Compressor surge detection system

Information

  • Patent Grant
  • 4940391
  • Patent Number
    4,940,391
  • Date Filed
    Monday, November 7, 1988
    36 years ago
  • Date Issued
    Tuesday, July 10, 1990
    34 years ago
Abstract
In a compressor system, surge detection is provided by sensing the current of the motor driving the compressor and reading oscillations of the current about the average thereof and above and below two opposite threshold levels from such average current and counting the polarity changes within a time interval containing so many samples of the current and defining a sliding window.
Description

BACKGROUND OF THE INVENTION
The invention relates in general to compressor control, and more particularly to the prevention of the consequences of "surging" in such compressor control. "Surging" is a phenomenon caused by the compressor operating below a given amount of horse power or gas volume and the resulting occurrence of nonsteady state condition compounded by erratic control and an uncontrolled output.
The problem of surging and the prevention thereof have been explained in "Centrifugal and Axial Compressor Control by Gregory K. McMillan (The Instrument Society of America, 67 Alexander Drive, P.O. Box 12277, Research Triangle Park, NC 27709), 1983.
The present invention aims at an early detection of the occurrence of a surge, thereby providing an immediate opportunity of preventing the consequence thereof. Therefore, the invention distinguishes itself from the prior art methods by preventing the consequences of a surge by the timely and accurate detection of an occurring surge.
The invention is based on the magnitude of the current of the motor driving the compressor at the critical moment. The prior art has recognized the relationship existing between the volume of the compressed gas, or the pressure of the compressor, to the energy of the motor in operation, or its current. However, the object, there, has been to control the compressor as a function of the sensed current in order to prevent a surge, not to detect the imminence of a surge, or in order to control the compressor in relation to the gas volume while observing the surge limit. See, for instance, U.S. Pat. Nos. 3,778,695; 4,519,748 and 3,380,650, and also: "Surge Control For Multistage Centrifugal Compressors" by David F. Baker in Chemical Engineering May 31, 1982, pp. 117-122; "Improved Surge Control for Centrifugal Compressors" by N. Staroselsky and L. Ladin in Chemical Engineering May 21, 1979, pp. 175-184; and "Surge Control for Centrifugal Compressors" in Chemical Engineering Dec. 25, 1972, pp. 54-62.
SUMMARY OF THE INVENTION
The invention provides for the detection of the imminence of a surge in a motor driven compressor by sampling the motor current within a predetermined time interval and counting the occurrence of a predetermined minimum number of changes of polarity of the derived samples within such time interval.
Such changes of polarity represent sharp deviations of the motor current in opposite directions from normal operation current magnitude, these deviations being due to motor loading behavior with the compressor and manifesting themselves as a result of a cyclical change of operation by the motor between motoring and regenerating modes, upon the occurrence of a surge. More specifically, such deviations are measured by reference to the average current as sensed under normal operation and they are converted into an error condition when beyond a predetermined minimum deviation used as a threshold. The average current used as a reference is obtained by sampling the actual current and averaging the samples through a sliding window. Each error condition indicative of the threshold being exceeded is stored with an indication of its polarity. Changes of polarity are counted through a predetermined time interval as an indication of a surge condition. Protective measures are immediately triggered within the compressor system upon the detection of a surge condition.





BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is a block diagram of the compressor control system according to the present invention;
FIG. 2 is a graphic representation of the occurrence of a surge translated into polarity changes about two reference levels encompassing the average motor current, and as detected according to the present invention;
FIG. 3 is a block diagram of the surge detection system according to the invention which is part of the compressor control system of FIG. 1;
FIG. 4 illustrates with curves the occurrence of the control signals during the operation of the system of FIG. 3;
FIG. 5 is a block diagram of the internal organization of the twin summer used in FIG. 4 for the derivation of the thresholds to the comparator also of FIG. 4;
FIG. 6 is a block diagram illustrating the internal organization of the comparator and the trip signal derivation as taken from FIG. 3;
FIGS. 7A to 7E are operational amplifier implementations illustrating the hardware interconnection between the elements of FIGS. 5 and 6;
FIG. 8 is a flow chart explaining the steps involved in the operation of a microprocessor implementation of the circuit of FIG. 3;
FIG. 9 shows the solid state implementation of the circuit for the derivation of a signal representing the motor current;
FIGS. 10A to 10E are illustrating the solid state implementation of the circuit of FIG. 3 when built around a microcomputer.





DESCRIPTION OF THE INVENTION
Referring to FIG. 1, a compressor CMP is shown connected between the duct from the inlet and the duct to the outlet for a system using pressurized fluid. There is a separate duct having a normally closed valve BV used in an emergency to blowoff the outlet pressure and override the load for protection. The inlet valve IV is placed at the inlet to control the incoming fluid, especially during startup. The compressor is driven by an electric motor, typically an AC motor supplied on three phases, A,B,C from the AC power lines L1,L2,L3 through a contactor CNT. The system is controlled by a unit MCP including a computer, for instance a microcomputer, associated with digital circuitry including A/D and D/A converters interfaced with analog inputs and outputs, respectively. According to the present invention, motor current is sensed (from lines 50) and, after rectification and filtering (within block CNV), a signal representative of the rms of the current I of the motor is derived on line 40. This signal is inputted into the computer control unit MCP, from which is obtained an emergency trip signal on line 70 to a compressor control unit CCU in case alternate polarity changes in sufficient number have been detected, thereby denoting the occurrence of a surge. In response to such trip signal the compressor control unit will, by line 19, provide control for closing of the inlet valve IV and, by line 29, transmit a command to open the bypass valve BV. Control of the valves per se, when required, are well known in the compressor control field. They are controlled from the MCP unit, under normal control of the compressor unit CCU generally known, for instance for startup, pressure regulation, or merely to stop the overall operation.
Referring to FIG. 2, the occurrence of a surge is characterized by the motor current Imot erratically and cyclically oscillating between extreme values, up and down, as shown by curve (C). The gist of the present invention is to detect such oscillation and recognize a surge when oscillations occur a predetermined number of times. As explained hereinafter, high and low thresholds THi and THl are established above and below a normal operation representative signal selected to be the average current Iav occurring during a predetermined period, the latter chosen illustratively to be 4 seconds. When there is an oscillation, such threshold levels are exceeded (as shown between CD for the upper one, and EF for the lower one), alternatively one way and the other, and so many times. Indeed, should the value exceed one threshold more than once without encountering the opposite threshold, there would be no change of polarity, and therefore no count of an oscillation will take place. Imot (the sensed instantaneous value of the current), Iav (the average current for the period), THi (the higher threshold) and THl (the lower threshold) are determined within block MCP from the value of the current sensed and derived on line 40 of FIG. 1. Illustratively, the thresholds THi and THl are at equal distance above and below line AVG or ordinate Iav, by an amount of +TH and -TH (represented in analog terms by voltages vhi and -vlo in FIGS. 7A to 7E as explained hereinafter). The current is sampled and N successive samples are held through a sliding window, the samples therein being on the basis of the well known "first in and last out" method. See for instance: U.S. Pat. Nos. 4,463,432 and 4,229,795. For the purpose of describing how a sliding window is implemented, these two patents are hereby incorporated by reference. Illustratively, it is assumed that 40 samples are collected within a period of 4 seconds, i.e. at a rate of 10 samples per second. When, within such predetermined time interval, so many alternative changes of polarity have been taking place (N* being the critical count in the illustration of FIGS. 3 and 4, hereinafter), the conclusion is that a surge is occurring and that measures have to be taken immediately in order to prevent undesirable consequences for the compressor. Commands for these measures are illustratively shown in FIG. 1 to be control for closing of the inlet valve IV, by line 19, and for opening of the bypass valve BV, by line 29, as generally known.
Referring to FIG. 3, a block diagram illustrates the internal organization, in terms of functions, of the microprocessor unit MCP in its operation as a surge detector according to the present invention. The motor current is sensed from line 50 with respect to one phase of the motor (C), then rectified by an AC/DC converter CNV to provide on line 40 a signal I representative of the rms value of the motor current. From line 40, the motor current representative signal is applied to a sampler SMP controlled by a clock signal CLK1 derived on line 41 from a timer TMR, so as to so many times sample the motor current Imot, typically every 1/10 of a second. The sampling pulses are shown in FIG. 4 under (a), ten sampling pulses appearing upon each passing of a second. These samples are passed by line 42 through a delay line SW, typically designed to contain 40 samples from the oldest to the latest received, thus, for a duration of 4 seconds. A clock signal CLK2 (shown under (c) in FIG. 4), also derived from timer TMR, is applied by line 81 so as to command latching into two registers RG1 and RG2 of data from the sampling window, via lines 52 for register #1 (RG1), via lines 52' for register #2 (RG2). Every second the registers are reset by line 80 from the timer. Therefore, every second within delay line SW, the 40 samples are being circulated so as to lose 10 samples at one end and to gain 10 samples at the other end, this will appear for the data latched into the registers.
FIG. 4 shows under (b) three trains of 40 samples displaced by 1 second, or 10 samples, for successive instants (n-1), n, and (n+1). Register #1 is enabled by line 81 so that data be passed by line 54 to an averager AVG. The average motor current is Iav, obtained on line 55 of FIG. 3. After a time lag of .DELTA.T (typically of .DELTA.t=200 .mu.s) following the instant that line 81 enables register #1 and the averager AVG, by line 82 timer TMR enables register #2, with the result that a sampled value of the motor current Imot is derived on line 53.
Iav is shown under (d) in FIG. 4, the value being valid for a period of 1 second. The enabling signals of lines 81 and 82 are shown under (e) and (f), respectively. The time delay .DELTA.T, is required in order to allow the operation of the twin summeer TWS in generating the reference values applied on lines 60 and 62 ahead of the derivation of Imot on line 53, the three lines being applied to comparator COMP. Twin summer TWS receives the two value of +TH and -TH of FIG. 2, and this leads to the ordinates of thresholds THi and THl. The latter appear on lines 60 and 62, respectively. Comparator COMP relates Imot from line 53 to those two thresholds, and positive and negative polarity changes are generated on lines 66 and 67, one for the positive polarity being in relation to THi, the other for the negative polarity, thus, in relation to THl. A flip-flop FLP, reset by line 80, shifts from one state to the other when there is a change of polarity, and such occurrences are counted by a counter CNT responsive to line 68. A trip signal is generated on line 70 whenever the count N from line 68 reaches and exceeds a reference count N* applied on line 74. The counter is reset by line 73 upon each time interval of 1 second, thus, after each counting of Imot samples for a given Iav value under (d) in FIG. 4. FIG. 4 also shows under (g) how a count of N in such time interval may exceed the value N*. When this is the case, a logic 1 replaces the logic 0 on line 70 (under (h)) and a trip signal is applied to the compressor control unit CCU of FIG. 1.
FIG. 5 shows an analog implementation of the twin summer TWS first responsive to the Iav signal of lines 55 and 58 and to the +TH bias signal of line 57 with an output on line 62 of the signal -Iav.sup.31, and secondly, responsive to the Iav signal of lines 55 and 56, and (after inversion between line 57 and line 59) to the -TH bias signal of line 59 with an output Iav+ on line 60. The signal of line 55 is assumed to be negative (-Iav) in order to match the circuitry of FIGS. 7A and 7B hereinafter. FIG. 6 shows a hardware implementation of comparator COMP.
The circuits of FIGS. 5 and 6 are further illustrated in FIGS. 7A to 7E by an operational amplifier implementation. In FIG. 7A, for one half twin summer, signal -Iav of line 56 combined with a voltage -v.sub.hi representing +TH applied on line 59. Therefore, on line 60 at the output of the operational amplifier OA is the resulting signal (Iav+v.sub.hi)=Iav.sup.+. Similarly, for the other half twin summer, on line 58 of FIG. 7B is applied the signal -Iav, and on line 57 is applied the voltage +v.sub.lo representing -TH of FIG. 2. Therefore, on line 62 at the output of the operational amplifier OA is the resulting signal -(-Iav+v.sub.lo)=-Iav.sup.-. FIG. 7C shows the signal Imot of line 53 being inverted by device I so as to provide on line 53' the opposite value -Imot. Considering now FIGS. 7D and 7E which belong to comparator COMP of FIG. 6, the outputted values of FIGS. 7A, 7B and 7C become here inputs, one leading in FIG. 7D to -(Iav.sup.+ -Imot)=A as output on line 61 for the corresponding operational amplifier OA, and the other leading in FIG. 7E to +(Iav.sup.- -Imot)=B as output on line 63 for the corresponding operational amplifier OA. From the preceding, it appears that whenever on line 53' the sample -Imot is smaller than Iav.sup.+, that is Imot<(Iav+TH), like at point C on curve (C) of FIG. 2, the output A, on line 61, is negative. Therefore, on line 66 into the flip-flop FF1 of FIG. 6, the logic is low and the flip-flop is not triggered. This is because, for point C, Imot does not exceed the threshold THi. For point D, however, the quantity A is positive and line 61 becomes high, thereby triggering flip-flop FF1. In the same way, for point E on the curve of FIG. 2, line 53 has Imot smaller than -Iav-, that is Imot>(Iav-TH) and the output B, on line 63, is negative. This is because point E does not exceed the lower threshold. The logic of line 63 is low and flip-flop FF2 is not triggered. For point F, however, the logic of line 63 becomes high and flip-flop FF2 is triggered, because the threshold TH1 has been exceeded. Indeed, having triggered one flip-flop there will be a count of a change of polarity on line 68 of FIG. 3 only if such triggering happens from one flip-flop to the other.
From FIGS. 2, 3, 5, 6 and 7A to 7E it appears that, according to the present invention, motor current is sampled every 0.1 second and a profile (or array) SW is maintained containing the latest 40 samples of motor current. The array is examined once each second to determine if a surge condition exists. A surge condition is determined by first calculating a "sliding" average value Iav of motor current based on the latest 40 samples of motor current stored in SW. The individual samples are then sequentially examined from the oldest to the newest with respect to the sliding average. A "polarity" change is maintained by defining a "positive" change in polarity as a motor current value in excess of the "sliding" average by a predetermined threshold amount and a "negative" change being a motor current value below the "sliding" average by a predetermined threshold. The number of times N that the "polarity" changes over this four second interval (40 samples) is then recorded and compared to a reference value N*. If the number N exceeds or equals the value N*, then a surge condition exists.
FIG. 6 shows comparator CMP associated with the flip-flops FF1 and FF2. The counter CONT here runs with the pulses of line 68 applied thereto, and the digital count (6 bits for instance) is compared with the reference count N* of lines 74 (also 6 bits). When the limit count N* has been reached, lines 72 at the output will have all Zero, so that a ONE will appear on line 70 at the output of the NOR device, indicating a surge.
FIG. 8 is a flow chart explaining the software steps involved with the computer system MCP of FIG. 1. The steps are as follows:
At 101 the question is raised whether the compressor is compressing air. If No, by line 104 the system goes to Return. If Yes, by 102 the next step is at 103 to ascertain what the average motor current (Iav) is for the present period. Then, by 105 the system goes to 106 where is determined what are the upper and lower thresholds (THi and THl, which are assumed to be +TH and -TH in FIG. 2, and this is also what has been translated into voltages vhi and vlo for the operational amplifier inputs of FIGS. 7A and 7B). Thereafter, by line 107, the system goes to 108 where the polarity counter (CNT in FIGS. 3 and 6) is set to zero for the start. Moreover at such initial step, the polarity is assumed to be the "positive" one. This means that the polarity last encountered by the flip-flops is assumed to be for a point on curve (C) of FIG. 2 which was lastly above the upper threshold THi but is still above the lower threshold. This being done, by line 109 the system goes to block 110 where the question is: "whether for the oldest sample: Imot<(Iav-TH)?" This amounts to testing whether the polarity assumed to be "positive" is truly so. The question raised is whether the operative point is below the lower threshold rather than above. There are two possibilities. One is that the operative point for Imot remains above the lower threshold and there is no change of polarity, thus, the answer is NO on line 115. The other is that Imot has passed below the lower threshold, this means a change of polarity (from positive to negative): Therefore, by line 111 the system goes to block 112 where the polarity will be changed to the opposite one (the former assumption being wrongly to be positive). Such change of polarity is accounted for by counter CNT, at block 114, reached via line 113. As earlier stated, if at 110 the answer to Imot<(Iav-TH) is negated, by line 115 the system goes directly to line 116 without passing through a count at 114. Thereafter, by 120 the system goes to 121 where the next sample is taken into consideration. This will be done for all samples during the period (since there are 40 samples, it takes 4 second maximum to count at 114 the polarity changes and to see whether they reach and exceed the reference number N*), and it will be determined at 117 whether all samples have been tested. With each sample at 121, by line 122 the system goes to 123 where the test is whether Imot>(Iav+TH). This is, like for quantity A in FIG. 7D, assuming that the operative point on curve (C) exceeds the upper threshold THi and, therefore, that the polarity is positive. If it is so, by line 126 the next step is to check whether the last polarity recorded was also positive. If it is so, the conclusion is that there is no polarity change and by line 131, the system bypasses block 112, going directly to block 121 by lines 116 and 120. Otherwise, by line 130 (which means a NO) a change of polarity is effected at 112 followed via line 113 by a count at 114. If there is a NO on 124, from block 123 the test becomes at block 125 whether Imot<(Iav-TH). This is like under FIG. 7E with quantity B. If this is true, by line 127 there should be a negative polarity. If at 129 there is indeed already a negative polarity, no change of polarity (triggering) need to take place and the system by line 131 goes directly to lines 116, 120 onto block 121 for the next sample of Imot. If the actual polarity was positive, block 129 will say NO on line 130, thereby indicating a change of polarity (triggering). Such change is acknowledged and counted at 114. This goes until the last sample acknowledged at 117. Then, by line 118 the test becomes at 119 whether N* such changes have been taking place. If so, by line 132 (with a YES) and via line 134 (if at 133 a surge has not already been detected) the system goes to block 135 where the flag is set to indicate a "surge". Then, by lines 138 and 104 the system goes back to RETURN. The inlet valve and bypass valve control software logic at 141 is the one which by line 139 initiates at 101 the routine just considered. When a "set surge flag" logic has been received from line 138 and passed to RETURN by line 104, this is acknowledged by line 140 within block 141. The fact is established at 143 in another routine which by line 70 generates the command to the compressor control unit CCU of FIG. 1. Otherwise, that is, if there has been no surge flag set in the course of the time interval defined at 117 and as ascertained at 133, normal control will take place from block 143 to block CCU. The questions and answers between these two routines, namely between lines 139 and 140 is about 3 ms. If there has been no surge detection, there is a NO from block 119 with a return by line 136, and if there has been a surge already detected, there will be a return from block 133 via line 137. If the reference count N* has not been reached at 119, while incrementing counter CNT at 114, the decision is to return via line 104. If this is the first time that polarity changes have equated or exceeded N*, the command is to reset the surge detection in 3 second, to set the surge flag, and to set the other bits. This is done at 135 and the system returns by 104 for further testing of a possible surge occurrence.
The flow chart steps are performed by the computer control unit MCP illustrated by FIGS. 9 and 10A to 10E. The computer control unit includes a microcomputer MICRO (in FIG. 10A). FIG. 9 shows the derivation of the motor current representative signal, on line 40, after conversion by converter CNV of the AC current of the motor. The motor current signal is passed thereafter on line 240 toward line 240' of FIG. 10C at the input of the A/D converter section of the computer control unit. AD converter AD1 there generates on line 203 the ADIN data for treatment by the microcomputer. Two RAM devices are shown in FIG. 10B, one volatile, the other non-volatile. Line 203 of FIG. 10C goes to a latch LTC1 (FIG. 10B) which passes data, via data bus 202 (ADBUS), to a RAM device (RAM1) that is addressed by address bus 201 (ADDRESS BUS) from the MICRO (FIG. 10A). One of these two RAM devices is equivalent to registers RG1 and RG2 of FIG. 3. FIG. 10B also shows a latch LTC2 which is passing data from RAM1, as well as from another RAM device (RAM2), on line 204 (DAOUT) onto digital to analog converters DA1 and DA2 outputting analog signals on lines 219' and 229' (FIG. 10D) when line 70 (FIG. 1) commands tripping of the inlet valve and the bypass valve (IV and BV), respectively. The commands of lines 219' and 229' are passed on lines 219 and 229, respectively, which are similar to lines 19 and 29 of FIG. 1. The internal operation of the microcomputer MICRO handling data in digital form as received or as generated is according to the flow chart of FIG. 8. The solid state devices used for such software operation which have been labelled in the Figures are well understood as to their nature and mode of operation. Thus, the motor current analog value AIN carried on line 240 of FIG. 9 is brought via lines 240 and 240' onto the A/D converter AD1 to become its digital counter part ADIN of line 203 of FIG. 10C. From there ADIN is stored in latch LTC1 of FIG. 10B. The microcomputer MICRO of FIG. 10A receives or generates data in its various operations which are transmitted along ADBUS line 202 from or onto the ROM of FIG. 10A, the registers (RG1 and RG2) implemented as RAM's (RAM1 in this case) and/or the latches LTC1, LTC2 of FIG. 10B, the address bus (line 201) making the selection of the locations in the devices. A signal +vREF is shown on line 205 which is used for scaling of the value of the positioning signals for the inlet valve (IV) on line 219 (which matches line 19 of FIG. 1) and for the bypass valve (BV) on line 229 (which matches line 29 of FIG. 1).
For completion of the description of FIGS. 8, 9, and 10A to 10E, reference is made to the listings placed in Appendix thereafter. ##SPC1##
Claims
  • 1. In a compressor system including a compressor driven by an electric motor, a surge detection system comprising the following combination:
  • means for sensing the motor current to derive a current representative signal;
  • means responsive to said current representative signal for sampling over a sliding window a plurality of successive and consecutive samples thereof;
  • means for deriving one sample at a time as an instantaneous representation of the motor current magnitude;
  • means responsive to said plurality of samples for continuously deriving therefrom and through successive windows a signal representative of the average thereof;
  • comparator means responsive to a threshold signal and to said sample average for detecting a polarity signal representative of an excess in magnitude above and below said sample average;
  • means responsive to said polarity signal and to said one sample for detecting the occurrence of successive polarity changes in said polarity signal as an indication of an impeding surge; and
  • means for generating a surge indicative signal to be applied to the compressor system when said polarity changes have lasted during a predetermined time period.
  • 2. The surge detection system of claim 1 wherein said one sample is derived after a time delay following the derivation of said average representative signal.
  • 3. The surge detection system of claim 2 with said predetermined time period being equal to the duration of said sliding window.
  • 4. The surge detection system of claim 3 wherein said sliding window contains N successive and consecutive samples, said one sample being derived upon every shift of said sliding window from one number N of samples to a subsequent number N of samples therein, and wherein said average representative signal is derived in relation to each of said successive N samples.
US Referenced Citations (7)
Number Name Date Kind
3380650 Drummond et al. Apr 1968
3778695 Bauer, Jr. Dec 1973
4519748 Murphy et al. May 1985
4546618 Kountz et al. Oct 1985
4581900 Lowe et al. Apr 1986
4616483 Leonard Oct 1986
4686834 Haley et al. Aug 1987
Non-Patent Literature Citations (1)
Entry
Gregory K. McMillan, Centrifugal and Axial Compressor Control, pp. 57-72 (Section 6, Surge Control), Published by the Instrument Society of America, Aug. 1983.