3.2 Two-regime orifice flow model
3.2.7 Simulation of the cylinder drive circuit
The stability of simulation model was studied by simulating the practical circuit.
Two different simulation models were formed. One of the Simulink models con-tains the two-regime orifice flow model Equation (2.16) while the other concon-tains only the turbulent orifice flow model. The fixed-step Runge-Kutta integrator is used in the simulations.
Both Simulink-models are driven by using 3 different diameters for orifice 2.
The diameter values are defined as 1, 2 and 3 mm. The control signal for the directional valve is kept at its maximum during the simulation runs. The orifice 2 is chosen to be alternated to show the influence of orifice diameter on the function of the different orifice models and the system stability. The initial values of variables used in the simulations of the cylinder drive circuit are presented in Table 3.1.
3.2.8 Results: Cylinder drive - Two-regime flow orifice model
In the simulation runs pressure drops ∆p1 and ∆p2 are studied because the pressure drop over the orifice influences all the other parameters in the simulation model. Responses of pressure drops are shown,∆p1 in Figure 3.12 and∆p2 in Figure 3.13. In the figures the solid line represents the response when the two-regime orifice flow model is used and the dashed line represents the response when the pure turbulent orifice flow model is used.
From Figures 3.12 and 3.13 it can be seen that the two-regime orifice flow model gives a smooth response for pressure drop while the turbulent orifice model causes numerical noise. Figures 3.13 show that the increase of time step length increases the noise in response of turbulent model.
1.4 1.5 1.6 1.7 1.8 1.9 2 0
1 2 3 4 5 6 7
x 104
Simulation time [s]
Pressure drop ∆p1 [Pa]
RK4 / ∆t = 1 x 10−3 s
two−regime turbulent
Figure 3.12. Cylinder drive. Pressure drop ∆p1, when time step
∆t= 1×10-3s and orificedo2= 3×10-3m.
The influence of the integration time step length on the system stability was studied by simulating both models separately using a varied time step length with all 3 different orifice diameters. With this approach the longest usable integration time step is defined. If the integration time step is longer than the defined longest usable integration time step, the simulation run fails. Maximum time step lengths in terms of system stability are presented in Table 3.2. The results show that when using the two-regime orifice model the time step can be from 16 % to 32
% longer than the time step used in the turbulent orifice model.
3.3 Pseudo-dynamic solving method
In this section the cylinder drive circuit is studied as a numerical example of the pseudo-dynamic solving method. Some of the results shown in this section have previously been presented in reference [1].
The numerical example is identical to the one respresented in section 3.2 only
1.5 1.55 1.6 1.65 1.7 1.75 1.8 1.85 1.9 1.95 2
−1 0 1 2 3 4 5 6 7
x 104
Simulation time [s]
Pressure drop ∆p2 [Pa]
RK4 / ∆t = 2 x 10−3 s
two−regime turbulent
Figure 3.13. Cylinder drive. Pressure drop ∆p2, when time step
∆t= 2×10-3s and orificedo2= 3×10-3m.
initial parameter values are varied. Also two-regime orifice flow model is not used i.e. only turbulent orifice flow is described. To simplify the problem and to deminstrate the pseudo-dynamic solver it is assumed that the variation of the cylinder volume is negligible. It remains at least order of magnitude smaller than the volume between the orifices (volumeV1in Figure 3.2).
The pseudo-dynamic solving method is applied as represented in Section 2.3. In the pseudo-dynamic model the volumeV2is enlarged to the artificial volumeV. The initial values of the variables used in the simulations of the cylinder drive circuit are presented in Table 3.3.
3.3.1 Results: Cylinder drive - Pseudo-dynamic solving method
Figures 3.14 to 3.17 illustrate the behaviour of pseudo-dynamic model compared to reference model. Responses are reached by using integrator time step lengths
∆t= 1×10-4s and∆t= 1×10-3 s.
Table 3.2. Maximum integration time step length in terms of system stability.
Orifice 2 Time step (×10-3 s)
diameter (mm) Two-regime orifice model Turbulent orifice model
1 4.118 3.191
2 3.627 2.749
3 2.571 2.200
Table 3.3. Initial values of variables used in simulations of the cylinder drive circuit - Case: pseudo-dynamic solving method.
Be= 1×109Pa d1= 10×10-3m Cd= 0.6 k=18×106N/m V1=1×10-3m3 d2= 1×10-3 m ρ=900 kg / m3 m= 5000 kg V2=1×10-5m3 dp= 60×10-3m b = 500 Ns / m U= 10 V V =1×10-3m3 ps=100×105Pa p1= 0 Pa p2= 0 Pa
Figure 3.14 illustrates the cylinder position when∆t= 1×10-4 s is used. Re-sponses of pseudo-dynamic and reference model are identical.
Figure 3.15 illustrates the cylinder position when∆t= 1×10-3s is used. There exists divergence between responses of pseudo-dynamic and reference model.
Response of pseudo-dynamic model is identical to response in Figure 3.14 but the response of reference model differs significantly. The pseudo-dynamic model behaves more stable while in reference model there exists static deviation.
Figure 3.16 illustrates the pressure difference∆p2 between volumesV1 andV2
when ∆t= 1×10-4s is used. Responses of pseudo-dynamic and reference model are identical.
Figure 3.17 illustrates the pressure difference∆p2 between volumesV1 andV2
when ∆t= 1×10-3s is used. There exists divergence between responses of pseudo-dynamic and reference model. Response of pseudo-dynamic model is identical to response in Figure 3.16 but the response of reference model differs significantly. The pseudo-dynamic model behaves more stable while in reference model there exists numerical noise and static deviation.
The static deviation illustrated in Figures 3.15 and 3.17 can be explained by the numerical noise which appears in Figure 3.17. Numerical noise originates from too long integrator time step. By means of these observations the
pseudo-0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 1.2
1.3 1.4 1.5 1.6 1.7
x 10−3
Simulation time [s]
Cylinder piston position [m]
Cylinder drive
Pseudo RK4
Figure 3.14. Cylinder drive. Pseudo-dynamic solving method in com-parison with conventional method. Piston position, when time step length
∆t= 1×10-4s.
dynamic model can be noted to be more stable.
Figure 3.18 illustrates the increment in the error of pressurep2 when time step length∆tis varied. The error attained using pseudo-dynamic method increases moderately when time step length is increased. The error attained using con-vetional Runge-Kutta method starts to increase regularly after time step length is longer than 4 × 10 -4 s. This finding supports the statement above that the numerical noise is due to integrator time step length.
In Figure 3.19 the variation of the compressible volume∆V2 with respect to the time step length∆tis illustrated. The aim of studying the change in compressible volume is to observe the conservation of fluid mass. The fluid mass is calculated using the fluid volumeVf and fluid densityρas follows:
mf =ρVf. (3.10)
The first time derivative of fluid mass is
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 1.2
1.3 1.4 1.5 1.6 1.7
x 10−3
Simulation time [s]
Cylinder piston position [m]
Cylinder drive
Pseudo RK4
Figure 3.15. Cylinder drive. Pseudo-dynamic solving method in com-parison with conventional method. Piston position, when time step length
∆t= 1×10-3s.
˙
mf =ρV˙f, (3.11)
where
V˙f =Qin−Qout.
The change of compressible volume∆V is achieved using Meter-in - Meter-out method in which the difference of volume flows into and out of the volume is integrated (Equation 3.12). Thus, the variation of the compressible volume can be used as a measuring instrument for mass conservation.
∆V = Z
(Qin−Qout)dt. (3.12)
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Pressure drop ∆p2 [Pa]
Cylinder drive
Pseudo RK4
Figure 3.16. Cylinder drive. Pseudo-dynamic solving method in comparison with conventional method. ∆p2, when time step length
∆t= 1×10-4s.
On top of Figure 3.19 the situation where both loops of pseudo-dynamic solver have the same time step length is illustrated. The volume change in pseudo-dynamic method is much larger than in conventional method while time step length is shorter than 6 × 10 -4 s. In longer values of time step the pseudo-dynamic method is close to reference value and starts to approach it. The change is of positive sign, i.e. in volume exist compression. The compression is due to greater incoming flow compare to outgoing flow. This phenomenon can be explained by the iterations that are carried out in the inner loop of pseudo-solver. With short time step lengths due to the limited number of iteration rounds the pressure is not reaching its steady-state value and the unsaturated value is returned to the outer loop. This causes inaccuracy to pressures and inbalance between incoming and outgoing flows. The conventional method conserves the balance until time step length is shorter than 4 x 10 -4 s. After this point there starts to exist numerical noise and inbalance due to it.
As improvement to pseudo-dynamic solver the inner loop time step length∆tinner
0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
−8
−6
−4
−2 0 2 4 6 8 10x 105
Simulation time [s]
Pressure drop ∆p2 [Pa]
Cylinder drive
Pseudo RK4
Figure 3.17. Cylinder drive. Pseudo-dynamic solving method in compar-ison with conventional method. Pressure difference∆p2, when time step length∆t= 1×10-3s.
is increased up to 10 × ∆t. This enables the pressure to reach its steady-state value within the limited number of iteration rounds. The lower graph in Figure 3.19 presents that the pseudo-dynamic solver conserves the balance at the reference value through the tested time step range.
3.4 Reduced model based on singular perturbation theory
Some of the results shown in this section have previously been presented in reference [2].
This study was started by formulating a semi-empirical model of a two-way flow control valve [24] in MATLAB M-file. This code is used as a reference for the pseudo-dynamic model. The two-way flow control valve studied is represented in Section 3.1.3. The model is derived according to the semi-empirical modelling method [22].
2 3 4 5 6 7 8 9 10 x 10−4 0
2 4 6 8 10 12 14x 104
Time step ∆t [s]
ITAE of p2 [sPa]
Reference: Cylinder drive / RK4 / ∆t = 1 × 10−5 s
Pseudo RK4
Figure 3.18. Cylinder drive. Integrated absolute error of pressurep2when time step length is varied.∆tref = 1×10-5s
For the differential equations of pressures (p1 and p2), Equation (3.6), the ef-fective bulk modulusBe and volumes (V1 andV2) are defined. In the pseudo-dynamic model, volumeV1 is enlarged to artificial volumeV. The initial values are presented in Table 3.4.
In this study, a two-way flow control valve is modelled by using three different methods. The first method (model 1) is to use the semi-empirical model accord-ing to reference [24] while takaccord-ing the dynamic behaviour of the primary volume V1 into account. This causes numerical problems in small fluid volumes. The model is expressed in Equation (3.13):
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Figure 3.19. Cylinder drive. Variation of the compressible volume∆V2.
∆tref = 1×10-5s. for the compensator throttle, including the orifice geometry, discharge coeffi-cient, spool surface area and square root of the density of oil [m3/s√
Pa]. kt
is the variable for the control throttle, including the orifice geometry and area, discharge coefficient and square root of the density of oil [m3/s√
Pa]. ps, p1 and p2are the supply, primary and secondary pressures.
The second method (model 2) is described in section 3.4.1. The third method
is the pseudo-dynamic solver which is implemented based on the first method according to rules described in section 2.3. This method is compared with the semi-empirical and singular perturbation methods.
3.4.1 Singular perturbation theory in modelling two-way flow control valves The second method (model 2) is called as the singular perturbation method. It is developed from model 1 to avoid the integration of the pressure in primary volume V1, see reference [24]. In this reference, steady-state model of two orifices in series is used. It can be shown that the use of singular perturbation theory leads to exactly similar model. The drawback of this method is that this pressure is needed in the dynamic equation for the pressure compensator. It must then be reproduced with a static equation from the surrounding pressures and semi-empirical flow coefficients for the pressure compensator and control orifice.
By applying singular perturbation theory in Equation (3.13), the dynamic degree caused by the compressibility between the pressure compensator and control orifice can be neglected as follows.
By expressing Equation (3.13) into the following form:
K˙ = C5−p1+p2−(C1+C2(ps−p1))K
and substituting new variablesΨandεintroduced in section 2.1 such that
p1= Ψps, (3.15)
Then it can be written as
˙
p1= ˙ΨBeε . (3.17)
Then substituting Equation (3.17) into Equation (3.14) yields
Table 3.4. Initial values of variables used in system simulation.
According to singular perturbation theory we write Equation (3.18) when limε→0.
Consequently, the latter term in Equation (3.18) becomes 0 =K√
ps−p1−kt√
p1−p2. (3.19)
By taking the square on both sides of Equation (3.19) and solving for p1 we finally bring Equation (3.18) into the form
K˙ = C5−p1+p2−(C1+C2(ps−p1))K C3
p1= K2ps+kt2p2
K2+kt2 .
(3.20)
Allthough, now one degree and a small time constant is neglected, the use of Equilibrium (3.20) causes numerical problems when the term (K2 +kt2) ap-proaches zero.
The parameter values applied to this numerical example are presented in Ta-ble 3.4.
3.4.2 Results: Two-way flow control valve
The following results are achieved using a constant supply pressure of ps = 200×105 Pa. The directional control valve is closed (Qe2 = 0) momentarily between 0.4 . . . 0.7×simtime and opened again to create an impulse into the system during the simulation run. As a reference the model 1 is simulated using a main-loop integrator time step length of 5 × 10-6 s (right in Figure 3.20).
Its pressure response is selected to be the reference point to which all other responses are compared. Responses achieved using a main-loop integrator time step length of 3 × 10-5 s are illustrated in Figure 3.20. Pseudo-dynamic and singular perturbation methods are shown on the left and semi-empirical model using different time step lengths on the right. The responses of the pseudo-dynamic and singular perturbation models show identical smooth behaviour, while in the response of the semi-empirical model 1 physical oscillation occurs due to the dynamics of the small volumeV1. The oscillations are apparent also in reference response but on the different magnitude. The frequency of oscillations is different i.e. more numerical noise is involved to results when longer time step length is used.
The following results were achieved using a main-loop integrator time step length of 2 × 10-5 s. The stepwise supply pressure of ps = 100 . . . 200 × 105 Pa (Figure 3.21) is used to create an impulse in the system.
Responses of the three compared pairs are illustrated in Figure 3.22. The re-sponse of pseudo-dynamic model is stable and shows only a short-term peaks in the impulse point. In the response of model 1, physical oscillation occurs due to the dynamics of the small volumeV1. It also does not react properly to the impulse. The singularly perturbed model is highly unstable and its response shows long-term although damping noise after the impulse point. This can be seen from Figure 3.23
In Figure 3.24 the variation of the compressible volume∆V1is illustrated. The change in volume V1 is calculated similarly than in Section 3.3.1 is described (see Equation 3.12).
On top of Figure 3.24 in the response of the pseudo-dynamic solver there is a high peak in volume change at the impulse point. The response decreases close to zero quite rapidly and oscillates slightly. After the next impulse point at which the directional control valve is closed (Qe2 = 0) the response takes constant positive value after minor oscillation. This is due to that the output is more resistive. The response of singular perturbation method shows constant zero. This is due to the fact that in this method the incoming and outgoing
0.07 0.08 0.09 0.1 1.024
1.0245 1.025 1.0255 1.026 1.0265
x 106
Simulation time [s]
Pressure p1 [Pa]
Constant supply / ∆t = 3 × 10−5 s / ∆tinner = 10 × ∆t
Pseudo SPT
0.07 0.08 0.09 0.1 1.024
1.0245 1.025 1.0255 1.026 1.0265
x 106
Simulation time [s]
Pressure p1 [Pa]
Reference RK4
Figure 3.20. Two-way flow control valve. Comparison of responses of pressure p1 attained by three alternative methods to reference response.
∆tref = 5×10-6s.
flows are set to equal. I.e unfortunately the method is not comparable with this measure.
The lower graph in Figure 3.24 shows that the responses of semi-empirical model oscillate in a damping manner after the impulse which they are supposed to since the physical volume is taken into account in this method. The different amplitude originates from the different time step length.
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 1
1.2 1.4 1.6 1.8 2 2.2x 107
Simulation time [s]
Supply pressure ps [Pa]
stepwise input / RK4 / ∆t = 2 x 10−5 s
Supply
Figure 3.21. Two-way flow control valve. Stepwise supply pressure.
0.02 0.025 0.03 0.035 0.04 0.045 0.05
0.95 1 1.05
1.1x 106
Simulation time [s]
Pressure p1 [Pa]
Control throttle / Stepwise supply / ∆t = 2 x 10−5 s / ∆tinner = 10 x ∆t
Pseudo SPT RK4
Figure 3.22. Two-way flow control valve. Responses of pressure p1
attained by three alternative methods.
0.0246 0.0248 0.025 0.0252 0.0254 0.0256 0.0258 0.026 0.0262 1.014
1.016 1.018 1.02 1.022 1.024 1.026 1.028 1.03 1.032 1.034
x 106
Simulation time [s]
Pressure p1 [Pa]
Control throttle / Stepwise supply / ∆t = 2 x 10−5 s / ∆tinner = 10 x ∆t
Pseudo SPT RK4
Figure 3.23. More focused view of the Figure 3.22.
0.025 0.03 0.035 0.04
−1
−0.5 0 0.5 1
x 10−16
Simulation time [s]
Volume change ∆V1 [m3]
Meter−In − Meter−Out / ∆t = 2 x 10−5 s / ∆tinner = 10 x ∆t
Pseudo SPT
0.025 0.03 0.035 0.04
−1 0 1
x 10−12
Simulation time [s]
Volume change ∆V 1 [m3]
responses
RK4 Reference
Figure 3.24. Two-way flow control valve. Stepwise supply pressure.
Variation of the compressible volume∆V1.∆tref = 5×10-6s.
Conclusions
A numerically efficient, physically justified two-regime orifice model was pro-posed. The laminar and transition regions in the orifice model were described by a cubic spline function that gives a continuous first derivative of flow with respect to the pressure drop. The conventional turbulent orifice model was used to describe the turbulent region.
The model can be used in the semi-empirical modelling of fluid power compo-nents because it does not necessarily require any geometrical information about the orifice type. If physically adequate values for boundary pressure drops are desired, the equation for the Reynolds number can be used in calculating them. In that case, the transition value for the Reynolds number and the turbulent region value for the discharge coefficient must be used. It was shown that the model gives quite an accurate approximation of the discharge coefficient as a function of the square root of the Reynolds number. Steady-state responses of the volume flow and the first derivative of it with respect to the pressure drop were illustrated together with two alternative two-regime orifice models.
The dependency of the calculation error on the integrator time step and the selected boundary pressure drop was also investigated by simulating a simple hydraulic circuit. A physically justified transition pressure drop was used in simulation of the reference response. Integrated errors were illustrated for the proposed model and also for the alternative models. The behaviour of these alternative models does not differ widely from the proposed model. However, they require geometrical data of the orifice which is undesired in semi-empirical modelling of fluid power circuits.
61
The proposed two-regime orifice flow model was used also in the simulation of a practical fluid power circuit. It was shown that the two-regime model provides better integrator stability than the conventional turbulent model since larger integration time steps can be used. Also numerical noise apparent in responses calculated using the pure turbulent orifice model did not occur. It was shown that the proposed two-regime orifice flow description is more suitable for dynamic simulation.
A pseudo-dynamic solver for small volumes in fluid power system simulation was proposed. The pressures in small volumes were calculated as individual steastate solutions at each time step of the actual solver that solves the dy-namics of a complete circuit. Two different numerical examples of fluid power systems were studied: a cylinder drive and a fluid power circuit with a two-way flow control valve.
In the case of the cylinder drive circuit, the pseudo-dynamic model provided better integrator stability since larger integration time steps can be used. Also the numerical noise apparent in the responses of a reference model can be avoided.
The reference response was obtained by using the conventional fourth order Runge-Kutta integration routine with sufficiently short integration time step. The dependency of the accuracy on the integrator time step is illustrated in compar-ison to the conventional method. Also the dependency of mass conservation on the time step length was illustrated.
A two-way flow control valve was taken as an example to demostrate the ef-fectiviness of the pseudo-dynamic approach in finding a solution for the circuit with a small volume. The dynamics of a simple circuit including the two-way flow control valve was modelled and solved by applying three alternative methods. The pseudo-dynamic solver was compared to the reduced model by using singular perturbation theory and also to the semi-empirical method which takes the physical small volume into account. This model was used also to attain the reference response by using sufficiently short time step length. The responses
A two-way flow control valve was taken as an example to demostrate the ef-fectiviness of the pseudo-dynamic approach in finding a solution for the circuit with a small volume. The dynamics of a simple circuit including the two-way flow control valve was modelled and solved by applying three alternative methods. The pseudo-dynamic solver was compared to the reduced model by using singular perturbation theory and also to the semi-empirical method which takes the physical small volume into account. This model was used also to attain the reference response by using sufficiently short time step length. The responses