Neural Network for Obtaining True Borehole Sigma and True Formation Sigma
By using a combination of pre-programmed neural networks and dual-function detectors, the problem of difficulty in accurately measuring and compensating for near-wellbore environmental impact in the wellbore and formation sigma measurement in the prior art is solved, and the accurate measurement of wellbore and formation sigma is achieved. and automatic correction, improving measurement accuracy and efficiency.
Patent Information
- Application Number
- CN202280062275.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Priority Date
- 2021-09-16
- Filing Date
- 2022-09-16
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2042-09-16
AI Technical Summary
The prior art is difficult to simultaneously accurately measure and compensate for the near-wellbore environmental impact in wellbore and formation sigma measurements, resulting in limited accuracy and accuracy of measurement results.
Using computer program products with pre-programmed neural networks and trained neural networks, combining pulse neutron sources and multiple neutron and gamma ray dual-function radiation detectors, differentiating neutrons and gammas through pulse shape screening technology Rays, obtain the true Sigma values of the wellbore and formations, and automatically correct the near-wellbore environmental effects.
Simultaneous accurate measurement of wellbore and formation Sigma is achieved, reducing detector types and numbers, improving measurement accuracy and efficiency, and automatically compensates for near-wellbore environmental impacts, providing more accurate results.
Smart Images

Figure CN118265927B_ABST
Abstract
Description
Technical Field
[0001] The present disclosure generally relates to systems and methods for wireline or logging while drilling that use a pulsed neutron source and multiple dual-function neutron and gamma-ray radiation detectors, and utilize a trained neural network to predict and compare borehole constants and formation constants and borehole sigma and formation constant sigma, wherein the dual-function neutron and gamma-ray radiation detectors are preprogrammed to distinguish between neutrons and gamma rays by using pulse shape discrimination techniques. Background Art
[0002] 1. Overview
[0003] Neutron logging began in the early 1940s as an innovative method for measuring the relevant stratigraphic and geological structures of the subsurface. Neutron logging initially used chemical neutron sources, particularly radium-beryllium neutron sources, in combination with a single radiation detector, primarily to count gamma rays emitted when hydrogen and chlorine capture certain thermal neutrons. An example of such a tool was first introduced by Baker Atlas, a predecessor to the De Las Atlas Company, in the 1960s. This tool has proven very successful in determining water saturation in saline reservoirs. However, it was soon discovered that the wellbore has a significant effect on such tools, as hydrogen has the greatest effect on neutron transport. Therefore, a compensating tool must be used in combination, which uses an isotope source (usually americium-beryllium) and two neutron radiation detectors to measure the size of the neutron cloud by characterizing the neutron drop between the two detectors. Using neutrons has proven to be a better solution, as neutrons penetrate farther than gamma rays, minimizing the complexity of the above tools. Furthermore, the tool requires little alignment and does not press against the wellbore wall.
[0004] Later, further calculations were needed in order to measure the macroscopic thermal neutron capture cross section (sigma) of the formation surrounding the wellbore, and therefore the decay of the thermal neutron group in said formation was measured over a fixed time. Therefore, Dryas Atlas introduced a method in which a burst of high energy neutrons was measured and then divided into two equal groups based on the emission of such thermal neutrons, which were then used to calculate the rate of change over a selected time interval. The basic theory behind all this involves the emission of neutrons from a source, which interact with the surrounding rock formation in a variety of ways, depending on the local geology. In particular, water and hydrocarbons slow down (thermalize) the neutrons, which then scatter inelastically from certain elements, emitting gamma rays, and finally certain elements capture neutrons immediately or after the neutrons have been slowed down, emitting gamma rays and reducing the neutron flux.
[0005] Since then, other techniques have been developed to obtain more accurate measurements, including corrections for thermal neutron diffusion effects and near-wellbore environmental effects such as wellbore dimensions, casing size, wellbore fluids, cement thickness, etc., allowing pulsed neutron tools to operate in a variety of field applications. However, these environmental corrections were often (and still are) quite complex and not widely available for complete and perfect corrections in all field situations.
[0006] In particular, the two most common methods for pulse generator tools are wireline measurement and logging while drilling (LWD). In wireline measurement, the entire drill pipe is removed and the measurement tool string is inserted into the well, allowing bottom-up measurements. In contrast, in the LWD process, the measurement tool is contained in the drill pipe and the collected data is transmitted to the surface via telemetry.
[0007] Fast forward a few decades and now a typical modern Sigma logging tool consists of a deuterium-tritium (DT) pulsed neutron generator and two single-source detectors (i.e. thermal neutrons or gamma rays, but not both). The DT neutron generator seems to be chosen as the default source because the cross section of the DT nuclear fusion reaction is two orders of magnitude higher than that of the deuterium-deuterium (DD) reaction. Therefore, the neutron yield of a DT neutron generator is two orders of magnitude higher than that of a DD neutron generator for the same D ion current. As a result, the detector count rate is much higher, which leads to much smaller statistical uncertainties when using a DT neutron generator in the tool. Due to its computational shortcomings, the detector can only be programmed to detect the time decay of thermal neutrons by using two neutron detectors, or to detect the time decay of captured gamma rays. In fact, the latter requires the use of two dedicated gamma-ray detectors. For example, gas detectors such as He-3 tubes are used to detect neutrons, and various scintillation detectors such as NaI, CsI, GSO, BGO, LaBr3 scintillators and photomultiplier tubes are used to detect gamma rays. Likewise, these detectors can detect only neutrons or gamma rays, not both at the same time.
[0008] Nonetheless, the advantages of using neutron generators include: (a) outputting at least three times as many neutrons per second as the Am-Be neutron source it initially replaced; (b) causing less radiation dose to workers; (c) being easier to store because no external radiation is produced when turned off; (d) being less difficult to permit; (e) not requiring leak testing; (f) being easier to transport; and (g) being able to make more types of measurements because they can be timed. In general, the goal of oil and gas exploration is to identify the potential locations of hydrocarbons beneath the surface.
[0009] 2. Measurement and detection
[0010] Determining formation characteristics to obtain information indicating hydrocarbon content is a very difficult task. In a typical method, high-energy neutrons penetrate the subsurface formation and the energy of the neutrons is reduced as they interact with the subsurface formation. At low energy (thermal energy), the neutrons are captured by constituent nuclei. After being captured, the energetic nuclei release gamma rays, which can be detected by logging tools. Since the energy of gamma rays is very high and the average non-destructive path is relatively long, the amplitude and decay time characteristics of the detected gamma rays represent the volume average characteristics of the composition of the wellbore and surrounding formations. However, technicians in this field are concerned with the following two major formation characteristics: (a) the macroscopic thermal capture cross section (formation sigma, in units of cu) of the formation; and (b) the formation porosity (in units of pu). The macroscopic thermal neutron capture cross section (sigma) log is very important for formation evaluation. Therefore, the borehole sigma log can be used to examine near-wellbore information such as borehole salinity and its changes within the wellbore, changes in casing strings, etc. In contrast, the Formation Sigma tool is used to assess the water / oil / gas saturation, water / oil / gas contact locations, and remaining oil and gas reserves of downhole reservoirs.
[0011] A proposed method is to measure and fit the time decay curve of thermal neutrons or neutron capture gamma rays after a neutron pulse (or several neutron pulses) by using two exponential time decay functions, and obtain a time decay constant in an early time window that is more sensitive to materials in the borehole region, and another time decay constant in a late time window that is more sensitive to materials in the formation region. Using these two time decay constants, the borehole sigma and the formation sigma can be obtained. Using two detectors helps to correct for thermal neutron diffusion effects and obtain other measurements such as formation porosity.
[0012] A method is also proposed to determine the formation sigma value from the decay curve of the gamma ray count rate sensed by the processing detector, wherein in essence, the decay curve immediately after the gamma ray source pulse is used to eliminate the count effect caused by the interaction within the wellbore, so that the count caused by thermal capture in the formation dominates. Then, the decay time constant of the gamma counts generated by thermal neutron capture is used to calculate the formation sigma, which can adopt one of various known neutron-matter interaction modeling techniques, such as solving the neutron diffusion equation using the formation sigma as a parameter.
[0013] Recently, two scintillators have been developed: Cs2LiYCl6 (CLYC) and Cs2LiLaBr6 (CLLB) crystals, which are sensitive to both neutrons and gamma rays. See J. Glodo, R. Hawrami, KS Shah, "Development of Cs2LiYCl6 scintillator", Journal of Crystal Growth, Volume 379, 15 September 2013, Pages 73-78, and Kan Yang, Peter R. Meng, Julien Lejay, Vladimir Ouspenski, "Improving the Neutron and Gamma-ray Response of Cs2LiLaBr6: Ce 3+ ”,2013IEEE Nuclear Science Symposium & Medical Imaging Conference,Oct.27-Nov.2,2013,COEX,Seoul,Korea. Basically, these new scintillators are coupled with scintillation light sensitive elements such as photomultiplier tubes (PMTs) and have been shown to effectively detect neutrons and gamma rays.
[0014] The development of Cs2LiYCI6 (CLYC) and Cs2LiLaBr6 (CLLB) crystals has made it possible to detect neutrons and gamma rays simultaneously using one scintillation detector. This allows those skilled in the art to develop nuclear tools that can perform multiple measurements simultaneously. In addition, since different materials (wellbore fluids, casing, cement, etc.) attenuate thermal neutrons and gamma rays differently, the effects of near-wellbore materials on thermal neutron time decay and captured gamma ray time decay are also very different. Therefore, if multiple dual-function detectors are used to measure thermal neutron time decay and captured gamma ray time decay at different distances from the emission source, it is possible to automatically compensate for all environmental effects.
[0015] In recent years, there have also been significant improvements in the technology of neutron pulsers. Initially, most neutron pulsers were composed primarily of deuterium-tritium (DT), which, as is well known, requires special control. However, those skilled in the art began to consider the use of deuterium-deuterium (DD) in well logging situations, but the calculation methods lagged behind. This is mainly due to the low neutron yield of DD generators, which results in a significant reduction in the statistical uncertainty of sigma measurements. Therefore, the application of DD is mainly limited to experimental properties. However, since the neutron energy produced by DD generators is much lower than that produced by DT neutron generators (2.45 MeV vs. 14.1 MeV), the gamma rays produced by inelastic scattering of fast neutrons are greatly reduced, because 2.45 MeV neutrons do not produce as much inelastic scattering of fast neutrons on some elements as 14.1 MeV neutrons. Therefore, the background gamma rays produced by inelastic scattering of fast neutrons have a much reduced effect on the time decay measurement of captured gamma rays. Therefore, the background of neutron inelastic gamma rays has a particularly significant effect on the time decay measurement of early time windows.
[0016] In view of the above, and taking into account current methods and systems for obtaining borehole sigma and formation sigma measurements, as well as new advances in detector technology, a more advanced nuclear well logging computational approach is needed that has the potential to accurately make multiple measurements and automatically compensate for all environmental effects.
[0017] 3. Principles of formation sigma measurement
[0018] Formation sigma is usually measured by irradiating the formation with short fast neutron pulses from a DT neutron generator. After a pulse of successive inelastic and elastic scattering of fast neutrons with element nuclei in the surrounding medium, the thermal energy of the fast neutrons decreases rapidly. Then, thermal neutrons are captured by formation elements and emit gamma rays. The capture rate of thermal neutrons and the emission rate of captured gamma rays after the pulse depend on the macroscopic cross section of the formation for thermal neutrons, which in an infinite medium can be expressed as an exponential decay with time.
[0019] When the neutron pulse frequency emitted by the neutron generator is low (for example, 1 kHz) and the neutron duty time is short (for example, 30 μs), the longer interval time between neutron pulses (for example, from 30 μs to 1000 μs) allows technicians in this field to use the detector to obtain the thermal neutron time decay curve by directly detecting the thermal neutron time decay or indirectly detecting the captured gamma ray time decay.
[0020] In a typical logging situation, the materials in the borehole region (such as borehole fluid in an open hole or casing and cement in a cased hole) are usually very different from the formation. Therefore, the total time decay is not a singular exponential decay. If the thermal neutron diffusion effect is ignored, the thermal neutron flux rate near the borehole after the high-energy neutron burst and the detector count rate can be described as the sum of two exponential decays (one is the borehole decay that dominates in the early time window, and the other is the formation decay that dominates in the later time window) and a background decay. This can be expressed by Equation 1:
[0021]
[0022] Where C(t) is the count rate measured by the detector at time t, the unit of time is microseconds (μs), the unit of count rate is counts per second (cps), and the reference time is the start time of the neutron pulse; A b and A f is a constant, which can be interpreted as the initial amplitude of the borehole component and the formation component, in cps; ∑ b and∑ f are the macroscopic thermal neutron capture cross sections (sigma) of the wellbore and formation, respectively, with the unit of 1 / cm; v is the thermal neutron velocity, with the unit of cm / μs; B represents the background, with the unit of cps.
[0023] It is well known that when the neutron energy is less than 1eV, the absorption cross section increases in the proportion of 1 / v if the neutron is in equilibrium with the surrounding medium. This phenomenon is due to the longer interaction time of the nuclear force between the target nucleus and the neutron. The upper limit of the neutron energy of the 1 / v law depends on the weight of the element: the heavier the element, the lower the threshold. Therefore, for almost all elements, when the neutron energy is less than 1eV, the equation tends to the following:
[0024] ∑·v=k(constant) (2)
[0025] or
[0026]
[0027] Where k is the decay constant, in units of 1 / μs; τ is the average lifetime of a neutron in the material, in units of μs. Therefore, this allows equation (1) to be rewritten as follows:
[0028]
[0029] Among them, τ b and τ f are the average neutron lifetimes of materials in the borehole region and the formation region, respectively.
[0030] In the oil and gas industry, τ has long been referred to in the prior art literature as the average neutron lifetime and the time decay constant. Therefore, the average neutron lifetime and the time decay constant are used interchangeably. In fact, since equations (2) and (3) apply to all neutrons with energies in the 1 / v range, and the upper limit of neutrons in the 1 / v range for almost all elements is 1 eV, the average neutron lifetime is constant not only for thermal neutrons, but also for the surface neutrons in the measurement.
[0031] In addition, in the oil and gas industry, the unit of the macroscopic thermal neutron capture cross section Σ is the capture unit (cu), which is one thousandth of 1 / cm. Therefore, as long as the neutron decay time constant is measured, the thermal neutron capture cross section can be calculated using equation (5).
[0032]
[0033] Since the thermal neutron decay constant itself depends only on the density of the material, the formation sigma measurement is always accurate for a material of a specific density regardless of the temperature. As the temperature increases, the neutron velocity increases, but sigma decreases proportionally. Therefore, the product of sigma and the neutron velocity (the thermal neutron decay constant) does not change. Furthermore, according to equation (4), which uses a fixed thermal neutron velocity to obtain formation sigma, sigma should not change. When evaluating formation sigma, the only thing that a person of ordinary skill in the art needs to consider is the change in material density (such as formation matrix density, pore density) caused by temperature changes.
[0034] 4. Sigma measurement principle
[0035] Early pulsed neutron tools used a scintillation counter to measure the average thermal neutron lifetime. Formation sigma was then obtained using equation (5). However, modern pulsed neutron logging tools use two detectors when measuring sigma. The use of two detectors not only allows the development of correction algorithms for the effects of neutron diffusion and borehole environment on formation sigma measurements, but also allows other measurements to be made, such as obtaining ratio-based formation porosity.
[0036] There are two main types of pulsed neutron tools used for sigma measurements, where different types of radiation detectors are used to detect neutrons or gamma rays. One is a pulsed neutron-neutron (PNN) tool, where a He-3 proportional counter is used to directly detect the time decay of thermal neutrons. The other is a pulsed neutron capture (PNC) tool, where various scintillation detectors are used to detect the time decay of captured gamma rays, thereby indirectly obtaining the time decay of thermal neutrons. Both PNN and PNC tools have their own characteristics, making them attractive in specific applications.
[0037] In particular, PNN tools are less affected by background radiation, such as that of short-lived gamma rays produced by neutron activation, than PNC tools. This is why neutron measurements do not require quenching of the background period of the emitting source. Since thermal neutrons are more easily shielded than captured gamma rays, shielding and / or window designs in the tool enhance the formation component, making the tool sensitive only to thermal neutrons coming from specific angles. With this configuration, the detector response to neutrons in the borehole is minimized, which is particularly useful in wireline logging, where the borehole is usually much larger than the diameter of the tool. In addition, PNN tools have a slightly higher temperature rating than traditional scintillation detectors due to the use of gaseous H-3 detectors, which have a slightly higher temperature rating than traditional scintillation detectors. In addition, PNN tools are more sensitive to porosity changes when the source-detector distance is the same as that of PNC tools. Therefore, to obtain the same tool sensitivity to formation porosity, the source-detector distance must be shortened, thereby increasing the vertical resolution of the measurement. Together with the shallower exploration depth, this makes them more attractive for thin-bed stratigraphic determination.
[0038] On the other hand, PNC tools have a higher detector count rate, so the statistical uncertainty is better. In addition, the penetration range of captured gamma rays is longer than that of thermal neutrons. Therefore, PNC tools can survey deeper and measure more about the formation than the near-wellbore environment. Various technologies have been developed in the past, such as high-temperature scintillator crystals and high-temperature PMTs, background subtraction techniques, etc., making them very popular in field applications.
[0039] 5. Tools: Neutron generators, detectors, electronic systems and non-transitory computer readable memory devices for data processing
[0040] As mentioned above, a neutron generator is a neutron source device comprising a compact linear particle accelerator that produces neutrons by fusing isotopes of hydrogen together. In these devices, fusion reactions are carried out by accelerating deuterium, tritium, or a mixture of these two isotopes to a metal hydride target that also contains deuterium, tritium, or a mixture of these two isotopes. Deuterium nuclear fusion (D+D) produces a He-3 nucleus and a neutron with a kinetic energy of about 2.5 MeV. The fusion of deuterium and tritium nuclei (D+T) produces a He-4 nucleus and a neutron with a kinetic energy of about 14.1 MeV. See Reijonen, J. "Compact Neutron Generators for Medical, Homeland Security, and Planetary Exploration". Proceedings of 2005 Particle Accelerator Conference, Knoxville, Tennessee: 49-53.
[0041] When performing formation sigma measurements in the oil and gas industry, pulsed neutron tools typically use DT-type neutron generators because their neutron output is two orders of magnitude higher than that of DD-type neutron generators. However, if the neutron output is as high as that of current DT-type neutron generators, DD-type neutron generators would be preferred because DD reactions produce neutrons with much lower energy than DT reactions. As a result, it results in fewer gamma rays from inelastic scattering and less background from short-lived gamma-ray radiation from fast neutron activation, which affects the data processing of captured gamma-ray decays for detectors in PNC tools, especially in the early stages of decay.
[0042] Over the years, the design and manufacture of pulsed neutron generators have been continuously improved, and great progress has been made, which has greatly improved the neutron yield, reliability and service life of the generator. See D. Rose, T. Zhou et al, "An Innovative Slim Pulsed Neutron Logging Tool", SPWLA 56th Annual Logging Symposium, Long Beach, California, USA, July 18, 2015, and International Atomic Energy Agency, "Signal Processing and Electronics for Nuclear Spectrometry", Proceedings of a technical meeting, Vienna, 20-23 November 2007. In addition, the neutron duty cycle time and neutron pulse frequency schematic diagram can be adjusted and optimized through simulation and experiment to perform formation sigma measurement.
[0043] With regard to the radiation detectors used in these tools, one of ordinary skill in the art will quickly recognize that there are two main types of detectors used to detect neutrons or gamma rays in pulsed neutron tools used for formation sigma measurements. One is gas detectors such as He-3 tubes for neutrons, and the other is various scintillation detectors (such as NaI, CsI, GSO, BGO, LaBr3 scintillators, and photomultiplier tubes) for gamma rays. It can be seen that these detectors are designed to detect either neutrons or gamma rays, but not both. As mentioned earlier, this is not due to insufficient detector capability, but rather to insufficient development of computer algorithms to first distinguish between the two signals and then process them.
[0044] In the case of scintillation detectors, the scintillator converts the deposited energy of gamma rays into scintillation light, and the PMT converts the scintillation light into electrons and amplifies them to form electronic signals.
[0045] Conventional gamma-ray detectors are not sensitive to neutrons, and conventional neutron detectors are not sensitive to gamma rays. The recent development of new scintillators that are sensitive to both neutrons and gamma rays, such as Cs2LiYCI6 (CLYC) and Cs2LiLaBr6 (CLLB) crystals, has made it possible to design pulsed neutron tools that can detect both neutrons and gamma rays simultaneously, and therefore can provide more multiple measurements than a single PNN tool or a single PNC tool.
[0046] As far as electronic systems are concerned, those of ordinary skill in the art will quickly recognize that all nuclear logging tools contain electronic circuits and devices, commonly referred to as front-end electronics, that are used to receive and process the electrical signals generated by the radiation detectors. These front-end electronics consist of a series of signal processing subsystems that filter, amplify, shape, and digitize these electrical signals to ultimately produce digitally encoded information, such as the type, time, and energy of radiation that excited the radiation detector. The goal of the front-end electronics is to obtain as much information about the radiation as possible with the highest possible accuracy. Historically, front-end electronics have been composed of analog components. However, with the development of digital electronics, programmable logic, and digital signal processing, the performance of front-end electronics has continued to improve over time through the development and implementation of new and improved analog electronic devices and electronic designs, opening up new opportunities and providing new advantages that were not previously possible.
[0047] A similar situation arises when using non-transitory computer readable memory devices in wireline logging or logging while drilling (LWD) operations.
[0048] In wireline logging, the measurement data itself is recorded in an electronic data format at the surface or in the borehole, and a printed record or electronic document called a "log" is then provided to a skilled person along with an electronic copy of the original data. The data is recorded directly based on the measured depth. The memory data is provided based on time, and the depth data is measured simultaneously based on time. The two data sets are then merged using a common time reference to create a record of instrument response and depth. The measured cable depth can be derived from several different measurement methods, but is usually recorded based on a calibrated wheel counter, or (more accurately) using magnetic markers, the latter of which provide calibrated increments of cable length. The measurement must be corrected for elastic stretch and temperature.
[0049] In LWD, measurement data is transmitted to the surface in real time via pressure pulses in the mud column in the well. This method of mud telemetry provides a bandwidth of less than 10 bits per second, but because drilling through rock is a slow process, data compression techniques must be used to provide sufficient bandwidth for real-time transmission of information. Higher data sampling rates are recorded to memory and retrieved when the drill pipe is retracted for drill bit changes. High-precision downhole and subsurface information can be obtained through networks or wired drill pipe that can transmit memory-quality data in real time.
[0050] 6. Introducing neural network computing models in oil and gas exploration
[0051] Neural networks (NN) are generally considered to be nonlinear statistical modeling tools, in which complex relationships between inputs and outputs are modeled and regularities are found (see Haykin, Sm Neural Networks: A Comprehensive Foundation, 2nd edition, Prentice Hall, 1999, P842). In fact, those skilled in the art generally describe neural networks as computational models that simulate the structure and function of biological neural networks. Therefore, neural networks have an organizational structure similar to that of the human brain and are networks composed of processing elements called nodes. Nodes obtain data from surrounding nodes, which can be located in the same layer or in different layers to process certain computing algorithms, and then pass the results to other nodes in the same layer or different layers. The connections between nodes have weights associated with them. Each connection can transmit a signal from one node to another. The node that receives the signal can process it and then send a signal to other nodes connected to it. The information flowing through the network affects the structure of the neural network because the neural network changes according to the input and output, or in a sense, it is learning. Nevertheless, almost all neural networks are developed based on mathematical models with the following assumptions (see Mohaghegh, S., Virtual-Intelligence Applications in Petroleum Engineering: Part 1-Artificial, 09 / 01 / 2001, Society of Petroleum Engineers):
[0052] (1) Information is processed by elements called neurons or nodes;
[0053] (2) There are connection links between nodes, and information can be transmitted through them;
[0054] (3) Each connection link has its own weight; and
[0055] (4) Once a node receives an input, it applies an action function to determine the output.
[0056] In the oil and gas industry, neural networks are formed in a supervised manner, that is, nodes and layers of nodes are trained to perform certain algorithms before they are sent to the field. Among them, the first step in training a neural network for oil exploration generally involves collecting and selecting data to be input into the first input layer. Depending on the application, the input data can be selected from experimental testing, modeling, simulation, sensitivity analysis, or the opinions of ordinary technicians in the field. Once the data has been input into the first input layer, they are divided into three parts: training, testing, and validation. The training part is used to achieve the expected output, which is generally achieved by adjusting or calibrating the weights applied to each node. In the first training phase, errors in the input data are self-propagated through the network until calibration is achieved, at which point the neural network is considered trained. Calibration is achieved when there is no improvement in the network's generalization of the data. In some cases, the data needs to be normalized to increase the scale of data normalization to prevent the network from being overly strained and running indefinitely without achieving calibration. There are many data normalization methods in the field.
[0057] Another aspect of a trained neural network is determining the number of hidden layers and the number of nodes within each hidden layer. To do this, multiple iterations are performed starting with one hidden layer and one node until the network is found to have achieved the lowest possible error, which is usually done by using a series of error formula calculations such as the Total Average Absolute Deviation (TAAD) or Mean Squared Error (MSE). Once all values are satisfied and the neural network is fully trained, it is sent out into the field to perform its complex calculations. In the oil and gas industry, there is a large amount of historical data available that is often used to predict future outcomes to help make better decisions. However, future predictions are always challenging as there is a lot of uncertainty about the future. This is why there is a need for real-time predictions using computer-implemented methods to find the true wellbore and formation sigma, like the method disclosed in this article, which can provide decision makers with good accuracy to prepare and solve problems in advance. In fact, according to Table 2 of Alkinani, Husam & Al-Hameedi, Abo Taleb & Dunn-Norman, Shari & Flori, Ralph & Alsaba, Mortadha & Amer, Ahmed, (2019), Applications of Artificial Neural Networks in the Petroleum Industry: A Review, 2019, there is currently no application of neural networks in wireline or logging while drilling projects to find the true wellbore and true formation sigma near the wellbore.
[0058] 7. Conclusion
[0059] Given the above complexity, one of ordinary skill in the art will quickly realize that it is common for different companies to independently develop logging tools and data processing algorithms to better suit their projects or customer requirements. Nevertheless, as previously mentioned, the most accepted approach for real borehole and formation sigma measurements is to use two single-function (i.e., either thermal neutron or gamma ray) radiation detectors and fit the thermal neutron decay curve or the captured gamma ray decay curve using two exponential decays for each detector. It is just that the apparent borehole sigma value obtained from the near detector is a better estimate of the borehole sigma value, while the apparent formation sigma value obtained from the far detector is a better estimate of the formation sigma value. This requires further computational processing to correct the borehole and formation sigmas, as they need to be expressed as a function of the apparent borehole sigma and apparent formation sigma from the above two detectors, considering the lack of independence between them. The separation of these two time decay components is the main difficulty in today's techniques and methods for finding borehole and formation sigmas. Even if some methods are used, they still do not take into account the effects of the near-wellbore environment (e.g., wellbore dimensions, tool spacing, wellbore fluid salinity / density, casing, cement, etc.), which vary significantly and are superimposed on the thermal neutron diffusion effects, making the separation process more difficult. Therefore, these effects (on neutrons and gamma rays) cannot be fully compensated at the same time. On the other hand, the method disclosed in this article can compensate for all these effects and obtain two outputs (true wellbore sigma and true formation sigma) simultaneously without the need for any additional other sensors such as temperature, pressure, acoustic tool spacing, etc. Summary of the invention
[0060] In view of this, various embodiments of the present disclosure advantageously provide a computer-implemented method using a non-transitory computer-readable memory device with a pre-programmed neural network and a computer program product of a trained neural network to obtain true wellbore sigma and true formation sigma.
[0061] The disclosed multi-layer feedforward neural network has one or more hidden layers, which enables the network to model nonlinear and complex functions. The hidden layer is responsible for transferring meaningful information between the input and each level of the network through an effective method. In fact, the neural network interface predicts three sets of neutron-induced apparent borehole time decay constants and neutron-induced apparent formation time decay constants obtained from three dual-function detectors through the nodes or neurons of its input layer, as well as the obtained gamma-induced apparent borehole time decay constants and gamma-induced apparent formation time decay constants, which are transmitted or processed through one or more hidden layers to output the true borehole sigma and true formation sigma of its output layer. Each node of the hidden layer and the output layer carries an algorithm that indicates the internal activation phenomenon. The node output is usually achieved by transforming its input using the opposite transfer function. In the embodiment disclosed herein, the transfer function is a nonlinear regression algorithm, but as the data advances to the output layer, they become linear regression. It should be noted that each interconnection or node in the neural network has a strength, which is represented by a number called weight. Additionally, biases are additional inputs attached to nodes that are always less than one percent and are treated the same as other weights.
[0062] Embodiments of the present disclosure show that the computer-implemented method uses a blank pre-programmed neural network and a trained neural network. According to certain aspects of the present disclosure, both networks can be provided in a computer program product, which includes computer-readable operating instructions that, when run on or loaded into a non-transitory computer-readable memory device, or when executed by one or more processors of the non-transitory computer-readable memory device, perform a method of using a non-transitory computer-readable memory device with a computer program product of a pre-programmed neural network and a trained neural network (both of which have three layers) to perform operations for obtaining true wellbore sigma and true formation sigma. In certain aspects of the embodiments disclosed herein, both neural networks require a set of input values for the input layer, which can be experimental values or theoretical values from a tool model calibrated with experimental data. In this way, the neural network uses known inputs for processing and obtains outputs, which are generally weighted combinations of the outputs of the input layer nodes, but are compared with known values of true wellbore and formation sigma to complete the training phase.
[0063] After the above training phase, the weights are kept fixed and the neural network is then used to process multiple samples called test sets for the above fixed weights. If convergence is observed between the predicted values and the actual borehole sigma values and formation sigma values in the test set, the neural network is considered to have been trained and will be installed in the processor or data processing system of the bottom hole assembly on the surface for use in logging or drilling operations. If convergence between the neural network output and the actual true borehole sigma values and true formation sigma values is not found in the test set, further training is required. Therefore, compared with other methods currently used, this method not only can quickly solve complex field problems, but also requires less computing power (and time) to obtain output results.
[0064] In particular, one embodiment of the present disclosure includes a bottom hole assembly tool having a pulsed neutron source coupled to at least three dual-function radiation detectors, a high voltage supply source, and an electronic instrument having a non-transitory computer-readable memory device for performing processing, generating, and computing operations. The at least three dual-function radiation detectors can be arranged at one or both ends of the neutron source and maintain an optimal distance from the neutron source. The detectors are capable of detecting neutrons and neutron-induced gamma rays simultaneously. The signals of neutrons and gamma rays emitted by the detectors can be distinguished by pulse shape discrimination technology, and the thermal neutron time decay signals and captured gamma ray time decay signals measured at multiple detectors can be used to obtain borehole sigma and formation sigma.
[0065] The use of multiple dual-function detectors in the present disclosure can not only reduce the types and number of all detectors inside the tool, but also make more measurements simultaneously and automatically correct for near-wellbore environmental effects, thereby providing more accurate measurement results.
[0066] In certain embodiments of the present disclosure, the pulsed neutron source and the three dual-function radiation detectors may be arranged in the same radial direction or in different directions, i.e., have the same or different tool face angles when deployed in the formation. Nevertheless, other embodiments of the present disclosure may also have more than three detectors. For example, the system may have six detectors, which may be arranged relatively on the cross section of the logging tool, i.e., the tool face angle. With different tool face angles, the detectors may be arranged in a variety of ways to preferentially receive neutrons and gamma rays at specific incident angles from the formation. This may also improve the detection efficiency of neutrons and gamma rays by increasing the total count rate of all detectors. Similarly, the nuclear logging tool may have more than one neutron source, which may be turned on or off at the same time to increase the count rate of the at least three dual-function radiation detectors, thereby reducing the uncertainty of the statistical measurement.
[0067] For a pulsed neutron tool with three dual-function detectors, two exponential decays are used to curve fit the thermal neutron count rate time decay curve obtained by each detector to obtain two neutron-induced apparent time decay constants related to the early time decay component and the late time decay component (i.e., the apparent borehole time decay constant τ induced by neutrons from the near detector). bnn and the neutron-induced apparent formation time decay constant τ fnn ). Similarly, two exponential decays are used to curve fit the captured gamma-ray count rate time decay curve obtained by each detector to obtain two gamma-induced apparent time decay constants related to the early time decay component and the late time decay component (i.e., the gamma-induced apparent borehole time decay constant τ from the near detector) bgn and the gamma-induced apparent formation time decay constant τ fgn ). Using a non-transitory computer readable memory device having a computer program product with a trained neural network, a curve fit is performed on the time decay curves of all detectors immediately following a pulse from a neutron source, i.e., a deuterium-tritium (DT) neutron generator or a deuterium-deuterium (DD) neutron generator. The trained neural network then receives as inputs the neutron-induced apparent borehole time decay constant, the neutron-induced apparent formation time decay constant, the gamma-induced apparent borehole time decay constant, and the gamma-induced apparent formation time decay constant for each of at least three dual-function radiation detectors, which are then used for processing by weighted nonlinear and linear regression algorithms within hidden layers of the deep neural network. Once the last linear regression linear algorithm within the third of the three hidden layers is processed, the neural network generates a true borehole sigma and a true formation sigma using a weighted linear regression algorithm.
[0068] Both the pre-programmed neural network (i.e., blank neural network) and the trained neural network use a non-transitory computer readable memory device containing a computer program product executable on the device to perform operations or instructions for acquisition, training, separation, measurement, initialization, input, calculation, processing, transmission, and generation. The non-transitory computer readable memory device is connected to the memory resources and telemetry device via a communication bus for transmitting information. After the calculation process is finally completed, the embodiments of the present disclosure generate interim results, which are used by the computer program embodied in the non-transitory memory device to generate the results, which generally include the actual wellbore sigma ∑ b and the true formation sigma ∑ f. These results can be stored locally in the system or in a database of the application server. In addition, the test results show that the present disclosure using the above-mentioned system and method also has great potential in other practical applications, wherein the system and method use a pulsed neutron source and multiple dual-function radiation detectors of neutrons and gamma rays (pulse shape discrimination technology can be used to distinguish neutrons and gamma rays) to measure thermal neutron time decay signals and thermal neutron capture gamma ray time decay signals, thereby obtaining wellbore sigma and formation sigma.
[0069] Nevertheless, further details, embodiments and various aspects of the present disclosure will be described in more detail below with reference to the accompanying drawings of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS
[0070] The teachings of the present disclosure may be readily understood by considering the following detailed description in conjunction with the accompanying drawings. Therefore, the features and advantages of the present disclosure, as well as other obvious features and advantages, may be understood in more detail by referring to the embodiments illustrated in the accompanying drawings, which form a part of this specification. However, it should be noted that the accompanying drawings illustrate only some embodiments of the present disclosure and therefore should not be considered as limiting the scope of the present disclosure, as the present disclosure may also include other effective embodiments.
[0071] Figure 1 is a schematic diagram showing a cross section of a logging while drilling system having a bottom hole assembly located above an investigation area according to one embodiment of the present disclosure, which has a pulsed neutron tool and three dual-function radiation detectors (i.e., a near detector, an intermediate detector, and a far detector);
[0072] Figure 2A and Figure 2B A flow chart showing a computer-implemented method using a non-transitory computer-readable memory device with a pre-programmed neural network and a trained neural network, each having three layers, to perform operations for obtaining a true wellbore sigma and a true formation sigma, according to one embodiment of the present disclosure;
[0073] Figure 3 A computer program product of a trained neural network is shown for obtaining a true borehole sigma and a true formation sigma by using the apparent borehole sigma and the apparent formation sigma (or the apparent borehole time decay constant and the apparent formation time decay constant) from three detectors;
[0074] Figure 4 shows a pulse height discrimination technique according to one embodiment of the present disclosure, showing the separation between neutrons and gamma rays;
[0075] Figure 5A , Figure 5B and Figure 5C The thermal neutron count rate decay curve, the neutron intensity, and the gamma ray count rate decay curve for each of at least three dual-function detectors according to one embodiment of the present disclosure are shown;
[0076] Figure 6 is an electrical diagram in the form of a block diagram of an application server having a computer program product according to one embodiment of the present disclosure, the computer program product being embodied in a non-transitory computer-readable device for use in a distributed network for storing instructions which can be used by the device to perform a method for obtaining near-wellbore true borehole sigma and true formation sigma using a nuclear logging tool during oil and gas surveys. DETAILED DESCRIPTION
[0077] 1. Introduction
[0078] Reference will now be made in detail to several embodiments of the present disclosure, examples of which are shown in the accompanying drawings. It should be noted that, where applicable, the same or similar reference numerals are used in the accompanying drawings to represent the same or similar functions. These drawings depict embodiments of the present disclosure for illustrative purposes only. Those skilled in the art will readily appreciate from the following description that alternative embodiments of the structures, systems, and methods shown herein may be employed without departing from the principles described in the present disclosure.
[0079] The subject matter which is regarded as the present disclosure is particularly pointed out and distinctly claimed at the conclusion of the specification. However, the organization and method of operation of the disclosure, together with objects, features, and advantages thereof, will be best understood from the following detailed description and accompanying drawings.
[0080] Since most of the illustrated embodiments of the present disclosure can be implemented using electronic components and circuits known to those skilled in the art, in order to facilitate understanding and appreciation of the basic concepts of the present disclosure and also to avoid confusion or deviation from the teachings of the present disclosure, the explanation of details will not exceed the degree deemed necessary in the above-mentioned illustrations.
[0081] Any reference in the specification to a method should be applied mutatis mutandis to a system capable of performing the method and should be applied mutatis mutandis to a computer program product storing instructions which, once executed by the system, will cause the method to be performed.
[0082] In addition, the flowcharts and block diagrams in the figures illustrate the possible implementation architectures, functions, and operations of the systems, methods, and computer program products according to various embodiments of the present disclosure. It should also be noted that in some alternative implementations, the functions marked in the boxes may not appear in the order marked in the figures. For example, two boxes shown in succession may actually be executed substantially simultaneously, or sometimes the boxes may be executed in the opposite order, depending on the functions involved. It will also be noted that each box illustrated in the block diagram and / or flowchart, and the combination of boxes illustrated in the block diagram and / or flowchart, can be implemented by a hardware-based special-purpose system that performs a specified hardware function or action, or by a combination of special-purpose hardware and computer instructions.
[0083] 2. Terminology
[0084] In this specification and the appended claims, unless the context requires otherwise, the following terms shall have the meanings indicated:
[0085] "Memory resource" means a device or system for storing data. A memory resource may include one or more magnetic or magneto-optical or optical disk drives, solid-state memory devices, or magnetic tapes. For convenience, memory devices are sometimes referred to as "disks" or "hard disks." A data storage system may include the same or different types of memory devices having the same or different storage capacities.
[0086] The computer program product mentioned in this specification should be applied mutatis mutandis to a system capable of executing the instructions stored in the computer program product, and should be applied mutatis mutandis to a method executable by a system reading the instructions stored in a non-transitory computer-readable medium.
[0087] As used herein, "and / or" placed between a first entity and a second entity means one of (1) the first entity, (2) the second entity, and (3) the first entity and the second entity. Multiple elements listed using "and / or" should be interpreted in the same manner, i.e., "one or more" of the elements so combined.
[0088] It is easy to understand that each block and combination of blocks in the flowchart can be implemented by computer program instructions. The computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device can create a means for implementing the function / action specified in the block or the block.
[0089] The computer program product may also be stored in a non-transitory computer-readable memory device, which may instruct a computer or other programmable data processing device in the drilling assembly to operate in a specific manner so that the instructions stored in the non-transitory computer-readable memory device produce a manufactured product, which includes instruction means for implementing the functions / behaviors specified in the various blocks, flow charts and other structures illustrated and described herein.
[0090] Computer program instructions may also be loaded onto a computer or other programmable data processing device so that a series of operational steps can be executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device can provide steps for implementing the functions / behaviors specified in one or more blocks.
[0091] The "preprogrammed non-transitory computer readable memory device" or "computer program product or computer system device" on the ground processing system includes: direct actions that cause generation, and any indirect actions that facilitate generation. Indirect actions include: providing software to users, maintaining websites through which users can affect displays, hyperlinking to such websites, or cooperating or cooperating with entities that perform such direct or indirect actions. In this way, users can operate alone or in cooperation with third-party suppliers to enable the generation of reference signals on display devices. Output devices may include display devices and should be suitable for displaying the required information, such as but not limited to CRT monitors, LCD monitors, plasma devices, flat-panel devices, or printers. Display devices may include devices that have been calibrated using any conventional software intended for evaluating, correcting and / or improving display results (for example, a color monitor that has been adjusted using monitor calibration software). Instead of (or in addition to) displaying a reference image on a display device, a method consistent with the present disclosure may include providing a reference image to an object.
[0092] Software includes: any machine code stored in any storage medium (e.g., RAM or ROM), as well as machine code stored on other devices (e.g., non-transitory computer-readable media such as external hard drives or flash memory). Software may include: source code or object code, including any set of instructions that can be executed in a client machine, server machine, remote desktop or terminal.
[0093] A combination of software and hardware may also be used to provide enhanced functionality and performance for certain embodiments of the disclosed invention. An example is to build software functionality directly into a silicon chip. Therefore, a combination of hardware and software is also included in the definition of a computer system, and this disclosure contemplates it as possible equivalent structures and equivalent methods.
[0094] The data structure is a data organization that can be defined as an embodiment of the present disclosure. For example, the data structure can provide the organization of data, or the organization of executable code. The data signal can be carried by a non-transient transmission medium and store and transmit various data structures, so it can be used to transmit an embodiment of the present disclosure.
[0095] According to the preferred embodiments of the present disclosure, the description of specific hardware and software is described in detail, which is only an example embodiment and is not intended to limit the implementation structure of the disclosed embodiments. For example, although many internal and external components have been described, those of ordinary skill in the art will understand the known components and their interconnections. In addition, certain aspects of the disclosed invention may be embodied in software executed using one or more receiving systems, computer system devices, or non-transitory computer-readable memory devices. The program aspects of the technology can be considered as "products" or "articles of manufacture", which are usually carried or embedded in a machine-readable medium in the form of executable code and / or related data. Tangible non-transitory "storage" type media and devices include: any or all memories, or other storage for computers, processes, etc., or related modules including, for example, various semiconductor memories, tape drives, disk drives, optical disks or disks, etc., which can provide storage for software programming at any time.
[0096] It is noted that the term "survey area" as used herein refers to an area or volume of geological interest and may be associated with the geometry, posture and layout of the area or volume at any measurement scale. An area may have features such as folding, faulting, cooling, unloading and / or fracturing that have occurred therein.
[0097] In addition, the term "calculate" encompasses a variety of actions, including calculating, determining, processing, deriving, investigating, looking up (e.g., looking up in a table, database or other data structure), determining, etc. It can also include receiving (e.g., receiving information), accessing (e.g., accessing data in a memory), etc. In addition, "calculating" can include parsing, selecting, choosing, establishing, etc. On the other hand, the term "generating" includes the action of outputting a constant or sigma, variable, such as a borehole constant, a formation constant, a borehole sigma, or a formation sigma.
[0098] Capturing specific data may include creating or distributing a reference image to a subject by physical, telephone, or electronic delivery, providing access to a reference via a network, or creating or distributing software to a subject that is configured to run on a subject's workstation or computer that includes the reference image. In one example, providing the reference image may involve enabling the subject to obtain the reference image in the form of a hard copy via a printer. For example, information, software, and / or instructions may be transmitted (e.g., electronically, or physically via a data storage device or hard copy) and / or otherwise provided (e.g., via a network) to cause the subject to print the reference image in the form of a hard copy using a printer. In such an example, the printer may be a printer that has been calibrated using any conventional software intended for evaluating, correcting, and / or improving print results (e.g., a color printer that has been adjusted using color correction software).
[0099] In addition, modules, features, properties, methods and other aspects can be implemented in software, hardware, firmware or any combination thereof. Regardless of whether a component of the present disclosure is implemented as software, the component can be implemented as a stand-alone program, as part of a larger program, as multiple separate programs, as a static or dynamic link library, as a kernel loadable module, as a device driver, and / or in all and any other ways known to those skilled in the art of computer programming now or in the future. In addition, the present disclosure is not limited to implementation in any particular operating system or environment.
[0100] Some embodiments of the present disclosure relate to methods for monitoring the radiation output of a pulsed neutron generator using a dual-function radiation detector. The methods of the present disclosure may further include correcting the detector count rate for environmental effects (primarily lithology, porosity, and spacing) (also known as neutron induction), producing compensated (corrected, neutron induced) monitoring of source intensity. The corrected constant, variable, or sigma may be used to compensate measurements from other detectors to offset changes in source intensity caused by imperfections, power fluctuations, etc. of the pulsed neutron generator. The resulting source compensated measurements may improve the accuracy of formation logging.
[0101] 3. Attached photos
[0102] As mentioned above, there is no system and method that can provide an effective means for analyzing the three zones (i.e., the borehole zone, the shallow zone containing artificial components immediately surrounding the borehole, and the natural formation zone). This is particularly true because the depth of detection (DOI) of thermal neutrons is shallower than that of captured gamma rays. In addition, since the wellbore material and the artificial components surrounding the borehole have different attenuation effects on thermal neutrons and captured gamma rays, combining the two measurement methods provides a possible way to automatically compensate for the influence of these near-wellbore effects on the measurement, especially when multiple detectors are used, because the measurement will be affected by many near-wellbore parameters such as wellbore size, borehole salinity, casing thickness, etc. It can be seen that the main difficulty in obtaining the measurement of true borehole sigma and true formation sigma is how to separate the two time decay components, because the two time decays (borehole time decay and formation time decay) are not independent of each other. Therefore, for example, there is always a correlation between the apparent borehole sigma and the apparent formation sigma. In addition, the near-wellbore environmental effects (wellbore size, tool spacing, wellbore fluid salinity / density, casing, cement, etc.) vary greatly, and they also increase the thermal neutron diffusion effect, making separation more difficult. On the other hand, according to an embodiment of the computer-implemented method of the present disclosure, the neutron-induced apparent wellbore sigma, the gamma-induced apparent wellbore sigma, the neutron-induced apparent formation sigma, and the gamma-induced apparent formation sigma can be obtained from three detectors respectively, and finally the real wellbore sigma and the real formation sigma are obtained. The computer-implemented method of the present disclosure uses a trained neural network computer product that can compensate for all these effects and generate two outputs (real wellbore sigma and real formation sigma) at the same time, thanks to the proposed algorithm and almost effortless (only limited by time and computer resources) iterations to complete the training of the pre-programmed neural network.
[0103] Come and see Figure 1 , which shows a typical survey area 101, located above a land area, wherein a drilling rig 102 in which the present disclosure may be implemented is shown, which penetrates different types of formations 103, 104, 105, so that an embodiment of the present disclosure may be used therein. One of ordinary skill in the art will recognize that seismic surveys of an area can generate detailed images of the local geology in order to determine the location and size of possible hydrocarbon (oil and gas) reservoirs, and thus determine well locations. However, as Figure 1 As shown, when using a wireline or logging while drilling (LWD) downhole system 106 during directional drilling, the wireline or LWD downhole system 106 must deviate from a vertical downward trajectory to a trajectory that remains within a specified azimuth and inclination range to reach the well or reservoir. This degree of deviation is caused by countless situations, but is most likely caused by densely populated or obstructed areas.
[0104] In these survey areas 101, the drilling rig 102 may also be located above the wellbore to directly penetrate the underground formation. In addition, the lower end of the downhole system 106 includes a drill bit to help penetrate the underground formation and reach the desired reservoir. In addition, a drilling fluid or mud stored in the well pit 107 is formed at the well site, which is sent out through the mud channel 117 to lubricate the drill bit of the wired or logging while drilling (LWD) downhole system 106, and then used to carry the formation cuttings to the ground when returning to the well pit 107 for recirculation. More importantly, the wired or logging while drilling (LWD) downhole system of the present disclosure includes a bottom hole assembly (BHA), which is generally arranged near the drill bit and has the function of measuring, processing, calculating, generating and storing information, as well as the function of wireless (108) or wired communication with the ground processing system 109 for further processing. The measurement tools in the BHA include dual-function neutron and gamma ray radiation detector sensors, as well as acoustic and resistivity logging tools.
[0105] Figure 1 The internal components of the above-mentioned wireline or logging while drilling (LWD) downhole system 106, particularly the BHA portion, are shown below the ground 110. The BHA portion of the wireline or logging while drilling (LWD) downhole system 106 of the present disclosure further shows a pulsed neutron tool consisting of a pulsed neutron source 111 and three dual-function radiation detectors, namely a near detector 112 (denoted by the subscript letter n in the algorithm), a middle detector 113 (denoted by the subscript letter m in the algorithm), and a far detector 114 (denoted by the subscript letter f in the algorithm). For the present disclosure to work properly, the pulsed neutron source 111 can be a deuterium-tritium (DT) neutron generator or a deuterium-deuterium (DD) neutron generator. Which neutron source 111 (DT or DD) is used in the tool 106 depends mainly on the required measurement and the neutron yield of the neutron generator.
[0106] If measurements need to be made (e.g. spectroscopy) to obtain C / O logs for oil saturation, or to identify formation elements and obtain their concentrations in mineralogy determinations, then a DT neutron generator is a better choice, because not only do DT reactions produce much higher neutron energies than DD reactions, which is important for initiating inelastic scattering of neutrons for many elements (e.g. C, O, Fe, etc.), but also the neutron yield of DT reactions is two orders of magnitude higher than that of DD reactions for the same d-ion current, because the cross section of DT fusion reactions is larger. However, the frequency and pulse schemes need to be balanced or properly designed for all measurements. On the other hand, if formation sigma is the only thing that needs to be measured, and the neutron yield of the DD generator is as high as the current DT generators (in the region of 10 8n / s), then DD generators will be preferred in formation sigma measurements because DD generators do not emit as many gamma rays from neutron inelastic scattering as DT generators, and gamma rays can cause background in the decay curve very early after the neutron pulse, especially for systems using scintillation detectors to detect captured gamma rays. The frequency of DD neutron pulses is about 1 kHz, and the neutron duty cycle is about 50 μs. The time interval between neutron pulses is about 950 μs, which is long enough for fast neutrons to slow down to thermal neutrons and decay to background levels.
[0107] In an embodiment of the present disclosure, the nuclear logging tool of the BHA section may have more than one neutron source 111. In this embodiment, one neutron source 111 may be located at the proximal end of the BHA section and another at the distal end, and the radiation detectors 112, 113, and 114 may be alternately arranged between the neutron sources. For the neutron source 111 and the case where the remote detector 114 is replaced by another neutron source 111 (a shield is set between the neutron source and the detector), they can be alternately turned on or off, thereby alternately inducing neutrons and gamma rays in the formation, which are received by the radiation detectors 112, 113, and 114. Due to the four different source-to-detector distances (d1 to d4), the data generated by the two detectors 112 and 113 can better compensate for near-wellbore effects, such as wellbore size, tool spacing, mud weight and / or salinity, casing size, cement thickness, etc., compared with tools with only two or three source-to-detector distances. Therefore, the formation parameters obtained will be more accurate. However, when the source 111 and the detectors 112, 113 and 114 have different tool face angles, the data generated by the near detector 112, the middle detector 113 and the far detector 114 will reflect different sectors of the formation, and by comparing the measurements of the near detector 112, the middle detector 113 and the far detector 114, the differences between different formation sectors at any given time can be revealed.
[0108] With regard to the three detectors 112, 113, and 114, the embodiments of the present disclosure only use detectors that can detect neutrons and gamma rays simultaneously. The crystals used in the detectors may be Cs2LiYCI6 (CLYC) or Cs2LiLaBr6 (CLLB), which are sensitive to both neutrons and gamma rays. However, Cs2LiLaBr6 (CLLB) crystals perform better at higher temperatures and are therefore more preferred. Therefore, there is no need to use a cooling device (such as a flask) used in the prior art to install these detectors. By coupling the above-mentioned detectors with a scintillation photosensitive element (such as a photomultiplier tube PMT), such a scintillator can be effectively utilized to detect neutrons and gamma rays. The three dual-function radiation detectors 112, 113, and 114 are pre-programmed using a pulse shape discrimination technique via a non-temporary computer-readable memory device, such as Figure 4 As shown. The three detectors 112, 113 and 114 can be arranged at one end of the pulsed neutron source, or one detector can be arranged at one end of the pulsed neutron source and the other two detectors can be arranged at the other end of the pulsed neutron source. Nevertheless, those skilled in the art will soon recognize that certain embodiments of the present disclosure can also arrange the pulsed neutron source 111, the near detector 112, the middle detector 113 and the far detector 114 in the same radial direction or in different directions, that is, they have the same or different tool face angles when deployed in the formation. In other embodiments of the present disclosure, the wireline or logging while drilling (LWD) downhole system 106, especially the BHA part, can have more than three detectors. For example, there can be six detectors, wherein the detectors are arranged at angles of 0 degrees and 180 degrees relative to each other on the cross section of the logging tool (i.e., the tool face), while the remaining detectors are arranged at an angle of 120 degrees on the cross section of the wireline or logging while drilling (LWD) downhole system 106. The positioning of the detectors not only enables one of ordinary skill in the art to use different tool face angles, but also enables the detectors to preferentially receive neutrons and gamma rays at specific incident angles from the formation. In addition, it also improves the detection efficiency of neutrons and gamma rays by increasing the total count rate of all detectors.
[0109] Other embodiments of the present disclosure have four radiation detectors and two neutron sources in the BHA portion of a wireline or logging while drilling (LWD) downhole system 106. In this embodiment, the detectors are arranged in pairs at substantially equal distances from different neutron sources. As previously described, having more than one detector within a certain distance can increase the count rate within that distance, thereby allowing the use of less powerful neutron sources. The count rate of a single remote detector may be too low to provide reliable measurement data, so by using two or more remote detectors, the count rate will be significantly increased, thereby allowing reliable measurements to be obtained by processing data from multiple remote detectors.
[0110] In addition, the embodiment of the present disclosure shows a logging tool BHA portion of a wireline or logging while drilling (LWD) downhole system 106, which has multiple shielding covers 115, 116 that can absorb neutrons and gamma rays. The shielding covers need to be arranged between the neutron source 111 and the detectors 112, 113 and 114 in the logging tool BHA portion of the wireline or logging while drilling (LWD) downhole system 106, and between the neutron source 111 and other internal equipment 116, 119 and 120, so that the detectors can receive neutrons and gamma rays from the formation, and these rays will not pass through the logging tool itself. Alternatively, the detectors can also be partially shielded by shielding materials to absorb neutrons and gamma rays from certain directions. The shielding covers 115 and 116 are made of one or more materials that can effectively attenuate thermal neutrons and gamma rays, or contain these materials. The materials of the shielding covers 115 and 116 can be selected from heavy elements with high thermal neutron absorption cross-sections, including metals such as gadolinium (Gd) and samarium (Sm), metal oxides such as Gd2O3, Sm2O3, B2O3, alloys containing Gd or Sm with other heavy metals (such as iron, lead or W), or materials containing boron and other heavy elements (such as tungsten boride (WB, WB2, etc.)).
[0111] With regard to other internal components of the BHA portion of the logging tool arranged within the wireline or logging while drilling (LWD) downhole system 106, the figure further shows a mud channel 117, a high pressure supply source 118, electronic instruments 119 and a telemetry system 120. The mud channel 117 is also used to transmit neutron and gamma measurement data during logging while drilling to the surface, while the onboard electronic instruments 119 are also used to process, calculate, generate and store certain data of the wireline or LWD tool. For the latter, the data typically includes all information captured, processed, calculated, generated and stored during the disclosed method, and can also be retrieved when the tool returns to the surface. In any case, the basic data will still be further processed by the surface processing system 109, which is used in Figure 6 Electronics 119 assists in amplifying the neutron and gamma ray signals from detectors 112, 113, and 114, which are also distinguished from each other using pulse shape discrimination (PSD) techniques via an embedded non-transitory computer readable memory device of electronics 119, such as Figure 4As shown. The digital signal processing system in the electronic instrument 119 is a preferred embodiment of the present disclosure because it can provide more accurate measurements. A telemetry subassembly is also included in the BHA portion of the logging tool of the wired or logging while drilling (LWD) downhole system 106, which is mainly used for data and control communications with the surface. This device can be of any suitable type, such as a mud pulse (pressure or acoustic wave) telemetry system, which receives the output signal of the data measurement sensor and transmits a coded signal representing such output to the surface, where the signal is detected, decoded, and applied to a processor and / or recorder in the surface processing system 109, such as Figure 6 As further described in . The BHA portion of the logging tool deployed in the wireline or logging while drilling (LWD) downhole system is calibrated by using laboratory formations. Through these specially constructed formations, the BHA portion of the logging tool deployed in the wireline or logging while drilling (LWD) downhole system 106 can be responsive to characterize different borehole sigmas, formation sigmas, borehole dimensions, borehole fluids, casing and cement, etc. In any given laboratory formation, the basic measurements obtained from the tool are the neutron-induced apparent borehole time decay constant, the neutron-induced apparent formation time decay constant, the gamma-induced apparent borehole time decay constant and the gamma-induced apparent formation time decay constant for each of the three detectors. Once the apparent borehole and formation time decay constants are obtained in all laboratory formations, as well as the known real borehole and formation time decay constants, they can be used as pre-programmed neural networks for subsequent training.
[0112] See also Figure 2A and Figure 2B , reference numeral 201 shows a flowchart of a computer-implemented method and instructions used by a wired or LWD tool, which uses a pulsed neutron source and multiple dual-function radiation detectors of neutrons and gamma rays (pulse shape discrimination technology may be used first to distinguish the radiation of neutrons and gamma rays) so as to generate wellbore sigma and formation sigma using a computer program product of a trained neural network. The computer-implemented method 201 adopts the configuration of the pulsed neutron tool and three dual-function detectors as described above. The thermal neutron count rate time decay curve and the captured gamma ray count rate time decay curve obtained by each detector are processed to obtain two sets of apparent wellbore attenuation curves and apparent formation attenuation curves induced from thermal neutrons and captured gamma rays, which are related to the early time decay component and the late time decay component, respectively. Therefore, six pairs of apparent time decay constants can be obtained from laboratory measurements or generated from actual field lithology readings in the survey area, namely (τ bnn ,τ fnn ), (τ bnm ,τ fnm ), (τ bnf ,τ fnf), (τ bgn ,τ fgn ), (τ bgf ,τ fgf ) and (τ bgf ,τ fgf ), and further use them to generate the real wellbore sigma (∑ b ) and true formation sigma (∑ f ).
[0113] The thermal neutron count rate time decay curve and the captured gamma ray count rate time decay curve obtained by each detector can also be processed to obtain two sets of apparent borehole sigma and apparent formation sigma induced from thermal neutrons and captured gamma rays related to the early time decay component and the late time decay component, respectively. Therefore, six pairs of apparent sigma can be obtained from laboratory measurements or generated from actual field lithology readings in the survey area, namely (∑ bnn ,∑ fnn )、(∑ bnm ,∑ fnm )、(∑ bnf ,∑ fnf )、(∑ bgn ,∑ fgn )、(∑ bgf ,∑ fgf ) and (∑ bgf ,∑ fgf ), and further use them to generate the real wellbore sigma (∑ b ) and true formation sigma (∑ f ).
[0114] In particular, method 201 is a two-part computer-implemented method in which two subroutines have been included. These subroutines primarily include a blank pre-programmed neural network (NN) that uses known values to obtain a trained neural network (NN) computer program product. The trained NN computer program product is a second subroutine that is initiated when a non-transitory computer readable memory (NTCRM) device having a pre-programmed neural network is stored in memory resource 603 and then replaced by a message hook subroutine on non-transitory computer readable memory device 605.
[0115] The computer-implemented method 201 begins when a non-transitory computer readable memory (NTCRM) device having a pre-programmed neural network 302 in the surface processing system 109 is initialized at block 202. The above initialization may also be performed in conjunction with the initialization 203 of the wireline or LWD system 106, wherein the wireline or LWD system 106 is initialized at block 204 with at least two neutron pulses having a frequency of at least 500 Hz and a pulse width of at least 20 μs from the source 111. Thereafter, the non-transitory computer readable memory device having the pre-programmed neural network 302 sends a message hook to the neutron source 111 to begin irradiating the subsurface. Then, at block 205, at least three dual-function radiation detectors 112, 113, and 114 are used to measure thermal neutrons and neutron-induced gamma rays emitted by the neutron pulses emitted by the neutron source 111. Once all dual-function radiation detectors have measured the emitted thermal neutrons and neutron induced gamma rays, at block 206 the preprogrammed non-transitory computer readable memory device uses a pulse shape discrimination (PSD) technique to separate the thermal neutron and neutron induced gamma ray signals from each other. Figure 4 An exemplary separation output of the preprogrammed non-transitory computer readable memory device is shown. Then, at box 207, the non-transitory computer readable memory device with the preprogrammed neural network 302 sends a message hook to the underground telemetry system 120 to begin transmitting the mutually separated neutron and gamma ray signals from each detector to the telemetry host system 602 on the surface so that they can be further processed by the multi-channel analyzer (MCA) embedded in the surface processing system 109 and connected to the computer system device 608 via the communication bus 604. Once the MCA receives the above-mentioned signals, it generates a histogram or "time spectrum" representing the number of events and the arrival time of the events at box 208, which is then stored in the memory resource 603. The stored time spectrum can be displayed by the display 610 and analyzed by a person of ordinary skill in the art. Once the separated signals have been stored, the non-transitory computer readable memory device with a pre-programmed neural network receives a message hook from the MCA and begins generating a thermal neutron time spectrum and a capture gamma ray time spectrum for each of the three dual-function radiation detectors, wherein each thermal neutron time spectrum is essentially composed of a time decay curve of thermal neutrons, and the capture gamma ray time spectrum is essentially composed of a time decay curve of capture gamma rays, both of which follow the neutron pulse.
[0116] The non-transitory computer readable memory device with the pre-programmed neural network 302 then begins to calculate curve fits of the generated time decay curves for thermal neutrons and captured gamma rays using two exponential decays at boxes 209 and 210 to obtain the neutron-induced apparent borehole time decay constant, the neutron-induced apparent formation time decay constant, the gamma-induced apparent borehole time decay constant, and the gamma-induced apparent formation time decay constant for each of the three detectors. The "best fit" of the curves is determined by the non-transitory computer readable memory device with the pre-programmed algorithm 605 by varying at least five parameters in the fitting function Y(t) to bring Y(t) as close to the data as possible. Mathematically, this is achieved by having a least squares relative error regression L programmed on the non-transitory computer readable memory device with the pre-programmed algorithm 605 as follows:
[0117]
[0118] Where C(t) is the experimental count rate at time t in detector C. The fitting function Y(t) is a double exponential time decay function expressed as:
[0119]
[0120] Among them, A b and A f is a constant, which can be interpreted as the initial amplitude of the borehole component and the formation component, in cps. b and τ f are the time decay constants (average neutron lifetimes) of materials in the borehole region and the formation region, respectively. B represents the background in cps.
[0121] Thereafter, the non-transitory computer readable memory device with the pre-programmed neural network 302 reads L, where the smaller the value, the better the fit between Y(t) and the curve data. The function Y(t) here includes 5 parameters: A b , A f , τ b , τ fand B. The "best fit" is determined by the non-transitory computer readable memory device with pre-programmed neural network 302 by finding the values of these 5 parameters that will make L as small as possible. When the function L is minimized, the curve Y(t) will be as close to the data as possible. To determine how small L should be, the non-transitory computer readable memory device with pre-programmed algorithm 605 divides the value of L by the number of data points fitted to each detector. This number is called the least squares relative error for each data point in the computer-implemented method, and informs the non-transitory computer readable memory device with pre-programmed algorithm 605: how far the (average) relative error of the fit is from each data point. For an acceptable fit, the relative error of the data points should be less than 2%. Once the relative error is less than 2%, the non-transitory computer readable memory device with pre-programmed algorithm 605 has achieved an acceptable curve fit.
[0122] like Figure 5A , Figure 5B and Figure 5C As shown, the time decay curves of the near detector 501, the middle detector 502 and the far detector 503 are similar, but not identical. Specifically, the near detector ( Figure 5A ) has the largest signal amplitude, and the far detector ( Figure 5C ) has the smallest signal amplitude, which is due to the fact that the flux rate of thermal neutrons or captured gamma rays decreases with the distance from the source. Secondly, the slope of formation time decay is steepest in the near detector and lowest in the far detector. This is because the thermal neutron diffusion effect has a greater impact on the decay curve of the near detector than that of the far detector. The reason is that the higher the thermal neutron flux rate, the greater the diffusion effect of thermal neutron time decay and captured gamma ray time decay. It can be seen from the figures 502 and 510 that the thermal neutron flux rate is highest at the near detector and lowest at the far detector.
[0123] At block 211, the thermal neutron time decay curve is further processed using a non-transitory computer readable memory device having a pre-programmed neural network 302 to obtain a neutron-induced apparent borehole time decay constant and a neutron-induced apparent formation time decay constant for each of the at least three dual-function radiation detectors, assigning the following term to each of the at least three dual-function radiation detectors according to its location (i.e., proximal, intermediate, distal): bnn , τ bnm , τ bnf , τ fnn , τ fnm , τ fnfThe captured gamma-ray time decay curve is further processed at block 211 to obtain a gamma-induced apparent borehole time decay constant and a gamma-induced apparent formation time decay constant for each of the at least three dual-function radiation detectors, and the following terms are assigned to each of the at least three dual-function radiation detectors according to its location (i.e., near, intermediate, far): τ bgn , τ bgm , τ bgf , τ fgn , τ fgm , τ fgf Thereafter, with the inputs collected and the outputs known, the non-transitory computer readable memory device with the pre-programmed neural network 302 will send a message to the rest of the system to stop any further data collection until the neural network has completed training. Thus, the first step in training the non-transitory computer readable memory device with the pre-programmed neural network 302 is to send a message to the rest of the system to stop any further data collection until the neural network has completed training. Figure 3 The input layer 3021 of the present invention provides a set of input values. The outputs 3231 and 3232 of the output layer 3204 are compared with the known values of the borehole sigma and formation sigma corresponding to this set of input values for the sample. These input and output values can be experimental values or theoretical values from the experimental data calibration tool model. In the training phase, this training is performed on a set of samples called the "training set", and the weights at each processor are adjusted to improve the match between the outputs 3231 and 3232 and the known borehole sigma values and formation sigma values for each set of input values in the training set. When convergence is reached, the training process ends. Convergence is reached when the incremental change in the weight is very small, or the difference between the output and the true value is very small. The middle layer of the non-transitory computer readable memory device with a pre-programmed neural network 302 is a hidden layer 3023, which can include at least three additional hidden layers (also known as a deep neural network system). The processors of the above-mentioned intermediate layers are weighted from nonlinear regression to linear regression according to the depth of the hidden layer. The first layer of the non-transitory computer readable memory device with the pre-programmed neural network 302 is trained at box 212 taking into account the neutron-induced apparent borehole time decay constant, the neutron-induced apparent formation time decay constant, the gamma-induced apparent borehole time decay constant and the gamma-induced apparent formation time decay constant acquired at each node 3210, 3211, 3212, ..., 3213. Thereafter, the non-transitory computer readable memory device with the pre-programmed neural network 302 sends these inputs to the hidden layer, which begins to be processed by different regression algorithms at box 213. In particular, the non-transitory computer readable memory device with the pre-programmed neural network 302 processes the apparent borehole time decay constant in its first hidden layer according to the following expression:
[0124]
[0125] Furthermore, in the same first hidden layer, the non-transitory computer readable memory device having the pre-programmed neural network 302 also processes the apparent formation time decay constant according to the following expression:
[0126]
[0127] Among them, w i is the weight, X i is the apparent borehole time decay constant or the apparent formation time decay constant in the input layer, where i = 1, 2, 3, ..., 12, and n = 3. Similarly, the second hidden layer of the non-transitory computer readable memory device having the pre-programmed neural network 302 continues to process the input values using nonlinear regression, and in particular processes the apparent borehole time decay constant according to the following expression:
[0128]
[0129] and the apparent formation time decay constant according to the following expression:
[0130]
[0131] Among them, w i is the weight, Y i is the apparent borehole time decay constant node value or the apparent formation time decay constant node value in layer 1, where i = 1, 2, 3, ..., 12, and n = 3. At this point, the non-transitory computer readable memory device with the pre-programmed neural network 302 will have detected that the number of nodes available for comparison with the data input value is different, or insufficient, and will transfer it to the third layer for training according to the simpler linear regression algorithm:
[0132]
[0133] Among them, w i is the weight, Z i is the apparent borehole time decay constant node value or the apparent formation time decay constant node value in hidden layer 2, where i = 1, 2, 3, …, 12, and n = 3.
[0134] The non-transitory computer readable memory device with the pre-programmed neural network 302 then sends the calculated non-linear and linear values of the apparent borehole time decay constant and the apparent formation time decay constant to the output layer 3024 to begin generating the true borehole and formation sigmas. The non-transitory computer readable memory device with the pre-programmed neural network 302 does so according to a certain pre-programmed algorithm, where wi and w k is the weight, U i is the apparent borehole time decay constant node value in hidden layer 3, U k is the apparent formation time decay constant node value in hidden layer 3, where j = 1, 2, 3, ..., 6, k = 1, 2, 3, ..., 6. In practice, the algorithm used by the non-transitory computer readable memory device with a pre-programmed neural network 302 to generate the true wellbore sigma is:
[0135]
[0136] Similarly, the algorithm used by the non-transitory computer readable memory device having the pre-programmed neural network 302 to generate the true formation sigma is:
[0137]
[0138] Nevertheless, the training of the non-transitory computer readable memory device having the pre-programmed neural network 302 is not yet complete at step 214, because it still needs to compare the output of the third layer (∑ b or f ) and the known values of the true borehole sigma and the true formation sigma. When comparing these values, the non-transitory computer readable memory device with the preprogrammed neural network 302 will evaluate whether the difference is less than 1%, and if not, it will send a message hook at box 215 to restart the process from step 204. The non-transitory computer readable memory device with the preprogrammed neural network 302 will consider the iteration complete when the relative difference is successfully achieved less than 1%. Once the non-transitory computer readable memory device with the preprogrammed neural network 302 achieves the relative difference, the weights will remain fixed and a plurality of samples, referred to as the "test set", will be further processed using the non-transitory computer readable memory device with the preprogrammed neural network 302 with the fixed weights. If good agreement between the predicted values and the actual borehole sigma and formation sigma values is observed in the test set, the neural network is considered to have been trained and a trained neural network computer program product 303 is generated at box 216 and then stored in the memory resource 603 at box 217 and subsequently replaced or installed in the electronic system or data processing system of the bottom hole assembly (BHA) of the surface processing system 109 at box 218 for use during logging or drilling operations. If good agreement between the neural network output and the actual true borehole sigma and true formation sigma is not found in the test set, further training is required, including repeating steps 204 to 215.
[0139] After replacing the pre-programmed neural network on the non-transitory computer readable memory device with the trained neural network, it now becomes the non-transitory computer readable memory device 605 with the trained neural network computer program product 303. The non-transitory computer readable memory device 605 sends information to the neutron pulse source 111 of the wired or LWD system 106, causing the deuterium-tritium (DT) neutron generator or the deuterium-deuterium (DD) neutron generator (neutron source) to generate at least two pulses at block 219. For the signals generated by the pulses, the neutrons and neutron-induced gamma rays are measured by at least three dual-function radiation detectors at step 220. Considering that the neural network has been trained at this stage, the input data will be the neutrons and neutron-induced gamma rays actually measured after each of the at least three dual-function radiation detectors emits at least two pulses. The measured neutron and neutron-induced gamma-ray signals still need to be separated into thermal neutrons and neutron-induced gamma-rays by the underground electronic instrument 119 with non-transitory computer readable memory device using pulse shape discrimination techniques at box 221. Since all of this separation occurs at the BHA underground, the underground telemetry system 120 sends a message hook to the surface telemetry system 602, which begins to send the separated neutron and gamma-ray signals from each detector at box 222 so that they can be further processed by the multi-channel analyzer (MCA) embedded in the surface processing system 109 and connected to the computer system device 608 via the communication bus 604. Once the MCA receives the above-mentioned signals, it will generate a histogram or "time spectrum" representing the number of events and the arrival time of the events at box 223, and then store it in the memory resource 603. The stored time spectrum can be displayed by the display 610 and analyzed by a person of ordinary skill in the art. Once the separated signals have been stored, the non-transitory computer readable memory device of the computer program product having the trained neural network receives a message hook from the MCA and begins generating a thermal neutron time spectrum and a capture gamma ray time spectrum for each of the three dual-function radiation detectors, wherein each thermal neutron time spectrum is substantially composed of a time decay curve of thermal neutrons and the capture gamma ray time spectrum is substantially composed of a time decay curve of capture gamma rays, both of which immediately follow the neutron pulse.
[0140] The non-transitory computer readable memory device with the trained neural network computer program product 303 then receives the message hook from the MCA and begins calculating a curve fit of the generated time spectrum at blocks 224 and 225, which fits the time decay curve of thermal neutrons and the time decay curve of captured gamma rays using two exponential decays. The "best fit" of the curve is determined by the non-transitory computer readable memory device with the trained neural network computer program product 303 in the same manner as previously when the neural network was not trained.
[0141] At block 226, the time decay curves are further processed using a non-transitory computer readable memory device having a computer program product 303 with a trained neural network to obtain a neutron-induced apparent borehole time decay constant and an apparent formation time decay constant for each of the at least three dual-function radiation detectors, assigning the following term to each of the at least three dual-function radiation detectors according to its location (i.e., proximal, intermediate, distal): bnn , τ bnm , τ bnf , τ fnn , τ fnm , τ fnf The captured gamma-ray time decay curve is further processed at block 226 to obtain a gamma-induced apparent borehole time decay constant and a gamma-induced apparent formation time decay constant for each of the at least three dual-function radiation detectors, and the following terms are assigned to each of the at least three dual-function radiation detectors according to its location (i.e., proximal, intermediate, distal): τ bgn , τ bgm , τ bgf , τ fgn , τ fgm , τ fgf Thereafter, once the input has been collected, the non-transitory computer readable memory device of the computer program product with the trained neural network 303 will begin processing. Thus, the first step of processing the input values using the non-transitory computer readable memory device of the computer program product with the trained neural network 303 involves inputting the set of input values obtained at block 226 into the non-transitory computer readable memory device of the computer program product with the trained neural network 303 at block 227. Figure 3The computer program product of the trained neural network then processes the neutron-induced apparent borehole time decay constant and apparent formation time decay constant obtained for each of the at least three dual-function radiation detectors in the first of the three second processing hidden layers 3033 at box 228 using a weighted nonlinear regression algorithm (Equations 8 and 9). Thereafter, the acquired time decay constants are further processed through the second of the three hidden layers at box 229 using a weighted nonlinear regression algorithm (Equations 10 and 11). Then, at box 230, the neutron-induced apparent borehole time decay constant and apparent formation time decay constant obtained for each of the at least three dual-function radiation detectors are processed using a weighted linear regression algorithm (Equation 12) in the third of the three second processing hidden layers. All time decay constants are processed through the three hidden layers. After successfully completing the processing of the three hidden layers, the non-transitory computer readable memory device of the computer program product 303 with the trained neural network will start to use the weighted linear regression algorithm (Equations 13 and 14) in the third output layer 3034 of the non-transitory computer readable memory device of the computer program product with the trained neural network to calculate the true borehole sigma 3331 and the true formation sigma 3332. Thereafter, the non-transitory computer readable memory device 605 of the computer program product 303 with the trained neural network will generate the true borehole sigma and the true formation sigma at box 232, and send a message to the memory resource 603 to start storing the true borehole sigma and the true formation sigma generated at box 232, and after successfully completing the storage process, it will send a signal to the non-transitory computer readable memory device of the computer program product 303 with the trained neural network to display the above two sigmas on the external display 610 of the computer system device 608. All of the above sigmas are assigned respective designations by the non-transitory computer readable memory device having a trained neural network 303, depending on the location of the probe in the BHA section (ie, proximal, intermediate, distal).
[0142] Thereafter, the memory resource 603 will receive a message hook from the non-transitory computer readable memory device 605 so that it can send information, data, constants, variables, and sigmas to the computer system device 608 for display on the display 610 and printing using the printer 613.
[0143] Figure 3Two neural networks are shown, a pre-programmed or blank neural network 302, and a trained neural network 303 where all weights are fixed. Each neural network shown in the figure has a single hidden layer 3023 or 3033 with multiple processors, but due to drawing limitations, the layers are actually deep neural network layers composed of at least three hidden layers. The input layers 3021 and 3031 include multiple processors 3210, 3310, 3211, 3311, 3212, 3312, ..., 3213, 3313, respectively. The inputs to the input layer may also include measurements such as wellbore dimensions, tool-to-wellbore wall spacing, formation mineralogy, wellbore fluid salinity, formation fluid salinity, wellbore fluid density, wellbore fluid temperature, formation temperature, wellbore fluid pressure, formation pore pressure, etc. The above list is not complete, as mentioned above, some of these measurements may be made during the drilling process, while other measurements may be a priori measurements based on knowledge of the subsurface or previously drilled wells. The output layers 3024 and 3034 show two independent processors for calculating the true borehole sigma 3231, 3331 and the true formation sigma 3232, 3332, respectively. In the example shown, the processors of the output layers 3024 and 3034 provide a weighted combination output of each output of the processors of the hidden layer. The output of each processor in the hidden layer is a weighted nonlinear or linear regression combination of each output of the processors 3210, 3310, 3211, 3311, 3212, 3312, ..., 3213, 3313 of the input layer, which can be a measured value or a function of a measured value.
[0144] Figure 4 How the pulse shape discrimination (PSD) technology pre-programmed in the non-transitory computer readable memory device of at least three dual-function radiation detectors can distinguish between neutrons and gamma rays is shown in the form of a graph 401. In particular, by coupling two scintillators composed of Cs2LiYCI6 (CLYC) and Cs2LiLaBr6 (CLLB) crystals that are sensitive to both neutrons and gamma rays with a scintillation photosensitive device such as a photomultiplier tube (PMT), at least three dual-function radiation detectors can effectively detect neutrons and gamma rays, respectively, as shown in curves 402 and 403, because the electronic signal of gamma rays decays faster than that of neutrons.
[0145] on the other hand, Figure 5A , Figure 5B and Figure 5CThe thermal neutron count rate decay curve, neutron intensity, and gamma ray count rate decay curve for the near dual-function radiation detector 502, the middle dual-function radiation detector 506, and the far dual-function radiation detector 510 are shown respectively. The neutron pulse frequency used is at least 500 Hz, and the pulse width is at least 20 μs, which is used for each of the three dual-function radiation detectors, namely the near detector 502, the middle detector 506, and the far detector 510. The thermal neutron count rate decay curve for each of the at least three dual-function radiation detectors is also shown (the near detector 503, the middle detector 507, and the far detector 511). The captured gamma ray count rate decay curve for each of the at least three dual-function radiation detectors is also shown (the near detector 505, the middle detector 509, and the far detector 513). After each neutron pulse (in the interval time between neutron pulses), the borehole attenuation is dominant in the early stage, while the formation attenuation is dominant in the later stage.
[0146] In fact, if Figure 6 As shown, the pre-programmed non-transitory computer readable memory device on the surface processing system is part of the computer system device 601, which is used to compile information from various application servers set in the surface processing system 109. As shown, the computer system device 601 on the surface processing system 109 generally includes a telemetry system 602, a memory resource 603 for storing data, a communication bus 604, a non-transitory computer readable memory device 605, and a computer system device 608 used as a user interface. The pre-programmed non-transitory computer readable memory device 605 (represented in the form of a functional block diagram) performs a series of operations and instructions of the computer-implemented method, which uses a pulsed neutron source and a plurality of dual-function radiation detectors of neutrons and gamma rays (pulse shape discrimination technology can be used to distinguish neutrons and gamma rays first) to measure thermal neutron time decay signals and thermal neutron capture gamma ray time decay signals, thereby obtaining wellbore sigma and formation sigma.
[0147] Memory resources 603 may include any of various forms of memory media and memory access devices. For example, memory devices 603 may include semiconductor RAM and ROM devices, as well as mass storage devices such as CD-ROM drives, disk drives, and tape drives.
[0148] The computer system device 608 serves as a user interface of the non-transitory computer-readable memory device 605 to input, set, select and execute the operations of retrieving, acquiring, calculating, generating, storing and repeating functions (collectively referred to as "message hook processes"). The computer system device 608 is also connected to the ground processing system 109, the memory resource 603 and the non-transitory computer-readable memory device 605 by wire and / or wirelessly through the communication device 604. The computer system device 608 also includes other devices, such as a central processing unit (CPU) 609, a display or monitor 610, a keyboard 611, a mouse 612 and a printer 613. One or more users can provide input to the pre-programmed non-transitory computer-readable memory device on the ground processing system through a set of input devices 611 or 612 of the computer system. However, those skilled in the art will soon realize that the input device can also include devices such as digitizing pads, trackballs, light pens, data gloves, eye direction sensors, head direction sensors, etc. Output devices 610 and 613 can also include devices such as projectors, head-mounted displays, plotters, etc.
[0149] In one embodiment of a preprogrammed non-transitory computer readable memory device on a surface processing system, it may include one or more communication devices (communication buses) 604, such as a network interface card for interfacing with a computer network. For example, seismic data collected at a remote site or underground may be transmitted to a preprogrammed non-transitory computer readable memory device on a surface processing system via a computer network using a telemetry system 602. The preprogrammed non-transitory computer readable memory device on the surface processing system may receive seismic data, coordinates, elements, source, and receiver information from an external computer network using a communication bus 604 (such as a network interface card) in addition to receiving data from a wired or LWD tool 106. In other embodiments, the preprogrammed non-transitory computer readable memory device on the surface processing system may include multiple computers and / or other devices coupled via a computer network, wherein the storage and / or implementation of the computations of the present embodiment may be distributed across the computers (and / or devices) as needed.
[0150] The pre-programmed non-transitory computer readable memory device on the ground processing system has firmware, kernel and software for providing connectivity and interoperability of multiple connected devices such as telemetry system 602, memory resources for storing data 603, communication bus 604, non-transitory computer readable device 605 and computer system device 608. The pre-programmed non-transitory computer readable memory device on the ground processing system includes an operating system, a set of pre-programmed message hook routines and system applications.
[0151] Furthermore, since performance and computational cost are always an important issue, the pre-programmed non-transitory computer readable memory device on the ground processing system uses the non-transitory computer readable device 605 to ensure that the steps of method 201 are not constrained by all of its I / O or any other network communications. In fact, a file distribution system like Apache Hadoop combined with appropriate data compression and intelligent file caching based on the data will ensure that the operations or instructions performed in method 201 are only limited by memory / cache speed and CPU / GPU computing power, and not other limitations.
[0152] The operating system embedded in the computer system device 601 and the preprogrammed non-transitory computer readable memory device 605 can be Microsoft's "WINDOWS" operating system, OS / 2 from IBM Corporation, UNIX, LINUX, Sun Microsystems, or Apple operating system, as well as countless embedded application operating systems (such as those available from Wind River, Inc.).
[0153] For example, the message hook process of the preprogrammed non-transitory computer-readable memory device 605 may represent an operation or command of the telemetry system 602 , the memory resource 603 , or the computer system device 608 , which may currently be executing a step process or subroutine of the computer-implemented method 201 .
[0154] The set of message hook processes may be initiated first by: (1) input from a user, who is a person of ordinary skill in the art, such as input of user-defined values or parameters; (2) operation of the computer system device 608; (3) processing operations in the non-transitory computer-readable memory device 605; or (4) automatically once specific data has been automatically stored or retrieved by the memory resource 603 or the non-transitory computer-readable memory device 605. Based on any of these input, process or operation events, the memory resource 603, the non-transitory computer-readable memory device 605 or the computer system device 608 generates a data packet transmitted via the communication bus 604, which indicates an event that has occurred and an event that needs to occur. When the memory resource 603, the non-transitory computer-readable memory device 605 or the computer system device 608 receives the data packet, it converts it into a message based on the event and executes the steps required by the computer-implemented method 201. This is achieved when the operating system checks the message hook list and determines whether any message hook process has registered itself with the operating system. If at least one message hook process has completed its own registration in the operating system, the operating system will pass the message to the first registered message hook process in the list via the communication bus 604. The called message hook executes and returns a value to the memory resource 603, the non-transitory computer readable memory device 605 or the computer system device 608, thereby instructing them to pass the message to the next registered message hook, and the memory resource 603, the non-transitory computer readable memory 605 and a pre-programmed non-transitory computer readable memory device 605 in the computer system device 608 continue to perform operations until all registered message hooks have been passed, indicating that the operations or instructions of the method 201 have been completed, and the final spatial variation wavelet set is generated and stored in the memory resource 603.
[0155] The non-transitory computer-readable memory device 605 is configured to read and execute program instructions, for example, program instructions provided on a storage medium, such as a set of one or more optical disks, and loaded into a semiconductor memory when executed. The non-transitory computer-readable memory device 605 can be connected to the memory resource 603 by wire or wirelessly via a communication bus 604 (or a set of buses). In response to the program instructions, the non-transitory computer-readable memory device 705 can operate on the data stored in the one or more memory resources 703. The non-transitory computer-readable memory device 603 may include one or more programmable processors (such as microprocessors).
[0156] Although the present disclosure has been described in conjunction with certain preferred embodiments thereof in the foregoing description, and many details have been set forth for illustrative purposes, the present disclosure is not overly limited to the foregoing contents set forth for illustrative purposes. On the contrary, various modifications and alternative embodiments will be apparent to those skilled in the art without departing from the true scope of the present disclosure, as defined by the claims set forth below. In addition, it should be understood that the structural features or method steps shown or described in any one embodiment herein may also be used in other embodiments.
[0157] Reference numerals list
[0158]
[0159]
Claims
1. A computer-implemented method using a non-transitory computer-readable memory device with a computer program product of a pre-programmed neural network and a trained neural network to perform operations of obtaining a true borehole sigma and a true formation sigma, wherein the pre-programmed neural network and the trained neural network each have three layers, the operations of the computer-implemented method comprising: Initializing a non-transitory computer readable memory device having a pre-programmed neural network in a ground processing system coupled to a memory resource, a ground telemetry system, a communication bus, a multi-channel analyzer, and a computing system device; Initializing a wireline or logging while drilling system for a nuclear logging tool, the system having a deuterium-tritium (DT) neutron generator or a deuterium-deuterium (DD) neutron generator neutron source coupled to at least three dual-function radiation detectors, wherein each dual-function radiation detector is pre-programmed via a non-transitory computer readable memory device and using a pulse shape discrimination technique, the system also having a high voltage supply source, an underground telemetry system, electronic instrumentation, and an underground non-transitory computer readable memory device; causing the deuterium-tritium (DT) neutron generator or the deuterium-deuterium (DD) neutron generator neutron source to pulse at least twice; measuring neutrons and neutron-induced gamma rays after each of the at least two pulses from each of at least three dual-function radiation detectors; separating the measured signals of thermal neutrons and neutron-induced gamma rays into thermal neutrons and neutron-induced capture gamma rays using at least three dual-function radiation detectors pre-programmed with pulse shape discrimination techniques via a non-transitory computer-readable memory device; transmitting the separated thermal neutron and neutron-induced capture gamma ray signals to a multi-channel analyzer using an underground telemetry system and using at least three dual-function radiation detectors pre-programmed by a non-transitory computer readable memory device; generating, using a preprogrammed neural network in the ground processing system, a thermal neutron time spectrum and a neutron induced capture gamma ray time spectrum for each of three dual-function radiation detectors, wherein the thermal neutron time spectrum consists essentially of thermal neutron time decay curves and the neutron induced capture gamma ray time spectrum consists essentially of neutron induced capture gamma ray time decay curves; Using a pre-programmed neural network in a ground processing system, a curve fit is calculated using two exponential decays immediately after a pulse from a deuterium-tritium (DT) neutron generator or a deuterium-deuterium (DD) neutron generator neutron source; using a pre-programmed neural network in a ground processing system to calculate a curve fit using two exponential decays of a neutron-induced capture gamma ray time decay curve immediately after a pulse from a deuterium-tritium (DT) neutron generator or a deuterium-deuterium (DD) neutron generator neutron source; deriving, using a preprogrammed neural network in a surface processing system, a neutron-induced apparent borehole time decay constant and an apparent formation time decay constant for each of at least three dual-function radiation detectors from a calculated curve fit of a thermal neutron time decay curve and a calculated curve fit of a neutron-induced captured gamma-ray time decay curve; training a first input layer of a non-transitory computer readable memory device having a pre-programmed neural network using the acquired neutron-induced apparent borehole time decay constant and apparent formation time decay constant as input; training three second processing hidden layers of a non-transitory computer readable memory device having a pre-programmed neural network using a weighted nonlinear regression algorithm and a weighted regression algorithm; training a third output layer of a non-transitory computer readable memory device having a pre-programmed neural network using a weighted linear regression algorithm to output an apparent wellbore sigma, training a third output layer of a non-transitory computer readable memory device having a pre-programmed neural network using a weighted regression algorithm to output an apparent formation sigma, and then comparing the outputs of the third layer to known values of true wellbore sigma and true formation sigma; Repeating the following operations: pulsing a deuterium-tritium (DT) neutron generator or a deuterium-deuterium (DD) neutron generator neutron source; measuring neutrons and neutron-induced gamma rays; separating the measured neutron signals and the measured neutron-induced gamma ray signals from each other; sending the separated thermal neutrons and neutron-induced capture gamma rays to a multi-channel analyzer; obtaining neutron-induced apparent borehole time decay constants and apparent formation time decay constants; training a first layer of a non-transitory computer readable memory device having a pre-programmed neural network; training three second processing hidden layers; and training a third layer until the relative difference between the output of the third layer and the known values of the true borehole sigma and the true formation sigma is less than 1%; generating a computer program product having a trained neural network with three layers using a non-transitory computer readable memory device having a preprogrammed neural network in a ground processing system; storing the generated computer program product of the trained neural network having three layers on a memory resource in a ground processing system; replacing a pre-programmed neural network in a non-transitory computer readable memory device in a ground processing system with a stored computer program product having a trained neural network with three layers; causing the deuterium-tritium (DT) neutron generator or the deuterium-deuterium (DD) neutron generator neutron source to pulse at least twice; measuring neutrons and neutron-induced gamma rays after each of the at least two pulses from each of at least three dual-function radiation detectors; separating the measured signals of thermal neutrons and neutron-induced gamma rays into thermal neutrons and neutron-induced capture gamma rays using at least three dual-function radiation detectors pre-programmed with pulse shape discrimination techniques via a non-transitory computer-readable memory device; transmitting the separated thermal neutrons and neutron-induced capture gamma rays to a multi-channel analyzer using an underground telemetry system and using at least three dual-function radiation detectors pre-programmed by a non-transitory computer readable memory device; generating, using a preprogrammed neural network in the ground processing system, a thermal neutron time spectrum and a neutron induced capture gamma ray time spectrum for each of three dual-function radiation detectors, wherein the thermal neutron time spectrum consists essentially of thermal neutron time decay curves and the neutron induced capture gamma ray time spectrum consists essentially of neutron induced capture gamma ray time decay curves; Using a non-transitory computer readable memory device having a computer program product with a trained neural network, calculating a curve fit using two exponential decays of a thermal neutron time decay curve immediately after a pulse from a deuterium-tritium (DT) neutron generator or a deuterium-deuterium (DD) neutron generator neutron source; Using a non-transitory computer readable memory device having a computer program product with a trained neural network, calculating a curve fit using two exponential decays of a neutron induced capture gamma ray time decay curve immediately after a pulse from a deuterium-tritium (DT) neutron generator or a deuterium-deuterium (DD) neutron generator neutron source; utilizing a non-transitory computer readable memory device having a computer program product having a trained neural network to derive a neutron-induced apparent borehole time decay constant and an apparent formation time decay constant for each of at least three dual-function radiation detectors from a calculated curve fit of a thermal neutron time decay curve and a calculated curve fit of a neutron-induced captured gamma-ray time decay curve; inputting the acquired neutron-induced apparent borehole time decay constant and apparent formation time decay constant for each of the at least three dual-function radiation detectors into a first layer of a non-transitory computer-readable memory device of a computer program product having a trained neural network; processing the acquired neutron-induced apparent borehole time decay constant and apparent formation time decay constant for each of the at least three dual-function radiation detectors in a first of three second processing hidden layers of a non-transitory computer-readable memory device of a computer program product having a trained neural network using a weighted nonlinear regression algorithm; processing the acquired neutron-induced apparent borehole time decay constant and apparent formation time decay constant for each of the at least three dual-function radiation detectors in a second of three second processing hidden layers of a non-transitory computer-readable memory device of a computer program product having a trained neural network using a weighted nonlinear regression algorithm; processing the acquired neutron-induced apparent borehole time decay constant and apparent formation time decay constant for each of the at least three dual-function radiation detectors in a third of three second processing hidden layers of a non-transitory computer-readable memory device of a computer program product having a trained neural network using a weighted nonlinear regression algorithm; In a third output layer of the non-transitory computer readable memory device of the computer program product having the trained neural network, using a weighted linear regression algorithm to calculate the true wellbore sigma and the true formation sigma; and A true wellbore sigma and a true formation sigma are generated using a weighted linear regression algorithm of the trained third output layer of a non-transitory computer readable memory device of a computer program product having a trained neural network.
2. The method according to claim 1, wherein: The obtained neutron-induced apparent borehole time decay constant for each of the at least three dual-function radiation detectors is processed in a first of three second processing hidden layers of a non-transitory computer-readable memory device of a computer program product having a trained neural network using a weighted nonlinear regression algorithm, further comprising the expression:
3. The method according to claim 1, wherein: The acquired apparent formation time decay constant for each of the at least three dual-function radiation detectors is processed in a first of three second processing hidden layers of a non-transitory computer readable memory device of a computer program product having a trained neural network using a weighted nonlinear regression algorithm, further comprising the expression:
4. The method according to claim 1, wherein: The obtained neutron-induced apparent borehole time decay constant for each of the at least three dual-function radiation detectors is processed in a second of three second processing hidden layers of a non-transitory computer-readable memory device of a computer program product having a trained neural network using a weighted nonlinear regression algorithm, further comprising the expression:
5. The method according to claim 1, wherein: The acquired apparent formation time decay constant for each of the at least three dual-function radiation detectors is processed in a second of three second processing hidden layers of a non-transitory computer readable memory device of a computer program product having a trained neural network using a weighted nonlinear regression algorithm, further comprising the expression:
6. The method according to claim 1, wherein: The acquired neutron-induced apparent borehole time decay constant and apparent formation time decay constant for each of the at least three dual-function radiation detectors are processed in a third of three second processing hidden layers of a non-transitory computer-readable memory device of a computer program product having a trained neural network using a weighted nonlinear regression algorithm, further comprising the expression:
7. The method according to claim 1, wherein: The method of generating the true wellbore sigma using a weighted linear regression algorithm of the trained third output layer of a non-transitory computer readable memory device of a computer program product having a trained neural network further comprising the expression:
8. The method according to claim 1, wherein: The method of generating the true formation sigma using a weighted linear regression algorithm of the trained third output layer of a non-transitory computer readable memory device of a computer program product having a trained neural network further comprising the expression:
Citation Information
Patent Citations
Methods for the preparation of organotin compounds
EP0002019A1
Neutron gamma density logging measuring device and method
CN111335886A
Formation sigma measurement from thermal neutron detection
EP0552073A2
Pulsed neutron decay tool for measuring gamma radiation energy spectra for fast neutron inelastic collisions and thermal neutron capture events
US5374823A
