Negative-pressure-wave-guided pipe leakage detection and positioning method based on unmanned aerial vehicle inspection

Abstract

A negative-pressure-wave-guided pipe leakage detection and positioning method based on unmanned aerial vehicle inspection, the method comprising the following steps: S1, collecting, in real time, pressure and flow data at the head end and the tail end of a pipe and corresponding location data by means of a dedicated collection apparatus for a pipe leakage, and on the basis of real-time pressure and flow changes, determining whether the pipe is leaking; S2, correcting an acoustic wave velocity in a tailings pipe to obtain a calculation formula for the corrected acoustic wave velocity, and then positioning a leakage point on the basis of a negative pressure wave; S3, converting the positioned leakage point into longitude and latitude coordinates, and transmitting the longitude and latitude coordinates to an unmanned aerial vehicle; S4, an unmanned aerial vehicle system planning an inspection path, and then the unmanned aerial vehicle inspecting the surroundings of the leakage point of the pipe, and capturing a photograph of a surrounding environment of the pipe having longitude and latitude coordinates; and S5, designing and training an automatic positioning algorithm for an ore pulp pipe leakage, performing automatic recognition on the captured photograph in real time, recognizing a pipe leakage, and positioning the location of the leakage point.

Description

A method for detecting and locating pipeline leaks using unmanned aerial vehicles (UAVs) guided by negative pressure waves. Technical Field

[0001] This invention belongs to the field of slurry pipeline detection technology, and specifically relates to a method for detecting and locating pipeline leaks using unmanned aerial vehicles (UAVs) guided by negative pressure waves. Background Technology

[0002] Pipeline transportation is widely used in the transportation of mineral concentrates and tailings due to its advantages such as strong continuity and large capacity. However, if a pipeline leaks or is deliberately sabotaged, it can seriously affect the safety of people in the surrounding area and the security of national property. Therefore, it is necessary to monitor the pipeline's operation and locate the leak point.

[0003] When a slurry pipeline leaks, a pressure difference typically occurs. This pressure difference causes the medium inside the pipeline to leak out rapidly, resulting in a decrease in fluid density and pressure at that point. This leads to a phenomenon where the density and pressure decrease sequentially from the leak point along both ends of the pipeline, similar to a negative pressure wave in hydraulics. This negative pressure wave can propagate along the pipeline for tens of kilometers or even longer due to the waveguide effect of the pipeline. The propagation speed of the negative pressure wave in the pipeline is approximately equivalent to the propagation speed of sound waves in the transported fluid, typically in the range of 1000 to 1200 m / s. Once the transient negative pressure wave carrying leakage information is captured by pressure sensors installed at both ends of the pipeline, the leak can be detected. The location of the leak can be calculated by the time difference between the arrival of the transient negative pressure wave at both ends of the pipeline. Current pipeline leak detection technology using negative pressure waves detects leaks by installing pressure sensors at both ends of the pipeline. However, in actual use, due to the long length of the slurry pipeline, the negative pressure wave signal will be lost during the propagation process inside the slurry pipeline. At the same time, it is also easily affected by other external factors, making it impossible to accurately detect the precise location of the slurry leak. It can only detect the approximate location of the leak in the slurry pipeline, and there are false detections and missed detections.

[0004] In recent years, drones have been gradually used for the inspection of oil and gas pipelines and urban hot water pipelines, which has solved the problems of low efficiency and untimely emergency response of manual inspection operations and achieved certain results. However, they still rely on personnel for identification and cannot perform real-time detection.

[0005] Summary of the Invention

[0006] The present invention aims to provide a method for detecting and locating pipeline leaks by unmanned aerial vehicle (UAV) inspection based on negative pressure wave guidance, in order to solve the problem that the existing technology cannot accurately detect the precise location of slurry leaks, but can only detect the approximate location of slurry pipeline leaks, resulting in low detection accuracy.

[0007] A kind of unmanned aerial vehicle inspection pipeline leakage detection positioning method based on negative pressure wave guidance, comprising the following steps:

[0008] S1, by pipeline leakage special acquisition device real-time acquisition pipeline head and tail pressure and flow data and corresponding position data, according to the real-time change of pressure and flow whether pipeline leaks are judged;

[0009] S2, the sound wave speed in tailing pipeline is corrected to obtain the sound wave speed calculation formula after correction, then according to negative pressure wave positioning leak point;

[0010] The sound wave speed calculation formula is as follows:

[0011] In the formula:

[0012] v 浆 The corrected sound wave propagation speed;K is the volumetric elastic coefficient of slurry;ρ is the density of slurry;E is the elastic modulus of pipeline material;D is the diameter of pipeline;e is the thickness of pipeline wall;C1 is the correction coefficient of pipeline;

[0013] The formula for positioning leak point according to negative pressure wave is:

[0014] x=t1×v 浆 (2)

[0015] L-x=t2×v 浆 (3)

[0016] In the formula:

[0017] x: the distance between pipeline leak point and upstream pressure measuring point of pipeline;L: the distance between upstream and downstream pressure sensor;v 浆 : the propagation speed of pressure wave in pipeline;t1: the time difference of upstream sensor receiving pressure wave, t2: the time difference of downstream sensor receiving pressure wave;

[0018] S3, convert the leak point into latitude and longitude coordinates, and transmit to unmanned aerial vehicle;

[0019] S4, the unmanned aerial vehicle system carries out inspection path planning, then carries out inspection around pipeline leak point periphery, and takes the pipeline periphery environment photo with latitude and longitude coordinates;

[0020] S5, design and train the ore slurry pipeline leakage automatic positioning algorithm to automatically identify the photographed photos in real time, identify pipeline leakage, and locate the leakage point position; the ore slurry pipeline leakage automatic positioning algorithm automatically identifies the photographed photos, which is an image change detection model constructed based on an EfficientNetV2 deep neural network, realizes two-time image input through the construction of a double-encoder path, replaces the classifier at the back of the network with a decoder with an upsampling process, and then obtains the change detection result of the two-time image input according to the two-time image input;

[0021] When the pipeline leakage is identified, the latitude and longitude coordinates of the photographed photos are obtained, that is, the accurate positioning coordinates of the ore slurry pipeline leakage point, and the latitude and longitude coordinates of the unmanned aerial vehicle are sent to the maintenance personnel to inform the maintenance personnel to perform maintenance; when the pipeline leakage is not identified, it is determined as a false detection.

[0022] Further, the pipeline leakage special collection device comprises high-pressure, wear-resistant ore slurry special electromagnetic flowmeters arranged at the two ends of the tailing pipeline.

[0023] Further, the step of judging whether the pipeline leaks in step S1 is:

[0024] S11, difference calculation is performed on the flow data at the two ends of the pipeline;

[0025] S12, when the flow difference between the two ends is maintained within a certain value range, that is, the flow of the upstream and downstream remains stable, and the pipeline does not leak, the data is continuously monitored;

[0026] When the flow difference between the two ends increases and the difference exceeds the set threshold value, that is, the upstream flow increases and the downstream flow decreases; further judge the pressure change at the two ends of the pipeline, when the upstream pressure decreases and the downstream pressure decreases, it is determined that the pipeline leaks.

[0027] Further, the specific method of step S3 is as follows:

[0028] S31, the latitude and longitude coordinates of the tailing pipeline along the line are measured by using a theodolite;

[0029] S32, the distance of the entire pipeline from the head-end pressure sensor and the measured latitude and longitude coordinates are established in a corresponding relationship;

[0030] S33, the leakage point positioning in step S2 is transmitted to the unmanned aerial vehicle, and the latitude and longitude coordinates of the positioning leakage point are obtained by a linear fitting method.

[0031] Further, the step of planning the unmanned aerial vehicle system inspection path in step S4 is as follows:

[0032] S41, obtain the latitude and longitude coordinate data of the pipeline positioning leakage point;

[0033] S42, obtain the latitude and longitude coordinate data of the no-fly zone around the pipeline;

[0034] S43, digitize the latitude and longitude coordinate data of the pipeline and the latitude and longitude coordinate data of the no-fly zone in the program, and equivalent to points;

[0035] S44, the unmanned aerial vehicle departs from the airport, avoids the no-fly zone and the no-signal zone, plans a flight path at a specified distance and speed, slows down when the distance is X before and after approaching the pipeline leakage point, and takes photos of the environment around the pipeline, each photo has the latitude and longitude coordinates of the shooting point;

[0036] S45, after rechecking is completed, the unmanned aerial vehicle returns to the airport according to the original path.

[0037] Further, in the step S5, the mine slurry pipeline leakage automatic positioning algorithm is designed and trained, and the photographed photos are automatically identified in real time, the pipeline leakage is identified, and the leakage point position is located; the mine slurry pipeline leakage automatic positioning algorithm automatically identifies the photographed photos, which is an image change detection model based on an EfficientNetV2 deep neural network, realizes two-time image input through a double-encoder path, replaces the classifier at the back of the network with a decoder with an up-sampling process, and then obtains the change detection result of the two-time image according to the input of the two-time image.

[0038] The specific steps are as follows:

[0039] S51, construct an encoder: in the encoder, the convolutional network is constructed with the optimal configuration through a neural network structure search method, the double-encoder path respectively converts two-time input into feature maps through convolution operation, and the picture becomes smaller; the parameters between the two-encoder paths are shared; the first 7 layers of convolution in the EfficientNetV2 are used in the encoder part, which respectively includes 1×1 convolution, depth convolution, squeeze excitation module and Dropout process, and the input image is feature extracted; each encoder layer corresponds to a decoder layer; the encoder part is composed of two parallel identical structure encoders, the convolution kernel parameters of which are shared, and each is used for feature extraction process of one time phase;

[0040] S52, construct a decoder: the up-sampling in the decoder restores the picture size reduction caused by the encoder step by enlarging the picture; in the up-sampling process, the features of the corresponding layers of the encoder are spliced and convolved again by using the cascade method; finally, the same size as the original image is obtained through the decoder stage. The final feature map; the decoder part is connected in a cascade manner, that is, the feature map corresponding to the encoder layer is spliced and connected with the output of the previous decoder layer as the input of the decoder layer through the jump connection.

[0041] S53, output layer: a fully connected layer is added at the end of the network, and the feature map is mapped to the change probability pixel by pixel, and finally a change detection result with the same size as the original input image is obtained;

[0042] S54, input the data set into the constructed EfficientNetV2 deep learning network, train the detection model, identify the changes of the ore pulp pipeline over time, including ore pulp leakage, pipeline deformation, pipeline rupture, etc.; in combination with the image geographical position, the coordinates of the leakage point are given; the latitude and longitude coordinate data taken from the unmanned aerial vehicle photo is the coordinate of the pipeline leakage point.

[0043] Further, in the step S54, the data set is periodically photographed along the pipeline by the unmanned aerial vehicle equipped with a camera according to the set fixed route, the video is frame extracted, registered and matched, the photo is taken as the change detection data set, and manual labeling is performed, and the training set and the test set are randomly divided.

[0044] In summary, the present application has the following beneficial effects:

[0045] 1、The present application can overcome the problem that the ore pulp pipeline is relatively long, the negative pressure wave signal will be lost to a certain extent during the propagation in the ore pulp pipeline, and at the same time, it will be easily affected by other external factors, so that the accurate positioning of the ore pulp leakage cannot be accurately detected, only the approximate position of the ore pulp pipeline leakage can be detected, and the detection precision is low, thereby avoiding the occurrence of safety hazards.

[0046] 2、The present application simplifies the route planning problem of unmanned aerial vehicle pipeline inspection and pipeline leakage detection into the shortest path problem under the target condition according to the characteristics of unmanned aerial vehicle pipeline inspection, and has the advantages of simple basic principle, high execution efficiency and wide application range, and has been widely applied in the field of computer technology.

[0047] 3、The unmanned aerial vehicle inspection of the present application also overcomes the problem that personnel are needed for identification and real-time detection cannot be performed, and through the design and training of the ore pulp pipeline leakage automatic positioning algorithm, the photographed photos can be automatically identified in real time, so that the ore pulp pipeline leakage point can be found in time and accurately, and large-area farmland pollution caused by ore pulp pipeline leakage can be avoided, which can not only avoid environmental pollution, but also avoid economic losses. BRIEF DESCRIPTION OF DRAWINGS

[0048] Fig. 1 is a flow chart of the method of the present application;

[0049] Fig. 2 is a change curve of pressure and flow before and after pipeline leakage;

[0050] Fig. 3 is a fusion flow chart of the unmanned aerial vehicle composite positioning method guided by the negative pressure wave of the present application;

[0051] Fig. 4 is a re-inspection flow chart of the unmanned aerial vehicle of the present application;

[0052] FIG. 5, the image change detection model based on the EfficientNetV2 deep neural network of the present application. DETAILED DESCRIPTION

[0053] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0054] Embodiment 1

[0055] An unmanned aerial vehicle inspection pipeline leakage detection and positioning method based on negative pressure wave guidance, as shown in FIG. 1, includes the following steps:

[0056] S1, pressure and flow leakage detection

[0057] 1) Install high-pressure, wear-resistant slurry special electromagnetic flow meters, pressure transmitters and GPS time correction modules at both ends of the tailings pipeline, and synchronously collect the pressures and flow rates at both ends of the pipeline, and transmit the pressure and flow data and the corresponding latitude and longitude to the leakage detection and positioning server at the same time.

[0058] 2) Perform leakage detection according to the changes in pressure and flow. The pressure and flow change curves before and after the pipeline leaks are shown in FIG. 2.

[0059] When the pipeline does not leak, the flow rates at the upstream and downstream remain stable, and the flow rate difference at both ends will remain near a certain value; but when the pipeline leaks, the upstream flow rate increases, the downstream flow rate decreases, and the flow rate difference at both ends will increase significantly. When the difference increases to exceed the set threshold, it can be basically determined that the pipeline has leaked. After determining the leakage event by the flow rate difference, the positioning is performed according to the pressure waveforms at the pipeline inlet and outlet. When the pipeline leaks, the upstream pressure decreases, and the downstream pressure decreases. When the above leakage characteristics are met at the same time, it can be determined that the pipeline has leaked.

[0060] S2, negative pressure leakage positioning

[0061] 1) First, correct the sound wave velocity in the tailings pipeline

[0062] The sound wave velocity generated by the slurry pipeline leakage depends on the elasticity of the slurry, the density of the slurry, and the material of the pipeline, and the design formula is corrected:

[0063] In the formula,

[0064] v浆 V is the sound wave propagation velocity; K is the bulk modulus of the slurry; p is the density of the slurry; E is the elastic modulus of the pipeline material; D is the diameter of the pipeline; e is the pipeline wall thickness; C1 is the correction coefficient of the pipeline;

[0065] 2) Locating the pipeline leakage point according to the negative pressure wave

[0066] The leakage locating formula according to the negative pressure wave is:

[0067] x = t1 x v 浆 (2)

[0068] L-x = t2 x v 浆 (3)

[0069] In the formula:

[0070] X: the distance from the pipeline leakage point to the upstream pressure measuring point of the pipeline; L: the distance between the upstream and downstream pressure sensors; v 浆 : the propagation velocity of the pressure wave in the pipeline; t1, t2: the time difference between the upstream and downstream sensors receiving the pressure wave, respectively.

[0071] S3, convert the negative pressure wave positioning result into latitude and longitude coordinates, and transmit it to the unmanned aerial vehicle; the unmanned aerial vehicle system plans the inspection path, and the unmanned aerial vehicle system plans the inspection path in real time, and inspects the surrounding environment of the pipeline leakage point in real time, and takes photos of the surrounding environment of the pipeline with latitude and longitude coordinates;

[0072] Although the flow velocity is corrected when locating the pipeline leakage point, the positioning result may still have false detection results, and the positioning accuracy is low. In order to accurately locate and improve the positioning accuracy, the negative pressure positioning method and the effect of rapid confirmation of the unmanned aerial vehicle are combined, and the negative pressure wave guided unmanned aerial vehicle composite positioning method is used for tailing pipeline leakage positioning, and the fusion process is shown in FIG. 3.

[0073] Specifically,

[0074] 1) Coordinate conversion

[0075] The negative pressure wave positioning result is x meters away from the first end pressure sensor, which needs to be converted into the latitude and longitude coordinates of the leakage point position, and the conversion steps and methods are as follows:

[0076] ① Use the theodolite to measure the latitude and longitude coordinates of the tailing pipeline along the line every 5 meters;

[0077] ② Establish a corresponding relationship between the distance of the entire pipeline from the first end pressure sensor and the measured latitude and longitude coordinates;

[0078] ③When the negative pressure wave positioning result is transmitted to the unmanned aerial vehicle, the longitude and latitude coordinates of the specific leakage point are obtained by linear fitting method.

[0079] 2) Unmanned aerial vehicle reinspection (process as shown in Figure 4)

[0080] The steps and methods of the unmanned aerial vehicle system inspection path planning are as follows:

[0081] ① Obtain the longitude and latitude coordinate data of the pipeline.

[0082] ② Obtain the longitude and latitude coordinate data of the no-fly zone around the pipeline.

[0083] ③ Digitize the longitude and latitude coordinate data of the pipeline and the no-fly zone longitude and latitude coordinate data in the program, and equivalent to points.

[0084] ④ When the pipeline leaks, the unmanned aerial vehicle departs from the airport, avoids the no-fly zone and the no-signal zone, and plans the flight path with the shortest distance and the fastest speed; when approaching the pipeline leakage point, the speed is slowed down and the pipeline leakage situation is photographed.

[0085] ⑤ Real-time automatic detection and identification of the photographed pipeline pictures along the line, when the leakage is identified, the longitude and latitude coordinates of the unmanned aerial vehicle are sent to the maintenance personnel, and the maintenance personnel are notified to repair.

[0086] ⑥ After the reinspection is completed, the unmanned aerial vehicle returns to the airport according to the original path.

[0087] S4, ore pulp pipeline leakage automatic positioning algorithm, real-time automatic identification of the photographed pictures, identification of pipeline leakage, and positioning of the leakage point position;

[0088] When the pipeline leakage is identified, the longitude and latitude coordinates of the photograph are taken, which is the accurate positioning coordinates of the ore pulp pipeline leakage point, and the longitude and latitude coordinates of the unmanned aerial vehicle are sent to the maintenance personnel, and the maintenance personnel are notified to repair; when the pipeline leakage is not identified, it is judged as a false detection.

[0089] Design and training of ore pulp pipeline leakage automatic positioning algorithm, real-time automatic identification of the photographed pictures, identification of pipeline leakage, and positioning of the leakage point position; is based on the image change detection model (as shown in Figure 5) constructed by EfficientNetV2 deep neural network, realizes the input of two time images through the construction of double encoder path, replaces the classifier at the back of the network with a decoder with up-sampling process, and obtains the change detection result of two time images according to the input of two time images, automatically identifies the pipeline leakage, and locates the leakage point position; specific steps include:

[0090] ①Build the encoder: In the encoder, the convolutional network is built with the optimal configuration through the neural network structure search method. By adjusting the input resolution, the depth of the network structure, and the channel width, it still has excellent representation ability with less parameter amount. The two-way encoder path respectively converts the two-phase input into feature maps through convolution operation, while the picture becomes smaller. The parameters between the two-way encoder are shared. The first 7 layers of convolution in EfficientNetV2 are used in the encoder part, which includes 1x1 convolution, depth convolution, squeeze excitation module, and Dropout process, for feature extraction of the input image; each encoder layer corresponds to a decoder layer. The encoder part is composed of two parallel identical structure encoders, which share the convolution kernel parameters and are used for feature extraction of each phase.

[0091] ②Build the decoder: The up-sampling in the decoder restores the picture size reduction caused by the encoder process by enlarging the picture. In the up-sampling process, the features of the corresponding encoder layers are concatenated, spliced, and convolved again. Finally, the same size as the original image is obtained through the decoder stage. The decoder part is connected in a cascading manner, i.e. the feature map of the corresponding encoder layer is spliced with the output of the previous decoder layer through jump connection and used as the input of the decoder.

[0092] ③Output layer: A fully connected layer is added at the end of the network to map the feature map pixel by pixel to the change probability, and finally obtain the same size change detection result as the original input image. Specifically, the role of the output layer is to map the final feature map obtained in the decoder stage to the change probability pixel by pixel. That is, each pixel point corresponds to a probability value, which represents the possibility of change of that pixel point. This probability matrix can be regarded as a heat map, where the area with higher probability value may indicate that a change has occurred. To convert these change probabilities into actual detection results, a threshold value needs to be set. If the probability value of a certain pixel point exceeds this threshold, it can be considered that the point detects a change. In addition, some post-processing techniques such as connected component analysis may be used to determine the boundary of the change area, so as to more accurately locate the change area.

[0093] ④Input the dataset into the built EfficientNetV2 deep learning network to train the detection model and identify the changes in the ore slurry pipeline over time, including ore slurry leakage, pipeline deformation, pipeline rupture, etc. Combined with the image geographical location, the coordinates of the leakage point are given. The latitude and longitude coordinate data carried by the unmanned aerial vehicle photo is the coordinate of the pipeline leakage point;

[0094] The above data set is periodically photographed along the pipeline by a drone equipped with a wide-angle camera following a set fixed flight path, a total of 4 time periods of 5 min video, after frame extraction, registration and matching, 3 time periods are selected, 948 pairs of images between each two time phases are used as change detection data set, and manual annotation is performed, 758 images are randomly divided as training set, and 190 images are randomly divided as test set.

[0095] Slurry leakage, pipeline deformation and pipeline rupture are several specific types of changes that may occur in slurry pipelines. In this application, the EfficientNetV2 deep learning network is trained, and the model learns how to identify the features of these changes from two-phase images. When the model outputs the change probability matrix, it is actually predicting the possibility of each pixel point in the image occurring these specific changes. By analyzing these probability values, it can be identified which areas are most likely to have slurry leakage, pipeline deformation or pipeline rupture.

[0096] Specifically, during the training process, the model learns image features related to slurry leakage, pipeline deformation and pipeline rupture. In practical applications, the model will output a change probability map, and analysts or automated systems can identify high-probability areas based on this probability map, and then combine other information (such as geographic location, image analysis, etc.) to determine the specific leakage point or deformation area. For example, if the probability value of a certain area in the probability map is very high, and this area matches the known pipeline location, it can be inferred that the area is likely to have slurry leakage or pipeline rupture.

[0097] In addition, combined with the geographical position information of the image, the specific coordinates of the leakage point can be further located. This is usually achieved by combining the latitude and longitude coordinate data contained in the photo taken by the unmanned aerial vehicle with the high probability area in the change probability map. In this way, specific conclusions such as "slurry leakage, pipe deformation, pipe rupture" can be obtained from the change probability map, and the accurate position of the leakage point can be given. The fusion process of the above-mentioned unmanned aerial vehicle composite positioning method for tailing pipeline leakage positioning using negative pressure wave guidance is briefly described as follows: real-time collection of pressure and flow data at both ends of the pipeline, real-time calculation by negative pressure wave and flow balance algorithm, judgment of whether the pipeline exists leakage, detection of leakage, real-time positioning by calling the negative pressure wave positioning model, conversion of the leakage location into latitude and longitude when the negative pressure wave locates the leakage point, triggering of the event to transmit the latitude and longitude coordinates to the unmanned aerial vehicle system. The unmanned aerial vehicle system plans the inspection path in real time, assigns the unmanned aerial vehicle to the target site to inspect the surrounding environment of the pipeline leakage point, and takes photos of the surrounding environment of the pipeline with latitude and longitude coordinates. The automatic positioning algorithm for slurry pipeline leakage is designed and trained, and the photos taken are automatically identified in real time. When the pipeline leakage is identified, the latitude and longitude coordinates taken by the photos are the accurate positioning coordinates of the slurry pipeline leakage point. At the same time, the alarm is started after the leakage and the leakage point are confirmed, and the accurate coordinates of the leakage point are sent to the maintenance personnel, informing the maintenance personnel to go to the pointed location for emergency repair. If no leakage point is found, it is judged as a false detection, and the alarm is eliminated.

[0098] The above is only an embodiment of the present application, and common technical solutions or characteristics in the scheme are not described in detail. It should be noted that for those skilled in the art, without departing from the technical solutions of the present application, a number of modifications and improvements can be made, which should also be considered as the protection scope of the present application, and these will not affect the effect and practicality of the present application. The protection scope claimed in this application should be subject to the content of its claims, and the specific implementation mode and the like recorded in the specification can be used to explain the content of the claims.

Claims

1. A method for detecting and locating pipeline leaks using unmanned aerial vehicles (UAVs) guided by negative pressure waves, characterized in that, Comprise the following steps: S1, through the pipeline leakage special collection device real-time acquisition pipeline head and tail two ends pressure and flow data and corresponding position data, according to the real-time change of pressure and flow to determine whether the pipeline leaks; S2, the tailing pipeline sound wave speed correction to get the corrected sound wave speed calculation formula, and then according to the negative pressure wave positioning leak point; The acoustic wave velocity calculation formula is as follows: In the formula: v 浆 where v is the corrected sound wave propagation velocity; K is the bulk modulus of the slurry; p is the density of the slurry; E is the modulus of elasticity of the pipe material; D is the diameter of the pipe; e is the pipe wall thickness; and C1is a correction factor for the pipe. The step of positioning leak point of the negative pressure wave is: x = t1 x v 浆 (2) L - x = t2x v 浆 (3) In the formula: x: distance from the upstream pressure measurement point to the pipe leakage point; L: distance between the upstream and downstream pressure sensors; v 浆 : propagation speed of the pressure wave in the pipe; ti: time difference for the upstream sensor to receive the pressure wave, t2: time difference for the downstream sensor to receive the pressure wave; S3, the positioning leak point is converted into latitude and longitude coordinates, and is transmitted to the unmanned aerial vehicle; S4, the unmanned aerial vehicle system carries out inspection path planning, and then carries out inspection around the pipeline leak point, and takes photos of the pipeline surrounding environment with latitude and longitude coordinates; S5, design and training of ore pulp pipeline leakage automatic positioning algorithm, real-time automatic identification of the photos, identification of pipeline leakage, and positioning of leak point position; the ore pulp pipeline leakage automatic positioning algorithm automatically identifies the photos, which is an image change detection model based on EfficientNetV2 deep neural network, realizes two time image input through the construction of double encoder path, replaces the classifier at the back of the network with a decoder with upsampling process, and then obtains the change detection result of the two time image input; When the pipeline leakage is identified, the latitude and longitude coordinates of the photos are taken, which are the accurate positioning coordinates of the ore pulp pipeline leak point, and the latitude and longitude coordinates of the unmanned aerial vehicle are sent to the maintenance personnel, informing the maintenance personnel to repair; when the pipeline leakage is not identified, it is judged as a false detection.

2. The UAV inspection pipeline leak detection and localization method based on negative pressure wave guiding according to claim 1, characterized in that, The pipeline leakage special collection device comprises high-pressure, wear-resistant ore pulp special electromagnetic flowmeter, pressure sensor and GPS time correction module arranged at the head and tail of the tailing pipeline.

3. The UAV inspection method of claim 1, wherein, The step of judging whether the pipeline leaks in step S1 is: S11, difference calculation is performed on the flow data at the head and tail of the pipeline; S12, when the flow difference between the two ends maintains a certain value range, that is, the flow of the upstream and downstream remains stable, the pipeline does not leak, and the data continues to be monitored; When the flow difference between the two ends increases and the difference exceeds the set threshold, that is, the upstream flow increases and the downstream flow decreases; then judge the pressure change at the head and tail of the pipeline, when the upstream pressure decreases and the downstream pressure decreases, it is determined that the pipeline leaks.

4. The UAV inspection method of claim 1, wherein, The specific method of step S3 is as follows: S31, the latitude and longitude coordinates of the tailing pipeline along the line are measured by the theodolite; S32, the distance of the whole pipeline from the head pressure sensor and the measured latitude and longitude coordinates are established in a corresponding relationship; S33, the positioning leak point in step S2 is transmitted to the unmanned aerial vehicle, and the latitude and longitude coordinates of the positioning leak point are obtained by linear fitting method.

5. The UAV inspection method of claim 1, wherein, The steps of the unmanned aerial vehicle system inspection path planning in step S4 are as follows: S41, obtain the latitude and longitude coordinate data of the pipeline leak point; S42, obtain the latitude and longitude coordinate data of the no-fly zone around the pipeline; S43, digitize the latitude and longitude coordinate data of the pipeline and the no-fly zone latitude and longitude coordinate data in the program, and equivalent to points; S44, the unmanned aerial vehicle departs from the airport, avoids the no-fly zone and the no-signal zone, plans a flight path at a specified distance and speed, slows down when the distance to the pipeline leakage point is X before and after approaching the pipeline leakage point, and takes photos of the environment around the pipeline, each photo having the latitude and longitude coordinates of the shooting point; S45, after the re-inspection is completed, the unmanned aerial vehicle returns to the airport according to the original path.

6. The UAV inspection method of claim 1, wherein, In the step S5, the ore slurry pipeline leakage automatic positioning algorithm is designed and trained, and the photographed photos are automatically recognized in real time to identify the ore slurry pipeline leakage and locate the leakage point position. The ore slurry pipeline leakage automatic positioning algorithm automatically recognizes the photographed photos, which is an image change detection model constructed based on an EfficientNetV2 deep neural network. The model realizes two-time image input through the construction of a double-encoder path, replaces the classifier at the back of the network with a decoder with an up-sampling process, and then obtains the change detection result of the two-time images according to the input of the two-time images. The specific steps are as follows: S51, constructing an encoder: in the encoder, the convolutional network is constructed in the optimal configuration through a neural network structure search method. The double-encoder path respectively converts the two-time input into feature maps through convolution operation, and the picture becomes smaller at the same time. The parameters between the two encoders are shared. The first 7 layers of convolution in the EfficientNetV2 are used in the encoder part, which respectively include 1x1 convolution, deep convolution, squeeze excitation module and Dropout process, and the input image is subjected to feature extraction. Each encoder layer corresponds to a decoder layer. The encoder part is composed of two parallel encoders with the same structure, and the convolution kernel parameters are shared. Each is used for feature extraction of one time; S52, constructing a decoder: the up-sampling in the decoder enlarges the picture to restore the picture size reduction caused by the encoder step. In the up-sampling process, the features of the corresponding layers of the encoder are spliced and convolved again through the cascade method. Finally, the same size feature map as the original image is obtained through the decoder stage; The decoder part is connected in a cascade manner, that is, the feature map corresponding to the encoder layer is spliced with the output of the previous decoder layer through a jump connection and used as the input of the decoder; S53, output layer: a fully connected layer is added at the end of the network to map the feature map pixel by pixel to the change probability, and finally obtain the change detection result with the same size as the original input image; S54, input the data set into the constructed EfficientNetV2 deep learning network to train the detection model, identify the changes of the ore slurry pipeline over time, including ore slurry leakage, pipeline deformation, pipeline rupture, etc. Combined with the geographical position of the image, the leakage point coordinates are given. The latitude and longitude coordinate data taken from the unmanned aerial vehicle photo is the coordinate of the pipeline leakage point.

7. The UAV inspection pipeline leak detection and localization method based on negative pressure wave guiding according to claim 6, characterized in that, In the step S54, the data set is periodically photographed along the pipeline by the unmanned aerial vehicle equipped with a camera according to the set fixed flight route. After frame extraction, registration and matching, the photos are used as the change detection data set and are manually labeled. The training set and the test set are randomly divided.

Images