A method and system for fault diagnosis of a neutron detector array

Abstract

This invention relates to the field of geophysical logging technology, specifically to a fault diagnosis method and system for neutron detector arrays. The method includes: acquiring the original count rate sequences and well temperature curves of near- and far-field detectors from neutron logging, and performing non-zero clamping, alignment, and statistical fluctuation smoothing. Based on the smoothed far-field detector data and well temperature, the theoretical count rate of the near-field detector is calculated using an array response prediction model with a temperature drift correction term. The prediction residual between the smoothed near-field detector data and the theoretical value is calculated, and the formation gradient value of the smoothed far-field detector data is calculated. A fault discrimination index is calculated by combining the prediction residual, formation gradient value, near-field detector data, and weighting coefficients. This index is compared with a threshold; if the continuity confirmation condition is met, the detector array is determined to be faulty, and the type is determined based on the well temperature. This invention, by introducing formation gradient weighting, can automatically identify formation interfaces, effectively suppressing false alarms due to resolution mismatch while maintaining sensitivity.

Description

Technical Field

[0001] This invention relates to the field of geophysical logging technology. More specifically, this invention relates to a fault diagnosis method and system for a neutron detector array. Background Technology

[0002] In oil exploration and development, pulsed neutron logging technology is applied to monitor residual oil saturation, identify lithology, and perform elemental analysis in cased wells. These instruments typically consist of a neutron generator and multiple axially distributed near-field, far-field, and ultra-far-field detectors, forming a linear detector array. Because logging instruments operate in extreme downhole environments, they face multiple challenges, including high-temperature and high-pressure conditions, high-flux radiation damage, and complex formation environments, which can easily lead to detector performance degradation or malfunction.

[0003] To address the issues of detector performance degradation or malfunction, existing fault diagnosis methods primarily rely on simple threshold discrimination or multi-probe ratio methods. Threshold discrimination mainly monitors the instrument's status by determining whether the count rate exceeds a certain fixed range. Multi-probe ratio methods, on the other hand, use the ratio of near-field and far-field detectors for verification, attempting to identify potential problems by analyzing the data relationships between detectors at different locations, thereby assisting well logging operations to some extent.

[0004] However, the aforementioned existing technologies have significant shortcomings in practical applications. First, for the threshold discrimination method, when the instrument passes through gas layers or tight layers, drastic changes in the normal count rate can easily trigger false alarms. Second, for the multi-probe ratio method, due to the different source distances of the near and far detectors, their detection depth and vertical resolution differ significantly. When the instrument passes through the formation interface, i.e., at abrupt lithological changes, the near detector responds first and changes rapidly, while the far detector responds later and changes slowly, resulting in a severe data mismatch at the interface. This mismatch makes it impossible for existing technologies to effectively distinguish between data fluctuations caused by formation changes and data anomalies caused by instrument malfunctions, thus easily generating a large number of false fault alarms at the formation interface, seriously affecting the efficiency of logging operations and the accuracy of data interpretation. Summary of the Invention

[0005] The purpose of this invention is to propose a fault diagnosis method and system for neutron detector arrays, in order to solve the technical problems of existing technologies that are prone to false alarms at the formation interface and cannot accurately distinguish between temperature drift and real faults; to this end, this invention provides solutions in the following two aspects.

[0006] In a first aspect, the fault diagnosis method for a neutron detector array provided by the present invention includes: acquiring the raw count rate sequence of the near detector, the raw count rate sequence of the far detector, and the downhole temperature curve uploaded by a neutron logging instrument; performing non-zero clamping, depth alignment correction, and statistical fluctuation smoothing on the raw count rate sequences of the near detector and the far detector to obtain smoothed near detector data and far detector data; calculating the theoretical predicted count rate of the near detector at the current depth point using an array response prediction model including a temperature drift correction term based on the smoothed far detector data and the downhole temperature curve; calculating the prediction residual between the smoothed near detector data and the theoretical predicted count rate, and calculating the formation gradient value of the smoothed far detector data; calculating a fault discrimination index based on the prediction residual, the smoothed near detector data, the formation gradient value, and a gradient suppression weight coefficient; comparing the fault discrimination index with a preset fault alarm threshold; if the fault discrimination index is greater than or equal to the fault alarm threshold and satisfies a preset continuity confirmation condition, then determining that the neutron detector array has a fault, and determining the fault type according to the downhole temperature curve.

[0007] The beneficial effects of this invention are as follows: By using the above technical solution, the array response prediction model including temperature drift correction is established by combining the relatively stable remote detector data with well temperature data. The theoretical value of the near detector under normal conditions is calculated, and the fault discrimination index is calculated. This index is weighted by introducing formation gradient values, which can automatically identify gradient changes generated by the formation interface. Thus, while ensuring the sensitivity of fault detection, it effectively suppresses false alarms caused by resolution mismatch due to the formation interface.

[0008] Preferably, the non-zero clamping, depth alignment correction, and statistical fluctuation smoothing of the raw count rate sequences of the near detector and the far detector include: detecting the raw count rate; if the count rate is zero or negative, forcing it to a preset positive integer; using the far detector as a reference, performing depth translation on the raw count rate sequence of the near detector according to the physical source distance difference between the near and far detectors so that they correspond to the same measurement depth; and using a moving average filter to denoise the aligned data.

[0009] The beneficial effects of this invention are: non-zero clamping avoids division by zero or square root errors in subsequent mathematical operations; depth alignment eliminates depth lag caused by different source distances; and smoothing reduces the interference of nuclear signal statistical fluctuations on fault diagnosis, providing a high-quality data foundation for subsequent model calculations.

[0010] Preferably, the array response prediction model satisfies the expression:

[0011] In the formula, For depth Theoretical predicted count rate of the proximity detector; For depth Smoothed remote detector data; 、 、 for Formation response coupling coefficient; For reference temperature; This is the temperature drift suppression constant; For depth The downhole temperature at that location.

[0012] The beneficial effects of this invention are: by constructing a predictive model that includes formation response and temperature drift characteristics, dynamic correlation of data between detectors is realized, which can decouple the influence of ambient temperature and formation changes on the count rate from a physical mechanism perspective, and provide an accurate dynamic benchmark for fault determination.

[0013] Preferably, the formation gradient value satisfies the expression:

[0014] In the formula, For depth The stratigraphic gradient value at that location, and These are smoothed remote sensor data from the current depth point and the previous depth point, respectively. The sampling depth interval.

[0015] The beneficial effects of this invention are: by calculating the formation gradient value, it is possible to quantify the degree of data change at the current depth point relative to the previous depth point, accurately characterize the existence of the formation interface, and provide key parameters for the subsequent introduction of gradient suppression factors in the fault discrimination index.

[0016] Preferably, the fault discrimination index satisfies the expression:

[0017] In the formula, For depth Fault discrimination index at the location; For depth Smoothed near-detector data, To predict the count rate theoretically; To predict residuals; This is the normalized term for Poisson noise; 1 represents the gradient suppression weight coefficient; 1.0 represents the anomaly prevention number. For depth The stratigraphic gradient value at that location.

[0018] The beneficial effect of this invention is that by introducing a gradient term into the denominator... ， At the formation interface, the gradient value increases, which increases the denominator and thus reduces the discrimination exponent, avoiding false high residual alarms caused by resolution mismatch. In homogeneous formations, the gradient term approaches zero, and the algorithm degenerates into a highly sensitive residual detection, ensuring the ability to capture minor faults. At the same time, the introduction of a constant term ensures the mathematical robustness of the algorithm under extreme data.

[0019] Preferably, the gradient suppression weight coefficient ranges from 2.0 to 5.0.

[0020] Preferably, satisfying the preset continuity confirmation condition includes: determining whether the fault discrimination index is within the continuous range. All depth points are greater than or equal to the fault alarm threshold; if so, a fault alarm is confirmed to be triggered; wherein, This refers to the number of sampling points included at a depth distance of at least 0.5 meters.

[0021] Preferably, determining the fault type based on the downhole temperature curve includes: when a fault is determined to exist, if the current downhole temperature exceeds a preset high temperature threshold, marking the fault type as a high temperature thermal decay warning; when a fault is determined to exist, if the current downhole temperature does not exceed the preset high temperature threshold, marking the fault type as a detector response inconsistency fault.

[0022] Preferably, the curve formed by the theoretically predicted count rate and the curve formed by the smoothed near-detector data are displayed on the same screen; the fault discrimination index is plotted as an independent quality control channel for output.

[0023] In the second aspect, the fault diagnosis system for the neutron detector array includes:

[0024] processor;

[0025] The memory stores computer instructions for fault diagnosis of the neutron detector array, which, when executed by the processor, cause the system to perform the aforementioned fault diagnosis method for the neutron detector array.

[0026] The beneficial effects of this invention are as follows:

[0027] 1. Significantly reduced the false alarm rate at the formation interface; by innovatively adding a gradient suppression term to the denominator of the discrimination formula, the false residuals caused by the inconsistency between near and far detector resolutions at abrupt changes in formation are automatically offset by mathematical logic, thus solving the problem of "interface false alarms" that has plagued the industry.

[0028] 2. Dynamic decoupling is achieved in high-temperature deep well environments; the solution integrates a temperature drift correction model based on physical laws, enabling the algorithm to automatically distinguish between the physical temperature drift of the detector and actual circuit faults, which is particularly suitable for operational support in high-temperature environments. Attached Figure Description

[0029] Figure 1 The flowchart illustrating the steps of the fault diagnosis method for the neutron detector array in this embodiment is shown in the schematic diagram.

[0030] Figure 2 This schematically illustrates a graph of depth alignment data for the array detectors.

[0031] Figure 3 The diagram illustrates a comparison of the fault detection effects of existing technologies and the present invention.

[0032] Figure 4 The diagram illustrates the deviation analysis between the dynamic benchmark prediction and the measured values. Detailed Implementation

[0033] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0034] like Figure 1 As shown, the fault diagnosis method for the neutron detector array in this embodiment includes the following steps:

[0035] Step S1: Obtain the near-detector raw count rate sequence, far-detector raw count rate sequence, and downhole temperature curve uploaded by the neutron logging instrument. Perform non-zero clamping, depth alignment correction, and statistical fluctuation smoothing on the near-detector raw count rate sequence and the far-detector raw count rate sequence to obtain smoothed near-detector data and far-detector data.

[0036] In an optional embodiment, the non-zero clamping, depth alignment correction, and statistical fluctuation smoothing of the raw count rate sequences of the near detector and the far detector include:

[0037] The raw count rate is detected. If the raw count rate is zero or negative, it is forced to a preset positive integer. Using the far detector as a reference, the raw count rate sequence of the near detector is depth-shifted according to the physical source distance difference between the near and far detectors so that they correspond to the same measurement depth. A moving average filter is used to denoise the aligned data.

[0038] Specifically, firstly, the total count rate sequence of the near detector, the total count rate sequence of the far detector, and the internal temperature curve of the downhole instrument are extracted from the remote transmission sub-decoding data of the ground system.

[0039] Then, a data validity pre-check is performed, and "non-zero clamping" is executed. Specifically, if the collected count rate is 0 or negative, it is set to a very small positive integer of 1 CPS to ensure the validity of subsequent mathematical operations.

[0040] Subsequently, depth alignment correction is performed. Due to the physical distance between the near-field and far-field detectors, they may measure different depths of strata at the same time. Specifically, using the far-field detector as a reference, the near-field detector data is depth-shifted, resulting in the aligned near-field detector data:

[0041] ；

[0042] in, The data is the aligned near-detector data; To obtain the total count rate of the near detector; This represents the current logging depth. This represents the physical source distance difference between the near-detector and the far-detector.

[0043] For example, such as Figure 2 As shown, Figure 2 This shows the raw count rates of the far and near detectors after depth alignment. From Figure 2 As can be seen, an abnormal dip appears in the near-detector data within the 3035-3045 meter range, representing a simulated "detector sensitivity drift" soft fault. Finally, statistical fluctuation smoothing is performed, and a 5-point moving average filter is used to denoise the aligned data, resulting in smoothed near-detector data. and remote detector data .

[0044] In this way, by using non-zero clamping, depth alignment, and smoothing, singularities, depth lag, and statistical noise in the data are eliminated, ensuring the quality and spatiotemporal consistency of the input data and providing a reliable foundation for subsequent correlation analysis.

[0045] Step S2: Based on the far-field detector data and the downhole temperature curve, the theoretical predicted count rate of the near-field detector at the current depth point is calculated using an array response prediction model that includes a temperature drift correction term.

[0046] In an optional embodiment, using the relatively stable far-field detector and well temperature data, the theoretically expected count rate of the near-field detector under normal operating conditions is calculated. A dynamic benchmark prediction model is then constructed based on physical laws, and the array response prediction model satisfies the expression:

[0047] ;

[0048] in, For depth Theoretical predicted count rate of the proximity detector; For depth Smoothed remote detector data; , , The formation response coupling coefficient is obtained by nonlinear least squares regression fitting of a known standard logging curve. For reference temperature, e.g., room temperature 25℃; This is the temperature drift suppression constant, obtained through high-temperature laboratory tests. For depth The downhole temperature at that location.

[0049] For example, the calculated depth under certain temperature and geological conditions. The theoretical predicted count rate of the proximity detector is 2696.4 CPS, indicating that under these temperature and formation conditions, if the proximity detector is functioning correctly, its count rate should be 2696.4 CPS. If the measured value deviates significantly from this value, a malfunction may exist.

[0050] Thus, by constructing a predictive model that incorporates formation response and temperature drift characteristics, dynamic correlation of data between detectors is achieved. This decouples the influence of ambient temperature and formation changes on the count rate from a physical mechanism perspective, providing an accurate dynamic benchmark for fault diagnosis.

[0051] Step S3: Calculate the prediction residual between the near detector data and the theoretical predicted count rate, and calculate the formation gradient value of the far detector data. Based on the prediction residual, the near detector data, the formation gradient value, and the gradient suppression weight coefficient, calculate the fault discrimination index.

[0052] Specifically, the first step is to calculate the formation gradient value, which is used to characterize the homogeneity of the current formation.

[0053] The formation gradient value satisfies the formula:

[0054] ;

[0055] In the formula, For depth The stratigraphic gradient value at that location, and These are smoothed remote sensor data from the current depth point and the previous depth point, respectively. The sampling depth interval is denoted as 0.1 meters.

[0056] Then, a fault discrimination index is constructed, satisfying the formula:

[0057] ;

[0058] In the formula, It is a dimensionless fault discrimination index; For depth Smoothed near-detector data, To predict the count rate theoretically; To predict residuals; This is the normalized term for Poisson noise; The gradient suppression weight coefficient is typically set between 2.0 and 5.0; 1.0 is the outlier prevention coefficient, used to prevent the calculation results from diverging when the denominator is close to zero, thus ensuring numerical stability. For depth The stratigraphic gradient value at that location.

[0059] When the strata are uniform, the changes in the remote detector are very small, and the corresponding fault discrimination index value is very small, far below the alarm threshold, it is judged as normal.

[0060] When the instrument passes through the formation interface, the different source distances cause the near detector to respond quickly and the far detector to respond slowly, resulting in a high residual between the near and far detectors. At this time, the data of the far detector will also change drastically. Although the residual between the near and far detectors is large, the existence of the gradient term results in a low discrimination index, which will not trigger an alarm, thus avoiding false alarms.

[0061] When a soft fault occurs in a homogeneous layer, the corresponding fault discrimination index value is very large, far exceeding the threshold, and the system will accurately issue an alarm.

[0062] Thus, by introducing a gradient suppression factor and anomaly prevention, the index maintains high sensitivity in homogeneous formations while automatically reducing sensitivity at formation interfaces. This cleverly solves the problem of false alarms at formation interfaces using mathematical logic, while ensuring the robustness of the algorithm.

[0063] Step S4: Compare the fault discrimination index with the preset fault alarm threshold. If the fault discrimination index is greater than or equal to the fault alarm threshold and the preset continuity confirmation condition is met, then it is determined that there is a fault in the neutron detector array, and the fault type is determined according to the downhole temperature curve.

[0064] For example, the fault alarm threshold is set to 6.0. If the fault discrimination index is less than the fault alarm threshold, it is determined to be normal; if the fault discrimination index is greater than or equal to the fault alarm threshold, the determination logic is entered.

[0065] In an optional embodiment, satisfying the preset continuity confirmation condition includes:

[0066] Determine whether the fault discrimination index is in a continuous state. The fault alarm threshold is greater than or equal to the fault alarm threshold at each depth point;

[0067] If so, then confirm that a fault alarm has been triggered; among which, This refers to the number of sampling points included at a depth distance of at least 0.5 meters.

[0068] This prevents single-point spike interference caused by cosmic rays or sudden electronic noise.

[0069] like Figure 3 As shown, in the prior art, a spike false alarm occurs at the formation interface at 3015 meters, while the curve of the present invention remains stable at this point. Obviously, the present invention effectively eliminates interface false alarms by introducing a gradient suppression factor. In the real fault zone of 3035-3045 meters, the curve of the present invention accurately breaks through the alarm threshold, achieving precise capture of real faults.

[0070] In this way, by threshold determination, continuity confirmation and multi-parameter joint analysis, the final qualitative and classification of instrument faults can be achieved, effectively eliminating random noise interference and providing operators with intuitive fault type indications and visual basis.

[0071] In addition, such as Figure 4 As shown, Figure 4 The diagram illustrates the comparison between the measured values ​​of the near-detector and the "dynamic benchmark prediction values" calculated by the algorithm of this invention. In the normal well section of 3000-3035 meters, the predicted values ​​and measured values ​​highly overlap, indicating that the model accurately learned the formation characteristics. In the fault zone of 3035-3045 meters, the measured values ​​are significantly lower than the predicted values, forming a clear red-filled area between them, intuitively revealing the existence and magnitude of the fault.

[0072] In an optional embodiment, determining the fault type based on the downhole temperature profile includes:

[0073] When a fault is detected, if the current downhole temperature exceeds the preset high temperature threshold, the fault type is marked as a high temperature thermal decay warning.

[0074] When a fault is detected, if the current downhole temperature does not exceed the preset high temperature threshold, the fault type is marked as a detector response inconsistency fault.

[0075] In an optional embodiment, the curve formed by the theoretically predicted count rate and the curve formed by the smoothed near-detector data are displayed on the same screen; the fault discrimination index is plotted as an independent quality control channel for output.

[0076] The present invention also provides a fault diagnosis system for a neutron detector array. The system includes a processor and a memory, the memory storing computer program instructions. When the processor executes the computer program instructions, it implements the fault diagnosis method for a neutron detector array according to the present invention.

[0077] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0078] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented by computer-readable / executable instructions stored or otherwise maintained on such a computer-readable medium.

[0079] In the description of this specification, "multiple" means at least two, such as two, three or more, etc., unless otherwise expressly and specifically defined.

[0080] While various embodiments of the invention have been shown and described in this specification, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention.

Claims

1. A fault diagnosis method for a neutron detector array, characterized in that, include: The raw count rate sequences of the near detector, the raw count rate sequences of the far detector, and the downhole temperature curve uploaded by the neutron logging instrument are acquired. Non-zero clamping, depth alignment correction, and statistical fluctuation smoothing are performed on the raw count rate sequences of the near detector and the far detector to obtain smoothed near detector data and far detector data. Based on the smoothed far-field detector data and downhole temperature curve, the theoretical predicted count rate of the near-field detector at the current depth is calculated using an array response prediction model that includes a temperature drift correction term; the array response prediction model satisfies the expression: ； In the formula, For depth Theoretical predicted count rate of the proximity detector; For depth Smoothed remote detector data; , , The formation response coupling coefficient; For reference temperature; This is the temperature drift suppression constant; For depth Downhole temperature at the location; Calculate the prediction residual between the smoothed near-detector data and the theoretical predicted count rate, and calculate the formation gradient value of the smoothed far-detector data. Based on the prediction residual, the smoothed near-detector data, the formation gradient value, and the gradient suppression weighting coefficient, calculate the fault discrimination index; the formation gradient value satisfies the expression: ； In the formula, For depth The stratigraphic gradient value at that location, and These are smoothed remote sensor data from the current depth point and the previous depth point, respectively. The sampling depth interval; The fault discrimination index is compared with a preset fault alarm threshold. If the fault discrimination index is greater than or equal to the fault alarm threshold and the preset continuity confirmation condition is met, the neutron detector array is determined to be faulty, and the fault type is determined according to the downhole temperature curve.

2. The fault diagnosis method for a neutron detector array according to claim 1, characterized in that, The non-zero clamping, depth alignment correction, and statistical fluctuation smoothing processing of the raw count rate sequences of the near detector and the far detector include: Detect the original count rate; if the original count rate is zero or negative, set it to a preset positive integer. Using the far detector as a reference, the original count rate sequence of the near detector is depth-shifted according to the physical source distance difference between the near detector and the far detector so that the two correspond to the same measurement depth. A moving average filter is used to denoise the aligned data.

3. The fault diagnosis method for a neutron detector array according to claim 1, characterized in that, The fault discrimination index satisfies the expression: ； In the formula, For depth Fault discrimination index at the location; For depth Smoothed near-detector data, To predict the count rate theoretically; To predict residuals; This is the normalized term for Poisson noise; 1 represents the gradient suppression weight coefficient; 1.0 represents the anomaly prevention number. For depth The stratigraphic gradient value at that location.

4. The fault diagnosis method for a neutron detector array according to claim 3, characterized in that, The gradient suppression weight coefficient ranges from 2.0 to 5.

0.

5. The fault diagnosis method for a neutron detector array according to claim 1, characterized in that, The conditions for satisfying the preset continuity confirmation include: Determine whether the fault discrimination index is in a continuous state. The fault alarm threshold is greater than or equal to the fault alarm threshold at each depth point; If so, then confirm that a fault alarm has been triggered; among which, This refers to the number of sampling points included at a depth distance of at least 0.5 meters.

6. The fault diagnosis method for a neutron detector array according to claim 1, characterized in that, The process of determining the fault type based on the downhole temperature curve includes: When a fault is detected, if the current downhole temperature exceeds the preset high temperature threshold, the fault type is marked as a high temperature thermal decay warning. When a fault is detected, if the current downhole temperature does not exceed the preset high temperature threshold, the fault type is marked as a detector response inconsistency fault.

7. The fault diagnosis method for a neutron detector array according to claim 1, characterized in that, Display the curve formed by the theoretically predicted count rate and the curve formed by the smoothed near-detector data on the same screen; The fault discrimination index is plotted as an independent quality control channel and output.

8. A fault diagnosis system for a neutron detector array, characterized in that, include: processor; A memory storing computer instructions for fault diagnosis of a neutron detector array, which, when executed by the processor, cause the system to perform the fault diagnosis method for a neutron detector array according to any one of claims 1-7.

Images