Unsteady filling of a vacuum tank

whiteskulleton

Here is the procedure that I used to get the governing equation:
From post #24 and the equation from the Wikipedia article
$$\dot{m}(t)=\left\{\begin{matrix} 0.007902C_d & \textrm{if } 0\leq t\leq 1.8963\\ C_dA \sqrt{2\, \rho_1\, p_{atm}\left ( \frac{\gamma}{\gamma-1} \right )\left [ \left ( \frac{p(t)}{p_{atm}} \right )^\frac{2}{\gamma}-\left ( \frac{p(t)}{p_{atm}} \right )^\frac{\gamma+1}{\gamma} \right ]} & \textrm{if }1.8963< t \end{matrix}\right. \tag{3}$$$$C_d=0.6740$$
I then plug the pressure from the experimental data into the piecewise function above in excel to get the mass flow rate.

I then get the mass at each data point by doing integration by the trapezoidal rule
$$\sum_{n=1}^{\infty}m_n(t_n)=\frac{\dot m(t_n)-\dot m(t_{n-1})}{2(t_n-t_{n-1})}+m(t_{n-1})\tag{4}$$From post #7
$$m_0=n_iM=\frac{P_iV_TM}{RT_i}=0.02350\, \textrm{kg}\tag{5}$$
I then calculate the pressure at each data point using equation 2 from post 17:
$$p=\frac{RT_i}{V_T}(\gamma n-(\gamma-1)n_i)\tag{6}$$
$$\sum_{n=1}^{\infty}p_n(t_n)=\frac{RT_i}{V_T}(\gamma \frac{m_n(t_n)}{M}-(\gamma-1)\frac{m_0}{M})\tag{7}$$Finally, I look at the percent error at each data point to see how well this model fits to the experimental data:
$$\textrm{% Error}=\left |\frac{p_n(t_n)-p_{exp}(t_n)}{p_{exp}(t_n)} \right |\cdot 100$$
Looking at the data, I got the max error to be 2.62% with an average of 1.064%. Thus, I feel that these equations accurately model the experimental data. I did try the other equation that you recommended with the expansibility factor but it lead to more error.
Thus, the governing equations for $p<53530.58\mathrm{Pa}$:
$$\frac{dm}{dt}=C_dA\sqrt{\gamma \rho_{atm}P_{atm}\left (\frac{2}{\gamma+1}\right )^{\frac{\gamma+1}{\gamma-1}}}\tag{8}$$Integrating and plugging $m_0$ from eq. 5 into yields:
$$m(t)=C_dA\sqrt{\gamma \rho_{atm}P_{atm}\left (\frac{2}{\gamma+1}\right )^{\frac{\gamma+1}{\gamma-1}}}t+\frac{P_iV_TM}{RT_i}\tag{9}$$ Which is then put into equation 6 giving the final governing equation for $p<53530.58\mathrm{Pa}$:
$$p(t)=\frac{RT_i}{V_T}\left (\frac{\gamma C_dA}{M}\sqrt{\gamma \rho_{atm}P_{atm}\left (\frac{2}{\gamma+1}\right )^{\frac{\gamma+1}{\gamma-1}}}t+\frac{\gamma P_iV_T}{RT_i}-(\gamma-1)n_i \right )\tag{10}$$ To get the other half, we have to plug in the other equation but that is a function of pressure and I'm not quite sure how to integrate that.

Starting this experiment, I had no idea the complexity involved with finding the governing equations and thus far it has been an amazing learning experience.

whiteskulleton

This sort of brings me to the next real world application. We are testing the holding force of suction cups using this tank. The standard vacuum pressures that we recommend should be above 40% vacuum or $p<60795\,\mathrm{Pa(abs)}$. We test the suction cups by connecting the tank to a pressure regulator and then slowly decreasing the pressure until the suction cup drops a known weight. The relationship between holding force and pressure is linear. However, if the surface that the suction cup is holding on to is cardboard, there is a leakage rate. What we are trying to do is figure out the pressure drop as a function of leakage rate of different surfaces. I plan on running a test in the next few days where I connect the suction cup directly to the tank and to a known weight in a cardboard box and then let the tank drain drain until the box falls. Since the pressure of test will be less than 53530 kPa (abs), that should mean that the flow is throttled as the leakage rate is significantly less than this past experiment I ran and thus means that the leakage area is less than the tubing area. Or $A_L<A$.

JBA

With respect to the above, I have created the below graph for the filling mass flow vs tank pressure for tank pressures from 4.7 to 13.7 psia (the same graph with more refined values can be provided if desired) that illustrates the two conditions, sonic and subsonic nozzle flow that exist for this problem. This graph does not necessarily reflect the actual rates because, for lack of a confirmed value, it is for a nozzle coefficient of "1.0" (but I can revise it for any other coefficient you might desire). Because of the wide range of pressure differentials during the filling both of the above flow conditions that have substantially different profiles exist during the filling and make developing an accurate single curve fitting difficult. In such cases, for programming purposes, I have resorted to breaking the total curve into separate sonic and subsonic curves using the sonic pressure ratio for the gas as an (if,then,or value) function to select the appropriate curve for for a curve fitting for each flow region.
The excel program I use to obtain the graph's values calculates the mass flow for a Pin air of 14.7 psia & 70°F (R critical = .5283) for the values for each tank pressure point and I used those to develop the shown graph (the program actually calculates values for both regions simultaneously and then uses the (if, then,or) function to present a final value selected. P tank (psia) lb/sec (lb/sec) 4.7​ 0.0175​ 5.7​ 0.0175​ 6.7​ 0.0175​ 7.7​ 0.0175​ 8.7​ 0.0174​ 9.7​ 0.0168​ 10.7​ 0.0159​ 11.7​ 0.0144​ 12.7​ 0.0123​ 13.7​ 0.0091​

Chestermiller

Mentor
Here is the procedure that I used to get the governing equation:
From post #24 and the equation from the Wikipedia article
$$\dot{m}(t)=\left\{\begin{matrix} 0.007902C_d & \textrm{if } 0\leq t\leq 1.8963\\ C_dA \sqrt{2\, \rho_1\, p_{atm}\left ( \frac{\gamma}{\gamma-1} \right )\left [ \left ( \frac{p(t)}{p_{atm}} \right )^\frac{2}{\gamma}-\left ( \frac{p(t)}{p_{atm}} \right )^\frac{\gamma+1}{\gamma} \right ]} & \textrm{if }1.8963< t \end{matrix}\right. \tag{3}$$$$C_d=0.6740$$
I then plug the pressure from the experimental data into the piecewise function above in excel to get the mass flow rate.

I then get the mass at each data point by doing integration by the trapezoidal rule
$$\sum_{n=1}^{\infty}m_n(t_n)=\frac{\dot m(t_n)-\dot m(t_{n-1})}{2(t_n-t_{n-1})}+m(t_{n-1})\tag{4}$$From post #7
$$m_0=n_iM=\frac{P_iV_TM}{RT_i}=0.02350\, \textrm{kg}\tag{5}$$
I then calculate the pressure at each data point using equation 2 from post 17:
$$p=\frac{RT_i}{V_T}(\gamma n-(\gamma-1)n_i)\tag{6}$$
$$\sum_{n=1}^{\infty}p_n(t_n)=\frac{RT_i}{V_T}(\gamma \frac{m_n(t_n)}{M}-(\gamma-1)\frac{m_0}{M})\tag{7}$$Finally, I look at the percent error at each data point to see how well this model fits to the experimental data:
$$\textrm{% Error}=\left |\frac{p_n(t_n)-p_{exp}(t_n)}{p_{exp}(t_n)} \right |\cdot 100$$
Looking at the data, I got the max error to be 2.62% with an average of 1.064%. Thus, I feel that these equations accurately model the experimental data. I did try the other equation that you recommended with the expansibility factor but it lead to more error.
Thus, the governing equations for $p<53530.58\mathrm{Pa}$:
$$\frac{dm}{dt}=C_dA\sqrt{\gamma \rho_{atm}P_{atm}\left (\frac{2}{\gamma+1}\right )^{\frac{\gamma+1}{\gamma-1}}}\tag{8}$$Integrating and plugging $m_0$ from eq. 5 into yields:
$$m(t)=C_dA\sqrt{\gamma \rho_{atm}P_{atm}\left (\frac{2}{\gamma+1}\right )^{\frac{\gamma+1}{\gamma-1}}}t+\frac{P_iV_TM}{RT_i}\tag{9}$$ Which is then put into equation 6 giving the final governing equation for $p<53530.58\mathrm{Pa}$:
$$p(t)=\frac{RT_i}{V_T}\left (\frac{\gamma C_dA}{M}\sqrt{\gamma \rho_{atm}P_{atm}\left (\frac{2}{\gamma+1}\right )^{\frac{\gamma+1}{\gamma-1}}}t+\frac{\gamma P_iV_T}{RT_i}-(\gamma-1)n_i \right )\tag{10}$$ To get the other half, we have to plug in the other equation but that is a function of pressure and I'm not quite sure how to integrate that.

Starting this experiment, I had no idea the complexity involved with finding the governing equations and thus far it has been an amazing learning experience.
I would have done this somewhat differently. From our previous development, you have an equation for the mass flow rate in terms of the rate of change of pressure given by $$\dot{m}=\frac{V_TM}{\gamma RT_i}\frac{dp}{dt}$$ So you could substitute this for the mass flow rate in your flow equation and obtain a differential equation entirely in terms of the pressure, with the time derivative of pressure directly expressed as a function of pressure itself. You could then solve this ordinary differential equation numerically (say by forward Euler) to obtain the pressure as a function of time. You could then compare the entire curve with the experimental data.

You have split the behavior into two parts, representing the initial sonic part and the subsequent sub-sonic part. And you have used the theoretical equations in Wiki for determining the mass flow rate. I prefer working with the ISO standards recommendations (approximations) because they have extensively been applied and validated in practice. I also like the idea of using a continuous description of the compression parameter so that no judgments need to be made about when to switch from one representation to the other.

What I'm going to do is use Eqn. 1 and 2 in post #21 to solve for the pressure in your experiment as a function of time. I have already started doing this using a value of 0.48 for C, which was based on the initial rate of pressure decline from your data, together with a value for $\epsilon$ of 0.8 using Eqn. 2 for the initial conditions. This, of course, worked well at predicting the pressure variation at short times (< 2 seconds), but it deviated from the data a bit at larger times, and ended up predicting a filling time of about 11 seconds (compared to the observed value of about 9 seconds). My next iteration will be to use your value of 0.67 for C, and to adjust $\epsilon$ so it again matches the rate of pressure decrease at short times. To do this, I will have to modify the parameterization of the compression factor $\epsilon$ so that it matches properly at the two end points. The form I will be using will be:
$$\epsilon=\left[0.649+0.351\left(\frac{p}{p_{atm}}\right)^{1/\gamma}\right]^{2.5}\tag{2-mod}$$
Together with C = 0.67, this parameterization will match your "choked flow" mass flow rate at short times and will match the condition that, as p approaches atmospheric, $\epsilon$ approaches unity. With this change, I am confident that the model will do a much better job of matching the entire pressure profile as well as the fill time.

I'll keep you posted.

Chet

Chestermiller

Mentor
I completed the model calculation I described in my previous post. The new calculation again matches the rate of pressure decline and choke mass flow rate at short times, but also matches the observed fill time of about 9 seconds. I'll present comparison graphs later.

Chet

Chestermiller

Mentor The circles are the experimental data. The line is the model calculation as I described in my most recent posts.

whiteskulleton

I just ran another test in the same exact geometry conditions as before. The starting pressure was at -510 mmHg or 33330 Pa(abs). However I stopped the flow suddenly at 57462 Pa(abs). It then dropped to a final pressure of 54929 Pa(abs). Since the change in pressure occurs in a closed system, it is an isochoric process. Thus:$$\frac{T_1}{P_1}=\frac{T_2}{P_2}$$ Solving gets me $T_1=309 \,K$. Based on the adiabatic assumption, the temperature of the air when I stopped it should be approximately 336 K. That's an error of 8.7%. Do you think that this is cause for concern for the models we made?

Chestermiller

Mentor
I just ran another test in the same exact geometry conditions as before. The starting pressure was at -510 mmHg or 33330 Pa(abs). However I stopped the flow suddenly at 57462 Pa(abs). It then dropped to a final pressure of 54929 Pa(abs). Since the change in pressure occurs in a closed system, it is an isochoric process. Thus:$$\frac{T_1}{P_1}=\frac{T_2}{P_2}$$ Solving gets me $T_1=309 \,K$. Based on the adiabatic assumption, the temperature of the air when I stopped it should be approximately 336 K. That's an error of 8.7%. Do you think that this is cause for concern for the models we made?
What was the time interval over which the pressure dropped?

JBA

If using a adiabatic solution when filling from the 14.7 psia pressure air into the lower pressure tank there is a expansion cooling for that gas mass relative to the transitive pressure ratio during the fill. Only the air mass that is heated by the compression by using the initial 4.85 and final 14.7 pressures is the air mass in the vessel before filling can be this equation. The relative percentage of each gas mass and its BTU contribution must be used to determine the actual vessel air temperature of the mixture after filling.
The additional heat transfer into or out of the tank is transitive during the fill due the changing internal temperature as well and also dependent upon the fill time. This combination can lead to a significant difference between the calculated and actual finished contained air temperatures. The best method for determining a reasonably accurate final temperature is by measurement at the completion of your fill test.

I learned these lessons when responding to a request to develop a program that would "accurately determine the number of bottle fills of a given size and desired pressure from a supply tank of a given size and pressure" relative to accurate test results from multiple and consistent test fillings with accurate and calibrated instruments. In the end, the above combination of temperature factors for the bottles being filled defeated the project after some six months of intense work.

Chestermiller

Mentor
If using a adiabatic solution when filling from the 14.7 psia pressure air into the lower pressure tank there is a expansion cooling for that gas mass relative to the transitive pressure ratio during the fill. Only the air mass that is heated by the compression by using the initial 4.85 and final 14.7 pressures is the air mass in the vessel before filling can be this equation. The relative percentage of each gas mass and its BTU contribution must be used to determine the actual vessel air temperature of the mixture after filling.
There is also significant viscous heating occurring in the air flow through the entrance hole, essentially the same as that present with the Joule Thomson effect. So even though the gas enters the tank at a lower pressure than atmospheric, its expansion cooling is fully offset by the viscous heating in flow through the hole, so that its temperature essentially does not change in passing through the hole. It then compresses within the tank, and its temperature rises as a result. This is the essence of the first law of thermodynamics analysis presented early in this thread. Either one accepts the first law of thermodynamics or one doesn't.

JBA

This is not consistent with the bottle temperatures in the prior project where the air was being injected into 14.7 psia & 54°F atmospheric pressure bottles and being increased to 4514.7 psia with only a consistent 102°F measured final temperature in testing and verified by an equally limited pressure drop upon cooling to ambient. See the below test data example which is consistent over multiple such fillings.

 Pressure (psig) Pressure (psia) Temp (°F) 0 14.7 54 500 514.7 61 1000 1014.7 67.2 1500 1514.7 74 2000 2014.7 79.9 2500 2514.7 83.4 3000 3014.7 87.7 3500 3514.7 92.3 4000 4014.7 96.4 4500 4514.7 101.9

Chestermiller

Mentor
This is not consistent with the bottle temperatures in the prior project where the air was being injected into 14.7 psia & 54°F atmospheric pressure bottles and being increased to 4514.7 psia with only a consistent 102°F measured final temperature in testing and verified by an equally limited pressure drop upon cooling to ambient. See the below test data example which is consistent over multiple such fillings.

 Pressure (psig) Pressure (psia) Temp (°F) 0 14.7 54 500 514.7 61 1000 1014.7 67.2 1500 1514.7 74 2000 2014.7 79.9 2500 2514.7 83.4 3000 3014.7 87.7 3500 3514.7 92.3 4000 4014.7 96.4 4500 4514.7 101.9
I'm not familiar with the details of the experiment you describe, so I can't comment intelligently. But, if you can identify something wrong with how the first law of thermodynamics was applied (by two different members using two different equivalent versions of the first law) in the first several posts of the current thread to the problem at hand, please point it out.

whiteskulleton

I just ran the test again but this time filmed it. Below is the raw video. I don't have time right now to analyze and extract the data. It's filmed at 240 fps but I like to upload it and slow it down by 8x using this website:
https://ezgif.com/video-speed

Also, attached is a pdf of the geometry of the inside of the flow. That is, it's the geometry of the actual air itself. It looks like the pressure was stopped at -342 mmHg and then fell to -360 mmHg.

Attachments

• 14.3 KB Views: 11

Chestermiller

Mentor
COMPARISON OF EXPANDABILITY PARAMETERIZATIONS

My goal in the present post is to provide a comparison of the expandability parameterizations that we have been using in our models. This is to show that they are virtually indistinguishable.

Even though you didn't officially employ the expandability form of the equation in your model, it is possible to derive the expandability factor $\epsilon$ from your equations. For the choked case of $p/p_{atm}<0.528$, you employed $$\dot{m}=C_dA\sqrt{\gamma \rho_{atm}P_{atm}\left (\frac{2}{\gamma+1}\right )^{\frac{\gamma+1}{\gamma-1}}}$$We can show that this translates into an expandability factor of $$\epsilon=\sqrt{\frac{\gamma}{2}\left(\frac{2}{\gamma+1}\right)^{\frac{\gamma+1}{\gamma-1}}\frac{1}{\left(1-\frac{p}{p_{atm}}\right)}}$$

And for the non-choked case of $p/p_{atm}>0.528$, you used $$\dot{m}=C_dA \sqrt{2\, \rho_1\, p_{atm}\left ( \frac{\gamma}{\gamma-1} \right )\left [ \left ( \frac{p}{p_{atm}} \right )^\frac{2}{\gamma}-\left ( \frac{p}{p_{atm}} \right )^\frac{\gamma+1}{\gamma} \right ]}$$
This translates into an expandability factor of
$$\epsilon=\sqrt{\left(\frac{\gamma}{\gamma-1}\frac{(p/p_{atm})^{2/\gamma}-(p/p_{atm})^{(\gamma+1)/\gamma}}{1-(p/p_{atm))}}\right)}$$
The corresponding relationship I've been using over the full range of pressure ratios has been:
$$\epsilon=\left[0.649+0.351\left(\frac{p}{p_{atm}}\right)^{1/\gamma}\right]^{2.5}$$

I have made a plot of how closely these two parameterizations compare: In the figure, the circles are the parameterization you have been using, involving the choked flow region, and the solid curve is the parameterization I have been using. As you can see, the match is remarkable. Both will result in an excellent fit to your experimental pressure vs time data (as we've already seen in post #31. An equally acceptable parameterization would be the simple linear fit $\epsilon=0.385+0.615p/p_{atm}$ which would again span both the choked and unchoked regions.

What do you think?

JBA

@Chestermiller, I continue to be astounded by the depth and width of your knowledge of the deep technical physics of many processes; and, in the event that we diverge on a subject, if ask by someone which path to take I would generally recommend that your's, by far, has the highest probability of being correct.

I find the above to be amazing. In all of my reviewing of reference texts available to me on the analysis and physics of fluid and gas flow the focus was on Bernoulli's equations and the standard array P/T = PT etc using k = 1.4 etc. I have never seen anything with as much depth as presented here. That is why for the majority of the thread I was satisfied to be a "fly on the wall" observer.

On the expansion cooling issue, after reviewing my prior project I realize there is a disconnect between that system and the one here with regard to expansion cooling of the source air being delivered to the tank. In this case, there is an infinite supply of ambient temperature air provided during the filling. In my tank to bottle filling project there is a finite initial high pressure supply in the supply tank. As a result, as that tank is depressurized it goes through and expansion cooling of the remaining air in the tank; and, therefore, the air delivered to the bottle is progressively colder as the filling progresses and it has nothing to do with the expansion thru the flow valve to the bottle. I also now realize, if the viscous heating you mentioned occurs, the delivered air can then be subject to compression heating from its initial feed reduced temperature, all of which is something I need to investigate.

At the same time, I still have a problem resolving the nozzle flow ΔT/ΔP first law issue simply because in years of flow testing, I have not observed the degree of temperature change the classical first law predicts; and, I am most experienced with pressures in the 1 to 10 thousand psig range where there are can be extreme ΔP changes, i.e. such as the 14.7 to 4500 psia compression as exists in the scuba filling tests or flowing valve and nozzle certification tests with the reverse condition.

I am not taking a hard stand on this issue, it is just that I am having difficulty reconciling my prior observations with the first law results.

Above you stated:
There is also significant viscous heating occurring in the air flow through the entrance hole, essentially the same as that present with the Joule Thomson effect. So even though the gas enters the tank at a lower pressure than atmospheric, its expansion cooling is fully offset by the viscous heating in flow through the hole, so that its temperature essentially does not change in passing through the hole.
I am really interested in being able to read any reference material that you may recommend that discusses this issue. I have not heard or read this before and all of my prior references have only made vague statements about the Joule Thomson effect and have never covered it to any depth.

Chestermiller

Mentor
@Chestermiller, I continue to be astounded by the depth and width of your knowledge of the deep technical physics of many processes; and, in the event that we diverge on a subject, if ask by someone which path to take I would generally recommend that your's, by far, has the highest probability of being correct.

I find the above to be amazing. In all of my reviewing of reference texts available to me on the analysis and physics of fluid and gas flow the focus was on Bernoulli's equations and the standard array P/T = PT etc using k = 1.4 etc. I have never seen anything with as much depth as presented here. That is why for the majority of the thread I was satisfied to be a "fly on the wall" observer.

On the expansion cooling issue, after reviewing my prior project I realize there is a disconnect between that system and the one here with regard to expansion cooling of the source air being delivered to the tank. In this case, there is an infinite supply of ambient temperature air provided during the filling. In my tank to bottle filling project there is a finite initial high pressure supply in the supply tank. As a result, as that tank is depressurized it goes through and expansion cooling of the remaining air in the tank; and, therefore, the air delivered to the bottle is progressively colder as the filling progresses and it has nothing to do with the expansion thru the flow valve to the bottle. I also now realize, if the viscous heating you mentioned occurs, the delivered air can then be subject to compression heating from its initial feed reduced temperature, all of which is something I need to investigate.

At the same time, I still have a problem resolving the nozzle flow ΔT/ΔP first law issue simply because in years of flow testing, I have not observed the degree of temperature change the classical first law predicts; and, I am most experienced with pressures in the 1 to 10 thousand psig range where there are can be extreme ΔP changes, i.e. such as the 14.7 to 4500 psia compression as exists in the scuba filling tests or flowing valve and nozzle certification tests with the reverse condition.

I am not taking a hard stand on this issue, it is just that I am having difficulty reconciling my prior observations with the first law results.

Above you stated:

I am really interested in being able to read any reference material that you may recommend that discusses this issue. I have not heard or read this before and all of my prior references have only made vague statements about the Joule Thomson effect and have never covered it to any depth.
Thank you for your kind thoughts. A few years ago, another PF member and I were very puzzled about the Joule Thomson effect, and were trying to get an understanding of why, when an ideal gas experiences an adiabatic pressure drop in passing through a valve or porous plug, its enthalpy per unit mass and temperature do not change as a result of expansion cooling. So we collaborated on this and analyzed the problem in private conversations for several weeks. We finally came to the (correct) conclusion that the viscous heating was exactly cancelling out the expansion cooling.

In the case of an incompressible fluid that passes through a porous plug or valve, it experiences a temperature rise equivalent to the rate of frictional flow work done on the fluid divided by the product of mass flow rate and heat capacity. This is what happens when there is no expansion cooling. In both cases, the change in enthalpy per unit mass is equal to zero (neglecting changes in kinetic energy). The difference is in the equation of state for an incompressible fluid compared to and ideal gas. I hope that this helps.

I have never seen any discussion of these mechanistics in any treatment in the open literature. But I am fully confident that it is correct. I'll try to locate the original link to the thread where we reported back the results of our deliberations.

Chet

Chestermiller

Mentor
• BvU

JBA

Thank you for taking the time to review and reply to my extended post. I have read your summary report and while it does not provide any specific supporting analyses; and, because of my confidence in your rigorous approach to problem solving as well as the fact that I no longer have access to a required flow testing facility to investigate it in that manner I am fully prepared to accept the validity of your conclusions.

At the same time, I am afraid I am back again with my concerns on the 1st Law issue. This time it relates directly to the issue addressed in this thread as an example of my concerns.

Using this as an example, using the 1st Law equation as presented in post#8, I to calculate the final Tank temperature, ignoring tank heat transfer, to be as follows:

 P i = 4.85​ psia T I = 70​ °F T I = 530​ °R P f = 14.7​ psia k air = 1.4​ T f = 655​ °R T f = 281​ °F

Which would indicate that even with an large amount of Tank heat loss during the 9 sec filling time, there would be an additional amount of filling time required as the Tank consumed the air required to maintain its target 14.7 psia pressure level; or, if the fill valve were to be closed, a substantial post fill pressure drop would be expected as the tank cools to ambient temperature.

If you are willing to continue consuming your time with my ongoing inquiries I would really appreciate your feedback on this issue; if not, I fully understand because it getting to one of those extended "OK, but what about" situations.

Jack

Chestermiller

Mentor
Thank you for taking the time to review and reply to my extended post. I have read your summary report and while it does not provide any specific supporting analyses; and, because of my confidence in your rigorous approach to problem solving as well as the fact that I no longer have access to a required flow testing facility to investigate it in that manner I am fully prepared to accept the validity of your conclusions.

At the same time, I am afraid I am back again with my concerns on the 1st Law issue. This time it relates directly to the issue addressed in this thread as an example of my concerns.

Using this as an example, using the 1st Law equation as presented in post#8, I to calculate the final Tank temperature, ignoring tank heat transfer, to be as follows:

 P i = 4.85​ psia T I = 70​ °F T I = 530​ °R P f = 14.7​ psia k air = 1.4​ T f = 655​ °R T f = 281​ °F

Which would indicate that even with an large amount of Tank heat loss during the 9 sec filling time, there would be an additional amount of filling time required as the Tank consumed the air required to maintain its target 14.7 psia pressure level; or, if the fill valve were to be closed, a substantial post fill pressure drop would be expected as the tank cools to ambient temperature.

If you are willing to continue consuming your time with my ongoing inquiries I would really appreciate your feedback on this issue; if not, I fully understand because it getting to one of those extended "OK, but what about" situations.

Jack
655 R = 195 F, not 281 F. This agrees roughly with the results of my calculations and those of the OP.

JBA

My error, checking my conversion program confirms that 195°F is indeed correct; but, that still does not address my concerns regarding the issue of a pressure drop associated with the return of the tank's temperature to ambient or does your calculation method account for this as well.
If so, this type of calculation with a single equation that crosses the Pcr point and accounts for the pressure vs temperature reduction is something I would very much like to learn and understand.

Chestermiller

Mentor
My error, checking my conversion program confirms that 195°F is indeed correct; but, that still does not address my concerns regarding the issue of a pressure drop associated with the return of the tank's temperature to ambient or does your calculation method account for this as well.
If so, this type of calculation with a single equation that crosses the Pcr point and accounts for the pressure vs temperature reduction is something I would very much like to learn and understand.
The analysis in this thread (so far) assumes that the process takes place adiabatically, and does not include heat transfer from the gas to the tank walls and surroundings. This, of course, can be included, but, for a 9 second process, its inclusion does not seem worthwhile.

With regard to your question regarding the critical pressure, I don't quite follow. However, if changes are occurring at high pressures, we would need to include consideration of non-ideal gas effects. For the present problem, that wasn't necessary.

JBA

With regard to my question regarding the critical pressure, it wasn't focused upon the critical pressure; but, on the fact that, if I understand it correctly, you appear to have a method to calculate the flow rates for the full range of pressure ratios including both the critical and sub critical pressure regions without having to calculate the two regions separately and sum the results; and, if so, I really want to understand that method.

Chestermiller

Mentor

With regard to my question regarding the critical pressure, it wasn't focused upon the critical pressure; but, on the fact that, if I understand it correctly, you appear to have a method to calculate the flow rates for the full range of pressure ratios including both the critical and sub critical pressure regions without having to calculate the two regions separately and sum the results; and, if so, I really want to understand that method.
OK. The calculations involve integrating an equation expressing the time derivative of pressure as a function of pressure itself. The key parameter in the calculation is the expansivity factor $\epsilon$, which is a function of the ratio of the pressure in the tank to the outside pressure. There are two regions of functionality, as you indicated, critical and sub critical. The expansivity factor is continuous at the transition between these regions.

The OP and I got our information on the expansivity factor from a Wiki link on Oriface Plates: https://en.wikipedia.org/wiki/Orifice_plate

We have experimented with different forms of the relationship, as discussed in post #39. Three of these forms give nearly identical results for the functionality: the theoretical form recommended on the link, a continuous approximation that I proposed, and a linear approximation, also continuous over the entire range. All three of these functionalities give nearly identical results for the tank pressure vs time, and match the OP's experimental data on pressure vs time astonishingly well.

JBA

Thanks for the reference, until this point I was surprised that I had not encountered any of the type of analysis that you were using but I now understand because in our pressure relief industry and products we are required by the customers to design our products with high efficiency nozzles, generally in the Cd = 97-98 range; and, in our flow testing and ASME certification facility we use ASME certified Cd sonic nozzles for flow measurement. As a result, all of my technical references and investigations have specifically focused on nozzle design where beta ratio focused theories are not an issue , as opposed to orifice plate flow measurement technology.

"Unsteady filling of a vacuum tank"

Physics Forums Values

We Value Quality
• Topics based on mainstream science
• Proper English grammar and spelling
We Value Civility
• Positive and compassionate attitudes
• Patience while debating
We Value Productivity
• Disciplined to remain on-topic
• Recognition of own weaknesses
• Solo and co-op problem solving