Mud Pump Noise Cancellation System and Method

1. Field of the Invention

This invention relates to communication systems, and more particularly, to systems and methods for receiving and interpreting data signals being transmitted to the surface of the earth in a logging-while-drilling system.

2. Prior Art

Logging-while-drilling (LWD) or measurement-while-drilling (MWD) involves the transmission to the earth's surface of downhole measurements taken during drilling. The measurements are generally taken by instruments mounted within drill collars above the drill bit. Indications of the measurements must then be transmitted uphole to the earth's surface. Various schemes have been proposed for achieving transmission of measurement information to the earth's surface. For example, one proposed technique transmits logging measurements by means of insulated electrical conductors extending through the drill string. This scheme, however, requires adaptation of drill string pipes including expensive provision for electrical connections at the drill pipe couplings. Another proposed scheme employs an acoustic wave that is generated downhole and travels upward through the metal drill string; but the high levels of interfering noise in a drill string are a problem in this technique.

The most common scheme for transmitting measurement information utilizes the drilling fluid within the borehole as a transmission medium for acoustic waves modulated to represent the measurement information. Typically, drilling fluid or "mud" is circulated downward through the drill string and drill bit and upward through the annulus defined by the portion of the borehole surrounding the drill string. The drilling fluid not only removes drill cuttings and maintains a desired hydrostatic pressure in the borehole, but cools the drill bit. In a species of the technique referred to above, a downhole acoustic transmitter known as a rotary valve or "mud siren", repeatedly interrupts the flow of the drilling fluid, and this causes a varying pressure wave to be generated in the drilling fluid at a frequency that is proportional to the rate of interruption. Logging data is transmitted by modulating the acoustic carrier as a function of the downhole measured data.

One difficulty in transmitting measurement information via the drilling mud is that the signal received is typically of low amplitude relative to the noise generated by the mud pumps which circulate the mud, as the downhole signal is generated remote from the uphole sensors while the mud pumps are close to the uphole sensors. In particular, where the downhole tool generates a pressure wave that is phase modulated to encode binary data, such as is disclosed in U.S. Pat. No. 4,847,815 and assigned to the assignee hereof, and where the periodic noise sources are at frequencies which are at or near the frequency of the carrier wave (e.g. 12 Hz), difficulties arise.

Mud pumps are large positive displacement pumps which generate flow by moving a piston back and forth within a cylinder while simultaneously opening and closing intake and exhaust valves. A mud pump typically has three pistons attached to a common drive shaft. These pistons are one hundred and twenty degrees out of phase with one another to minimize pressure variations. Mud pump noise is caused primarily by pressure variations while forcing mud through the exhaust valve.

The fundamental frequency in Hertz of the noise generated by the mud pumps is equal to the strokes per minute of the mud pump divided by sixty. Due to the physical nature and operation of mud pumps, harmonics are also generated, leading to noise peaks of varying amplitude at all integer values of the fundamental frequency. The highest amplitudes generally occur at integer multiples of the number of pistons per pump times the fundamental frequency, e.g., 3F, 6F, 9F, etc. for a pump with three pistons.

Mud pumps are capable of generating very large noise peaks if pump pressure variations are not dampened. Thus, drilling rigs are typically provided with pulsation dampeners at the output of each pump. Despite the pulsation dampeners, however, the mud pump noise amplitude is typically much greater than the amplitude of the signal being received from the downhole acoustic transmitter. To reduce or eliminate the mud pump noise so that the downhole signal can be recovered, different techniques have been proposed, such as may be found in U.S. Pat. Nos. 3,488,629 to Claycomb, 3,555,504 to Fields, 3,716,830 to Garcia, 4,215,425 to Waggener, 4,215,427 to Waggener et al., 4,262,343 to Claycomb, 4,590,593 to Rodney, and 4,642,800 to Umeda. What is common to all of the techniques is that they try to eliminate the mud pump noise by adding the mud pump noise to an inverted version of itself. Most of the techniques utilize two sensors in the mud stream (usually two pressure sensors) and take the difference of signals in an attempt to cancel the mud pump noise without canceling the data signal. Various of the techniques require particular physical arrangements.

The Umeda U.S. Pat. No. 4,642,800 takes a slightly different approach to eliminating mud pump noise. Umeda teaches that an average pump signature may be found by obtaining the pump signatures in the presence of data over a certain number of pump cycles. The updated average pump signature is corrected by interpolation to match the current pump cycle length and is subtracted from the current pump signature to provide the residual data signal. While the technique disclosed in Umeda may be effective for particular arrangements, it has several drawbacks. First, because Umeda averages pump signatures which include data pulses, unless the effect of the data signal over any averaging period is zero (i.e. non-carrier frequency systems), the data signal which is to be recovered will tend to be undesirably subtracted from itself. Second, because Umeda uses only a single strobe per pump cycle, estimates (e.g. interpolations) are utilized which can introduce significant error. Third, Umeda does not disclose in detail how to treat a multi-pump system. In particular, if Umeda assumes that the pump signature for each pump of a multi-pump system is the same as it would be for a single pump system, large errors are introduced in attempting to cancel out the pump noise, as pumps which are working in multi-pump systems will have different signatures than they would it they were working in a single pump system. In addition, because estimates are required for each pump in the multi-pump system, additional error in the multi-pump system is introduced.

It is therefore an object of the invention to provide methods and systems for accurately recovering data signals introduced into drilling mud in the presence of mud pump noise.

It is another object of the invention to provide methods and systems for accurately recovering logging-while-drilling (LWD) or measurement-while-drilling (MWD) information which is modulated in drilling mud by correlating mud pump piston positions to a mud pressure signature in a calibration procedure.

It is a further object of the invention to provide methods and systems for accurately obtaining LWD or MWD information in multiple mud pump systems by allocating noise attributable to each mud pump and by tracking the mud pump piston position of each mud pump.

Another object of the invention is to provide method and systems for recovering LWD or MWD information transmitted through drilling mud by varying the pressure of the drilling mud regardless of the manner in which the information is coded.

In accord with the objects of the invention, methods for recovering a LWD or MWD data signal in the presence of mud pump noise are provided, and generally comprise calibrating the drilling mud pressure as a function of the mud pump piston position, and then tracking the piston position during transmission of the LWD or MWD data signal and using the calibration information to subtract out the mud pump noise. More particularly, calibration is accomplished in the absence of the LWD or MWD data signal to provide a correlation between mud pump piston position and the drilling mud pressure; i.e., the pressure signature as a function of mud pump piston position is obtained. Then, when the LWD or MWD data signal is being provided, the mud pump piston position is tracked such that the pressure due to the pump can be subtracted; i.e., by knowing the mud pump piston position, the pressure due to the mud pump is found and subtracted from the total received signal to provide the LWD or MWD signal. Where a plurality of mud pumps are used, calibration is accomplished by running the mud pumps together in the absence of the LWD or MWD data signal, and processing the received mud pressure signals in the Fourier domain to allocated respective portions of the mud pressure signals to respective mud pumps such that each mud pump is provided with a signature as a function of its own piston position. With the piston position of each mud pump being tracked, the sum of the mud pressure signals generated by the mud pumps based on their piston positions is subtracted from the total received signal to provide the LWD or MWD signal.

According to a preferred aspect of the invention, the calibration procedure is periodically repeated, e.g., each time additional pipe is added to the drill string, thereby eliminating the effects of depth and mud property variation on the system.

A better understanding of the invention, and additional objects and advantages of the invention will become evident to those skilled in the art upon reference to the detailed description taken in conjunction with the accompanying drawings.

FIG. 1 is a schematic diagram showing the present invention in use in conjunction with a downhole pressure pulse signaling device.

FIGS. 2a and 2b are schematic diagrams of exemplary mud pump piston position sensors utilized in practicing the invention.

FIG. 3 is a graph illustrating how mud pump piston position correlates to mud pump noise for a given set of operating conditions.

FIG. 4 is a flow chart of the mud pump calibration procedure for a system utilizing one mud pump.

FIG. 5 is a flow chart of the noise cancellation procedure for a system utilizing one mud pump.

FIGS 6a and 6b are respectively mud pump noise signals prior to and after noise cancellation in a one pump system.

FIG. 7 is a diagram showing the relationship between FIGS. 7a and 7b.

FIG. 7a and 7b together comprise a flow chart of the mud pump calibration procedure for a system utilizing multiple mud pumps.

FIGS. 8a, 8b, and 8c are respectively the total pump signal, and the signals from pump one and pump two in the multiple pump system calibrated according to FIGS 7a and 7b.

FIGS. 9a, 9b, and 9c are respectively the real parts of the signals of FIGS. 8a, 8b, and 8c as shown in the Fourier domain.

FIG. 10 is a flow chart of the noise cancellation procedure for a system utilizing multiple mud pumps.

FIGS. 11a and 11b are respectively drilling mud signals prior to and after noise cancellation in a multiple pump system.

Referring to FIG. 1, the operation of the present invention in a typical drilling arrangement is illustrated schematically. Drilling mud 10 is picked up from mud pit 11 by one or more mud pumps 12 which are typically of the piston reciprocating type. The mud 10 is circulated through mud line 13, down through the drill string 14, through the drill bit 15, and back to the surface of the formation via the annulus 16 between the drill stem and the wall of the well bore 29. Upon reaching the earth's surface 31, the mud is discharged through line 17 back into the mud pit 11 where cuttings of rock or other well debris are allowed to settle out before the mud is recirculated.

A downhole pressure pulse signaling device 18 is incorporated in the drill string for transmission of data signals derived during the drilling operation by the measurement instrument package 19. Signaling device 18 may be of the valve or variable orifice type which generates pressure pulses in the drilling fluid by varying the speed of flow. A preferred signaling device which generates sinusoidal signals is disclosed in U.S. Pat. No. 4,847,815 assigned to the assignee hereof. Data signals are encoded in a desired form by appropriate electronic means in the downhole tool. Arrows 21, 22, and 23 illustrate the path taken by the pressure pulses provided by the downhole signaling device 18 under typical well conditions. Pump 12 also produces pressure pulses in the mud line 13 and these are indicated by arrows, 24, 25, 26 and 26a which also illustrate the flow of the mud through the annulus 16.

In order for the downhole pressure pulse signals to be recoverable at the surface, some means must be provided to remove or substantially eliminate the portion of the mud pressure signal due to the mud pumps. Subsystem 30, including pressure transducer 32, mud pump piston position sensors 34, and computer or processor 36, comprises such a means.

The preferred pressure transducer 32 of subsystem 30 is a piezoelectric pressure transducer which provides an analog signal which is preferably bandpass filtered by a filter (not shown) or by the computer 36. The preferred mud pump piston position sensor 34 may either comprise an LVDT which utilizes a linear position transducer, or an RVDT which utilizes a rotary position transducer. The LVDT, as shown in FIG. 2a, has an arm 40a, a rod 42a, and a linear position transducer 44a with leads 46a. Arm 40a is coupled to one of the piston rods 47 of the mud pump 12 as well as to rod 42a of the LVDT. Rod 42a moves coaxially within the linear position transducer 44a, which provides a high precision digital indication of the location of piston 48 in the mud pump 12. The RVDT, as shown in FIG. 2b, has an arm 40b, a cable 42b, and an encoder or rotary position transducer 44b with a spring loaded sheave takeup reel 45b. The RVDT also includes leads 46b. Arm 40b of the RVDT of FIG. 2b is coupled to one of the piston rods 47 of the mud pump 12 as well as to the cable 42b of the RBDT. As arm 40b moves with the pump piston rod 47, the cable 42b is let out or reeled onto the takeup reel 45b takeup reel. The rotation of the takeup reel 45b provides a high precision digital indication of the location of piston 48 in the mud pump 12.

Testing has shown that the drilling mud pressure generated by the mud pump 12 is determined by the position of the mud pump piston for a given set of operating conditions. FIG. 3 illustrates how mud pump piston position correlates to mud pump noise. By coupling the linear position transducer 44a or rotary position transducer 44b to the piston rod 47 of the mud pump, a calibration can be performed that measures the pressure generated as a function of piston position.

The preferred calibration procedure for correlating mud pressure generated as a function of piston position for a single mud pump system is seen in FIG. 4. After the pump noise stabilizes in the system, and before the LWD and MWD tool turns on (i.e. before the data signal starts), the signals output by the position sensor 34 and the signals output by the pressure transducer 32 which are bandpass filtered at 39 are preferably recorded at 52 as related position and pressure arrays 55, 57 in the computer (e.g. in computer memory). Preferably, approximately eight seconds of data (e.g., five to ten pump cycles) are accumulated. Then, averages of the pressure as a function of position are calculated (thereby reducing random pressure variations) at 58 to produced a single position vs. pump noise calibration array 59. Indications of the average calibration array or the inverse thereof are stored and used for canceling mud pump noise as is hereinafter described.

The noise cancellation procedure according to the invention is set forth in FIG. 5. Upon the turning on of the downhole tool and the transmission of LWD or MWD data (hereinafter referred to simply as LWD data for sake of brevity), the position sensor 34 and pressure transducer 32 continue to provide indications of piston location and mud pressure; except that the piston position data is used in real time to determine the electrical signal (based on the calibration array 59) which must be subtracted from the composite LWD/noise signal to cancel the noise component of the signal and leave only the LWD signal. Thus, as shown in FIG. 5, the position sensor signal is sampled at 62 (i.e. based on the position sensor signal, the average calibration array is accessed and a corresponding pump noise is provided), and the corresponding pump noise pressure 64 is subtracted at 66 from the real time sensed pressure 32 which was bandpass filtered at 67 to eliminate high frequency components. The difference between the real time sensed pressure and the pump noise pressure provides an indication of the LWD data signal 68.

Test results of a real time sensed pressure pump noise signal are seen in FIG. 6a, where the amplitude of the signal as expressed in dB (in 10 dB increments) is plotted versus the frequency expressed in Hz (in 4 Hz increments). As seen in FIG. 6a, the noise signal includes several peaks having amplitudes between -10 dB and 0 dB, and even includes a peak having an amplitude exceeding 10 dB. The noise signal of FIG. 6a was then subjected to the noise cancellation procedure of FIG. 5. The noise signal remaining after mud pump noise cancellation is seen in FIG. 6b, and shows that the calibration and noise cancellation procedures reduced noise considerably. In fact, the largest remaining noise peak found at about 5 Hz, has an amplitude of approximately -15 dB, which is more than 25 dB less than the largest peak seen in FIG. 6a prior to noise cancellation.

Turing to FIGS. 7, 7a and 7b, a flow chart of the mud pump calibration procedure for a system utilizing two mud pumps is seen. After the pump noise stabilizes in the system, and before the LWD tool turns on (i.e. before the data signal starts), the signals output by each position sensor 34a, 34b and the signal output by the pressure transducer 32 and filtered at 39 by a bandpass filter which measures composite pump noise are recorded as related position arrays 55a, 55b and pressure array 57 in the computer (e.g. in computer memory). Preferably, approximately twelve seconds of data are accumulated in computer memory at 52; FIG. 8a showing an example of the analog pressure signal which is digitized and stored as part of the array. A fast Fourier transform (FFT) of the composite pump noise signal is then conducted at 70 by the computer. As a result of the FFT, the amplitude and phase of all frequencies contained in the composite mud pump noise signal is obtained at 70 (see FIG. 9a). Utilizing the operating speed of each pump which can be computed from the position sensor of each mud pump, the fundamental frequency and harmonics for each pump are calculated at 72. The, at 75, the amplitude and phase information for each fundamental and harmonic frequency are extracted from the FFT and assigned to its source (i.e. a particular one of the mud pumps) to provide results as seen in FIGS. 9b and 9c. Taking an inverse Fourier transform of the frequency spectra of FIGS. 9b and 9c at 76a and 76b, signals attributable to each of the pumps are obtained as seen in FIGS. 8b and 8c. As indicated in FIG. 7b at 58a and 58b, the position of each mud pump position sensor is related to the mud pressure generated by the respective mud pump, and an average of the pressure as a function of position is calculated for each mud pump to produce two position vs. pump noise calibration arrays 59a and 59b. Indications of the average calibration arrays are stored in computer memory and used for canceling mud pump noise as is described above with reference to FIG. 10.

Referring now to FIG. 10, the noise cancellation procedure for a system using multiple mud pumps is seen. Upon the turning on of the downhole tool and the transmission of LWD data, the position sensors 34a and 34b and pressure transducer 32 continue to provide indications of piston location and mud pressure; except that the piston position data is used in real time to determine the electrical signal (based on the calibration arrays 59a and 59b) which must be subtracted from the composite LWD/noise signal to cancel the noise component of the signal and leave only the LWD signal. Thus, as shown in FIG. 10, the position sensor signals are sampled at 62a and 62b (i.e. based on the position sensor signals, the average calibration arrays 59a and 59b are accessed and corresponding pump noises are provided), and the corresponding pump noise pressures 64a and 64b are subtracted at 66 from the real time sensed pressure 32 which was bandpass filtered at 67 to eliminate high frequency components. The difference between the real time sensed pressure and the pump noise pressures provides an indication of the LWD data signal 68. That signal is then decoded according to techniques known in the art which are not part of the present invention.

Test results of a real time sensed pressure containing pump noise for two mud pumps is seen in FIG. 11a where amplitude is plotted against frequency. As seen in FIG. 11a, numerous noise peaks having amplitudes of -20 dB or higher are seen, with the largest peak of about -5 dB at 5 Hz. The pressure signal obtained after utilizing the calibration and noise cancellation steps of FIGS. 7 and 10 in order to substantially cancel mud pump noise from the signal of FIG. 10a is seen in FIG. 10b. As seen in FIG. 10b, the remaining noise is substantially reduced relative to the noise of FIG. 10a, with the largest peak of about -18 dB occurring at approximately 18 Hz.

There have been described and illustrated herein methods and apparatus for canceling mud pump noise in order to recover a logging while drilling signal. While particular embodiments of the invention have been described it is not intended that the the invention be limited exactly thereto, as it is intended that the invention be as broad in scope as the art will allow. Thus, while particular pressure transducers, position sensors, pump-types, computers, FFT programs, and the like have been disclosed, it will be appreciated that other equipment and programs can be utilized effectively. Similarly, while certain preferred data gathering time periods were disclosed prior to running the LWD or MWD tool, it will be appreciated that other time frames could be utilized. Also, while the invention was described with reference to LWD and MWD procedures, it will be appreciated that the terms LWD and MWD are intended to include any other data signaling procedure where data is transmitted in drilling mud in the presence of mud pump noise. Further, while the invention was disclosed with reference to systems utilizing one or two mud pumps, it will be appreciated that the teachings equally apply to systems utilizing additional mud pumps. All that is required is that the pressure signature of each mud pump relative to its piston position be obtained via transforming the total signal into the Fourier domain, dividing the Fourier response among the various mud pumps based on their fundamental and harmonic frequencies, and converting the responses back into respective pressure signatures. It will be understood, of course, that where two mud pumps are working in unison (i.e. at the same frequency), their signatures can be treated together. Therefore, it will be apparent to those skilled in the art that other changes and modifications may be made to the invention as described in the specification without departing from the spirit and scope of the invention as so claimed.