. Przedstawiono procedury wyznaczania dwóch uniwersalnych funkcji wagowych laminarno- turbulentnych (wg.Zielke iVardy-Brownaorazwg.Zielke iZarzyckiego).Woparciuo te funkcje przedsta- wionometodęmodelowania(symulacji)przebiegówprzejściowychwprzewodachciśnieniowychwszerokim zakresie liczb Reynoldsa. Eliminuje on problemy natury numerycznej związane z przechodzeniem min. z przepływu laminarnegow turbulentny i odwrotnie. Podanoprzykład zastosowaniaopracowanejmetody do symulacji uderzenia hydraulicznego.Wyniki obliczeń porównano z wynikami eksperymentalnymi wg. Berganta i Simpsona, otrzymując dobrą zgodność. Manuscript received August 1, 2011; accepted for print November 10, 2011