CN-121980966-A - Logging instrument drilling fluid outlet flow monitoring system and method based on machine vision

CN121980966ACN 121980966 ACN121980966 ACN 121980966ACN-121980966-A

Abstract

The invention discloses a system and a method for monitoring the flow rate of a drilling fluid outlet of a logging instrument based on machine vision, belonging to the technical field of drilling fluid detection, and comprising the following steps of constructing a graph structure; the method comprises the steps of data acquisition, graph convolution processing, feature extraction, visual mode visibility probability calculation, flow mode probability calculation, pressure mode probability calculation, comprehensive flow break probability determination based on multi-mode probability fusion, and risk level determination and hierarchical response mechanism based on the comprehensive flow break probability. By integrating three heterogeneous modes of vision, flow and pressure, the invention solves the problem that the reliability of the prior art still cannot meet the real-time decision requirement under complex working conditions, improves the system reliability under severe working conditions, and realizes the technical effects of reducing false alarm and missing report rate.

Inventors

HUANG YUQI
HUANG GE
ZENG ZHUO
RAO YINGBO
XIAO QINGHUA
LIU JIYUAN
Mei Zilai
SU CHAO
ZHU YUPING
LIANG LEI

Assignees

四川恒溢石油技术服务有限公司
四川越盛能源集团有限公司

Dates

Publication Date: 20260505
Application Date: 20260403

Claims (10)

1. The method for monitoring the drilling fluid outlet flow of the logging instrument based on machine vision is characterized by comprising the following steps of: constructing a graph structure; Collecting data; Performing graph convolution processing; Extracting features; Calculating the visibility probability of the visual mode; Calculating the flow modal probability; Calculating pressure modal probability; Determining a comprehensive cut-off probability based on the multi-modal probability fusion; the risk level and hierarchical response mechanism are determined based on the integrated outage probability.
2. The method for monitoring the flow rate of the drilling fluid outlet of the logging instrument based on machine vision according to claim 1, wherein the graph convolution processing mode is as follows: ; Wherein, the Is the first A layer feature matrix is provided which is a layer feature matrix, In order to activate the function, To the-1/2 power of the degree matrix, In the form of a self-loop adjacency matrix, Is the first A layer feature matrix is provided which is a layer feature matrix, Is the first A layer weight matrix.
3. The machine vision based logging tool drilling fluid outlet flow monitoring method of claim 2, wherein the feature extraction comprises: Extracting global features; Extracting local features; The global feature extraction method comprises the following steps: ; Wherein, the As a global feature vector of the object, As a total number of nodes, Is a node In the first place A feature vector obtained by rolling the layer diagram; The local feature extraction method comprises the following steps: ; Wherein, the Is a subset of the key nodes.
4. The method for monitoring the flow rate of the drilling fluid outlet of the logging instrument based on machine vision as claimed in claim 3, wherein the mode of calculating the visibility probability is as follows: ; Wherein, the In order to be a probability of visibility, The function is activated for Sigmoid, As a matrix of the global weights, As a global feature vector of the object, As a matrix of local weights, Is a local feature vector.
5. The method for monitoring the flow rate of the drilling fluid outlet of the logging instrument based on machine vision according to claim 4, wherein the flow rate modal probability is calculated in the following manner: ; Wherein, the As a probability of a flow anomaly, The function is activated for Sigmoid, As a parameter of the weight-bearing element, As a parameter of the bias it is possible, Is a normalized flow deviation; The calculation mode of the standardized flow deviation is as follows: ; Wherein, the For the current moment of time of the flow value, For the average flow value over the past 5 moments, In order to design the flow rate, The pressure modal probability calculation method comprises the following steps: ; Wherein, the As a probability of a pressure anomaly, The function is activated for Sigmoid, As a parameter of the weight-bearing element, As a parameter of the bias it is possible, Is a normalized differential pressure; The normalized differential pressure The calculation mode of (a) is as follows: ; Wherein, the In order to pump out the pressure of the pump, For the inlet pressure of the pipe, Is the average value at the time of the history of normal pressure differences, Is the standard deviation of the historical differential pressure.
6. The machine vision based logging tool drilling fluid outlet flow monitoring method of claim 5, wherein the determining the integrated outage probability based on the multi-modal probability fusion comprises: determining a visual disturbance intensity based on the image blur; determining a flow disturbance intensity based on the vibration value; Determining a pressure disturbance intensity based on the fluctuation rate; Determining dynamic weights of all modes based on the interference intensities of all modes; the comprehensive outage probability is determined based on the visual modality visibility probability, the traffic modality probability, the pressure modality probability, and the modality weights.
7. The method for monitoring the flow rate of the drilling fluid outlet of the logging instrument based on machine vision according to claim 6, wherein the visual disturbance intensity is calculated by the following steps: ; Wherein, the In order to provide the intensity of the visual disturbance, For the two norms of the image gradient, As the light intensity coefficient of the light, Is a gradient normalization constant; The flow interference intensity is calculated by the following steps: ; Wherein, the In order to vibrate the RMS value of the wave, Is a vibration threshold; the calculation mode of the pressure interference intensity is as follows: ; Wherein, the In order to be able to adapt the intensity of the pressure disturbance, In order to provide a pump stroke rate, Is the volatility threshold.
8. The method for monitoring the flow rate of the drilling fluid outlet of the logging instrument based on machine vision as claimed in claim 7, wherein the calculation mode of the dynamic weights of all modes is as follows: ; Wherein, the Is of a mode shape Is used for the normalization of the weight coefficient of (c), Is of a mode shape Is used for the sensitivity coefficient of the (c), Is of a mode shape Is used for the interference strength of the (c) signal, To sum variables; the calculation mode of the comprehensive flow breaking probability is as follows: ; Wherein, the In order to integrate the probability of a cut-off, Is of a mode shape Is used for the normalization of the weight coefficient of (c), Is of a mode shape Is an abnormal probability of (2); ; Wherein, the In order to be a probability of visual anomalies, Is a visibility probability.
9. The machine vision based logging tool drilling fluid outlet flow monitoring method of claim 1, wherein the risk level and hierarchical response mechanism can be confirmed by comparing the integrated outage probability with a risk threshold; The risk level is determined to be normal when the comprehensive cut-off probability is smaller than the early warning threshold, the risk level is determined to be early warning when the comprehensive cut-off probability is larger than or equal to the early warning threshold and smaller than the alarm threshold, the risk level is determined to be alarm when the comprehensive cut-off probability is larger than or equal to the alarm threshold and smaller than the emergency threshold, and the risk level is determined to be emergency when the comprehensive cut-off probability is larger than or equal to the emergency threshold.
10. A machine vision based tool drilling fluid outlet flow monitoring system for performing the machine vision based tool drilling fluid outlet flow monitoring method of any one of claims 1-9, comprising: The building module is used for building a graph structure; the acquisition module is used for acquiring data; The processing module is used for graph convolution processing; The extraction module is used for extracting the characteristics; the first calculation module is used for calculating the visibility probability of the visual mode; the second calculation module is used for calculating the traffic modal probability; the third calculation module is used for calculating the pressure modal probability; the first determining module is used for determining the comprehensive flow breaking probability; and the second determining module is used for determining the risk level and the hierarchical response mechanism.

Description

Logging instrument drilling fluid outlet flow monitoring system and method based on machine vision Technical Field The invention relates to the technical field of drilling fluid detection, in particular to a system and a method for monitoring the flow of a drilling fluid outlet of a logging instrument based on machine vision. Background The drilling fluid circulation system is a core component of petroleum drilling engineering, and the stable operation of the drilling fluid circulation system is directly related to drilling safety and operation efficiency. Drilling fluid is pumped to the bottom of a well through a drill rod by high-pressure pumping, and returns to the ground from an annulus after carrying rock debris, so that a complete circulation loop is formed. In the process, the real-time monitoring of the outlet flow of the drilling fluid plays a key role in identifying abnormal working conditions such as kick, lost circulation, gas invasion and the like, and the sudden flow drop or flow break often indicates serious underground accident risks. Traditional drilling fluid flow monitoring mainly relies on contact sensors such as electromagnetic flowmeters and ultrasonic flowmeters, and flow quantification is achieved through single-point measurement. However, the drilling site has severe environment, high temperature, high pressure, strong vibration and strong electromagnetic interference easily cause sensor drift or failure, in addition, the drilling fluid has high solid content and mixed bubbles, which often causes fluctuation or even blockage of the readings of the flowmeter, and the long-term stability is difficult to ensure. On the other hand, the camera is used for observing the returning state of the drilling fluid, so that the visual characteristics of liquid level, color, bubbles and the like are obtained in a non-contact mode, and the defects of the traditional sensor are effectively overcome. However, visual monitoring is easily interfered by environmental factors such as rain and fog, steam, oil stains and the like, and the reliability is still difficult to meet the real-time decision requirement under complex working conditions. Disclosure of Invention In order to solve the above problems, the first aspect of the present invention provides a method for monitoring the flow rate of a drilling fluid outlet of a logging tool based on machine vision, comprising the following steps: constructing a graph structure; Collecting data; Performing graph convolution processing; Extracting features; Calculating the visibility probability of the visual mode; Calculating the flow modal probability; Calculating pressure modal probability; Determining a comprehensive cut-off probability based on the multi-modal probability fusion; the risk level and hierarchical response mechanism are determined based on the integrated outage probability. Preferably, the graph convolution processing mode is as follows: ; Wherein, the Is the firstA layer feature matrix is provided which is a layer feature matrix,In order to activate the function,To the-1/2 power of the degree matrix,In the form of a self-loop adjacency matrix,Is the firstA layer feature matrix is provided which is a layer feature matrix,Is the firstA layer weight matrix. Preferably, the feature extraction includes: Extracting global features; Extracting local features; The global feature extraction method comprises the following steps: ; Wherein, the As a global feature vector of the object,As a total number of nodes,Is a nodeIn the first placeA feature vector obtained by rolling the layer diagram; The local feature extraction method comprises the following steps: ; Wherein, the Is a subset of the key nodes. Preferably, the means for calculating the visibility probability is: ; Wherein, the In order to be a probability of visibility,The function is activated for Sigmoid,As a matrix of the global weights,As a global feature vector of the object,As a matrix of local weights,Is a local feature vector. Preferably, the flow modal probability is calculated by: ; Wherein, the As a probability of a flow anomaly,The function is activated for Sigmoid,As a parameter of the weight-bearing element,As a parameter of the bias it is possible,Is a normalized flow deviation; The calculation mode of the standardized flow deviation is as follows: ; Wherein, the For the current moment of time of the flow value,For the average flow value over the past 5 moments,In order to design the flow rate, The pressure modal probability calculation method comprises the following steps: ; Wherein, the As a probability of a pressure anomaly,The function is activated for Sigmoid,As a parameter of the weight-bearing element,As a parameter of the bias it is possible,Is a normalized differential pressure; The normalized differential pressure The calculation mode of (a) is as follows: ; Wherein, the In order to pump out the pressure of the pump,For the inlet pressure of the pipe,Is the avera