Charging pump maintenance auxiliary system based on point cloud data superposition

Abstract

The invention relates to the technical field of charging pumps, and particularly discloses a charging pump maintenance auxiliary system based on point cloud data superposition, which is used for solving the problems of inaccurate point cloud data, low defect recognition rate and uneven maintenance quality in the traditional charging pump maintenance, and comprises an auxiliary maintenance module and a maintenance data recording and analyzing module, the auxiliary maintenance module comprises an intelligent maintenance guidance module, an auxiliary measurement module and a digital result storage module, the auxiliary maintenance module is connected with a data analysis module, the data analysis module comprises a point cloud data superposition comparison module, a measurement data comparison module and a model data comparison module, and the data analysis module is connected with a data processing module. The data processing module comprises an AI defect identification module and an AI fault prediction module; the point cloud data superposition and comparison module configures an improved dynamic registration algorithm; through point cloud data superposition, a dynamic registration algorithm and a deep learning model, impeller defects are accurately recognized, the fault prediction capability is improved, and misrecognition and omission are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of charging pumps, and more specifically, to a charging pump maintenance assistance system based on point cloud data superposition. Background Art

[0002] The charging pump in nuclear power is used for supplying water to the main shaft pump seal, making up water for the volume control of the chemical volume control system, and supplying water when the high-pressure safety injection system fails. Its disassembly and maintenance mainly include steps such as charging pump disassembly, measurement, repair or replacement, re-measurement, reinstallation, testing, and function re-verification. It has characteristics such as large disassembly difficulty, many measurement parameters, and delicate reinstallation work. Once a failure occurs, it will cause abnormalities in the reactor cooling system, trigger serious safety accidents, and cause huge economic losses. Traditional charging pump maintenance mainly relies on manual experience and simple tools for detection and maintenance. Manual detection is difficult to detect some subtle wear, corrosion and other defects, and for complex internal structures, the detection means are limited. The maintenance process lacks standardized and digital guidance, and it is easy to make operation errors, resulting in uneven maintenance quality. At the same time, the maintenance data recording and management methods are backward, making it difficult to conduct effective data analysis and fault prediction. In particular, the charging pump has a complex structure, with easily occluded areas such as the interface between the impeller and the rotating shaft. Traditional detection methods are difficult to comprehensively and accurately obtain information in these areas, resulting in incomplete detection results and potential fault hazards being missed. The charging pump maintenance steps are cumbersome, involving the disassembly and assembly of multiple components. Without intuitive and detailed operation guidance, it is easy to make operation mistakes, affecting the maintenance quality and progress. In the past, the maintenance data records were scattered, lacking effective classification and indexing, making it difficult to conduct comprehensive analysis of historical data, unable to provide strong support for fault prediction and maintenance decision-making, unable to predict the possible faults of the charging pump in a timely and accurate manner, and only being able to perform maintenance after the fault occurs, increasing the maintenance cost and safety risk. Summary of the Invention

[0003] In order to overcome the above-mentioned defects of the prior art, the present invention provides a charging pump maintenance assistance system based on point cloud data superposition. Through point cloud data superposition, dynamic registration algorithms, and deep learning models, it accurately identifies impeller defects and improves the fault prediction ability, effectively reducing misidentifications and omissions, significantly improving the detection efficiency and reliability, and providing a reliable technical guarantee for the maintenance of charging pumps.

[0004] During the daily maintenance and overhaul of nuclear power plants, the charging pump is transported to a dedicated inspection area for disassembly and repair. The maintenance team is generally composed of 4 to 8 people. To prevent human damage and misoperation, there are procedural requirements during the repair process, and operators need to operate step by step according to the procedures. However, paper work orders cannot constrain the on-site operation process, and workers operate according to their own understanding, resulting in problems such as unqualified repair quality and secondary damage to equipment. The key dimensions of the diameter of the charging pump impeller and the thickness of the blades are within the preset tolerance range. However, when the data fluctuates, the measurement results show that the diameter of the impeller has decreased slightly compared to the previous repair, and the blade thickness has also decreased to a certain extent. Although the change range is still within the allowable tolerance, without combining the change trend of long-term maintenance data and the superposition comparison of past measurement data, it is impossible to intuitively and clearly discover the areas where local material loss occurs on the surface of the impeller blades, resulting in the inability to fully analyze the reasons for this phenomenon. Without the comparison of the original design model, it is also impossible to discover the impact of this change in the blades on the impeller. There are great challenges in traditionally relying on manual and empirical analysis of the data in this scenario. The incompleteness of the data and the lack of experience will both reduce the accuracy of the analysis results, affect the hidden danger inspection of the charging pump during maintenance and overhaul, increase the risk of sudden failures after the repair of the charging pump, and cannot provide strong auxiliary support for the repair of the charging pump.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A maintenance assistance system for a charging pump based on point cloud data superposition, comprising an auxiliary maintenance module and a maintenance data recording and analysis module. The auxiliary maintenance module is connected to the maintenance record and analysis module. The auxiliary maintenance module includes an intelligent maintenance guidance module, an auxiliary measurement module, and a digital result storage module. The auxiliary measurement module is sequentially connected to the intelligent maintenance guidance module and the digital result storage module. The auxiliary maintenance module is connected to a data analysis module. The data analysis module includes a point cloud data superposition and comparison module, a measurement data comparison module, and a model data comparison module. The data analysis module is connected to a data processing module. The data processing module includes an AI defect identification module and an AI fault prediction module. The auxiliary measurement module is connected to the point cloud data superposition and comparison module. The point cloud data superposition and comparison module is configured with an improved dynamic registration algorithm, introducing a dynamic weight factor to adaptively adjust the impeller curvature difference, identifying size thinning, crack, and wear areas at a set threshold level. The AI defect identification module locates and marks the defect area based on a deep learning model that fuses point cloud data features at three voxelization scales of 0.1 mm, 0.5 mm, and 1.0 mm. The AI fault prediction module constructs a dynamic trend model based on historical point cloud data and current measurement results, real-time monitors the changes in blade thickness and impeller diameter, combines gradient features and current cycle features, predicts future faults, and issues early warnings. In the point cloud superposition and comparison module, the dynamic registration algorithm allows the alignment of point cloud data under different scanning perspectives, introducing a dynamic weight factor adapted to the impeller shape of the charging pump to adaptively adjust the registration strategy according to the impeller curvature difference. The formula for the dynamic registration algorithm is:

[0007]

[0008] In the formula: E is the total registration error, i is the index of the point in the point cloud data, n is the total number of points in the point cloud data, ω i is the weight of point p i , dynamically adjusted according to the curvature size of the area where the point is located, are the coordinates of the i-th point of the two point cloud datasets A and B respectively, is the Euclidean distance between the i-th points of the two point cloud datasets A and B, j is the index of the point in the local area of the impeller voxel model, m is the total number of points in the local area of the impeller voxel model, C j is the curvature feature of the j-th point in the local area of the impeller voxel model point cloud data, α is the regularization parameter, ||C j || 2 is the curvature deviation of the j-th point in the local area of the impeller model.

[0009] As a further solution of the present invention, the auxiliary measurement module includes a load rotating table, on which a measurement head is provided. The measurement head includes a binocular lens, a projection head, a projection light source device, a grating generator, an ambient light filtering module, a calibration plate and a vibration adaptive module. The measurement head is connected to a point cloud synthesis computer through a data transmission line, and the point cloud synthesis computer is connected to a data segmentation module. At the bottom of the load rotating table, there is a cross-wheel type tripod, which includes a main shaft and a horizontal cross bar. The length of the main shaft is greater than 2 meters, and the length of the horizontal cross bar is greater than 0.8 meters.

[0010] As a further solution of the present invention, the measurement head operates in a continuous scanning mode. The maximum single scan range is 1000m, and the single scan time is less than or equal to 0.2 seconds. It is located directly above the load rotating table. The binocular lens includes a left camera and a right camera, which are located at the front of the measurement head and face the object to be scanned at a set angle. The projection head is located in front of the binocular lens and is 0.3 cm higher than the base camera lens. The projection light source device is located 4.2 cm below the midline position between the projection head and the base double camera lens. The projection light source device uses LEN blue light with a wavelength range of 440 - 480nm as the projection light source. The grating generator cooperates with the projection head to generate a digital structured cross grating pattern and projects it onto the target surface, which coincides with the field of view of the binocular lens. The deformed grating pattern is captured by the binocular lens and embedded in the projection head. The ambient light filtering module is used to ensure that the red laser irradiation has no impact on the surface of the measured object. The interface of the calibration plate provides graphic and text guidance for the calibration of the measurement head and the setting of the lens. The vibration adaptive module is used to detect the vibration during the measurement process. When the vibration exceeds the preset range, it issues an instruction to rescan 3 to 5 times, and automatically continues to scan after the vibration stops.

[0011] As a further solution of the present invention, the measurement head is wirelessly connected to an optical probe. The optical probe is a ruby probe, which real-time displays the position of the probe ball center and performs real-time dot detection. And the point data and the scanned model are in the same coordinate system, which is used to detect the geometric dimensions of cylinders, round holes, oblong holes, cones, spherical curved surface points and planes in the blind spot area.

[0012] As a further solution of the present invention, the scanning perspectives of the measurement head include: the binocular lens perspective, the left scanning perspective, and the right scanning perspective. The binocular lens perspective is the scanning perspective of the binocular lens, the left scanning perspective is the scanning perspective composed of the left camera and the projection head, and the right scanning perspective is the scanning perspective composed of the right camera and the projection head.

[0013] As a further solution of the present invention, the data analysis module further includes an inverse projection module connected to both the point cloud data overlay comparison module, the measurement data comparison module, and the model data comparison module, which is used to annotate the three-dimensional coordinates of the measurement data and the point cloud comparison detection data on the workpiece image of the upper filling pump element model.

[0014] As a further solution of the present invention, the process of the intelligent maintenance guidance module performing visual intelligent maintenance guidance is as follows:

[0015] Step 1, Visual Guidance: Based on the 3D model of the charging pump, provide a 3D animation of the charging pump assembly and disassembly demonstration. The demonstration animation includes the assembly and disassembly sequence, operating tools, operating steps, and operating precautions of the charging pump components. The demonstration process is accompanied by text, images, videos, and voice to explain the details and key points of each step;

[0016] Step 2, assisted maintenance: Provide a digital maintenance method based on the on-site maintenance process and steps, with the order of component disassembly and assembly as the node, guide the maintenance personnel to scan the charging pump components one by one, save the digital results of measurement data, point cloud data, and model data, compare the scanned model data with the component model, and if it does not match the node model, it will prompt that the steps are abnormal, and guide the maintenance personnel to scan the charging pump components according to the correct steps. Each guided step includes demonstration operation, voice introduction to guide scanning, and saving data. Let the maintenance personnel watch the operation demonstration first, and then guide scanning and saving data after they are familiar with the steps, and monitor the components of the maintenance process;

[0017] Step 3, record maintenance data: record and save maintenance start time, maintenance personnel, disassembly and assembly sequence, and charging pump component scan data, support label division according to unit, charging pump number, and maintenance time, save maintenance data, form a multi-time point maintenance database based on the charging pump, form a dual-dimensional data index of the charging pump number and maintenance time, and manage maintenance data based on the number and equipment;

[0018] Step 4, generate maintenance report: automatically generate maintenance summary report according to the last shift of maintenance project.

[0019] As a further solution of the present invention, the process of the AI ​​defect recognition module identifying and locating the defective area of ​​the charging pump impeller includes:

[0020] Step 1: Data collection and preprocessing: Obtain the point cloud data of the charging pump impeller after alignment using the improved dynamic registration algorithm, perform data cleaning, process the point cloud data according to the predetermined voxelization scale, and convert it into a structure that is suitable for deep learning model input;

[0021] Step 2: Feature extraction: According to different voxelization scales, different levels of texture, shape and depth feature information are extracted from the point cloud data. The voxelized point cloud data is input into the deep learning model, and local features are extracted through the convolution layer. The point cloud data at each scale passes through three convolution layers to extract the feature data of the impeller surface.

[0022] Step 3: Feature fusion: Fuse the data from different scales, perform weighted fusion based on the importance of each scale’s features to defect identification, and splice the features of different scales in the feature dimension;

[0023] Step 4, Defect Classification and Prediction: Defect classification is performed through the fully connected layer of the deep neural network. The feature vector of each point cloud is matched with predefined defect categories, and the probability value of whether each point is a defect is output. When the probability is greater than the set threshold, it is marked as a defect area. According to the probability map output by the model, the location of the defect is determined, and the defect area is marked, and the coordinate range of the defect is determined by the regression method;

[0024] Step 5, Defect Area Annotation and Post-processing: After the point cloud area with defects is detected, further region extraction is performed to calibrate the shape and size of the defects. Smoothing processing is performed through a Gaussian smoothing filter. The misidentified areas in the marked defect areas are corrected based on the point cloud data overlap technology and the improved dynamic registration algorithm. The defect areas are highlighted in the 3D point cloud map through a 3D visualization tool, and the wear and damage degree of the impeller is analyzed through the defect recognition results.

[0025] As a further solution of the present invention, in Step 5, the point cloud data of the first scan and the current scan are used, and the misidentified areas in the marked defect areas are corrected based on the point cloud data overlap technology and the improved dynamic registration algorithm by using the diameter and blade thickness characteristics of the impeller. The misidentification area correction formula is:

[0026] P cor (p i ) = P smo (p i )·[1 - I(p i )•E(p i )•F(p i )]

[0027] In the formula: P cor (p i ) is the position coordinate of the i-th point after misidentification area correction, P smo (p i ) is the position coordinate of the i-th point after smoothing processing, I(p i ) is the misidentification correction index function. If the point p i belongs to the misidentified area, then I(p i ) = 1, otherwise it is 0. E(p i ) is the error threshold function, which is used to measure whether the error of the point p i exceeds the set threshold, and is determined based on the point cloud registration error and the geometric feature change amount. F(p i ) is a function for judging whether the point p i belongs to the normal area based on the impeller diameter and blade thickness. If the point p i does not exceed the allowable range of the impeller blade diameter and thickness, F(p i) = 1 if it is, otherwise 0.

[0028] As a further solution of the present invention, the process of the AI fault prediction module for impeller fault prediction includes:

[0029] Step I: Fusing historical data and current data: Fuse historical point cloud data with current impeller blade thickness and impeller diameter data, and construct a dynamic trend model through time series analysis to show the changing trends of impeller blade thickness and impeller diameter over time;

[0030] Step II, Obtaining gradient features and periodic features: Calculate the change gradients of impeller blade thickness and impeller diameter, analyze their change gradients, and perform periodic analysis of the data;

[0031] Step III, Real-time monitoring and fault prediction: Real-time monitor the parameter changes of the impeller recorded during previous repairs and overhauls, combine with the dynamic trend model to perform machine learning prediction on the current data. According to the comparison between the prediction result output by the model and the preset threshold, when the prediction result exceeds the preset threshold, identify potential faults of the component, and issue a warning through the intelligent maintenance guidance module in the auxiliary maintenance module to remind the maintenance personnel to take preventive maintenance measures.

[0032] Technical effects of a maintenance assistance system for a charging pump based on point cloud data superposition proposed by the present invention:

[0033] By applying point cloud data overlapping technology, an improved dynamic registration algorithm, and a defect identification and trend prediction model based on deep learning, and through the precise registration and analysis of point cloud data, the present invention not only improves the accuracy and reliability of maintenance operations, but also reduces misidentifications and omissions caused by traditional reliance on experience judgment, improves detection efficiency and fault prediction ability, can discover potential problems of impeller components in advance, preventively reduce the occurrence of sudden failures, and provides a reliable technical guarantee for the maintenance and operation of the charging pump. Description of the Drawings

[0034] Figure 1 It is a disassembly interface and a model disassembly interface diagram of the charging pump equipment in the present invention;

[0035] Figure 2 It is a schematic diagram of intelligent maintenance of the charging pump in the present invention;

[0036] Figure 3 It is an analysis interface for charging pump maintenance data in the present invention;

[0037] Figure 4 It is a report viewing interface in the present invention;

[0038] Figure 5 It is a disassembly interface diagram of the charging pump coupling in the present invention;

[0039] Figure 6 It is a data analysis diagram for the maintenance process;

[0040] Figure 7 It is a diagram of the maintenance management interface;

[0041] Figure 8 It is a comparison diagram of the measured data and model data of the charging pump assembly;

[0042] Figure 9 It is a physical diagram of the measuring head;

[0043] Figure 10 It is a model diagram of the tripod;

[0044] Figure 11 It is a physical diagram of the load rotating table;

[0045] Figure 12 It is a schematic diagram of the divided area of the load rotating table;

[0046] Figure 13 It is an overlapping effect diagram of the detection of the measuring head and the actual image, a display diagram of the back-projection function, and a dynamic tracking function diagram;

[0047] Figure 14 It is a schematic diagram of the use of the optical probe. Specific implementation manners

[0048] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0049] To achieve the above object, the present invention provides the following technical solutions:

[0050] Embodiment 1

[0051] As Figures 1 to 8As shown in the figure, a charging pump maintenance assistance system based on point cloud data superposition proposed by the present invention includes an auxiliary maintenance module and a maintenance data recording and analysis module. The auxiliary maintenance module is connected to the maintenance record and analysis module. The auxiliary maintenance module includes an intelligent maintenance guidance module, an auxiliary measurement module, and a digital result storage module. The auxiliary measurement module is sequentially connected to the intelligent maintenance guidance module and the digital result storage module. The auxiliary maintenance module is connected to a data analysis module. The data analysis module includes a point cloud data superposition comparison module, a measurement data comparison module, and a model data comparison module. The data analysis module is connected to a data processing module. The data processing module includes an AI defect identification module and an AI fault prediction module. The auxiliary measurement module is connected to the point cloud data superposition comparison module. The point cloud data superposition comparison module is configured with an improved dynamic registration algorithm, which introduces a dynamic weight factor to adaptively adjust the impeller curvature difference, and identifies the size thinning, crack, and wear areas of the set threshold level. The AI defect identification module locates and marks the defect area based on the deep learning model that fuses the point cloud data features of three voxelization scales of 0.1mm, 0.5mm, and 1.0mm. The AI fault prediction module constructs a dynamic trend model based on the historical point cloud data and the current measurement results, monitors the changes in blade thickness and impeller diameter in real time, combines the gradient feature and the current cycle feature, predicts the future occurrence of faults, and issues early warnings. In the point cloud superposition comparison module, the dynamic registration algorithm allows the point cloud data under different scanning perspectives to be aligned, and introduces a dynamic weight factor adapted to the impeller shape of the charging pump to adaptively adjust the registration strategy according to the impeller curvature difference. The formula of the dynamic registration algorithm is:

[0052]

[0053] In the formula: E is the total registration error, i is the index of the point in the point cloud data, n is the total number of the point cloud data, ω i is the weight of point p i , which is dynamically adjusted according to the curvature size of the area where the point is located, are the coordinates of the i-th point of the two point cloud data sets A and B respectively, is the Euclidean distance between the i-th points of the two point cloud data sets A and B, j is the index of the point in the local area of the impeller voxel model, m is the total number of points in the local area of the impeller voxel model, C j is the curvature feature of the j-th point in the local area of the impeller voxel model point cloud data, α is the regularization parameter, ||C j || 2 is the curvature deviation of the j-th point in the local area of the impeller voxel model.

[0054] To clearly and explicitly illustrate the technical effects of the dynamic registration algorithm in the above technical solution, the data of Test Group 1, Test Group 2, and Test Group 3 are selected for specific illustration. The test equipment of the test groups uses the auxiliary measurement module described in the technical solution, and the scanning environment maintains stable temperature and humidity conditions. Test Group 1 is based on the first overhaul data of the upper filling pump impeller, focusing on detecting whether there are early wear or thinning areas on the impeller surface. During the first overhaul, the impeller diameter is within the standard tolerance range, local roughness begins to appear at the blade edge, and it has not reached the stage of severe wear. The thickness of the blade edge is thinned by 0.3 mm, and the local surface roughness increases, but it has not reached the failure threshold. Test Group 2 is based on the first major overhaul data. At this time, some areas of the impeller blades have been repaired, and damaged blades have been replaced. The slight warping at the blade joint and the surface texture mismatch in the repaired area will cause changes in local hardness or surface finish. The dynamic registration algorithm is used to identify the geometric deformation in the joint area and check the consistency of the surface texture after repair. Test Group 3 is based on the historical data after the second major overhaul, focusing on detecting the fatigue and potential cracks of the impeller after long-term use. Microcracks appear at the blade edge, and the cracks expand radially and the curvature changes violently. Through the point cloud superposition technology and the dynamic registration algorithm, the crack position and morphology are accurately identified, and the impact of the cracks on the overall structure of the impeller is evaluated. The comparison group uses traditional manual experience detection methods and simple registration techniques, and selects the total registration error value E and curvature deviation of the four comparison groups, ||C j || 2 , as shown in Table 1:

[0055] Table 1 Comparison Table of Key Parameters in the Detection Area

[0056]

[0057] The reduction of the curvature deviation indicates that the dynamic registration algorithm significantly improves the alignment accuracy of the point cloud data, especially when dealing with complex shapes (such as the impeller surface), reducing the errors caused by curvature differences.

[0058] The changes in parameters such as the curvature change rate, curvature deviation, and area of the thickness thinning area indicate that by improving the algorithm, the defect areas of the impeller can be more accurately identified and calibrated, avoiding the errors caused by roughness or uneven surfaces in traditional methods.

[0059] Overall, the reduction of E (registration error) and curvature deviation improves the accuracy of the defect area, avoids the risk of missed detection or misjudgment, and enhances the reliability of fault prediction and maintenance assistance.

[0060] Example 2

[0061] As Figures 9 to 12 shown, the auxiliary measurement module includes a load rotating table ( Figure 11Show the overall rotating stage. The rotating stage is divided into several regions (such as Figure 12 shown). It can place several upper filling pump components on the rotating stage together and complete the scanning of all components on the rotating stage at one time. After the scanning is completed, it is automatically divided into several scanning parts. There is a measuring head on the rotating stage (such as Figure 9 shown). The measuring head includes a binocular lens, a projection head, a projection light source device, a grating generator, an ambient light filtering module, a calibration plate and a vibration adaptive module. The measuring head is connected to a point cloud synthesis computer through a data transmission line. The point cloud synthesis computer is connected to a data segmentation module. There is a cross-wheel type tripod at the bottom of the rotating stage (such as Figure 10 shown). The cross-wheel type tripod includes a main shaft and a horizontal cross bar. The length of the main shaft is greater than 2 meters, and the length of the horizontal cross bar is greater than 0.8 meters. It can move freely indoors, which is convenient for measurement.

[0062] The measuring head is in a continuous scanning working mode. The maximum single scan range is 1000m, and the single scan time is less than or equal to 0.2 seconds. It is located directly above the rotating stage. The binocular lens includes a left camera and a right camera, which are located at the front of the measuring head and face the object to be scanned at a set angle. The projection head is located in front of the binocular lens and is 0.3 cm higher than the base camera lens. The projection light source device is located 4.2 cm below the midline position between the projection head and the base binocular camera lens. The projection light source device uses LEN blue light with a wavelength range of 440-480 nm as the projection light source. There is no need to spray powder on the label points for scanning. The test data is safe and reliable. During the scanning process, noise points are effectively reduced or controlled, and the orange peel phenomenon of the scanning components is avoided. The light projected by this technology measuring head is uniform in brightness throughout the range, and the four corners cannot be darker than the middle. It is audited by comparison. The grating generator cooperates with the projection head to generate a digital structured cross grating pattern and project it onto the target surface, which coincides with the field of view of the binocular lens. The deformed grating pattern is captured by the binocular lens and embedded in the projection head. The ambient light filtering module is used to ensure that the red laser irradiation on the surface of the measured object has no influence. The interface of the calibration plate provides graphic and text guidance for measuring head calibration and lens setting. The vibration adaptive module is used to detect the vibration during the measurement process. When the vibration exceeds the preset range, it issues an instruction to rescan 3 to 5 times and automatically continuously scans after the vibration stops. The working temperature range of the measuring head is +5°C to +35°C, without condensation. When operating at different temperatures, the temperature of the calibration plate is identified with a thermometer and input, and automatic temperature compensation is performed. The image signal and control signal are transmitted through a 10-GigE data link established by an optical fiber cable.

[0063] It should be noted that the measuring head is connected to an optical probe through a wireless connection method (such as Figure 14As shown in the figure, the optical probe is a ruby probe. The probe rod of the optical probe is a pre-calibrated group of label points (the relationship between the label points and the center of the sphere). By measuring the group of label points on the probe rod, the center-of-sphere measurement points are automatically established, which has the characteristics of dynamic reference, can display the position of the probe center in real time, and perform real-time dot detection. Moreover, the point data and the scanned model are in the same coordinate system, which is used to detect the geometric dimensions of cylinders, round holes, oblong holes, cones, spherical curved surface points and planes in the blind spot area. The scanning perspectives of the measuring head include: the binocular lens perspective, the left scanning perspective, and the right scanning perspective. The binocular lens perspective is the scanning perspective of the binocular lens, the left scanning perspective is the scanning perspective composed of the left camera and the projection head, and the right scanning perspective is the scanning perspective composed of the right camera and the projection head. Such a triple-perspective scanning technology can ensure that more occluded area features of the complex structure of the upper filling pump assembly are measured, and the detailed structure of the part can be scanned better.

[0064] It should be noted that the data analysis module also includes an inverse projection module connected to the point cloud data overlay comparison module, the measurement data comparison module, and the model data comparison module, which is used to annotate the three-dimensional coordinates of the measurement data and the point cloud comparison detection data on the workpiece image of the upper filling pump primitive model to enhance the recognition of the component (as Figure 13 shown).

[0065] The auxiliary measurement module combines full-area contact scanning and contact 3D point measurement through an optical probe. Through component alignment and positioning, it is transmitted to the real world in a virtual alignment manner to achieve dynamic tracking. The measurement data is inverse projected onto the component to achieve inverse projection measurement. During the scanning process, the screen changes of the component are displayed in real time. Measurements are adjusted at different angular positions, and position and volume parameters assist the scanning in real time, and it supports laser focusing with scanning.

[0066] The interface for detecting the upper filling pump blades of the auxiliary detection module is provided with an airfoil detection work area, a toolbar with detection process guidance, and an I-Inspect button that includes the function of detecting the airfoil of the profile section, such as stacking points, centerlines, geometric dimensions of the inlet and outlet edges, chord lines, supports the detection of the profile shape and position, is provided with an offsettable centerline, and expert parameter settings for the detection of the profile shape and position. It also has the detection of the profile edge thickness, the scanning and detection of the blisk, the detection of the bow and tilt, and analyzes the stacking axis, the mid-arc of the stacking points, the centroid of the profile, the profile shape and position, the profile boundary points, the profile boundary circles, the profile boundary ellipses, the profile chord lines (axial chord, double tangent chord, maximum chord, aerodynamic chord), the maximum profile thickness, and the profile boundary thickness.

[0067] It should be noted that as Figure 2 shown, the process of the intelligent maintenance guidance module for visual intelligent maintenance guidance is as follows:

[0068] Step 1, Visual Guidance: Based on the 3D model of the charging pump, provide a 3D animation of the charging pump assembly and disassembly demonstration. The demonstration animation includes the assembly and disassembly sequence, operating tools, operating steps, and operating precautions of the charging pump components. The demonstration process is accompanied by text, images, videos, and voice to explain the details and key points of each step;

[0069] Step 2, assisted maintenance: Provide a digital maintenance method based on the on-site maintenance process and steps, with the order of component disassembly and assembly as the node, guide the maintenance personnel to scan the charging pump components one by one, save the digital results of measurement data, point cloud data, and model data, compare the scanned model data with the component model, and if it does not match the node model, it will prompt that the steps are abnormal, and guide the maintenance personnel to scan the charging pump components according to the correct steps. Each guided step includes demonstration operation, voice introduction to guide scanning, and saving data. Let the maintenance personnel watch the operation demonstration first, and then guide scanning and saving data after they are familiar with the steps, and monitor the components of the maintenance process;

[0070] Step 3, record maintenance data: record and save maintenance start time, maintenance personnel, disassembly and assembly sequence, and charging pump component scan data, support label division according to unit, charging pump number, and maintenance time, save maintenance data, form a multi-time point maintenance database based on the charging pump, form a dual-dimensional data index of the charging pump number and maintenance time, and manage maintenance data based on the number and equipment;

[0071] Step 4, generate maintenance report: automatically generate maintenance summary report according to the last shift of maintenance project.

[0072] The auxiliary maintenance module of the charging pump also has a charging pump fault tree database, spare parts management function, and maintenance management function, such as Figure 8 As shown, a comparison chart of the measured data and the model data is displayed, which can intuitively show the comparison results between the measured data and the plain model data in each step of the standard maintenance process of each component in the maintenance process.

[0073] It should be noted that the process of AI defect recognition module identifying and locating defective areas of charging pump impeller includes:

[0074] Step 1: Data collection and preprocessing: Obtain the point cloud data of the charging pump impeller after alignment using the improved dynamic registration algorithm, perform data cleaning, process the point cloud data according to the predetermined voxelization scale, and convert it into a structure that is suitable for deep learning model input;

[0075] Step 2: Feature extraction: According to different voxelization scales, different levels of texture, shape and depth feature information are extracted from the point cloud data. The voxelized point cloud data is input into the deep learning model, and local features are extracted through the convolution layer. The point cloud data at each scale passes through three convolution layers to extract the feature data of the impeller surface.

[0076] Step 3, Feature Fusion: Perform feature fusion on data from different scales, perform weighted fusion according to the importance of features at each scale for defect recognition, and splice features at different scales in the feature dimension;

[0077] Step 4, Defect Classification and Prediction: Perform defect classification through the fully connected layer of the deep neural network, match the feature vector of each point cloud with the predefined defect categories, output the probability value of whether each point is a defect, when the probability is greater than the set threshold, mark it as a defect area, determine the location of the defect according to the probability map output by the model, and label the defect area, and determine the coordinate range of the defect through the regression method;

[0078] Step 5, Defect Area Annotation and Post-processing: After detecting the point cloud area with defects, further perform area extraction, calibrate the shape and size of the defects, perform smoothing processing through a Gaussian smoothing filter, correct the misidentified areas of the marked defect areas based on the point cloud data overlap technology and the improved dynamic registration algorithm, highlight the defect areas in the three-dimensional point cloud map through a three-dimensional visualization tool, and analyze the wear and damage degree of the impeller based on the defect recognition results.

[0079] By combining the dynamic registration algorithm and deep learning technology, the defect areas of the charging pump impeller can be accurately identified and located. The multi-scale method of feature extraction and feature fusion ensures the comprehensive retention of details at different levels, and accurately classifies and predicts defect areas such as micro-cracks and thinning. Combining Gaussian filtering smoothing and three-dimensional visualization annotation not only optimizes the data post-processing, but also reduces the misidentified areas, provides intuitive and reliable data support for maintenance decisions, and improves the accuracy, efficiency and maintenance safety of the overall detection.

[0080] It should be noted that in Step 5, the point cloud data of the first scan and the current scan are used, and the misidentified areas of the marked defect areas are corrected based on the point cloud data overlap technology and the improved dynamic registration algorithm using the diameter and blade thickness features of the impeller. The misidentified area correction formula is:

[0081] P cor (p i )=P smo (p i )·[1-I(p i )·E(p i )·F(p i )]

[0082] In the formula: P cor (p i ) is the position coordinate of the i-th point after correcting the misidentified area, P smo (p i) is the position coordinate of the i-th point after smoothing processing, I(p i ) is the misrecognition correction index function. If the point p i belongs to the misrecognition area, then I(p i ) = 1, otherwise it is 0. E(p i ) is the error threshold function, which is used to measure whether the error of the point p i exceeds the set threshold, and is determined based on the point cloud registration error and the geometric feature change amount. F(p i ) is a function for judging whether the point p i belongs to the normal area based on the impeller diameter and blade thickness. If the point p i does not exceed the allowable range of the impeller blade diameter and thickness, F(p i ) = 1, otherwise it is 0.

[0083] Due to the complex structure of the upper charging pump impeller and the high maintenance difficulty, traditional methods cannot accurately identify local defect areas. Especially in the case of errors in point cloud data and significant differences in impeller curvature, it is easy to cause mislabeling. By combining the point cloud data of the first scan and the current scan with the diameter and blade thickness characteristics of the impeller to correct the misrecognition area, the improved dynamic registration algorithm can be used to adaptively adjust the registration weight, accurately exclude mislabeled points, ensure more accurate annotation of the impeller surface defect area, and thus avoid subsequent analysis and maintenance quality problems caused by data loss or mislabeling. By using the point cloud data of the first scan and the current scan, and performing misrecognition correction through the improved dynamic registration algorithm that combines the impeller diameter and blade thickness characteristics, it can significantly reduce the mislabeled points caused by curvature differences and registration errors, reduce omissions or false detections in defect area annotation, improve the alignment accuracy of point cloud data, optimize the accurate calibration of defect areas, effectively improve the reliability of data analysis and maintenance quality, ensure that subtle defects on the complex surface of the impeller can be fully discovered, provide more reliable and efficient data support for the maintenance decision-making of the upper charging pump, reduce the operation risk, and enhance the safety and efficiency of maintenance work.

[0084] It should be noted that the process of the AI fault prediction module for impeller fault prediction includes:

[0085] Step I: Fuse historical data and current data: Use historical point cloud data and current impeller blade thickness and impeller diameter data for fusion, and construct a dynamic trend model through time series analysis to show the change trends of impeller blade thickness and impeller diameter over time;

[0086] Step II, Obtain gradient features and periodic features: By calculating the change gradients of impeller blade thickness and impeller diameter, analyze their change gradients, and perform periodic analysis of the data;

[0087] Step III, Real-time Monitoring and Fault Prediction: Real-time monitor the parameter changes of the impeller recorded during previous repairs and overhauls, combine with the dynamic trend model to perform machine learning prediction on the current data, and according to the comparison between the prediction result output by the model and the preset threshold, when the prediction result exceeds the preset threshold, identify potential faults of the component, and issue a warning through the intelligent maintenance guidance module in the auxiliary maintenance module to remind the maintenance personnel to take preventive maintenance measures.

[0088] In summary, by applying the point cloud data overlapping technology, the improved dynamic registration algorithm, and the defect identification and trend prediction model of deep learning, and through the precise registration and analysis of the point cloud data, the present invention not only improves the accuracy and reliability of maintenance operations, but also reduces the misidentification and omission caused by traditional experience-based judgment, improves the detection efficiency and fault prediction ability, can discover potential problems of the impeller component in advance, and preventively reduce the occurrence of sudden failures, providing a reliable technical guarantee for the maintenance and operation of the charging pump.

[0089] The above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present application, and all should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0090] Finally: The above is only the preferred embodiment of the present invention and is not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A charging pump maintenance assistance system based on point cloud data superposition, comprising an auxiliary maintenance module and a maintenance data recording and analysis module, characterized in that, The auxiliary maintenance module includes an intelligent maintenance guidance module, an auxiliary measurement module, and a digital result storage module. The auxiliary measurement module is successively connected to the intelligent maintenance guidance module and the digital result storage module. The auxiliary maintenance module is connected to a data analysis module. The data analysis module includes a point cloud data overlay and comparison module, a measurement data comparison module, and a model data comparison module. The data analysis module is connected to a data processing module. The data processing module includes an AI defect identification module and an AI fault prediction module. The auxiliary measurement module is connected to the point cloud data overlay and comparison module. The point cloud data overlay and comparison module is configured with an improved dynamic registration algorithm, introducing a dynamic weight factor to adaptively adjust the impeller curvature difference, identifying the dimension thinning, crack, and wear areas of the set threshold level. The AI defect identification module locates and marks the defect area based on the deep learning model that fuses the point cloud data features at three voxelization scales of 0.1mm, 0.5mm, and 1.0mm. The AI fault prediction module constructs a dynamic trend model based on the historical point cloud data and the current measurement results, real-time monitors the changes in blade thickness and impeller diameter, combines the gradient features and the current cycle features, predicts the future occurring faults, and issues early warnings. In the point cloud overlay and comparison module, the dynamic registration algorithm allows the alignment of point cloud data under different scanning perspectives, introducing a dynamic weight factor that adapts to the shape of the upper charging pump impeller to adaptively adjust the registration strategy according to the impeller curvature difference. The formula of the dynamic registration algorithm is: Where: E is the total registration error, i is the index of the point in the point cloud data, n is the total number of points in the point cloud data, ω i is the weight of point p i , which is dynamically adjusted according to the curvature of the area where the point is located, are the coordinates of the i-th points of two point cloud datasets A and B respectively, is the Euclidean distance between the i-th points of two point cloud datasets A and B, j is the index of the points in the local area of the impeller element model, m is the total number of points in the local area of the impeller element model, C j is the curvature feature of the j-th point in the local area of the impeller element model point cloud data, α is the regularization parameter, ||C j || 2 is the curvature deviation of the j-th point in the local area of the impeller element model.

2. The auxiliary system for upper charging pump maintenance based on point cloud data superposition according to claim 1, wherein, The auxiliary measurement module includes a load rotating table. A measuring head is provided on the load rotating table. The measuring head includes a binocular lens, a projection head, a projection light source device, a grating generator, an ambient light filtering module, a calibration plate, and a vibration adaptive module. The measuring head is connected to a point cloud synthesis computer through a data transmission line. The point cloud synthesis computer is connected to a data segmentation module. A cross-wheel type tripod is provided at the bottom of the load rotating table. The cross-wheel type tripod includes a main shaft and a horizontal cross bar. The length of the main shaft is greater than 2 meters, and the length of the horizontal cross bar is greater than 0.8 meters.

3. The charging pump maintenance assistance system based on point cloud data superposition according to claim 2, characterized in that, The measuring head is in a continuous scanning working mode. The maximum single scan range is 1000m, and the single scan time is less than or equal to 0.2 seconds. It is located directly above the load rotating table. The binocular lens includes a left camera and a right camera, which are located at the front of the measuring head and face the scanned object at a set angle. The projection head is located in front of the binocular lens, 0.3 cm higher than the base camera lens. The projection light source device is located 4.2 cm below the midline position between the projection head and the base binocular camera lens. The projection light source device uses LEN blue light with a wavelength range of 440 - 480nm as the projection light source. The grating generator cooperates with the projection head to generate a digital structured cross grating pattern and projects it onto the target surface, which coincides with the field of view of the binocular lens. The deformed grating pattern is captured through the binocular lens and embedded in the projection head. The ambient light filtering module is used to ensure that the red laser has no influence when irradiating the surface of the measured object. The interface of the calibration plate provides graphic and text guidance for the calibration of the measuring head and the setting of the lens. The vibration adaptive module is used to detect the vibration during the measurement process. When the vibration exceeds the preset range, it issues an instruction to rescan 3 to 5 times and automatically continuously scans after the vibration stops.

4. The charging pump maintenance assistance system based on point cloud data superposition according to claim 3, characterized in that, The measuring head is connected to an optical probe via a wireless connection. The optical probe is a ruby ​​probe, which displays the probe sphere center position in real time and performs real-time point detection. The point data and the scanning model are in the same coordinate system. It is used to detect the geometric dimensions of cylinders, circular holes, oblong holes, cones, spherical surface points and planes in the blind spot area.

5. The upper charging pump maintenance assistance system based on point cloud data superposition according to claim 4, characterized in that, The scanning angle of the measuring head includes: binocular lens angle, left scanning angle, and right scanning angle. The binocular lens angle is the scanning angle of the binocular lens, the left scanning angle is the scanning angle composed of the left camera and the projection head, and the right scanning angle is the scanning angle composed of the right camera and the projection head.

6. The charging pump maintenance assistance system based on point cloud data superposition according to claim 4, characterized in that, The data analysis module also includes a back projection module connected to the point cloud data overlay comparison module, the measurement data comparison module, and the model data comparison module, which is used to inject the three-dimensional coordinates of the measurement data and the point cloud comparison detection data onto the workpiece image of the upper filling pump plain mold.

7. The charging pump maintenance assistance system based on point cloud data superposition according to claim 1, wherein The process of visual intelligent maintenance guidance by the intelligent maintenance guidance module is as follows: Step 1, Visual Guidance: Based on the 3D model of the charging pump, provide a 3D animation of the charging pump assembly and disassembly demonstration. The demonstration animation includes the assembly and disassembly sequence, operating tools, operating steps, and operating precautions of the charging pump components. The demonstration process is accompanied by text, images, videos, and voice to explain the details and key points of each step; Step 2, assisted maintenance: Provide a digital maintenance method based on the on-site maintenance process and steps, with the order of component disassembly and assembly as the node, guide the maintenance personnel to scan the charging pump components one by one, save the digital results of measurement data, point cloud data, and model data, compare the scanned model data with the component model, and if it does not match the node model, it will prompt that the steps are abnormal, and guide the maintenance personnel to scan the charging pump components according to the correct steps. Each guided step includes demonstration operation, voice introduction to guide scanning, and saving data. Let the maintenance personnel watch the operation demonstration first, and then guide scanning and saving data after they are familiar with the steps, and monitor the components of the maintenance process; Step 3, record maintenance data: record and save maintenance start time, maintenance personnel, disassembly and assembly sequence, and charging pump component scan data, support label division according to unit, charging pump number, and maintenance time, save maintenance data, form a multi-time point maintenance database based on the charging pump, form a dual-dimensional data index of the charging pump number and maintenance time, and manage maintenance data based on the number and equipment; Step 4, generate maintenance report: automatically generate maintenance summary report according to the last shift of maintenance project.

8. The auxiliary system for upper charging pump maintenance based on point cloud data superposition according to claim 1, characterized in that, The process of AI defect recognition module identifying and locating defective areas of charging pump impeller includes: Step 1: Data collection and preprocessing: Obtain the point cloud data of the charging pump impeller after alignment using the improved dynamic registration algorithm, perform data cleaning, process the point cloud data according to the predetermined voxelization scale, and convert it into a structure that is suitable for deep learning model input; Step 2: Feature extraction: According to different voxelization scales, different levels of texture, shape and depth feature information are extracted from the point cloud data. The voxelized point cloud data is input into the deep learning model, and local features are extracted through the convolution layer. The point cloud data at each scale passes through three convolution layers to extract the feature data of the impeller surface. Step 3, Feature Fusion: Perform feature fusion on data from different scales, perform weighted fusion according to the importance of features at each scale for defect recognition, and splice features at different scales in the feature dimension; Step 4, Defect Classification and Prediction: Perform defect classification through the fully connected layer of the deep neural network, match the feature vector of each point cloud with the predefined defect categories, output the probability value of whether each point is a defect, and when the probability is greater than the set threshold, mark it as a defect area. Determine the location of the defect based on the probability map output by the model, mark the defect area, and determine the coordinate range of the defect through the regression method; Step 5, Defect Area Annotation and Post-processing: After detecting the point cloud area with defects, further perform area extraction, calibrate the shape and size of the defects, perform smoothing processing through a Gaussian smoothing filter, correct the misidentified areas of the marked defect areas based on the point cloud data overlap technology and the improved dynamic registration algorithm, highlight the defect areas in the 3D point cloud map through a 3D visualization tool, and analyze the wear and damage degree of the impeller based on the defect recognition results.

9. The a charging pump maintenance assistance system based on point cloud data superposition according to claim 1, characterized in that, In In Step 5, use the point cloud data of the first scan and the current scan, and correct the misidentified areas of the marked defect areas based on the point cloud data overlap technology and the improved dynamic registration algorithm by using the diameter and blade thickness features of the impeller. The misidentification area correction formula is: P cor (p i ) = P smo (p i )·[1 - I(p i )•E(p i )•F(p i )] Where: P cor (p i ) is the position coordinate of the i-th point after the misrecognition area is corrected, P smo (p i ) is the position coordinate of the i-th point after smoothing, I(p i ) is the misrecognition correction index function. If the point p i belongs to the misrecognition area, then I(p i ) = 1, otherwise it is 0. E(p i ) is the error threshold function, which is used to measure whether the error of the point p i exceeds the set threshold, and is determined based on the point cloud registration error and the geometric feature change amount. F(p i ) is a function for judging whether the point p i belongs to the normal area based on the impeller diameter and blade thickness. If the point p i does not exceed the allowable range of the impeller blade diameter and thickness, F(p i ) = 1, otherwise it is 0.

10. The a charging pump maintenance assistance system based on point cloud data superposition according to claim 1, characterized in that, The process of the AI fault prediction module for impeller fault prediction includes: Step I: Fuse historical data and current data: Fuse historical point cloud data with the current impeller blade thickness and impeller diameter data, and construct a dynamic trend model through time series analysis to show the change trends of the impeller blade thickness and impeller diameter over time; Step II, Obtain gradient features and periodic features: Calculate the change gradients of the impeller blade thickness and impeller diameter, analyze their change gradients, and perform periodic analysis of the data; Step III, Real-time Monitoring and Fault Prediction: Real-time monitor the parameter changes of the impeller recorded in previous inspections and overhauls, perform machine learning prediction on the current data in combination with the dynamic trend model, and compare the prediction results output by the model with the preset threshold. When the prediction result exceeds the preset threshold, identify potential faults in the components, and issue a warning through the intelligent maintenance guidance module in the auxiliary maintenance module to remind the maintenance personnel to take preventive maintenance measures.

Images