CN-122017282-A - Training data acquisition device of AI multiphase flow detection system and application method thereof

CN122017282ACN 122017282 ACN122017282 ACN 122017282ACN-122017282-A

Abstract

The invention provides a training data acquisition device of an AI multiphase flow detection system, which comprises an integral frame for supporting and fixing, a transparent container, a motion module for controlling the motion of a rotary drum assembly, the rotary drum assembly for simulating disperse phases with different speeds, an image acquisition module for acquiring multiphase flow image information and a data processing module for identifying and processing the disperse phases in the multiphase flow image information. According to the invention, different radiuses of the rotating drum are matched with different rotating speeds of the motor, the motion speed of the disperse phase can be simulated according to any rotating speed, the disperse phase is simulated to move in the horizontal direction and the vertical direction simultaneously, different liquids can be filled in the transparent container for measurement and correction, the influence of different physical and chemical properties of the liquids on a measurement result is avoided, and the identification result obtained through detection can be marked and stored as a data set for training of an AI intelligent multiphase flow detection system.

Inventors

LI MINGHAI
ZHANG YANAN
YANG SHUNLIN
LIN JUN

Assignees

南京师范大学

Dates

Publication Date: 20260512
Application Date: 20260120

Claims (6)

1. An AI multiphase flow detection system training data acquisition device, comprising: the integral frame comprises a device bracket for supporting and fixing and a detection probe bracket for supporting and fixing the image acquisition module; the side wall of the transparent container is provided with a mounting hole; the motion module comprises a vertical motion module for controlling the rotary drum assembly to move up and down and a rotary motion module for driving the rotary drum assembly to rotate; The rotary drum assembly is arranged in the transparent container, is installed with the motion module, and consists of a plurality of rotary drums with different diameters and different identification patterns, and is used for simulating disperse phases with different speeds; The image acquisition module is used for acquiring multiphase flow image information containing the identification pattern on the rotary drum and is arranged in the mounting hole and comprises an industrial camera, an optical fiber sensor and a laser sensor; The data processing module is used for identifying the disperse phase in the multiphase flow image information, simultaneously obtaining the image information of the simulation rotating drum, correcting the detection result by taking the image information of the simulation rotating drum as a reference, and finally obtaining the AI multiphase flow detection system training data which can be used for intelligent detection.
2. The training data acquisition device of the AI multiphase flow detection system of claim 1, wherein the vertical movement module comprises a stepping motor, a screw rod and a sliding table, the screw rod is vertically fixed with a device support, an output end of the stepping motor is installed with the screw rod, the sliding table is slidably installed on the screw rod, a connecting plate horizontally extends towards the upper side of a transparent container, the rotary movement module comprises a rotary motor and a coupler, the rotary motor is installed on the connecting plate, and an output end of the rotary motor is installed downwards through the coupler and a rotary drum assembly.
3. The training data acquisition device of the AI multiphase flow detection system as recited in claim 2, wherein the identification pattern on the drum is composed of a plurality of micropores, and the diameter of the micropores is 100-1000 μm.
4. The AI multiphase flow detection system training data acquisition device of claim 3, wherein the transparent container is a transparent organic glass container, and the laser sensor, the optical fiber sensor and the industrial camera are installed at the mounting hole.
5. The AI multiphase flow detection system training data acquisition device of claim 4, wherein limit switches are mounted at the front end and the rear end of the screw rod.
6. The AI multiphase flow detection system training data acquisition device of claim 5, wherein the training data acquisition is performed using the device as follows: s1, installing a selected rotary drum and a rotary motor through a coupler, and installing a rotary motion module integrated with the rotary drum on a sliding table of a vertical motion module; s2, placing a scale in the transparent container, acquiring a scale image by using an installed image acquisition module, and establishing a proportional relation between image pixels and actual physical dimensions by using a data processing module; s3, setting the rotating speeds of a stepping motor and a rotating motor according to the target motion state of the disperse phase to be simulated, so that the rotating drum simultaneously rotates and vertically moves in the working solution; s4, processing the acquired multiphase flow image information, and identifying micropore patterns on the rotary drum, calculating the real size of a disperse phase simulated by each micropore based on the proportional relation established in the step S2, and calculating the real movement speed of the simulated disperse phase based on the displacement and time interval of the same identification pattern in continuous frame images; s5, storing the disperse phase data marked with the real size and the real movement speed obtained in the step S4 as a data set for training of an AI model, comparing the real size and the real movement speed serving as reference values with measured values of other sensors under the same condition, and calculating a correction coefficient of the sensor.

Description

Training data acquisition device of AI multiphase flow detection system and application method thereof Technical Field The invention belongs to the technical field of industrial process detection and artificial intelligence, and particularly relates to an AI multiphase flow detection system training data acquisition device and a use method thereof. Background In many fields such as chemical engineering, multiphase flow measurement is always difficult, and because each phase has different physical and chemical properties in a flowing state, a phase interface is complex and changeable, and accurate measurement of the size and the movement speed of a disperse phase has great challenges. Common test methods are high-speed imaging, optical tomography, nuclear magnetism, particle Image Velocimetry (PIV), laser Doppler Velocimetry (LDV), and the like. The measuring precision of the methods is influenced by various factors, the content and the speed of each phase are measured with high precision, the calibration and the verification are difficult, standard verification means are lacked, and some methods are expensive and complex in instruments, have safety limitations and are difficult to apply in online industry. An important development trend of the multiphase flow detection technology at present is to make up for the defect of a single measurement technology by a method of multi-sensor information fusion and artificial intelligent data driving, and improve the robustness and reliability of a system. While AI presents unprecedented advantages in multiphase flow detection, there are also a number of significant challenges, including the scarcity of high quality labeling data, the extremely high difficulty of acquiring data that covers all flow patterns and is precisely labeled under real industrial conditions, poor generalization capability, and the possibility of severely degrading performance when applied to different pipe diameters, inclinations, media or sensor configurations for models trained in specific devices and fluid properties. Disclosure of Invention Aiming at the problems, the invention provides an AI multiphase flow detection system training data acquisition device and a use method thereof, which are used for overcoming the defects that the acquisition of multiphase flow detection data is difficult, a sensor lacks a reliable correction standard, AI training data is scarce and the labeling cost is high in the prior art. The specific scheme is as follows: An AI multiphase flow detection system training data acquisition device, comprising: the integral frame comprises a device bracket for supporting and fixing and a detection probe bracket for supporting and fixing the image acquisition module; the side wall of the transparent container is provided with a mounting hole; the motion module comprises a vertical motion module for controlling the rotary drum assembly to move up and down and a rotary motion module for driving the rotary drum assembly to rotate; The rotary drum assembly is arranged in the transparent container, is installed with the motion module, and consists of a plurality of rotary drums with different diameters and different identification patterns, and is used for simulating disperse phases with different speeds; The image acquisition module is used for acquiring multiphase flow image information containing the identification pattern on the rotary drum and is arranged in the mounting hole and comprises an industrial camera, an optical fiber sensor and a laser sensor; The data processing module is used for identifying the disperse phase in the multiphase flow image information, simultaneously obtaining the image information of the simulation rotating drum, correcting the detection result by taking the image information of the simulation rotating drum as a reference, and finally obtaining the AI multiphase flow detection system training data which can be used for intelligent detection. The vertical movement module comprises a stepping motor, a screw rod and a sliding table, wherein the screw rod is vertically fixed with the device support, the output end of the stepping motor is installed with the screw rod, the sliding table is slidably installed on the screw rod and horizontally extends towards the upper side of the transparent container to form a connecting plate, the rotary movement module comprises a rotary motor and a coupler, the rotary motor is installed on the connecting plate, and the output end of the rotary motor is installed downwards through the coupler and the rotary drum assembly. Further, the identification pattern on the rotary drum is composed of a plurality of micropores, and the diameter of each micropore is 100-1000 microns. Further, the transparent container is a transparent organic glass container, and the laser sensor, the optical fiber sensor and the industrial camera are installed at the installation hole. Further, limit switches are arranged at the front end and