In another study, Kostov (1990) obtained an estimate of the drill-bit source signature from the recorded seismic data, without placing an accelerometer on the drill string. A 1-D array containing up to 90 geophones was used, and the drill-bit signal was separated from other noise sources by virtue of its spatial coherency. While Kostov was able to detect direct arrivals from the drill bit, the strongest noise sources seen were surface sources located off the line, which could not be suppressed with the 1-D array. These noise sources complicated processing considerably.
I have previously studied ambient seismic noise in a variety of environments (Cole and Vanyan, 1990; Nichols et al, 1989) using 2-D arrays with hundreds of channels. Such arrays offer improved ability to determine the arrival direction of incident energy and to suppress unwanted noise sources. In November 1990, I used a 240 channel 2-D array to record data during the drilling of a well in a producing oilfield in Wyoming. A diagram of the array is shown in Figure . The 240 channels were selected from a larger array of 480 channels that was previously used to study ambient noise in the same oilfield (Cole and Vanyan, 1990). The array consisted of 11 lines of geophones, with an average inline spacing of 100 feet and an average crossline spacing of 150 feet. On one line in the middle of the array a smaller spacing of 50 feet was used.
240 channel array used for drill-bit study. Dots are geophone locations, asterisk at top is location of drilling rig. Geophones were cemented in place at a depth of 30 feet.
Data were recorded over a period of three days, during which the drill bit covered a range of depths from approximately 800 to 1400 feet. Operational constraints limited us to this fairly narrow range. A total of 64 records, each 99 seconds long and with a sampling interval of two milliseconds, were taken during that time. During most of the recording time, many strong surface noise sources were active, including drilling-related equipment and 10 to 12 producing pumps located throughout the array. We were fortunate to have been able to turn off these pumps at one point during the recording. Eight records were taken during this quiet period. As will be seen later in the paper, first results from this quiet period are much superior to the results obtained from other records.
In this paper, I begin the study of the data by searching for sources of energy in 3-D. I construct a cube of possible source locations and then, given an estimate of the velocity model, stack the data (or compute semblance) along the moveout trajectory corresponding to each source location. To begin, I use a constant velocity. When applied to data from the ``quiet'' period where pumps located within the array were turned off, this technique reveals what appear to be direct arrivals from the drill bit. Two different records from that time period give a consistent picture. In contrast, data from outside the quiet period present a noisy background against which it is not possible, at least with this simple method, to detect the drill bit signal.
While stacking with a constant velocity appears to do a good job, it would be useful to obtain velocity information directly from the data, rather than arriving at a good stacking velocity by trial and error. A velocity analysis can be performed at the drill rig location; I scan for hyperbolic events whose tops lie beneath the surface location of the drill rig. The parameters are stacking velocity and the depth of the source. This analysis also reveals the drill bit signal, when applied to data from the quiet recording period, and concludes that the constant velocity value that seemed to position the drill bit energy well in 3-D was in fact a good choice.
I also perform this velocity analysis at the locations of other sources, including pumps and a steam injection well, to see if these sources can also be detected. The results are less promising.
Finally, having found via constant velocity stacks what I believe to be energy from the drill bit, I stack the data along the correct moveout trajectory to come up with an estimate of the drill bit source signature, analogous to the source signal estimate obtained by Rector et al. from an accelerometer placed on the drill string. Crosscorrelation of the data with this estimate enhances drill bit energy, preserves the moveout present in the data, and compresses the continuous-time drill-bit signal to a narrow window centered around the zero lag of crosscorrelation.