The field of the invention relates to a technique predicting maintenance needs of a pump, including large diameter slurry pumps.
Pumps need to be maintained during the course of their lifetime. Pumps that are well maintained have a tendency to operate longer and more efficiently than pumps that are not well maintained.
By way of example, maintenance may be carried out to prolong the lifetime of a pump, or to increase the efficiency of the pump, or both. An example of lifetime prolonging maintenance is the dynamic balancing of impellers which reduces vibration and thereby increases the lifetime of a pump. An example of maintenance geared to increase efficiency is the replacement of worn parts.
It is known that an economic trade-off exists between increasing the efficiency of a pump, thereby reducing power usage, and decreasing maintenance needs and thus lowering the cost of non-productive down time and other costs related to performing the maintenance, including the cost of the parts, the man hours to perform the maintenance, etc. This trade-off exists, e.g., because the maximum efficiency of a pump can be achieved by reducing the pump's internal clearances and thereby making the internal pressure seal of the pump tighter. However, in doing so, the shear rates in the clearance spaces typically will increase, and thus the wall friction will typically increase as well leading to increased wear when the pumped fluids contain abrasive particulate material.
There is a need for a better way for maintaining a pump, including a better way for predicting the maintenance needs of a pump, including, e.g., large diameter slurry pumps.
Predicting maintenance needs of a pump, including, e.g., large diameter slurry pumps, is a task that may be approached using a variety of tools and techniques. The present invention looks to optimize and maintain the trade-off that exists between increasing the efficiency of a pump, thereby reducing power usage, and decreasing maintenance needs and thus lowering non-productive down time and other costs related to performing the maintenance. In order to do so, the present invention provides a new way to monitor the decrease in pump efficiency with wear, such that it becomes feasible to adjust the maintenance interval to the rate of wear, and vice versa the wear rate to the maintenance interval, so that the total cost of ownership is substantially minimized. Adjusting the rate of wear in a pump, of course, can be accomplished by increasing or decreasing internal clearances using a predictable maintenance schedule based at least partly on historical data that is tracked over time.
For example, one way of tracking pump efficiency is to compare over time, the hydraulic power generated by the pump and the electric power consumed by the motor driving the pump. Assuming that bearings and seals are well maintained and lubricated where necessary the wear of these elements might be negligible compared to the wear of wetted pump parts subjected to abrasive internal flow. If this assumption holds, then the ratio of hydraulic to electric power can be a useful number or indicator to track as it will allow striking a balance between efficiency and pump wear so as to make the cost of operating the pump as low as possible.
The present invention provides a new way to track pump efficiency over time and is based at least partly on a technique developed to track and analyze pump efficiency over a long (238 days) time period for two large diameter centrifugal slurry pumps feeding a battery of cyclone separators. The method is reasonably straightforward in theory and in its implementation, although it requires detailed attention to data quality.
According to some embodiments, the present invention may be implemented in the form of a signal processor comprising one or more signal processor modules configured to:
According to some embodiments of the present invention, the one or more signal processor modules may be configured to compare over time the hydraulic power generated by the pump and electric power consumed by a motor driving the pump, including tracking a ratio of hydraulic to electric power.
According to some embodiments of the present invention, the one or more signal processor modules may be configured to predict maintenance needs based on an algorithm that takes into account a trade-off between the cost of power consumed by the motor driving the pump and the cost associated with the pump maintenance in order to substantially minimize the total cost of operating the pump.
According to some embodiments of the present invention, the corresponding signaling containing information predicting maintenance needs of the pump may include information about one or more of the following: dynamically adjusting pump clearances, dynamically balancing an impeller and replacing worn parts.
According to some embodiments of the present invention, the historical data related to the performance of the pump may include the discharge flow and pressure, intake level, fluid density, pump speed, or motor current, or some combination thereof.
According to some embodiments of the present invention, the one or more signal processor modules may be configured to use non-dimensional quantities to infer the performance of the pump and the trending of the efficiency of the pump, where the non-dimensional quantities may include: the head coefficient, the flow coefficient, the efficiency, the power coefficient, the speed coefficient and the specific speed of the pump.
According to some embodiments of the present invention, the one or more signal processor modules may be configured to adjust the maintenance interval to the rate of wear, or vice versa, so that the total cost of operating the pump is substantially minimized.
According to some embodiments of the present invention, the one or more signal processor modules may be configured to receive the signaling containing information about historical data related to the performance of the pump; to store the signaling containing information about historical data related to the performance of the pump in a historical data database; or both.
According to some embodiments of the present invention, the one or more signal processor modules may be configured to analyze the signaling containing information about historical data related to the performance of the pump using a n-day moving window, e.g., a 2-day moving window.
According to some embodiments of the present invention, the one or more signal processor modules may be configured to receive information about maintenance performed on the pump, and to store associated signaling containing information about the maintenance performed in the historical data database.
According to some embodiments of the present invention, the one or more signal processor modules may be configured to use an algorithm that is based at least partly on prolonging the lifetime of the pump, increasing the efficiency of the pump, or both.
According to some embodiments of the present invention, the one or more signal processor modules may be configured to predict maintenance needs based at least partly on an algorithm that takes into account that the fact that maximum efficiency of the pump may only be achieved by reducing the pump's internal clearances and thereby making the internal pressure seal of the pump tighter, and that by doing so the shear rates in the clearance spaces increase and thus the wall friction will increase as well leading to increased wear when the pumped fluids contain abrasive particulate material.
According to some embodiments of the present invention, the one or more signal processor modules may be configured to extrapolate information related to the performance of the pump to a Best Efficiency Point, thus providing an efficiency measure at a common point.
The present invention may also be implemented in the form of a method comprising: comparing signaling containing information about historical data related to the performance of a pump; and providing corresponding signaling containing information about predicted maintenance needs of the pump on a real time basis based at least partly on the signaling compared containing information about the historical data related to the performance of the pump. According to some embodiments, the method may be implemented to include one or more of the features set forth above in relation to the signal processor.
According to some embodiments, the present invention may be implemented in apparatus taking the form of a computer-readable storage medium having computer-executable components for performing the aforementioned method, when executed on a signal processor running on a computer device.
The drawing, which includes
By way of example, the one or more signal processor modules 12 may be configured to compare over time one signal containing information about the hydraulic power generated by the pump and another signal containing information about electric power consumed by a motor driving the pump, including tracking a ratio of hydraulic to electric power. Techniques for determining the hydraulic power generated by the pump and the electric power consumed by the motor driving the pump, as well as comparing the associated signaling as a whole, are known in the art, and the scope of the invention is not intended to be limited to the manner in which the hydraulic and electric power related to the pump are determined or compared. Moreover, a detailed description of pump hydraulics, as well as pump data consistency, pump performance analysis, etc., is set forth herein, by way of example, that may be used to make these determinations in whole or in part.
Based at least partly on the aforementioned comparison, the one or more signal processor modules 14 may also be configured to predict maintenance needs of the pump based at least partly on an algorithm or procedure that takes into account a trade-off between the cost of power consumed by the motor driving the pump and the cost associated with the pump maintenance in order to substantially minimize the total cost of operating the pump over the course of its lifetime. For example, the algorithm may be provided the cost of power, e.g. the cost of electricity per kilowatt hour. The algorithm may also be provided the cost of a variety of factors related to or associated with the pump maintenance needed, including the cost of various parts that may be replaced, e.g. the impeller, seals, etc., the cost related to one or more adjustments of parameters like impeller clearance that may be needed, as well as the cost of performing the maintenance, including the production down-time and man hours to perform the maintenance needed, e.g. replacing one or more parts or adjusting one or more parameters. The cost of production down time may include, e.g., the cost of product not being made or processed while the maintenance is performed. The cost of power and the variety of factors may be stored in a database, and updated from time to time as the cost changes, e.g. due to inflation. With this information, the algorithm or procedure may be implemented to determine the power consumed by the motor driving the pump based at least partly on its current configuration, and to determine if it is likely to be cost effective to replace a worn down part or make an adjust based at least partly on the current pump performance and the likelihood of an improved pump performance that may result from the maintenance being performed. A person skilled in the art can formulate such an algorithm or procedure to implement the present invention consistent with that shown and described herein without undue experimentation, and the scope of the invention is not intended to be limited to any particular type or kind of implementation.
The corresponding signaling containing information predicting maintenance needs of the pump may include, by way of example, information about one or more of the following: dynamically adjusting pump clearances, dynamically balancing an impeller and replacing worn parts. However, the scope of the invention is not intended to be limited to the type or kind of part being replaced, or the type or kind of adjustment being made when the predicted maintenance needs are performed, and is intended to include other types or kinds of parts or adjustments either now known or later developed in the future.
The historical data related to the performance of the pump may include, by way of example, the discharge flow and pressure, intake level, fluid density, pump speed, or motor current, or some combination thereof. Moreover, the detailed description of the pump hydraulics, pump data consistency, pump performance analysis, etc., set forth herein may be used to make the determinations and/or comparisons in whole or in part based at least partly on the historical data. However, the scope of the invention is not intended to be limited to the type or kind of historical data being used to make the predicted maintenance needs, and is intended to include other types or kinds of historical data either now known or later developed in the future.
The one or more signal processor modules 14 may be configured to use non-dimensional quantities to infer the performance of the pump and the trending of the efficiency of the pump, where the non-dimensional quantities may include: the head coefficient, the flow coefficient, the efficiency, the power coefficient, the speed coefficient and the specific speed of the pump. Moreover, the detailed description of pump hydraulics, pump data consistency, pump performance analysis, set forth herein may be used to make the determinations and/or comparisons in whole or in part based at least partly on the historical data. However, the scope of the invention is not intended to be limited to the type or kind of non-dimensional quantities being used to make the predicted maintenance needs, and is intended to include other types or kinds of non-dimensional quantities either now known or later developed in the future.
The one or more signal processor modules 14 may be configured to adjust the maintenance interval to the rate of wear, or vice versa, so that the total cost of operating the pump is substantially minimized. The adjustment of the maintenance interval may be implement based at least partly on an algorithm that takes into account a relationship between, e.g., the maintenance interval of a particular part that needs to be maintained and the rate of wear of the particular part, or vice versa, so that the total cost of operating the pump is substantially minimized. Information related to these parameters may be provided to the signal processor, stored in a historical database, etc., and used to formulate or determine the adjustment to the maintenance interval or the rate of wear, or vice versa, so that the total cost of operating the pump is substantially minimized. Embodiments are also envisioned in which these parameters are stored in a database and a table look-up technique is used. A person skilled in the art can formulate such an algorithm or procedure to implement the present invention consistent with that shown and described herein without undue experimentation, and the scope of the invention is not intended to be limited to any particular type or kind of implementation.
The one or more signal processor modules 14 may be configured to receive the signaling containing information about historical data related to the performance of the pump; to store the signaling containing information about historical data related to the performance of the pump in a historical data database; or both. However, the scope of the invention is not intended to be limited to the manner in which the signaling is received or stored, and is intended to include other types, kinds or techniques for receiving or storing signaling either now known or later developed in the future.
The corresponding signal containing information predicting maintenance needs of the pump may take the form of audio or visual information, including a report that can be printed out on a printing device or displayed on a video monitor, as well as an audio announcement or warning. The corresponding signal containing information predicting maintenance needs of the pump may also take the form of a wireless signal that is transmitted to a remote location to further processed, and may also take the form signaling provided to a website on the Internet that can be downloaded by a remote computer desktop terminal, laptop, personal digital assistant (PDA), Blackberry device, cell phone, etc.
Signal processing technology is known and available in the art for implementing the functionality of the one or more signal processing modules 14. By way of example, the functionality of the one or more signal processor modules 14 may be implemented using hardware, software, firmware, or a combination thereof. In a typical software implementation, the one or more signal processor modules 14 would include one or more microprocessor-based architectures having a microprocessor, a random access memory (RAM), a read only memory (ROM), input/output devices and control, data and address buses connecting the same. A person skilled in the art would be able to program an algorithm or procedure in such a microprocessor-based implementation to perform the functionality described herein without undue experimentation. The scope of the invention is not intended to be limited to any particular type or kind of signal processing technology either now known or later developed in the future, and embodiments are envisioned using other types or kinds of signal processing technology either now known or later developed in the future.
The scope of the invention is intended to include the signal processor module 14 being implemented as a stand alone component or module or implemented as multiple modules.
The functionality of the one or more modules may also be implemented as apparatus taking the form of a computer-readable storage medium having computer-executable components for performing the steps of the aforementioned method or technique described herein.
The following is a summary of a detailed description of a pump performance analysis that was carried out on two large diameter cyclone feed pumps. The detailed description is set forth below in sections 1 through 5.
The analysis was carried out using historical data of discharge flow and pressure, intake sump level, fluid density, pump speed and motor current. Six non-dimensional quantities, the head coefficient, the flow coefficient, the efficiency, the power coefficient, the speed coefficient and the specific speed of the pump were arranged in a four plot to analyse and infer pump performance and efficiency trending. The analysis was carried out using a 2 day moving time window over a period of some 240 days spanning one impeller change of each pump.
For pump 1 a clear downward trend in the efficiency was found during the pre-impeller change. Post-impeller change in the efficiency is not increased until some time later when a second downward trend was observed. The delayed increase in efficiency after an impeller replacement could be due to pump clearance readjustment.
For pump 2 no clear trend was detected. Post-impeller replacement a slight but significant increase in efficiency was seen.
The performance analysis using non-dimensional quantities may be used on a real time basis in order to trade-off the cost of power (lowest at high efficiency) and the cost of pump maintenance (lowest at low wear rates) by adjusting pump clearances dynamically such that the rate of pump wear (as indicated by a decreasing trend in efficiency) is chosen to minimize the total cost.
In one particular example, the trade off exists because maximum efficiency of a pump can typically be achieved by reducing the pump's internal clearances and thereby making the internal pressure seal of the pump tighter.
One way of tracking pump efficiency is to compare over time, the hydraulic power generated by the pump and the electric power consumed by the motor driving the pump. Assuming that bearings and seals are well maintained and lubricated where necessary the wear of these might be negligible compared to the wear of wetted pump parts subjected to abrasive internal flow. Based on this assumption, the ratio of hydraulic to electric power can be a useful number or indicator to track as it should allow striking a balance between efficiency and pump wear so as to make the cost of operating the pump as low as possible.
The following describes a way to track pump efficiency over a fairly long (238 days) period for two large diameter centrifugal slurry pumps feeding a battery of cyclone separators. The method is straightforward in theory but in practice it requires detailed attention to data quality.
The purpose of a pump is to increase fluid pressure at a certain rate of flow. Centrifugal pumps achieve this by imparting kinetic energy to the fluid by means of a rotating impeller. Subsequently by decelerating the fluid in the pump's snail house the pressure is increased. The theory of hydro kinetic machinery, including centrifugal pumps is explained in detail by A. C. Walshaw and D. A. Jobson, Mechanics of Fluids, 3rd printing, Longmans, London, 1967.
As a measure of a pump's efficiency the head coefficient CH is often used which is a dimensionless number defined as:
In this equation Δp is the pressure differential created, ρ the density of the fluid, ω the circle frequency of the impeller and D the diameter of the impeller. Under ideal conditions the fluid at the tip of the impeller leaves the impeller at the tangential speed of the impeller which is equal to the product of the circle frequency and the impeller radius. Thus the denominator in equation {1} measures the dynamic head of the fluid leaving the impeller whereas the numerator measures the static head. The ratio of the two is a measure of how well the pump converts the impeller's kinetic energy into pressure.
Ideal conditions only apply when the fluid enters the pump without axial or tangential velocity components and when the fluid leaves the impeller without radial or axial velocity components. To account for non ideal conditions it is useful and instructive to relate the head coefficient to a dimensionless value of the rate of flow sustained:
Where Q is the volumetric rate of flow. The denominator in equation {2} is proportional to the volume swept by the impeller per revolution. Therefore the flow coefficient CQ relates the flowrate through the pump to the pump size and speed.
It is often assumed that there is a quadratic relationship between the head coefficient and the flow coefficient:
C
H
=a+bC
Q-cCQ2 {3},
where the coefficients a, b and c are all positive. Note that this relation is purely empirical. There is absolutely no hydraulic pump theory backing it up. Theory only stipulates a relation between the head and the flow coefficient but theory does not predict the shape of such a relation. The occurrence of a constant offset (a) in the empirical relation equation {3} may be interpreted as the pressure a pump can hold at zero flow. Physically, of course, this is not realizable as it would stipulate a pump which completely seals the outlet from the inlet without any clearances. Such a pump would not be practical because of the mechanical friction that the impeller would impart on the pump casing.
The hydraulic power generated by the pump is simply the product of the rate of flow and the pressure differential created. By comparing this to the power consumed by the (electric) motor driving the pump shaft an overall efficiency may be defined as:
This efficiency definition assumes that the motor is a one phase motor with a power factor equal to the cosine of the phase shift between the current I and the voltage V. By using the empirical relation between the head and flow coefficient it follows that the efficiency will be a cubic in the flow coefficient:
η=pCQ+qCQ2+rCQ3 {5},
where it should be noted that the coefficients p, q and r are not equal to a, b and c because the efficiency is also related to the (independent) pump current. It does follow, however that the efficiency should be zero at zero flow, i.e. there is no offset constant in equation {5}.
Using the same equation for hydraulic power a dimensionless power coefficient may be defined that is independent of the pump's speed:
A dimensionless speed coefficient is easily derived from the head coefficient as:
Clearly this is related to the impeller's tangential speed and the fluid velocity in the pump outlet.
All dimensionless quantities so far defined are related in some manner to the size of the pump expressed in terms of the impeller diameter D. A specific speed Ns may be defined as:
which is now a number that for geometrically similar pumps will be the same and will be constant at the maximum efficiency point.
It follows that the power and speed coefficient are related as:
C
P
=u+wC
S {9}.
Whereas the specific speed is related to the speed coefficient as:
Note that it is an error to assume that the offset constants in equation {9} and in equation {10}, u and x respectively, are zero.
The maximum efficiency of a pump may be calculated from equation {5} once the coefficients p, q and r have been determined by e.g. curve fitting of actual pump data. In terms of these coefficients the flow coefficient at the Best Efficiency Point (BEP) equals:
The corresponding quantities for the head coefficient, the speed coefficient, the power coefficient and the specific speed all at the BEP can now be easily derived:
Calculating the power coefficient, speed coefficient and specific speed this way should result in points that are coincident with the regression lines as defined by equation {9} and equation {10}.
Because of changing operational load on a pump, the pump is not always (almost never) running at its BEP; and, therefore, the present invention may be implemented so as to extrapolate information related to the performance of the pump to the BEP when required, thus providing an efficiency measure at a common point making a comparison of efficiencies over time practical.
The Cyclone Feed Pump (CFP) data set consists of data retrieved from an OSI PI data historian. The following was retrieved for a period of about 1 year for two pumps. The data is delivered in Excel spread sheets and is distributed over several workbook each having several tab sheets. A small routine in VBA serves to consolidate all data per pump in a single text file.
The data is time stamped every minute. It should be noted here that a data historian like PI does not normally sample data at regularly spaced intervals. Data from an instrument is saved only when the difference between the last stored point and a new point is outside a certain predefined hysteresis band. Subsequently upon a database query, when data points are required where there are no direct timestamped data available, the values are linearly interpolated. An example of the process is shown in
Obviously the process leads to a much smaller number of points to be stored. The price one pays for this is that the distribution of data is altered as all interpolated points are now dependent on each other.
In the event of a communication outage, hardware unavailability or other type of error a string is timestamped and stored. The currently known errors include the ones in the table below. Upon reading the data files such errors are replaced by missing value indicators (NA) and subsequently for further analysis the complete record (timestamped lines) with missing values is dropped from the data set.
As a result of dropping the missing values the contiguous timestamping is lost. It is not in general possible to overcome this by interpolation on non-missing points as the intervals containing missing values can be rather long.
As it will turn out, missing values are not the only issue. At times unrealistic measurement values are returned such as sump level values higher than 100% or density values below 0%. Such unphysical values may be measurement error but they may also be the result of a process condition that is not representative of normal operations. To be on the safe side, the data is filtered as per the table below.
In effect all conditions that are filtered are dropped from the data set. This reduces the number of available data points. For pump 1 the reduction is from 344144 points to 287622 points, which equates to some 39 days lost. Most of the missing data, some 15101 total, originates from the cyclone valves position not being known. If we exclude this, i.e. include all valid data when only the valve positions are not known, then 801 points are lost for pump 1. Pump 2, 321824 points contains 14382 missing valve position points. When filtering is applied 291855 points remain a loss of about 21 days.
In order to assess the cyclone feed pump performance quantitatively by using the dependence of the efficiency of the pump on the flow coefficient, it is necessary to first calculate the differential pressure over the pump.
The reference level for (hydrostatic) pressure is the center line of the pump 52. All vertical distances will be measured from this level. The pressure of the cyclone battery 56 is measured at the top of the battery 56 by a pressure gauge P measuring relative to atmosphere. The sump level in the tank 54 produces a positive intake pressure also dependent on the barometric pressure. Hence the difference is not dependent on the barometric pressure.
The sonar-based flowmeter 58 in the riser to the battery 56 measures the pump's discharge flow rate. By way of example, the sonar-based flowmeter technology may include the GH-100 and/or GVF-100 meter developed by the assignee of the instant patent application. By way of example, the sonar-based flowmeter 58 is disclosed in whole or in part in U.S. Pat. Nos. 7,165,464; 7,134,320; 7,363,800; 7,367,240; and 7,343,820, which are all incorporated by reference in their entirety.
Since there is no direct pump differential pressure available, the differential pressure over the pump 52 is calculated. The pump's outlet pressure is calculated from the measured battery pressure by adding the frictional and hydrostatic pressure losses to it but neglecting acceleration-related pressure losses or losses in pipe bends. The pump's inlet pressure is calculated by subtracting frictional losses from the hydrostatic head generated by the level in the sump tank 54.
The losses in bends (e.g. element 52c) are not taken into account because the radii of the bends are not known. The acceleration-related losses are dependent on the diameter increase from the pipe line to the battery which is similarly unknown. The hydrostatic losses however are by far the largest; an estimate of the other terms shows that at maximum pump rate the losses due to acceleration-related effects, sharp bends or sump tank outflow are each less than 3% of the hydrostatic loss. Note that density has no impact on this estimate because all losses scale linearly with density.
Table 4 lists the assumed values for the piping layout. Note that these are mostly estimates as precise measurements are not available. Using these hydraulic quantities we can calculate the density of the slurry being pumped as:
The solids volume fraction follows from the calculated slurry density by
Which in turn gives the viscosity of a slurry, according to the Einstein formula:
The pump's inlet pressure follows as:
The pump's outlet pressure equals:
In both of the above the Fanning friction factor f can be calculated from the Reynolds number Re by:
This last equation was first published by S. W. Churchill, Friction factor equation spans all fluid flow regimes, Chem. Eng. 7, November 1977, p. 19. It covers both rough and smooth pipes over the entire range of Re numbers including the transition region from laminar to turbulent flow.
After calculating the inlet and outlet pressure of the pump, the differential pressure is easily found. This enables the calculation of the six non dimensional quantities introduced in 2.1. The head coefficient and efficiency are plotted as a function of the flow coefficient. The power coefficient and specific speed are plotted versus the speed coefficient. This gives a four plot presenting different views of pump performance each with its own merit.
This cross plotting is repeated for each pair of days, moving forward in time by day. The determined BEP can thus be seen as an average over 2 days of data. Note that the ‘days’ as used here need not be consecutive calendar days because the data set had to be filtered heavily to remove non-representative and bad data.
Note as well that the very slight deviation from a straight line in the first, top right panel of the head versus flow coefficient does not at all mean that the efficiency curve is equally flat. In fact, the efficiency curve here has a very well defined maximum.
The data fits are carried out by a least squares method which leads to a design matrix in normal form. The fits are unconstrained however which allows the coefficients of the quadratic fit for the head curve and the cubic fit for the efficiency curve to take any value.
Under certain circumstances, notably when there is a small spread in the flow rate data, the head curve fit is then no longer found to be an inverted parabola. This is shown in
Obviously this type of behaviour is not desired. There are two potential remedies. First one may argue that if the flow range covered is too narrow to infer curvature in the head curve the fitted model should be linear rather than quadratic and the fit attempted should be redone with a linear model if the values of the coefficients of a quadratic model indicate a non inverted parabola. This in turn implies that the efficiency curve is an inverted parabola going through origin.
A different, yet somewhat more complicated, approach would be to constrain the fits by setting up a system of equations expressing the value ranges for the fitted coefficients and solving this problem by what is known as a quadratic programming model.
Note that in
For the second pump, similar behaviour was found. An example of a good, yet very different behaviour is shown in
The second example of this pump shows a normal head curve; a slightly downward curving line. Yet, the efficiency curve shows an upward curving tail at high flow coefficients which is not possible in a real pump. This serves to show that it is not sufficient to rely on the head curve in order to define the order of the polynomial used for the efficiency curve. After all the efficiency curve does use a completely independent piece of data namely the motor current.
Thus the procedure outlined in the previous paragraph must be invoked for both the efficiency curve and the head curve. Even if the head curve quadratic is accepted, inspection of the coefficients of the efficiency curve is still necessary.
See also the large dots at coordinates of about 0.013, 0.10 (
The main result of course is the value of the BEP. If one plots the BEP value versus the actual calendar day for both pumps, then one sees that for pump 1 this results in a downward trend in BEP whereas for pump 2 no clear trend is discernible.
Vertical red lines mark the days on which pump maintenance was carried out. The mid period line indicates an impeller replacement only, the two outer period lines represent a complete overhaul. The blue lines depict the evolution of the 95% prediction interval on the value of the BEP.
An unusual event occurs in the middle of the first period on pump 1. This event occurs on 21 Oct. 2008 and gives an efficiency of about 1.8. The entire spike around that period, which disturbs the downward trend of the pump efficiency which is otherwise visible is suspect. The detailed analysis of the 21st is given by a four plot as given in
It turns out that the efficiency data is broken up in 3 distinct groups which are not or only to a much smaller extent visible in the other three graphs of the four plot. This indicates that the grouping is almost certainly caused by a wide difference in the pump's motor current. Motor current is not a factor in either of the other three quantities, head coefficient, power coefficient or specific speed. It is possible that the current probe may have been replaced around that day and was not properly recalibrated until some days later. The effect of the recalibration is then immediately visible here as three distinct groups.
Obviously, if the explanation above is correct, similar events may occur because of replacement or recalibration of the flow meter or the pressure gauge. Contrariwise, those events will be visible in all of the four plot and may thus be much harder to identify.
The main result obtained as presented in
For example, on pump 1 there is no clear increase in efficiency after the mid period overhaul but there is a clear increase some time later in January. This increase is again followed by a down ward trend at about (at least visually) the same rate as before. It must be understood here that the efficiency of a pump is strongly dependent on the various clearances between impeller and casing. If those clearances increase due to wear there will be increased internal recirculation inside the pump which causes the pressure and discharge flow to be less than optimum. This is in effect what the analysis presented above attempts to detect.
However, if the clearances are high to begin with then pump wear is reduced because of the lower shear rate in the fluid between casing and impeller and the efficiency will be less to begin with. It is possible that upon a pump overhaul the clearances are not set correctly or, conversely, the clearances are adjusted whilst the pump is running or stopped for a short period of time. Some pumps are fitted with clearance adjustment screws specifically designed to allow just that.
In effect this allows the operator of the pump to choose between different modes. The operator can choose to run the pump at a low rate of wear, as for instance shown in pump 2 here but a reduced efficiency. Alternatively the operator can try to achieve a high efficiency of the pump, thereby lowering the cost of electric power traded of against a higher maintenance cost because now the pump will wear faster.
Long periods of pump historical data are sufficiently consistent to use in a statistical analysis geared towards inferring pump performance changes. The data can be filtered reliably to remove outliers. Where high pairwise correlations are expected they indeed occur such as between motor current and flow rate.
It is possible to estimate the generated pressure differential over the pump by back calculation from a downstream measured pressure (battery pressure) and an upstream level measured in the tank from which the pump draws fluid (sump tank).
Having a direct measurement of pump differential pressure will increase the reliability and accuracy. Presently losses in pressure due to friction are estimated, whilst pressure losses due to bends and pipe diameter changes are neglected.
The model equations for the head and efficiency (a quadratic for an inverted parabola and a cubic passing through the origin) do in most cases describe the pump performance adequately. When the range of flow rates covered within an analysis period is too small, it may be necessary to reduce the order of the polynomials by one or to perform constrained fits.
The non-dimensional coefficients remove the effect of varying pump speed thereby allowing the analysis to proceed regardless of the pump speed. Non-overlaying data can thus only be due to measurement data that is no longer in line with past data, possibly because of gauge repair or recalibration.
In rare cases the non-dimensional analysis shows non-overlaying features in the efficiency curve. Since the non overlay occurs only in the efficiency curve the suspected gauge is the current clamp used to monitor motor current as this is the unique measurement in the efficiency.
The final result for pump 1 shows a very clear downward trend in pump efficiency from a maintenance period to the next. From thereon the efficiency varies but no clear trend with time is discernible. Mid second period however the efficiency does increase by a significant amount. This could be due to pump clearance re-adjustment. Thereafter the efficiency trends down again.
The final result for pump 2 does not show a downward trend in efficiency at all. A slight increase in efficiency is observed though right after the maintenance period.
A person skilled in the art could implement VBA scripts and R scripts that were used as part of this analysis without undue experimentation to perform the functionality described herein in relation to the present invention, including the R scripts that do the following: check data consistency; create model plots of dimensionless numbers; process in batch, suppress all model plots and write results to files; and do final post processing of results files.
Accordingly, the invention comprises the features of construction, combination of elements, and arrangement of parts which will be exemplified in the construction hereinafter set forth.
It will thus be seen that the objects set forth above, and those made apparent from the preceding description, are efficiently attained and, since certain changes may be made in the above construction without departing from the scope of the invention, it is intended that all matter contained in the above description or shown in the accompanying drawing shall be interpreted as illustrative and not in a limiting sense.
This application claims benefit to provisional patent application Ser. No. 61/186,502, filed 12 Jun. 2009, which is hereby incorporated by reference in its entirety.
Number | Date | Country | |
---|---|---|---|
61186502 | Jun 2009 | US |
Number | Date | Country | |
---|---|---|---|
Parent | 13377083 | Feb 2012 | US |
Child | 16694323 | US |