A method for detecting the air tightness of a welding seam of an automobile axle housing

Abstract

The application relates to the technical field of air tightness detection, and particularly discloses a method for detecting air tightness of a welding seam of an automobile axle housing, which first uses an automatic pressure drop method to efficiently preliminarily screen the air tightness of the axle housing, so as to efficiently determine whether the workpiece has a leakage. Based on the determination result, a leakage signal is taken as a trigger mechanism to dynamically activate a thermal imaging system, so that the transient local low-temperature characteristics formed due to the throttling effect of high-pressure gas at the leakage point are captured in a non-contact manner. Further, the collected thermal imaging sequence is deeply analyzed, and is spatially correlated and mapped with a digital three-dimensional model of the workpiece, so that a complete automatic process from macroscopic determination of the leakage phenomenon to accurate positioning of the three-dimensional space coordinates of the leakage source is finally formed. In this way, the efficient automatic screening and the accurate defect positioning capability can be organically combined, so that the intelligent level and the closed-loop efficiency of the whole production quality control link are significantly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of air tightness detection, and more particularly, to a method for detecting air tightness of a welding seam of an automobile axle housing. BACKGROUND

[0002] In the modern automobile industry, the safety, reliability and durability of vehicles are the core indicators to measure their quality, and these performances depend largely on the manufacturing quality of their key components. The automobile axle housing, as a core component of the bearing and transmission system, its main function is to support the body weight, protect the internal differential and transmission shaft and other precision components, and contain lubricating oil to ensure the normal operation of the transmission system under harsh working conditions. The axle housing is usually composed of multiple stamped or cast metal parts combined by welding process, therefore, the quality of the welding seam directly determines the structural strength and sealing performance of the entire axle housing assembly. Once there is a tiny crack or pore in the welding seam, not only will it cause the leakage of lubricating oil, leading to serious wear and even failure of the transmission components due to poor lubrication, but also may allow the intrusion of water, dust and other impurities from the outside into the housing, polluting the lubricating oil and accelerating the corrosion of machine parts, ultimately posing a serious threat to driving safety. Therefore, implementing strict and reliable air tightness detection of the axle housing welding seam is an indispensable quality control link in the automobile manufacturing process.

[0003] In the prior art, to adapt to large-scale and fast-paced production requirements, the pressure drop method or differential pressure method with high automation is generally used for air tightness rapid screening, by filling a specific pressure of gas into the inside of the workpiece to be tested, and monitoring the pressure change within a certain time, it can quickly and objectively determine whether the workpiece has leakage, so as to realize the effective elimination of unqualified products. However, the existing automatic pressure drop method or differential pressure method has a significant limitation, it can only provide a macroscopic judgment of "qualified" or "unqualified", but cannot reveal the specific location of the leakage. This means that once unqualified products are detected, the production line still needs to rely on inefficient manual means (such as observing bubbles by immersion or applying soap water) for secondary investigation to find and locate the tiny leakage point, before subsequent repair processing. This disconnection between detection and positioning greatly reduces the efficiency of automated detection, severely restricts the automation level and overall production rhythm of the repair link, forming an efficiency fault between detection and positioning.

[0004] Therefore, an optimized method for detecting air tightness of the welding seam of the automobile axle housing is expected. SUMMARY

[0005] In order to solve the above technical problems, the present application is proposed. The embodiments of the present application provide a method for detecting the gas tightness of the welding seam of an automobile axle housing. The method first uses an automatic pressure drop method to efficiently preliminarily screen the gas tightness of the automobile axle housing, so as to efficiently determine whether the workpiece has a leakage. Based on the determination result, the leakage signal is taken as a trigger mechanism to dynamically activate a thermal imaging system, so as to capture, in a non-contact manner, a transient local low-temperature feature formed due to the throttling effect of high-pressure gas at the leakage point. Further, through deep analysis of the collected thermal imaging sequence and spatial correlation mapping of the thermal imaging sequence and a digital three-dimensional model of the workpiece, a complete automatic process from macroscopic determination of the leakage phenomenon to accurate positioning of the three-dimensional spatial coordinates of the leakage source is finally formed. In this way, the efficient automatic screening and accurate defect positioning capability can be organically combined, so as to significantly improve the intelligent level and closed-loop efficiency of the entire production quality control link.

[0006] According to an aspect of the present application, a method for detecting the gas tightness of the welding seam of an automobile axle housing is provided, which comprises:

[0007] filling compressed gas into the inside of the clamped and sealed to-be-detected automobile axle housing until the pressure reaches a target detection pressure, and recording the initial pressure, the initial time, the end pressure and the end time in the pressure stabilization time stage and the pressure detection time stage;

[0008] calculating the pressure drop rate and the leakage flag based on the initial pressure, the initial time, the end pressure and the end time;

[0009] in response to the leakage flag being true, collecting a thermal imaging sequence of the to-be-detected automobile axle housing;

[0010] performing image positioning and three-dimensional coordinate mapping of the leakage point based on the thermal imaging sequence of the to-be-detected automobile axle housing to obtain the three-dimensional coordinates of the leakage point.

[0011] Compared with the prior art, the method for detecting the gas tightness of the welding seam of an automobile axle housing provided by the present application first uses an automatic pressure drop method to efficiently preliminarily screen the gas tightness of the automobile axle housing, so as to efficiently determine whether the workpiece has a leakage. Based on the determination result, the leakage signal is taken as a trigger mechanism to dynamically activate a thermal imaging system, so as to capture, in a non-contact manner, a transient local low-temperature feature formed due to the throttling effect of high-pressure gas at the leakage point. Further, through deep analysis of the collected thermal imaging sequence and spatial correlation mapping of the thermal imaging sequence and a digital three-dimensional model of the workpiece, a complete automatic process from macroscopic determination of the leakage phenomenon to accurate positioning of the three-dimensional spatial coordinates of the leakage source is finally formed. In this way, the efficient automatic screening and accurate defect positioning capability can be organically combined, so as to significantly improve the intelligent level and closed-loop efficiency of the entire production quality control link. BRIEF DESCRIPTION OF DRAWINGS

[0012] The above and other objects, features and advantages of the present application will become more apparent from the following detailed description when taken in conjunction with the accompanying drawings in which:

[0013] Figure 1 Flow chart of the method for detecting the gas tightness of the welding seam of the automobile axle housing according to the embodiment of the present application.

[0014] Figure 2 Flow chart of the sub-step S4 of the method for detecting the gas tightness of the welding seam of the automobile axle housing according to the embodiment of the present application.

[0015] Figure 3 Data flow diagram of the sub-step S4 of the method for detecting the gas tightness of the welding seam of the automobile axle housing according to the embodiment of the present application.

[0016] Figure 4 Flow chart of the sub-step S41 of the method for detecting the gas tightness of the welding seam of the automobile axle housing according to the embodiment of the present application.

[0017] Figure 5 Flow chart of the sub-step S43 of the method for detecting the gas tightness of the welding seam of the automobile axle housing according to the embodiment of the present application.

[0018] Figure 6 Flow chart of the sub-step S44 of the method for detecting the gas tightness of the welding seam of the automobile axle housing according to the embodiment of the present application. DETAILED DESCRIPTION

[0019] As used in the description of the application and the appended claims, the singular forms "a", "an" and "the" include plural referents unless the context clearly dictates otherwise. Thus, for example, reference to "a component" or "the component" can include a plurality of such components, and so forth.

[0020] While the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are illustrative only, and different aspects of the system and method can use different modules.

[0021] Flowcharts are used in this application to illustrate the operations performed by systems according to embodiments of the present application. It should be understood that the operations in the figures do not necessarily have to be performed in the precise order shown. Rather, various steps can be handled in reverse order or simultaneously, as desired. Other operations can also be added or removed from the processes, or one or more steps can be added or removed from the processes.

[0022] In the following, example embodiments according to the present application will be described in detail with reference to the accompanying drawings. It is obvious that the described embodiments are only some of the embodiments of the present application, and the present application is not limited to the described embodiments but can be implemented in various ways. Thus, the embodiments are merely illustrative and should not be considered restrictive in any manner.

[0023] To solve the technical problems described in the above background, the present application proposes a method for detecting the gas tightness of a vehicle axle housing weld. First, an automatic pressure drop method is used to efficiently preliminarily screen the gas tightness of the vehicle axle housing, so as to efficiently determine whether the workpiece has a leak. Based on this determination result, the leak signal is used as a trigger mechanism to dynamically activate a thermal imaging system to capture, in a non-contact manner, the transient local low-temperature characteristics formed due to the throttling effect of high-pressure gas at the leakage point. Further, by deeply analyzing the collected thermal imaging sequence and spatially correlating and mapping it with the digital three-dimensional model of the workpiece, a complete automatic process is finally formed from macroscopic determination of the leakage phenomenon to precise positioning of the three-dimensional spatial coordinates of the leakage source. In this way, efficient automatic screening and precise defect positioning capability are organically combined, thereby significantly improving the intelligent level and closed-loop efficiency of the entire production quality control link.

[0024] Figure 1 A flowchart of the method for detecting the gas tightness of a vehicle axle housing weld according to an embodiment of the present application is shown in FIG. 1. As shown in FIG. 1, the method for detecting the gas tightness of a vehicle axle housing weld includes the following steps. Figure 1 S1, compressed gas is filled into the inside of the clamped and sealed vehicle axle housing to be tested until the pressure reaches the target detection pressure, and the initial pressure, the initial time, the end pressure and the end time are recorded during the pressure stabilization time period and the pressure detection time period; S2, based on the initial pressure, the initial time, the end pressure and the end time, the pressure drop rate and the leakage flag are calculated; S3, in response to the leakage flag being true, a thermal imaging sequence of the vehicle axle housing to be tested is collected; S4, based on the thermal imaging sequence of the vehicle axle housing to be tested, a leakage point image positioning and three-dimensional coordinate mapping are performed to obtain the three-dimensional coordinates of the leakage point.

[0025] In the above method for detecting the gas tightness of the automobile axle shell weld, the step S1 is to fill compressed gas into the inside of the clamped and sealed automobile axle shell to be detected until the pressure reaches the target detection pressure, and to record the initial pressure, the initial time, the end pressure and the end time in the pressure stabilization time period and the pressure detection time period. It should be understood that, since the tiny defects of the automobile axle shell weld, such as pinholes and incomplete penetration, are hidden and cannot be identified by conventional visual detection, and the gas leakage amount is directly related to the pressure difference. At the same time, the pressure fluctuation caused by temperature change, pipeline deformation and other factors during the gas filling process directly affects the detection accuracy. Therefore, the present application fills compressed gas into the inside of the clamped and sealed automobile axle shell to be detected to the target detection pressure, and records the pressure and time parameters in the pressure stabilization period and the detection period, respectively, so as to eliminate the interference of instantaneous pressure fluctuation and accurately judge the weld tightness by the pressure change amount. In this way, the pressure decay caused by tiny leakage can be effectively captured, the unqualified products caused by detection error can be avoided to flow into the subsequent process, the pressure sealing performance of the automobile axle shell in actual use can be ensured, and the safety risk caused by weld leakage during vehicle driving can be reduced.

[0026] In the process of specific implementation, first, the clamping and sealing operation is performed on the automobile axle shell to be detected, the two end flange surfaces and the process holes thereof are completely closed by a customized tool, and it is ensured that there is no other gas leakage passage except the weld. Then, the connecting pipeline of the gas tightness detector is connected with the preset interface of the automobile axle shell, and the detection program is started. The device fills dry compressed gas into the inside of the automobile axle shell at a set rate until the pressure in the cavity reaches the target detection pressure. After the pressure charging is completed, the pressure stabilization period is entered, at which time the detector monitors the pressure change in real time through the built-in sensor. When the pressure fluctuation amount in the continuous 30 seconds is reduced to below the set threshold value, the initial pressure value and the initial time at this moment are recorded. After the stabilization period ends, the pressure detection period is automatically entered, the detection system is kept in a sealed state, and the cavity pressure data is continuously collected. At the end of the preset detection time, the current end pressure value and the end time are recorded. By comparing the pressure change curves of the two periods, if the pressure decay amount does not exceed the specified limit value, it is determined that the weld gas tightness is qualified.

[0027] In the method for detecting the air tightness of the automobile axle shell weld seam, the step S2 is to calculate the pressure drop rate and the leakage sign based on the initial pressure, the initial time, the end pressure and the end time. It should be understood that, since the initial pressure, the initial time, the end pressure and the end time are only isolated detection data, the leakage state of the automobile axle shell weld seam cannot be directly reflected, the simple pressure difference lacks the time dimension reference, the pressure change under different detection durations cannot be compared horizontally, and the explicit determination logic is easy to cause subjective misjudgment. Therefore, the application further calculates the pressure drop rate by the ratio of the pressure difference and the time difference based on the four data, and then generates the leakage sign in combination with the preset standard, so as to convert the abstract data into quantifiable and determinable indexes, and establish the standardized leakage evaluation logic. In this way, the welding leakage rate can be accurately quantified, the interference of the detection duration difference on the result can be eliminated, the misjudgment or omission caused by subjective experience can be avoided, a clear trigger signal can be provided for subsequent start of the thermal imaging positioning, the detection process automation connection can be ensured, the automobile axle shell meeting the sealing requirement can be ensured to enter the next production link, the vehicle operation risk can be reduced from the quality control source. In one specific example of the application, the step S2 includes calculating the pressure drop rate based on the initial pressure, the initial time, the end pressure and the end time by the following formula:

[0028] ;

[0029] wherein, is the initial pressure, is the end pressure, is the end time, is the initial time, is the pressure drop rate. That is, the pressure drop rate is calculated by the ratio of the pressure difference and the time difference in combination with the initial time and the end time, so as to convert the discrete pressure data into the pressure change rate within the unit time, and establish the quantifiable leakage degree index. In this way, the influence of the detection duration on the leakage evaluation can be eliminated, the real rate of the welding leakage can be accurately reflected, objective and comparable quantitative basis can be provided for subsequent determination, the welding leakage grade can be avoided to be misjudged due to only focusing on the pressure difference, and the welding leakage characteristics of the automobile axle shell under different detection periods can be evaluated by the unified standard.

[0030] Specifically, the step S2 further comprises: comparing the pressure drop rate with a leakage determination threshold value, if the pressure drop rate is greater than the leakage determination threshold value, setting the leakage flag to true, otherwise, setting the leakage flag to false. Specifically, the application compares the pressure drop rate with the preset leakage determination threshold value matched with the vehicle type, sets the leakage flag according to the result, so as to establish an objective determination standard matched with the characteristics of the axle housing, and converts the quantitative index into an explicit pass or fail result. In this way, it can be ensured that the determination standard is consistent with the actual use requirements of the axle housing, and false or missed judgments are avoided, and at the same time, the leakage flag is used to guide the subsequent process, i.e. the qualified parts are directly transferred to the next process, and the unqualified parts trigger the thermal imaging positioning, so as to realize closed-loop management of the detection process, and improve the production efficiency and quality control accuracy.

[0031] In the process of specific implementation, first, the model information obtained by scanning the code of the axle housing is used to call the corresponding leakage determination threshold value from the system parameter library, and the threshold value is set in advance according to the axle housing design document (such as the maximum allowable leakage), the relevant automobile axle housing detection standard of GB / T and the real vehicle working condition verification result, such as the threshold value of the heavy truck rear axle housing is set to 0.01 kPa / s, and the threshold value of the passenger car front axle housing is set to 0.008 kPa / s. Secondly, the pressure drop rate value is obtained, and the units of the two are ensured to be kPa / s through format conversion. Then, the comparison program is started, if the pressure drop rate is greater than the threshold value, the leakage flag field of the detection task is immediately set to true, and the preliminary result of unqualified is generated; if the pressure drop rate is less than or equal to the threshold value, the leakage flag is set to false, and the preliminary result of qualified is generated. Finally, the leakage flag and the result are synchronized to the detection report, the qualified report is pushed to the next process terminal, the unqualified report is pushed to the repair work station terminal, and the comparison details of the pressure drop rate and the threshold value are attached, so as to facilitate the workers to confirm the determination basis, and at the same time, the result is uploaded to the MES system to complete the process closed loop.

[0032] In the above method for detecting the air tightness of the automobile axle shell weld, the step S3, in response to the leakage flag being true, collects the thermal imaging sequence of the measured axle shell. It should be understood that since the leakage flag being true only indicates that there is a leak in the axle shell, but the specific position of the leak cannot be determined, the manual positioning method (such as observing bubbles by immersion, and applying soap water) is low in efficiency and is easily affected by human judgment, and is difficult to adapt to the beat demand of the automatic production line, and may cause rust on the metal surface of the axle shell due to liquid residue, increasing the subsequent processing cost. Therefore, when the leakage flag is true, the thermal imaging device is started to collect the thermal imaging sequence of the measured axle shell, so as to capture the local temperature change characteristics of the leakage point caused by the throttling effect of the high-pressure gas, and to provide continuous and dynamic image data support for the subsequent accurate positioning of the leakage point. In this way, the disadvantages of manual positioning can be avoided, the automatic data collection of the leakage point positioning can be realized, the seamless connection with the previous pressure drop detection process can be ensured, the high-quality image foundation for the subsequent two-dimensional positioning and three-dimensional coordinate mapping of the leakage point can be laid, the processing cycle from detection to repair of unqualified products can be effectively shortened, and the overall quality control efficiency and product qualification rate of the production line can be improved.

[0033] In the process of specific implementation, first, after confirming that the leakage flag is true, a locking instruction is immediately sent to the clamping station where the measured axle shell is located, so that the clamping tool remains in the current clamped state, ensuring that the position of the axle shell does not shift during the collection process, and avoiding the misalignment of the thermal imaging area and the actual position of the weld due to the displacement of the workpiece. Secondly, the detection system retrieves the thermal imaging collection parameters corresponding to the model of the measured axle shell from the pre-set parameter library according to the previously obtained model information of the measured axle shell, including the visual angle coverage range of the thermal imaging device, the imaging sensitivity level, and the frame collection interval, and then controls the thermal imaging device to adjust its installation height, horizontal shooting angle, and lens focal length through a mechanical adjustment mechanism, until the field of view of the device can completely cover all the weld areas of the axle shell, including the connecting welds of the bridge pipe and the flanges at both ends, the butt welds of the reducer shell and the bridge pipe, and other key sealing positions. Then, the thermal imaging device starts the continuous collection mode according to the retrieved parameters, continuously captures the temperature distribution changes of the weld area, generates a thermal imaging sequence composed of multiple thermal images, and monitors the clarity and temperature resolution of each image in real time during the collection process. If the image quality is reduced due to temperature fluctuations in the workshop environment or air flow interference, the imaging sensitivity parameter will be automatically adjusted to ensure that each thermal image can clearly reflect the temperature details of the weld area. Finally, the collected thermal imaging sequence is associated and bound with the unique detection session ID of the measured axle shell, and is stored in the special image database of the detection system. At the same time, the detection system sends a data ready signal to the subsequent leakage point image positioning link, ensuring that the thermal imaging sequence can be directly used for subsequent differential image generation, cumulative cooling map construction, and other processing, realizing the coherent execution of data collection and subsequent analysis links without any process breakpoints.

[0034] In the method for detecting the weld air tightness of the automobile axle housing, the step S4 comprises the steps of: S41, performing time series difference and temperature drop signal extraction on the thermal imaging sequence of the to-be-tested axle housing to obtain a temperature drop signal sequence; S42, performing time domain accumulation and spatial filtering on the temperature drop signal sequence to obtain a cumulative temperature drop map; S43, performing abnormal area segmentation and leakage point positioning on the cumulative temperature drop map to obtain a leakage point two-dimensional coordinate; and S44, performing ray projection and three-dimensional coordinate calculation on the leakage point two-dimensional coordinate to obtain a leakage point three-dimensional coordinate. Figure 2 The flow chart of the sub-step S4 of the method for detecting the weld air tightness of the automobile axle housing according to the embodiment of the present application. Figure 3 The data flow schematic diagram of the sub-step S4 of the method for detecting the weld air tightness of the automobile axle housing according to the embodiment of the present application. Figure 2 And Figure 3 As shown in the figures, the step S4 comprises the steps of: S41, performing time series difference and temperature drop signal extraction on the thermal imaging sequence of the to-be-tested axle housing to obtain a temperature drop signal sequence; S42, performing time domain accumulation and spatial filtering on the temperature drop signal sequence to obtain a cumulative temperature drop map; S43, performing abnormal area segmentation and leakage point positioning on the cumulative temperature drop map to obtain a leakage point two-dimensional coordinate; and S44, performing ray projection and three-dimensional coordinate calculation on the leakage point two-dimensional coordinate to obtain a leakage point three-dimensional coordinate.

[0035] Specifically, the step S41, performing time series difference and temperature drop signal extraction on the thermal imaging sequence of the to-be-tested axle housing to obtain a temperature drop signal sequence. It should be understood that, in the thermal imaging sequence of the to-be-tested axle housing, the temperature drop signal of the weld leakage point caused by the throttling effect of the high-pressure gas is extremely weak, and is easily covered by static interference signals such as the temperature fluctuation of the workshop environment, the noise of the thermal imager itself, and the reflection of the workpiece surface. Single-frame thermal image or simple frame-by-frame comparison cannot effectively distinguish the effective temperature drop signal and the interference signal, so that the leakage point is difficult to be identified. Therefore, the present application further performs difference operation in the time dimension on the thermal imaging sequence, and extracts the temperature drop signal, so as to amplify the inter-frame dynamic temperature change, offset the influence of the static background temperature, and retain the temperature drop information caused only by the leakage. In this way, the weak leakage signal originally submerged in the noise can be highlighted, high-quality effective data can be provided for the subsequent time domain accumulation and spatial filtering link, leakage point missed judgment caused by insufficient signal recognition can be avoided, and the subsequent processing link can be focused on the temperature change information related to the real leakage, so as to improve the overall positioning accuracy. Among them, Figure 4Flowchart of sub-step S41 of the method for detecting the gas tightness of the automobile axle housing weld according to the embodiment of the present application. As shown in Figure 4 S41, the step S41 comprises steps of: S411, calculating the difference images between every two adjacent frames in the thermal imaging sequence of the automobile axle housing to be detected to obtain a difference image sequence; S412, performing threshold processing on the difference image sequence to obtain a temperature drop signal sequence.

[0036] More specifically, the step S411 calculates the difference images between every two adjacent frames in the thermal imaging sequence of the automobile axle housing to be detected to obtain a difference image sequence. It can be understood that, since a single frame image in the thermal imaging sequence of the automobile axle housing to be detected can only reflect the temperature distribution at a certain time, it cannot intuitively present the dynamic change of temperature, and the static background temperature (such as the constant temperature of the workshop environment and the temperature of the workpiece after stable heat dissipation) remains consistent in each frame, which will mask the weak temperature change signal of the leakage point over time, resulting in that the leakage feature cannot be captured by comparing single frame or random frame. Therefore, the present application further calculates the difference images of adjacent two frames in the thermal imaging sequence of the automobile axle housing to be detected and forms a difference image sequence, so as to eliminate the interference of the static background temperature, intuitively present the temperature change between frames in the form of images, and highlight the dynamic temperature change area. In this way, the dynamic temperature change signal of the leakage point originally hidden in the static background can be converted into observable difference image features, a clear analysis object is provided for subsequent targeted extraction of the temperature drop signal, dynamic signal omission caused by static background interference is avoided, and the foundation for identifying the leakage point is laid.

[0037] In the process of specific implementation, first, the complete thermal imaging sequence of the automobile axle housing to be detected is called from the system data cache area, the integrity of each frame image is confirmed by an image verification algorithm, that is, there is no pixel loss and no transmission error, and the total number of frames and the frame interval of the sequence are recorded (pre-set according to the detection requirements to ensure that the time span of the temperature change of the leakage point can be captured). Secondly, the frame-by-frame difference operation is started, taking the nth frame thermal image as the reference frame and the n+1th frame thermal image as the comparison frame, the temperature values of the same pixel coordinate points of the two frames are subtracted, that is, the temperature value of the comparison frame is subtracted from the temperature value of the reference frame, to obtain the difference image of the adjacent frames, and the operation is sequentially cycled until the operation of all adjacent frames is completed, to generate a difference image sequence containing (total number of frames-1) difference images. Then, each difference image generated is preprocessed, the image contrast is adjusted by a gray scale stretching algorithm to make the temperature change difference between frames more obvious, and the pixel area with a temperature change value exceeding the normal noise range is marked to preliminarily exclude extreme noise points. Finally, the preprocessed difference image sequence is stored in the original time sequence, the detection session ID of the automobile axle housing to be detected is associated with the weld area annotation information, to ensure that the automobile axle housing weld position can be accurately corresponded when the temperature drop signal is extracted subsequently, and data misplacement is avoided.

[0038] More specifically, the step S412, the differential image sequence is thresholded to obtain the cooling signal sequence. It should be understood that, since the differential image sequence contains not only the cooling signal (temperature difference is negative) caused by the leakage point, but also the positive temperature difference signal generated by the local environmental warming (such as the local temperature rise of the workpiece caused by the heat dissipation of the workshop equipment), the instantaneous noise of the thermal imager (such as the random temperature drift of the pixel points), and the like, these interference signals will be mixed with the cooling signal, resulting in that the real leakage-related temperature change information cannot be directly distinguished. Therefore, the differential image sequence is further thresholded to filter out the negative temperature difference signal meeting the leakage characteristics, eliminate the positive interference signal and meaningless small noise signal, and focus on the real cooling area. In this way, the pure cooling signal can be extracted from the mixed differential image sequence, forming a signal sequence related only to the leakage, avoiding the misdirection of the interference signal to the subsequent processing, ensuring that the subsequent time domain accumulation link can accurately superimpose the temperature change information of the leakage point, and improving the accuracy of the leakage point positioning.

[0039] In the process of specific implementation, first, the preprocessed differential image sequence is called, and the threshold parameter matching the model of the to-be-detected axle shell is called from the system parameter library. The threshold is preset based on the thermal conductivity coefficient of the axle shell material, the detection gas pressure, and the throttling effect temperature change rule, so as to ensure that the leakage cooling and noise can be accurately distinguished. Secondly, the threshold value of each pixel in the differential image sequence is judged, and if the temperature difference of a pixel is less than the set threshold value, that is, the temperature drop amplitude meets the leakage characteristic standard, the temperature difference information of the pixel is retained; if the temperature difference of the pixel is greater than or equal to the set threshold value, including positive warming and small noise, the temperature difference of the pixel is set to zero, so as to complete the threshold processing of a single differential image. Then, the edge smoothing processing is performed on each image after threshold processing, the median filtering algorithm is adopted to eliminate the influence of isolated zero-value pixels on the continuity of the cooling area, and the integrity of the cooling area contour is ensured. Finally, all the processed differential images are integrated according to the original time sequence to form a cooling signal sequence containing only effective cooling information, stored in the input port of the system image analysis module, and a data ready signal is sent synchronously to trigger the start of the subsequent time domain accumulation and spatial filtering link, ensuring the coherence of the process.

[0040] Specifically, the step S42, the time domain accumulation and spatial filtering are performed on the cooling signal sequence to obtain the accumulated cooling map. It should be understood that, since the cooling signal intensity of a single frame in the cooling signal sequence is still weak, and there may be isolated noise pixels, such as a single cooling pixel caused by the instantaneous error of the thermal imager, directly used for positioning is easy to cause misjudgment, and cannot form a clear leakage area contour. Therefore, the present application further performs signal accumulation in the time dimension on the cooling signal sequence, and then performs spatial filtering processing, so as to superimpose the cooling signal intensity of multiple frames, enhance the signal characteristics of the leakage area, eliminate isolated noise points, and form a clear cooling area image. In this way, the weak cooling signal dispersed in multiple frames can be concentrated and strengthened, the leakage area is clearly highlighted from the background, an explicit target is provided for subsequent abnormal area segmentation, the accuracy and reliability of the segmentation result are ensured, and positioning deviation caused by signal dispersion or noise interference is avoided.

[0041] In the process of specific implementation, first, the stored cooling signal sequence is called, the pixel size and signal validity of each frame in the sequence are confirmed, invalid frames are removed through quality detection in advance. Secondly, time domain accumulation processing is performed on all frames in the sequence, that is, for each pixel point in the image, the cooling signal values of the point in all frames are summed, so that the cooling signal of the leakage area is significantly enhanced due to the superposition of multiple frames, and the isolated noise points have very low accumulation values due to the discontinuity of the signal. Then, spatial filtering processing is performed on the accumulated image, a Gaussian filtering algorithm is used to smooth the image, and small noise pixels still existing after accumulation are filtered out, while the contour integrity of the leakage area is maintained, and excessive filtering is avoided to cause the region boundary to be blurred. Finally, the processed image is output as the accumulated cooling map, in which the leakage area presents as an obvious high brightness (or high gray value) area, and the background area remains low brightness, directly reflecting the position range of the leakage point, and stored for subsequent abnormal area segmentation.

[0042] Specifically, the step S43, the abnormal area segmentation and the leakage point positioning are performed on the accumulated temperature drop graph to obtain the leakage point two-dimensional coordinates. It should be understood that, although the accumulated temperature drop graph has enhanced the leakage-related temperature drop signal through time domain accumulation, there may still be local weak temperature drop areas in the graph caused by environmental noise, workpiece surface impurity reflection, etc. These areas are mixed with the real leakage area, and the real leakage area only presents as a continuous gray block without a specific leakage center point. Directly using the graph cannot provide accurate image reference for subsequent three-dimensional coordinate mapping, which is easy to cause positioning deviation. Therefore, the application further performs abnormal area segmentation on the accumulated temperature drop graph, and then locates the core position of the leakage point, so as to separate the real leakage area from the interference area and extract the unique leakage center point coordinates. In this way, the fuzzy gray area can be converted into accurate image coordinate system coordinate points, ensuring that there is a clear input reference in the subsequent three-dimensional mapping link, and avoiding positioning errors caused by fuzzy area boundaries or interference areas.

[0043] In the process of specific implementation, first, the accumulated temperature drop graph of the vehicle axle shell to be tested is called from the system data buffer area, and the weld area label graph of the vehicle axle shell of the vehicle is loaded, which is generated based on the CAD model in advance, and the corresponding range of the weld in the image is clear, so as to ensure that the analysis range focuses on the weld area and excludes invalid temperature drop signals in non-weld areas. Secondly, an abnormal area segmentation threshold is set, which is preset based on the heat conduction characteristics of the vehicle axle shell material and the detection gas throttling temperature variation rule, so as to ensure that only the real leakage area can be identified. The accumulated temperature drop graph is binarized, the area with a gray value exceeding the threshold is set as the foreground, i.e. the suspected leakage area, and the area with a gray value below the threshold is set as the background, so as to realize preliminary segmentation. Then, the binarized image is subjected to connected region analysis to identify all independent foreground areas. By calculating the area, gray mean value and overlap degree with the weld area of each area, the interference areas with too small area (less than the preset minimum leakage area), too low gray mean value (lower than the leakage temperature variation standard) or no overlap with the weld area are removed, and the only real leakage area is determined. Finally, the intensity weighted centroid of the leakage area is calculated, the gray value (representing the temperature drop intensity) of each pixel in the area is taken as the weight, and the image coordinates of the area center point are obtained through the coordinate weighted average algorithm. The coordinates are the leakage point two-dimensional coordinates. After accuracy verification (confirming that the coordinates are located at the geometric center of the leakage area), the coordinates are associated with the detection session ID for storage, so as to prepare for subsequent three-dimensional coordinate calculation.

[0044] It should be appreciated that the existing leakage positioning mechanism based on thermal imaging has inherent physical model simplification defects in the abnormal area segmentation and two-dimensional centroid positioning links. Specifically, the mechanism simply abstracts the complex thermal field formed by the gas leakage on the workpiece surface as a spot on a two-dimensional image, and assumes that the intensity weighted centroid of the spot is the leakage source point. However, in actual industrial scenarios, due to the complex geometry of the workpiece surface (such as welds, curved surfaces) and the scouring effect of the gas jet, the heat diffusion often presents anisotropy, forming an irregular ellipsoidal or comet-shaped morphology with obvious directionality, and the geometric center has a significant spatial deviation from the physical source point.

[0045] More importantly, the mechanism uses a hard threshold binarization method to segment the abnormal area, which is an information loss operation. It roughly reduces the continuous scalar field containing rich leakage intensity and morphology information to a binary area without distinction, completely losing the gradient, curvature and other key local geometric features of the temperature field. Physically, the true leakage source point is not only the minimum point of the temperature field, but also the point with the sharpest concave shape and the largest curvature. The existing mechanism does not distinguish between them, and only selects them by heuristic rules such as maximum area or maximum total intensity. Not only is the positioning accuracy limited, but its robustness and reliability are also difficult to guarantee when faced with multiple interference signals or complex background noise. This way of simplifying the complex field analysis problem into a primary morphological problem constitutes a core technical bottleneck.

[0046] In a preferred embodiment of the present application, to overcome the above-mentioned defects, an innovative precise positioning mechanism of the leakage source point is proposed. Instead of relying on the geometric centroid, the mechanism regards the cumulative temperature drop map as a two-dimensional continuous scalar field, and inversely analyzes the local curvature characteristics of the field to obtain the physically meaningful leakage source point.

[0047] Specifically, first, the cumulative temperature drop map is subjected to a probability field nonlinear mapping to obtain a leakage probability field. It should be appreciated that the traditional hard threshold segmentation method will cause serious loss of signal information, while a smooth and continuous probabilistic expression can better preserve the dynamic range of the signal and provide a function with excellent mathematical properties for subsequent differential operation. Specifically, the input cumulative temperature drop map A nonlinear Sigmoid function is applied for pixel-by-pixel transformation to convert the linear gray value into a probability value reflecting the probability of the pixel being a leakage point, strengthening the features of high probability areas and weakening the interference of low probability areas. The mathematical form of the transformation is:

[0048] ;

[0049] wherein, is a reference temperature, acting as a soft threshold or reference point, is a natural constant, is a conversion steepness coefficient, which controls the steepness of the conversion, is the temperature value of the cumulative temperature drop map at pixel , is the leak probability field at pixel .

[0050] Thus, the original cumulative temperature drop map with physical unit of temperature is converted into a dimensionless leak probability field with value range between (0, 1) , where is a reference baseline, which is used to distinguish the background noise from the potential leak signal. It can be set to a value slightly lower than the average background temperature of the weld area, for example, if the background temperature is stable at 25°C, it can be set to 24.5°C. controls the sensitivity of the conversion, The larger the value is, the steeper the probability curve changes near , which is close to hard threshold segmentation, The smaller the value is, the smoother the transition is. In practical applications, it can be set according to the sensitivity requirements of the detection, for example, it can be a constant between 0.5 and 2.0. The higher the intensity of the point in the cumulative temperature drop map, the closer the corresponding probability value in the probability field to 1, and vice versa. The entire field maintains continuity and differentiability, effectively suppressing background noise while enhancing the relative prominence of real signals. That is, by using the mathematical properties of the Sigmoid function, which changes steeply near the reference temperature and gently approaches the boundary far from the reference temperature, the small temperature changes caused by equipment heat dissipation and other disturbances in the bridge shell weld area are mathematically compressed into a low probability interval close to 0. The significant temperature drop caused by the continuous throttling of high-pressure gas (similar to the effect of the air nozzle near the tire when deflating) at the real leak point is mathematically mapped to a high probability interval close to 1, accurately quantifying the confidence of each pixel point belonging to the leak area. At the same time, the continuity and differentiability of this leak probability field in mathematics enable subsequent analysis of the local curvature characteristics of the leak probability field through the Hessian matrix field, thereby locking the leak source point and solving the problem of quantifying leak probability and spatial distribution characteristics with mathematical methods in linear grayscale images, significantly improving the anti-interference ability and positioning accuracy of bridge shell weld leak detection in complex workshop environments. For example, in bridge shell weld detection, the temperature of a certain section of weld decreases from 25°C to 22°C, and the temperature of another section decreases from 25°C to 18°C. The Sigmoid mapping can mathematically allow the leak probability of the former (e.g., 0.3) and the leak probability of the latter (e.g., 0.9) to present differences that conform to the actual leak severity.

[0051] Then, the Hessian matrix field of the leak probability field is calculated. The leak source point is physically represented as the sharpest concave point in the temperature field, and this geometric feature cannot be described by the simple intensity value, while the Hessian matrix field is a powerful mathematical tool for describing the local curvature of a multivariate function. The execution process is as follows: the second-order partial derivative of the leak probability field with respect to the horizontal coordinate of the pixel is calculated , which describes the second-order change rate of the leak probability field in the horizontal direction of the axle housing weld (such as the horizontal direction of the weld extension), and if the real leak point is horizontally sharp concave, the value will present a significant negative steep change; the first-order partial derivative of the leak probability field with respect to is calculated, and the partial derivative of is calculated , and the first-order partial derivative of the leak probability field with respect to is calculated, and the partial derivative of is calculated , which describes the second-order change correlation of the leak probability field in the horizontal and vertical (such as the vertical height direction of the weld) directions of the axle housing weld, and reflects whether the concave is a long pit along the weld extension or a local point-shaped deep pit; the second-order partial derivative of the leak probability field with respect to the vertical coordinate of the pixel is calculated, which describes the second-order change rate of the leak probability field in the vertical direction of the axle housing weld, corresponding to the sharpness of the concave in the vertical direction of the leak point. Finally, the above four second-order partial derivatives are organized into a 2x2 matrix to generate the Hessian matrix of each pixel , and finally the Hessian matrix field of the same size as the leak probability field is obtained. The mathematical expression is as follows:

[0052] ;

[0053] wherein is the partial derivative operation, is the second-order partial derivative of the probability field with respect to the horizontal coordinate of the pixel , is the first-order partial derivative of the probability field with respect to , and the partial derivative of , is the first-order partial derivative of the probability field with respect to , and the partial derivative of , is the second-order partial derivative of the probability field with respect to the vertical coordinate of the pixel , is the Hessian matrix field of the leak probability field at the pixel .

[0054] In this way, each pixel in the field obtains a mathematical entity that quantitatively describes the local surface shape of the point, thereby generating a Hessian matrix field of the same size as the probability field . The eigenvalues and eigenvectors of any matrix in this field completely encode the bending direction and bending degree of the probability surface at the point, providing a data basis for distinguishing real leakage signals from gentle background fluctuations from a geometric perspective. For example, the probability surface at the real leakage of the axle shell weld presents a steep pit, and the eigenvalues of the Hessian matrix at this point can reflect the sharp degree, and the eigenvectors can reflect the concave direction; the background fluctuations caused by the airflow in the workshop environment and the thermal radiation of the tooling make the probability surface a gentle slope, and the eigenvalues of the Hessian matrix at this point are significantly smaller. In this way, a data basis is provided for distinguishing real leakage signals from gentle background fluctuations from a geometric perspective, thereby accurately locating the leakage source point, so that the axle shell leakage detection not only conforms to the physical intuition of sharp temperature drop at the leakage point, but also has a rigorous mathematical tool to support positioning accuracy.

[0055] Finally, principal curvature analysis and source point coordinate optimization are performed on the leakage probability field and the Hessian matrix field to obtain the two-dimensional coordinates of the leakage point. It should be understood that, by using the aforementioned Hessian matrix field, a clear and robust criterion is needed to identify the unique point that best matches the physical characteristics of the leakage from the entire curvature field. The execution process is to construct the leakage point positioning problem as a mathematical optimization problem. The goal of this problem is to find a point , whose two eigenvalues and of the Hessian matrix field are both negative values, ensuring that the point is a local peak, and the sum of the absolute values of the two negative eigenvalues (i.e., the inverse of the average curvature) reaches the maximum value, under the premise of meeting a certain probability threshold . Specifically, first, determine the leakage probability threshold , which is set based on the probability value corresponding to the minimum detectable leakage amount of the axle shell weld, such as for a certain type of passenger car front axle shell, combined with historical detection data and air tightness standards, set , to ensure that the selected point has sufficient leakage confidence. Subsequently, perform principal curvature analysis on the leakage probability field and the Hessian matrix field, extract the two eigenvalues and of the Hessian matrix field at each pixel point , and verify whether the point meets the constraint condition, i.e. , to ensure that the point is a local peak corresponding to the most sharp concave feature in physics, and , to ensure that it is reliable in probability. Then, construct the objective function , which is the inverse of the average curvature, and by using a numerical optimization algorithm, such as gradient ascent method, the derivative information of the Hessian matrix is used to accelerate the optimization process, and the pixel size and computational efficiency requirements of the vehicle axle shell weld thermal image are adapted, among the pixel points that meet the constraint conditions, the pixel coordinates that maximize the objective function are found . Finally, the found pixel coordinates are sub-pixel interpolation optimized to obtain the two-dimensional sub-pixel coordinates of the leakage point . The mathematical expression of the optimization problem is: ;

[0056] Among them, is the pixel coordinate that maximizes the objective function , and are two eigenvalues of the Hessian matrix field of is the constraint condition, is the leakage probability threshold, is the two-dimensional sub-pixel coordinates of the leakage point.

[0057] In this way, the point that is most credible in probability and most acute in shape is accurately calculated from the entire field, which is defined as the leakage source point. Essentially, it is equivalent to outputting a two-dimensional coordinate with sub-pixel accuracy . The coordinate is no longer the geometric average center of a region, but the most likely projection of the real physical source point on the thermal image, and its positioning accuracy and adaptability to irregular signal shapes far exceed traditional centroid methods. For example, a certain section of the vehicle axle shell weld produces a small leakage due to welding defects, and the leakage area in the thermal image needs to meet both the condition that the leakage probability is higher than the threshold and the condition that it is the most acute depression in shape. Through this optimization solution, the real leakage point of such areas that are easily missed or misidentified by traditional methods can be accurately locked.

[0058] In summary, the preferred embodiment significantly improves the accuracy and robustness of the weld leakage point positioning based on thermal imaging. By introducing differential geometric analysis, this mechanism can accurately identify the physical leakage source point from continuous thermal field data full of noise and irregular shapes, rather than the geometric center of its diffusion, thereby reducing the positioning error by an order of magnitude. At the same time, since this method relies on the internal geometric structure of the signal rather than simply the intensity or area, it has stronger recognition ability and anti-interference for background noise, sensor artifacts, and complex scenes with multiple potential leakage points.

[0059] Specifically, the step S44, ray projection is performed on the two-dimensional coordinates of the leakage point and three-dimensional coordinate calculation is performed to obtain the three-dimensional coordinates of the leakage point. It should be understood that, since the two-dimensional coordinates of the leakage point only represent the pixel position thereof in the thermal imager image coordinate system, and are not directly related to the physical space of the vehicle axle shell entity to be measured, they cannot provide a reference for the repair worker to locate on the physical workpiece. Only relying on the image coordinates, the worker needs to repeatedly compare the image and the entity, which is prone to positioning deviation and low efficiency. Therefore, the present application further converts the two-dimensional coordinates into a spatial ray through ray projection, and then obtains the entity space coordinates by combining three-dimensional coordinate calculation, so as to establish a precise correspondence between the image information and the physical space of the vehicle axle shell, and clearly determine the specific position of the leakage point on the physical workpiece. In this way, errors and inefficiencies caused by manual comparison can be avoided, and three-dimensional coordinates that can be directly measured and accurately referenced are provided for the repair link, which significantly shortens the repair cycle and ensures that the positioning result can be linked with the digital model of the vehicle axle shell, adapting to the quality traceability and data management requirements in intelligent production. Among them, Figure 6 The flowchart of the sub-step S44 of the method for detecting the gas tightness of the vehicle axle shell weld according to the embodiment of the present application is shown in FIG. 4. Figure 6 As shown in FIG. 4, the step S44 includes the following steps: S441, pixel coordinate inverse normalization is performed on the two-dimensional coordinates of the leakage point to obtain the two-dimensional coordinates of the leakage point in the camera coordinate system; S442, a camera coordinate system ray is constructed based on the two-dimensional coordinates of the leakage point in the camera coordinate system and the camera optical center; S443, coordinate system transformation is performed on the camera coordinate system ray to obtain a world coordinate system ray; and S444, the first intersection point of the world coordinate system ray and the three-dimensional CAD model of the vehicle axle shell to be measured is calculated as the three-dimensional coordinates of the leakage point.

[0060] More specifically, the step S441, pixel coordinate inverse normalization is performed on the two-dimensional coordinates of the leakage point to obtain the two-dimensional coordinates of the leakage point in the camera coordinate system. It should be understood that, since the two-dimensional coordinates of the leakage point are coordinates in the thermal imager image pixels, they are affected by the camera optical system distortion (such as radial distortion and tangential distortion) and pixel arrangement characteristics, and the actual projection relationship between the coordinates and the camera optical center is nonlinear, so the coordinates cannot be directly used to construct an accurate spatial ray. If the pixel coordinates are directly used, it will cause deviation of the direction of the subsequent ray, and then cause errors in three-dimensional coordinate calculation. Therefore, the present application further performs inverse normalization processing on the pixel coordinates to eliminate the influence of optical distortion, convert the pixel coordinates into coordinates on the normalized plane in the camera coordinate system, and establish a linear projection relationship matching the camera optical characteristics. In this way, the direction of the subsequent spatial ray constructed can be accurate, and positioning deviation caused by optical distortion can be avoided, so as to provide a reliable camera coordinate system reference for three-dimensional coordinate calculation and ensure the accuracy of the final three-dimensional coordinates of the leakage point to meet the repair operation requirements.

[0061] In the process of implementation, first, the leakage point two-dimensional coordinates of the vehicle axle shell to be tested are called, the two-dimensional coordinates are pixel coordinates and the unit is pixel, and the intrinsic matrix of the thermal imager of the detection station is called from the system parameter library, which is calibrated in advance by Zhang's calibration method and contains focal length, principal point coordinates and distortion coefficient. Secondly, the leakage point two-dimensional coordinates are subjected to inverse distortion processing, the pixel coordinates are substituted into the inverse distortion model constructed based on the intrinsic distortion coefficient, the radial and tangential distortion caused by the camera optical system is eliminated, and the pixel coordinates without distortion are obtained. Then, according to the principal point coordinates (the position of the image coordinate system origin on the pixel plane) and the focal length in the intrinsic matrix, the normalized calculation is performed on the pixel coordinates without distortion, the pixel coordinates are subtracted by the principal point coordinates, and then divided by the focal length in the corresponding direction, and the two-dimensional coordinates of the normalized plane in the camera coordinate system are obtained. Finally, the normalized coordinates are verified for accuracy, the projection relationship with the intrinsic parameters of the thermal imager is confirmed, the pixel coordinates are calculated by inverse projection, and the error with the original two-dimensional coordinates is within the allowable range, and after verification, the two-dimensional coordinates of the leakage point in the camera coordinate system are stored as the camera coordinate system, which provides data support for subsequent ray construction.

[0062] More specifically, the step S442 constructs a camera coordinate system ray based on the leakage point two-dimensional coordinates in the camera coordinate system and the camera optical center. It should be understood that since the leakage point two-dimensional coordinates in the camera coordinate system are only a point on the normalized plane, the position of the point in the three-dimensional space cannot be determined alone, the point corresponds to all points in the space extending in the direction of the camera optical center to the point, and the space range is defined by the ray form, so as to be further associated with the three-dimensional model of the vehicle axle shell. Therefore, the camera optical center is further taken as the starting point, the normalized plane coordinate point is connected, the space ray in the camera coordinate system is constructed, so as to clearly point out the space direction of the leakage point from the camera view, and the two-dimensional point is expanded to the straight line range in the three-dimensional space. In this way, a clear space geometric carrier can be provided for subsequent coordinate system transformation and three-dimensional intersection, the space position uncertainty caused by relying on only single point coordinates is avoided, and it is ensured that the unique position of the leakage point in the three-dimensional space can be accurately locked based on the ray in the subsequent steps.

[0063] In the process of implementation, first, the origin of the camera coordinate system is determined, which is the camera optical center (0, 0, 0), which has been determined in the camera calibration process. Second, the two-dimensional coordinates of the leakage point in the camera coordinate system obtained in the previous step are called, which are located on the normalized image plane. Then, a three-dimensional direction vector is constructed based on the two-dimensional coordinates, which clearly shows the spatial direction from the camera optical center to the leakage point. Specifically, if the two-dimensional coordinates are (x, y), the corresponding three-dimensional direction vector can be defined as (x, y, 1), and the origin of the vector is the origin of the camera coordinate system. Then, the camera coordinate system ray in the camera coordinate system is constructed by combining the origin of the ray (the camera optical center) and the direction vector. The ray is uniquely determined in mathematics, which accurately describes the spatial path of the leakage point in the camera view. Finally, the ray is parameterized and the origin and direction vector information are stored, which provides accurate geometric data basis for subsequent coordinate system transformation.

[0064] More specifically, the step S443 transforms the camera coordinate system ray to obtain the world coordinate system ray. It should be understood that since the camera coordinate system and the world coordinate system (the coordinate system with the axle shell entity positioning reference as the origin) of the three-dimensional CAD model to be measured are independent of each other, the origins and coordinate axis directions of the two are different, if the camera coordinate system ray is directly used to intersect with the CAD model, the intersection point position will be wrong due to the non-uniform coordinate system, and the entity workpiece cannot be corresponded. Therefore, the application further converts the camera coordinate system ray to the world coordinate system through coordinate system transformation, so as to unify the spatial reference of the ray and the CAD model, and ensure that the ray can accurately reflect the direction from the camera to the leakage point in the world coordinate system. In this way, subsequent analysis and operation can be carried out in the same coordinate system, avoiding three-dimensional coordinate calculation errors caused by coordinate system deviation, and ensuring that the final obtained leakage point three-dimensional coordinates completely match the physical position of the axle shell entity.

[0065] In the implementation process, first, the extrinsic parameter data of the thermal imager relative to the world coordinate system is called from the system parameter library, wherein the world coordinate system is completely consistent with the three-dimensional CAD model coordinate system of the axle housing, ensuring the uniformity of the subsequent space calculation reference. The extrinsic parameter includes two types of core information, including rotation-related data for describing the posture conversion of the camera coordinate system to the world coordinate system, and translation-related data for describing the specific spatial position of the camera coordinate system origin in the world coordinate system, and all extrinsic parameters need to be calibrated in advance by a calibration method based on the three-dimensional CAD model of the axle housing to ensure that the data accuracy meets the space conversion requirements. Second, the origin of the camera coordinate system ray is converted to the world coordinate system, which is the camera optical center and is at the reference origin position in the camera coordinate system. During the conversion process, the position information of the camera optical center in the camera coordinate system is substituted into the preset world coordinate system conversion logic, combined with the rotation-related data and translation-related data in the extrinsic parameter, and the corresponding spatial position of the camera optical center in the world coordinate system is obtained through operation. Then, the direction vector of the ray is converted. Since the direction vector is only used to represent the extension direction of the ray and is not affected by the space translation operation, the translation-related data in the extrinsic parameter does not need to be introduced, and the direction vector of the ray in the camera coordinate system is directly processed by means of the rotation-related data in the extrinsic parameter, so that the direction vector in the world coordinate system can accurately reflect the extension direction of the ray. Finally, the converted origin spatial position and direction vector are combined to form a complete spatial ray in the world coordinate system. Then, the ray is verified for accuracy. The verification method is to obtain the corresponding camera coordinate system ray by reverse transformation, and compare the obtained ray with the original camera coordinate system ray to confirm that the deviation is within the preset allowable range. After verification, the spatial ray in the world coordinate system is stored, laying a data foundation for subsequent intersection operation with the three-dimensional CAD model of the to-be-measured axle housing.

[0066] More specifically, the step S444 calculates the first intersection point of the world coordinate system ray and the three-dimensional CAD model of the to-be-measured axle housing as the three-dimensional coordinates of the leakage point. It should be understood that since the world coordinate system ray only represents the spatial direction from the camera to the leakage point, the unique intersection point of the ray and the axle housing entity (represented by the three-dimensional CAD model) needs to be determined to determine the specific position of the leakage point on the entity workpiece. If the intersection step is missing, the ray is only an abstract spatial straight line and cannot be converted into specific three-dimensional coordinates that can be used for repair. Therefore, the present application further calculates the intersection point of the ray and the CAD model, and selects the first intersection point along the ray direction to exclude the interference of the internal cavity or back surface of the model, and lock the position of the first contact of the ray with the surface of the axle housing, i.e. the position of the leakage point on the entity surface. In this way, the unique and accurate three-dimensional coordinates of the leakage point can be obtained, ensuring that the coordinates directly correspond to the leakage position on the entity weld of the axle housing, providing a direct measurement and quick positioning operation reference for repair workers.

[0067] In the process of implementation, first, load the three-dimensional CAD model of the axle shell to be tested, which needs to be completely matched with the physical size of the axle shell entity workpiece, the model surface is formed by the preset discretization processing to form a continuous surface unit, and the world coordinate system ray obtained by the coordinate system transformation is called from the system data buffer, which contains the starting space position and the extension direction vector, to ensure that both are in the same world coordinate system for subsequent operation. Secondly, start the ray and model surface unit intersection method commonly used in the field of three-dimensional graphics, which can accurately calculate the spatial intersection point of the ray and the model surface of each discrete unit. During the operation, the system will traverse all the discrete units on the surface of the three-dimensional CAD model, and calculate the potential intersection point of the world coordinate system ray and each unit one by one, to ensure that no possible contact position is missed. Then, all the potential intersection points calculated are screened, and the screening logic is based on the starting point of the ray, the spatial distance between each intersection point and the starting point is calculated, and only the intersection point that first contacts the model surface in the direction away from the starting point of the ray is retained, which is the actual corresponding point of the ray and the entity surface of the axle shell. At the same time, two types of invalid intersection points are excluded, one type is the intersection point extending in the reverse direction of the ray and located on the other side of the starting point, and the other type is the intersection point extending in the forward direction of the ray but deep into the model or located on the back of the model. Finally, the only intersection point determined after screening is extracted, and the spatial coordinates of the intersection point are the three-dimensional coordinates of the leakage point. Then the coordinates are checked for validity, and the check is based on the pre-labeled weld area information in the three-dimensional CAD model to confirm that the coordinates fall within the weld area. After the check passes, the three-dimensional coordinates of the leakage point are associated and bound with the three-dimensional CAD model to generate a visual model view with coordinate annotations, which is pushed to the terminal device of the repair station in real time, and the repair worker can directly view the corresponding position of the coordinates on the entity axle shell through the terminal without additional conversion or comparison, and directly carries out weld repair operation on the position.

[0068] In summary, the method for detecting the gas tightness of the automobile axle shell weld based on the embodiments of the present application is clarified, which first uses the automatic pressure drop method to efficiently and preliminarily screen the gas tightness of the axle shell, to efficiently determine whether the workpiece has leakage. Based on this determination result, the leakage signal is used as a trigger mechanism to dynamically activate the thermal imaging system to capture the transient local low-temperature feature formed by the throttling effect of the high-pressure gas at the leakage point in a non-contact manner. Further, through in-depth analysis of the collected thermal imaging sequence and spatial correlation mapping with the digital three-dimensional model of the workpiece, a complete automatic process from macroscopic judgment of the leakage phenomenon to precise positioning of the three-dimensional spatial coordinates of the leakage source is finally formed. In this way, efficient automatic screening and precise defect positioning capability can be organically combined, thereby significantly improving the intelligent level and closed-loop efficiency of the entire production quality control link.

[0069] The above describes the basic principles of the present application in combination with specific embodiments, but it should be noted that the advantages, benefits, effects and the like mentioned in the present application are only examples and are not limiting, and these advantages, benefits, effects and the like cannot be considered as necessary for each embodiment of the present application. In addition, the specific details of the above embodiments are only for the purpose of illustration and understanding, and are not limiting, and the above details do not limit the present application to the above specific details.

[0070] In the above embodiments, the description of each embodiment has its own focus, and the parts not described or recorded in detail in a certain embodiment can be referred to the relevant description of other embodiments. In the several embodiments provided by the present application, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiments described above are only schematic, and the unit division is only a logical function division, and there can be other division ways in actual implementation. The unit described as a separate component can or can not be physically separated, and the components displayed as units can be or can not be physical units, i.e. they can be located in one place or distributed on a plurality of network units. In actual implementation, some or all of the units can be selected according to the actual needs to achieve the purposes of the embodiments.

[0071] It is obvious for those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application. Therefore, the embodiments should be considered as exemplary and non-limiting, and the scope of the present application is defined by the appended claims rather than the above description, and all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present application. Any reference signs in the claims should not be considered as limiting the claims.

[0072] In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. The plurality of units stated in the system claims can also be implemented by one unit through software or hardware.

[0073] Finally, it should be noted that the above description has been given for the purpose of illustration and description. In addition, the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application.

Claims

1. A method for testing the airtightness of weld seams in automotive axle housings, characterized in that, include: Compressed gas is injected into the sealed axle housing of the vehicle under test until the pressure reaches the target test pressure. The initial pressure, initial time, end pressure and end time are recorded during the pressure stabilization time and pressure test time. Calculate the pressure drop rate and leakage indicator based on the initial pressure, initial time, final pressure, and final time; In response to a true leakage indicator, a thermal imaging sequence of the axle housing of the vehicle under test is acquired; The leak point image is located and three-dimensional coordinates are mapped based on the thermal imaging sequence of the axle housing under test to obtain the three-dimensional coordinates of the leak point. Based on the thermal imaging sequence of the axle housing of the vehicle under test, the leak point is located and its three-dimensional coordinates are mapped to obtain the three-dimensional coordinates of the leak point, including: Time series difference and cooling signal extraction were performed on the thermal imaging sequence of the axle housing of the vehicle under test to obtain a cooling signal sequence; The cooling signal sequence is accumulated in the time domain and filtered spatially to obtain the cumulative cooling map; The cumulative cooling map is segmented into abnormal regions and the leak point is located to obtain the two-dimensional coordinates of the leak point. Ray projection and three-dimensional coordinate calculation are performed on the two-dimensional coordinates of the leak point to obtain the three-dimensional coordinates of the leak point; The thermal imaging sequence of the axle housing of the vehicle under test is subjected to time series difference and cooling signal extraction to obtain a cooling signal sequence, including: Calculate the difference image between each two adjacent frames in the thermal imaging sequence of the axle shell of the vehicle under test to obtain the difference image sequence; Thresholding is applied to the differential image sequence to obtain the cooling signal sequence; The cumulative cooling map is segmented into abnormal regions and the leak point is located to obtain the two-dimensional coordinates of the leak point, including: A probability field nonlinear mapping is performed on the cumulative cooling plot to obtain the leakage probability field; Calculate the Hessian matrix field of the leakage probability field; Principal curvature analysis and source point coordinate optimization were performed on the leakage probability field and Hessian matrix field to obtain the two-dimensional coordinates of the leakage point.

2. The method for detecting the airtightness of weld seams in automotive axle housings according to claim 1, characterized in that, Calculate the pressure drop rate and leakage indicator based on the initial pressure, initial time, final pressure, and final time, including: calculating the pressure drop rate using the following formula: ； in, As the initial pressure, To end the stress, End time, The initial time, This is the pressure drop rate.

3. The method for detecting the airtightness of weld seams in automotive axle housings according to claim 2, characterized in that, Based on the initial pressure, initial time, final pressure, and final time, calculate the pressure drop rate and leakage flag, including: comparing the pressure drop rate with a leakage judgment threshold; if the pressure drop rate is greater than the leakage judgment threshold, set the leakage flag to true; otherwise, set the leakage flag to false.

4. The method for detecting the airtightness of weld seams in automotive axle housings according to claim 3, characterized in that, The three-dimensional coordinates of the leak point are obtained by ray casting and three-dimensional coordinate calculation, including: The pixel coordinates of the two-dimensional coordinates of the leak point are denormalized to obtain the two-dimensional coordinates of the leak point in the camera coordinate system. Based on the two-dimensional coordinates of the leak point and the camera optical center in the camera coordinate system, a ray in the camera coordinate system is constructed; Perform a coordinate system transformation on the camera coordinate system ray to obtain the world coordinate system ray; The first intersection point between the ray in the world coordinate system and the three-dimensional CAD model of the axle housing under test is calculated as the three-dimensional coordinates of the leak point.

Images