CN-121997095-A - Automatic powder leakage defect identification method based on digital twin and AI video analysis

CN121997095ACN 121997095 ACN121997095 ACN 121997095ACN-121997095-A

Abstract

The invention relates to the technical field of coal mill powder pipes, in particular to an automatic powder leakage defect identification method based on digital twin and AI video analysis, which comprises the steps of obtaining operation parameters and environment parameters of a coal mill powder pipe, calculating a powder leakage risk characterization coefficient, and judging whether a powder leakage potential risk exists; the method comprises the steps of collecting image data of the surface and surrounding areas of a powder tube in an initial starting period through an AI video analysis module, extracting powder leakage characteristic parameters, drawing characteristic change curve segments, comparing the characteristic change curve segments with reference characteristic curve segments in similarity, judging whether a suspected powder leakage area exists, obtaining actual operation parameters and model prediction parameters of the powder tube in a stable operation period based on a digital twin model, calculating fluctuation degree of parameter deviation, combining a judging result of the suspected powder leakage area, comprehensively judging whether the powder leakage defect exists, carrying out grading alarm prompt according to the judging result, realizing early and accurate identification of the powder leakage defect, improving detection efficiency and precision, and reducing operation and maintenance cost and safety risk.

Inventors

LIN ZHOUYONG
WU YURU
KE YIHONG
GAO LIPING
LIN YONGJIANG
CHEN GUOBAO
Jiang Tianpei
Kang Xiaozhong
XU WENSHUN
WANG SHUHUI

Assignees

国能(泉州)热电有限公司

Dates

Publication Date: 20260508
Application Date: 20251225

Claims (10)

1. The automatic powder leakage defect identification method based on digital twin and AI video analysis is characterized by comprising the following steps: acquiring operation parameters and environment parameters of a coal mill powder pipe to calculate a powder leakage risk representation coefficient of the coal mill powder pipe, and judging whether the powder pipe has a powder leakage potential risk or not based on the powder leakage risk representation coefficient; Collecting image data of the surface and surrounding areas of the powder tube in a starting initial period by an AI video analysis module, extracting powder leakage characteristic parameters and drawing characteristic change curve segments, wherein the starting initial period is a preset duration period taking the starting moment of the coal mill as a time starting point; Determining the feature similarity of a powder leakage feature change curve segment and a reference feature curve segment corresponding to each image frame, and judging whether a suspected powder leakage area exists in the powder tube according to the fluctuation condition of the feature similarity; Acquiring actual operation parameters of the powder tube in a stable operation period and twin model prediction parameters based on a digital twin model, calculating fluctuation degree of parameter deviation, and comprehensively judging whether the powder tube has powder leakage defects according to the judging result of the suspected powder leakage area, wherein the stable operation period is a preset duration period after the starting initial period is ended; And carrying out grading alarm prompt on the judging result of the powder leakage defect of the powder pipe.
2. The automatic powder leakage defect identification method based on digital twin and AI video analysis as set forth in claim 1, wherein the powder leakage risk characterization coefficient of the coal mill powder tube is calculated according to the following formula: ; in the formula, For the powder leakage risk characterization coefficient, Is the deviation value of the actual wind pressure and the designed wind pressure in the powder tube, Is a preset wind pressure deviation reference value, Is the concentration of the coal powder in the powder pipe, Is a preset reference value of the concentration of the coal dust, Is the dust concentration in the surrounding environment of the powder tube, Is a preset reference value of the concentration of the environmental dust, Is the weight coefficient of the wind pressure deviation, Is the weight coefficient of the concentration of the coal powder, Is an environmental dust concentration weight coefficient.
3. The automatic powder leakage defect identification method based on digital twin and AI video analysis according to claim 2, wherein the determining process of whether the powder tube has the potential risk of powder leakage is as follows: comparing the powder leakage risk representation coefficient with a preset powder leakage risk representation coefficient threshold; and if the powder leakage risk representation coefficient is larger than the powder leakage risk representation coefficient threshold, judging that the powder pipe has the potential risk of powder leakage.
4. The automatic powder leakage defect identification method based on digital twin and AI video analysis of claim 1, wherein the image data acquisition process of the AI video analysis module comprises: And arranging high-definition industrial cameras on the surface of the powder tube and around a preset monitoring area, collecting continuous image frames in the starting initial period at a preset frame rate, preprocessing the image frames, and extracting powder leakage characteristic parameters through a trained powder leakage characteristic extraction model, wherein the powder leakage characteristic extraction model is a target detection model based on improvement YOLOv, and the powder leakage characteristic parameters comprise a powder leakage area contour area, a gray scale average value and a motion vector.
5. The automatic powder leakage defect identification method based on digital twin and AI video analysis as claimed in claim 4, wherein the characteristic change curve section takes a corresponding powder leakage characteristic parameter as a vertical axis, takes time as a horizontal axis, the length of the characteristic change curve section on the horizontal axis is the time length of the initial starting period, and the reference characteristic curve section is a powder tube surface characteristic parameter change curve in a powder leakage-free state simulated by a digital twin model.
6. The automatic powder leakage defect identification method based on digital twin and AI video analysis according to claim 5, wherein the determining of the fluctuation condition of the feature similarity includes: Calculating the average value of the feature similarity of all image frames in the starting initial period based on the feature similarity of the powder leakage feature change curve segment corresponding to the current image frame and the reference feature curve segment, and calculating the standard deviation of the feature similarity based on the average value of the feature similarity, wherein the standard deviation of the feature similarity is calculated according to the following formula: ; in the formula, As the standard deviation of the feature similarity, For the feature similarity of the kth frame image, And k is an integer greater than or equal to 1, and t is the total number of image frames in the starting initial period.
7. The automatic powder leakage defect identification method based on digital twin and AI video analysis according to claim 6, wherein the process of determining whether a suspected powder leakage area exists in the powder tube is as follows: comparing the characteristic similarity standard deviation with a preset characteristic similarity standard deviation threshold; If the feature similarity standard deviation is larger than the feature similarity standard deviation threshold, judging that a suspected powder leakage area exists in the powder tube, and recording the space coordinates and feature parameter change trend of the area.
8. The automatic powder leakage defect identification method based on digital twin and AI video analysis according to claim 1, wherein the process of constructing the digital twin model comprises: Based on a three-dimensional CAD model, material properties and operation boundary conditions of the powder pipe, a multi-physical-field digital twin model of the powder pipe is constructed, actual operation parameters of the powder pipe are synchronized through a real-time data interface, and prediction parameters of the twin model are wind pressure, flow and temperature of the powder pipe in a non-leakage powder state output by the model.
9. The automatic powder leakage defect identification method based on digital twin and AI video analysis of claim 8, wherein the degree of fluctuation of the parameter deviation is determined based on a standard deviation, the standard deviation being calculated according to the following formula: ; in the formula, As a standard deviation of the parameter in question, Respectively the actual wind pressure, flow and temperature at the mth moment in the stable operation period, And respectively predicting corresponding parameters of the twin model, wherein m is an integer greater than or equal to 1, and n is the total number of parameter acquisition times in the stable operation period.
10. The automatic powder leakage defect identification method based on digital twin and AI video analysis as set forth in claim 9, wherein the comprehensive determination process of whether the powder pipe has the powder leakage defect is as follows: Comparing the parameter deviation standard deviation with a preset deviation standard deviation threshold, if the parameter deviation standard deviation is larger than the deviation standard deviation threshold and the suspected powder leakage area exists, judging that the powder leakage defect exists in the powder tube, and triggering a grading alarm prompt according to the area of the powder leakage area and the parameter deviation degree.

Description

Automatic powder leakage defect identification method based on digital twin and AI video analysis Technical Field The invention relates to the technical field of coal mill powder pipes, in particular to an automatic powder leakage defect identification method based on digital twin and AI video analysis. Background In the industrial fields of thermal power generation, ferrous metallurgy and the like, a coal mill is used as core equipment for preparing coal dust, and the safe and stable operation of a powder pipe system directly influences the production efficiency and the service life of the equipment. The coal mill powder pipe is in the complex working condition environment with high dust and high wind pressure for a long time, and is influenced by factors such as coal dust scouring, pipeline abrasion, temperature stress and the like, so that the defect of powder leakage is easy to occur. The powder leakage not only can cause the waste of coal powder and environmental pollution, but also can cause safety accidents such as equipment corrosion, fire disaster and even explosion, so that the method is very important for early and accurate identification of the powder leakage defect of the powder pipe of the coal mill. The traditional powder leakage detection method mainly relies on manual inspection, and performs local state monitoring through visual observation, auscultation or a contact sensor, and has the problems of low detection efficiency, poor real-time performance, high powder leakage detection rate, incapability of covering complex working conditions comprehensively and the like. With the development of industrial intelligent technology, a state monitoring technology based on sensor data is gradually applied, but the monitoring range and the data dimension of a single sensor are limited, and the multi-physical field coupling characteristic of powder leakage defects is difficult to capture. In addition, fine features at the early stage of powder leakage are difficult to be effectively identified by the conventional method, resulting in a lag in defect diagnosis. The foregoing is provided merely for the purpose of facilitating understanding of the technical solutions of the present invention and is not intended to represent an admission that the foregoing is prior art. Disclosure of Invention The invention mainly aims to provide a powder leakage defect automatic identification method based on digital twin and AI video analysis, and aims to solve the problems that the existing coal mill powder pipe powder leakage defect detection has low manual inspection efficiency, the traditional technology detection dimension is single, the systematic detection method of multi-source data fusion is lack, and the like, and the technical problems of accurate identification of powder leakage potential risk early assessment, characteristic dynamic extraction and multi-mode collaborative judgment are difficult to realize. In order to achieve the above purpose, the invention provides a method for automatically identifying powder leakage defects based on digital twin and AI video analysis, which comprises the following steps: acquiring operation parameters and environment parameters of a coal mill powder pipe to calculate a powder leakage risk representation coefficient of the coal mill powder pipe, and judging whether the powder pipe has a powder leakage potential risk or not based on the powder leakage risk representation coefficient; Collecting image data of the surface and surrounding areas of the powder tube in a starting initial period by an AI video analysis module, extracting powder leakage characteristic parameters and drawing characteristic change curve segments, wherein the starting initial period is a preset duration period taking the starting moment of the coal mill as a time starting point; Determining the feature similarity of a powder leakage feature change curve segment and a reference feature curve segment corresponding to each image frame, and judging whether a suspected powder leakage area exists in the powder tube according to the fluctuation condition of the feature similarity; Acquiring actual operation parameters of the powder tube in a stable operation period and twin model prediction parameters based on a digital twin model, calculating fluctuation degree of parameter deviation, and comprehensively judging whether the powder tube has powder leakage defects according to the judging result of the suspected powder leakage area, wherein the stable operation period is a preset duration period after the starting initial period is ended; And carrying out grading alarm prompt on the judging result of the powder leakage defect of the powder pipe. Optionally, the powder leakage risk characterization coefficient of the powder tube of the coal mill is calculated according to the following formula: ; in the formula, For the powder leakage risk characterization coefficient,Is the deviation value of the actual win