Channel detection system and method

Abstract

Systems, methods, and computer-readable media are provided for detecting a channel behind casing and generating an image that represents the channel. An example method can include receiving data samples associated with at least one casing, each data sample representing channel information behind a representative casing, training a machine learning model using the data samples to generate a mapping between waveform information in each of the data samples and the channel information behind the representative casing, receiving acoustic data from a tool, the acoustic data representing a particular casing, and using the machine learning model to analyze the acoustic data from the tool and determine one of a presence and an absence of a channel behind the particular casing at a plurality of depths.

1 . A method comprising:
 receiving, by at least one processor, data samples associated with at least one casing, each data sample representing channel information behind a representative casing;   training, by the at least one processor, a machine learning model using the data samples to generate a mapping between integral amplitudes calculated from waveform information in each of the data samples and the channel information behind the representative casing, wherein at least one integral amplitude is calculated by rectifying the waveform information, determining a peaks-based envelope of the rectified waveform information, using the peaks-based envelope to modulate a sinusoid signal, and integrating a period of the sinusoid signal to obtain the integral amplitude;   receiving, by the at least one processor, acoustic data from a tool, the acoustic data representing a particular casing;   calculating a plurality of integral amplitudes based on the acoustic data from the tool; and   using, by the at least one processor, the machine learning model to analyze the calculated plurality of integral amplitudes from the acoustic data and determine one of a presence and an absence of a channel behind the particular casing at a plurality of depths.   
     
     
         2 . The method of  claim 1 , further comprising generating an azimuthal cement bond depth channel image that represents the presence and the absence of the channel behind the particular casing at the plurality of depths. 
     
     
         3 . The method of  claim 2 , wherein the channel image is a binary two-dimensional image. 
     
     
         4 . The method of  claim 3 , wherein the binary two-dimensional image represents a size and an azimuthal location of the channel behind the particular casing at the plurality of depths. 
     
     
         5 . The method of  claim 1 , wherein the tool comprises at least one array of receivers azimuthally arranged along a circumference of the tool. 
     
     
         6 . The method of  claim 5 , wherein the tool comprises a monopole transmitter that transmits waves having a frequency less than ultrasound frequencies. 
     
     
         7 . The method of  claim 5 , wherein the tool comprises a monopole transmitter, and wherein a ring of receivers is one of three feet and five feet from the monopole transmitter. 
     
     
         8 . The method of  claim 1 , wherein the machine learning model is a regression model based on random forest. 
     
     
         9 . A system comprising:
 an acoustic tool comprising at least one sensor;   at least one processor; and   at least one computer-readable storage medium having stored therein instructions, which when executed by the at least one processor cause the system to:   receive data samples associated with at least one casing, each data sample representing channel information behind a representative casing;   train a machine learning model using the data samples to generate a mapping between integral amplitudes calculated from waveform information in each of the data samples and the channel information behind the representative casing, wherein at least one integral amplitude is calculated by rectifying the waveform information, determining a peaks-based envelope of the rectified waveform information, using the peaks-based envelope to modulate a sinusoid signal, and integrating a period of the sinusoid signal to obtain the integral amplitude;   receive acoustic data from the tool, the acoustic data representing a particular casing;   calculate a plurality of integral amplitudes based on the acoustic data from the tool; and   use the machine learning model to analyze the calculated plurality of integral amplitudes from the acoustic data from the tool and determine one of a presence and an absence of a channel behind the particular casing at a plurality of depths.   
     
     
         10 . The system of  claim 9 , the at least one processor further to generate an azimuthal cement bond depth channel image that represents the presence and the absence of the channel behind the particular casing at the plurality of depths. 
     
     
         11 . The system of  claim 10 , wherein the channel image is a binary two-dimensional image. 
     
     
         12 . The system of  claim 11 , wherein the binary two-dimensional image represents a size and an azimuthal location of the channel behind the particular casing at the plurality of depths. 
     
     
         13 . The system of  claim 9 , wherein the acoustic tool comprises at least one array of receivers azimuthally arranged along a circumference of the tool. 
     
     
         14 . The system of  claim 13 , wherein the acoustic tool comprises a monopole transmitter that transmits waves having a frequency less than ultrasound frequencies. 
     
     
         15 . The system of  claim 13 , wherein the acoustic tool comprises a monopole transmitter, and wherein a ring of receivers is one of three feet and five feet from the monopole transmitter. 
     
     
         16 . The system of  claim 9 , wherein the machine learning model is a regression model based on random forest. 
     
     
         17 . A non-transitory computer-readable medium having instructions stored thereon that, when executed by at least one processor, cause the at least one processor to perform operations comprising:
 receiving data samples associated with at least one casing, each data sample representing channel information behind a representative casing;   training a machine learning model using the data samples to generate a mapping between integral amplitudes calculated from waveform information in each of the data samples and the channel information behind the representative casing, wherein at least one integral amplitude is calculated by rectifying the waveform information, determining a peaks-based envelope of the rectified waveform information, using the peaks-based envelope to modulate a sinusoid signal, and integrating a period of the sinusoid signal to obtain the integral amplitude;   receiving acoustic data from a tool, the acoustic data representing a particular casing;   calculating a plurality of integral amplitudes based on the acoustic data from the tool; and   using the machine learning model to analyze the calculated plurality of integral amplitudes from the acoustic data and determine one of a presence and an absence of a channel behind the particular casing at a plurality of depths.   
     
     
         18 . The non-transitory computer-readable medium of  claim 17 , the operations further comprising generating an azimuthal cement bond depth channel image that represents the presence and the absence of the channel behind the particular casing at the plurality of depths. 
     
     
         19 . The non-transitory computer-readable medium of  claim 18 , wherein the channel image is a binary two-dimensional image. 
     
     
         20 . The non-transitory computer-readable medium of  claim 19 , wherein the binary two-dimensional image represents a size and an azimuthal location of the channel behind the particular casing at the plurality of depths.