Method and system for detecting abnormal bolt installation of air pressure gun

Abstract

The invention relates to the technical field of air pressure gun detection, and discloses an air pressure gun bolt installation abnormity detection method and system, and the method comprises the steps: obtaining an actual fastening force value, air pressure fluctuation amplitude and frequency, and a surface feature image in an installation process; the dynamic deviation between the actual fastening force and the preset fastening force is calculated, and a fastening force deviation sequence is formed; performing multi-dimensional discrete analysis on the deviation sequence based on an adaptive clustering algorithm to obtain the number of abnormal discrete point clusters; if the number ratio exceeds a preset threshold value, the bolt installation is marked to be abnormal; whether the air pressure fluctuation amplitude and frequency of the air pressure gun exceed a preset range or not is compared to judge whether the air pressure is stable or not; if the surface image is not stable, inputting the surface image into a trained image recognition model, and outputting a leakage trace probability; and if the probability exceeds a preset threshold value, collecting three-dimensional flow field data of the leakage point, and calculating to obtain the gas leakage amount. The method can realize accurate detection and quantitative analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of air gun detection, and particularly to a detection method and system for abnormal installation of bolts of an air gun. Background Art

[0002] In the industrial field, bolts, as an important part of connecting and fixing key components, the installation quality thereof directly affects the overall stability and safety of equipment. Especially in pneumatic systems and high-precision equipment, the monitoring of the tightening force and installation status of bolts is particularly important.

[0003] Currently, the detection of abnormal installation of bolts of an air gun mainly adopts technologies such as tightening force monitoring, air pressure fluctuation analysis, and image recognition. Tightening force monitoring relies on a torque sensor to judge the installation quality by measuring the deviation of the tightening force of the bolt, but it is difficult to reflect the trend of the overall installation process. Air pressure fluctuation analysis uses an air pressure sensor to detect pressure changes, but the air pressure fluctuation is greatly affected by factors such as the external environment and the operating state of the equipment, resulting in unstable detection results. Image recognition technology is used to analyze the airflow characteristics on the surface of the air gun. Although it can assist in judging the abnormal area, it is difficult to accurately identify the leakage point due to the limitation of imaging conditions. At the same time, in terms of gas leakage detection, ultrasonic leak detection technology judges the leakage situation by analyzing high-frequency sound waves. However, due to background noise interference and sound wave attenuation, its detection results can often only identify the existence of leakage, and cannot quantitatively calculate the leakage amount. The existing technologies often can only provide qualitative judgments and cannot quantify the specific quantitative data of gas leakage. Many traditional detection methods can only identify the existence of airflow anomalies and leakage, but fail to provide accurate leakage amount data, which makes the maintenance and treatment decisions lack a clear basis. In summary, there is a problem of low detection accuracy for abnormal installation of bolts of an air gun in the existing technologies. Summary of the Invention

[0004] The present invention provides a detection method and system for abnormal installation of bolts of an air gun to achieve accurate detection and quantitative analysis.

[0005] In a first aspect, to solve the above technical problems, the present invention provides a detection method for abnormal installation of bolts of an air gun, including: Obtaining the actual tightening force value, air pressure fluctuation amplitude, air pressure fluctuation frequency, and surface feature image during the bolt installation process of the air gun; Performing dynamic deviation calculation on the actual tightening force value and a preset tightening force value to obtain a tightening force deviation sequence; Performing multi-dimensional discrete analysis on the tightening force deviation sequence based on an adaptive clustering algorithm to obtain the number of abnormal discrete point clusters; If the proportion of the number of abnormal discrete point clusters exceeds a preset threshold, it is marked as abnormal bolt installation; Compare the air pressure fluctuation amplitude and frequency of the pneumatic gun with abnormal bolt installation with the preset amplitude threshold and frequency range respectively to determine whether the air pressure is stable; Compare the air pressure fluctuation amplitude and frequency of the pneumatic gun with abnormal bolt installation with the preset amplitude threshold and frequency range respectively to determine whether the air pressure is stable; If the air pressure is unstable, input the surface feature image of the pneumatic gun with unstable air pressure into the trained image recognition model to output the leakage trace probability; Based on ultrasonic leak detection technology, extract acoustic wave features from the image recognition result to obtain acoustic wave frequency features, and determine whether there is gas leakage according to the acoustic wave frequency features; If the leakage trace probability is greater than the preset threshold, obtain the three-dimensional flow field data of the leakage point, and calculate the leakage amount according to the three-dimensional flow field data of the leakage point to obtain the gas leakage amount.

[0006] Preferably, the dynamic deviation calculation of the actual fastening force value and the preset fastening force value to obtain the fastening force deviation sequence includes: The fastening force deviation sequence is calculated by the following formula:

[0007] where, is the fastening force deviation sequence, is the actual fastening force value, is the preset fastening force value.

[0008] Preferably, the multi-dimensional discrete analysis of the fastening force deviation sequence based on the adaptive clustering algorithm to obtain the number of abnormal discrete point clusters includes: Based on the preset deviation direction judgment rule, determine the deviation direction of the fastening force deviation sequence to obtain the deviation direction distribution map; Based on the adaptive clustering algorithm, perform anomaly recognition on the deviation direction distribution map to obtain the number of abnormal discrete point clusters.

[0009] Preferably, the comparison of the air pressure fluctuation amplitude and frequency of the pneumatic gun with abnormal bolt installation with the preset amplitude threshold and frequency range respectively to determine whether the air pressure is stable includes: When the air pressure fluctuation amplitude does not exceed the preset amplitude threshold, it is determined that the air pressure is in a stable state; When the air pressure fluctuation amplitude exceeds the preset amplitude threshold, further compare the air pressure fluctuation frequency with the preset frequency range. If the air pressure fluctuation frequency is not within the preset frequency range, it is determined that the air pressure is in an unstable state.

[0010] Preferably, if the air pressure is unstable, the surface feature image of the air pressure gun with unstable air pressure is input into a pre-trained image recognition model, and the leakage trace probability is output, including: The image recognition model is obtained by training a ResNet-18 neural network model; Through the input layer of the image recognition model, the data format of the surface feature image of the air pressure gun is converted to obtain an image in the form of a three-dimensional tensor; Through the convolutional layer of the image recognition model, convolutional operations are performed on the three-dimensional tensor image to extract the local texture features of the image, and a first feature map is obtained; Through the activation function layer of the image recognition model, non-linear feature transformation is performed on the first feature map to obtain an enhanced second feature map; Through the pooling layer of the image recognition model, downsampling is performed on the second feature map to obtain a feature map with reduced size; Through the fully connected layer of the image recognition model, feature integration and classification processing are performed on the feature map with reduced size, and a classification feature result related to the leakage state is output; Through the output layer of the image recognition model, probability calculation is performed on the classification feature result, and the leakage trace probability is output.

[0011] Preferably, if the air pressure is unstable, the surface feature image of the air pressure gun with unstable air pressure is input into a trained image recognition model, and the leakage trace probability is output, including: The training process of the image recognition model includes: Based on the supervised learning algorithm, the pre-prepared training set is optimized to obtain an optimized training set; The optimized training set is input into the image recognition model, and the predicted leakage trace probability is output; The difference between the predicted leakage trace probability and the true leakage trace probability is calculated to obtain an output error value; Based on the output error value, the backpropagation algorithm is executed to update the load demand model parameters and optimize the image recognition model; Repeat the above process to continuously optimize the load demand model until the output accuracy of the load demand on the training set reaches the preset output accuracy requirement, and stop training.

[0012] Preferably, if the leakage trace probability is greater than a preset threshold, three-dimensional flow field data of the leakage point is obtained, and based on the three-dimensional flow field data of the leakage point, leakage amount calculation is performed to obtain the gas leakage amount, including: The three-dimensional flow field data includes: velocity vector, density, leakage start time, leakage end time, and leakage cross-sectional area; It is calculated by the following formula:

[0013] Wherein, Q is the gas leakage amount, t1 is the starting time of leakage, t2 is the ending time of leakage, V is the velocity vector, p is the density, and A is the leakage cross-sectional area.

[0014] In a second aspect, the present invention provides a detection system for abnormal installation of a pneumatic gun bolt, including: A data acquisition module, configured to acquire the actual tightening force value, air pressure fluctuation amplitude, air pressure fluctuation frequency, and surface feature image during the bolt installation process of the pneumatic gun; A tightening force deviation module, configured to perform deviation calculation on the actual tightening force value and a preset tightening force value to obtain a tightening force deviation value; A discrete point distribution module, configured to determine its deviation direction according to the tightening force deviation value to obtain the bolt tightening force discrete point distribution; A discrete point quantity module, configured to perform abnormal judgment on the discrete points of the bolt tightening force discrete point distribution based on a preset bolt tightening force discrete point distribution interval threshold to obtain the number of abnormal discrete points; An abnormal judgment module, configured to mark it as abnormal bolt installation if the number of abnormal discrete points is greater than or equal to a preset abnormal discrete point quantity threshold; An air pressure stability judgment module, configured to compare the air pressure fluctuation amplitude and air pressure fluctuation frequency of the pneumatic gun with abnormal bolt installation with a preset amplitude threshold and frequency range respectively to judge whether the air pressure is stable; An image recognition module, configured to input the surface feature image of the pneumatic gun with unstable air pressure into a pre-trained image recognition model to obtain an image recognition result if the air pressure is unstable; A gas leakage judgment module, configured to perform acoustic wave feature extraction on the image recognition result based on ultrasonic leak detection technology to obtain an acoustic wave frequency feature, and judge whether there is gas leakage according to the acoustic wave frequency feature; A gas leakage amount module, configured to obtain the gas flow data of the leakage point and perform leakage amount calculation according to the gas flow data of the leakage point to obtain the gas leakage amount if the pneumatic gun has gas leakage. In a third aspect, the present invention further provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the detection method for abnormal installation of a pneumatic gun bolt described in any one of the above.

[0015] Fourthly, the present invention also provides a computer-readable storage medium, which includes a stored computer program. When the computer program runs, it controls the device where the computer-readable storage medium is located to execute the detection method for abnormal installation of pneumatic gun bolts described in any one of the above.

[0016] Compared with the prior art, the present invention has the following beneficial effects: (1) By dynamically calculating the actual tightening force of the pneumatic gun during bolt installation and discretely analyzing the deviation sequence by combining an adaptive clustering algorithm, the present invention can accurately identify abnormal bolt installation situations and improve the intelligent level of installation quality control.

[0017] (2) The present invention constructs a stability judgment mechanism with the amplitude of air pressure fluctuation as the main criterion and the fluctuation frequency as the auxiliary criterion, and adopts a conditional cascade judgment method, which effectively reduces the misjudgment probability and improves the recognition reliability of the system for abnormal air pressure fluctuations.

[0018] (3) The present invention uses an image recognition model trained based on the ResNet-18 neural network model to extract deep features and classify and recognize the surface image of the pneumatic gun, and can effectively identify potential leakage traces, realizing high-precision image-level leakage detection.

[0019] (4) After identifying that the leakage probability exceeds the standard, the present invention accurately calculates the leakage amount by obtaining the three-dimensional flow field data of the leakage point and performing spatio-temporal integration in combination with density and velocity vectors, and has the ability to quantitatively evaluate the leakage degree, which is helpful for systematic diagnosis and risk warning.

[0020] In summary, the present invention provides a detection method for abnormal installation of pneumatic gun bolts that integrates multi-source signal analysis and deep learning recognition mechanism, realizing precise detection and quantitative analysis of abnormal installation of pneumatic gun bolts. Brief Description of the Drawings

[0021] Figure 1 is a schematic flowchart of the detection method for abnormal installation of pneumatic gun bolts provided by the first embodiment of the present invention; Figure 2 is a schematic diagram of the detection system for abnormal installation of pneumatic gun bolts provided by the second embodiment of the present invention. Detailed Embodiments

[0022] 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.

[0023] Refer to Figure 1 , a detection method for abnormal installation of a pneumatic gun bolt according to a first embodiment of the present invention includes the following steps: S11, obtaining the actual tightening force value, air pressure fluctuation amplitude, air pressure fluctuation frequency, and surface feature image during the bolt installation process of the pneumatic gun; S12, performing dynamic deviation calculation on the actual tightening force value and a preset tightening force value to obtain a tightening force deviation sequence; S13, performing multi-dimensional discrete analysis on the tightening force deviation sequence based on an adaptive clustering algorithm to obtain the number of abnormal discrete point clusters; S14, if the proportion of the number of abnormal discrete point clusters exceeds a preset threshold, it is marked as abnormal bolt installation; S15, if the air pressure fluctuation amplitude and air pressure fluctuation frequency of the pneumatic gun with abnormal bolt installation are respectively compared with a preset amplitude threshold and frequency range to determine whether the air pressure is stable; S16, if the air pressure is unstable, input the surface feature image of the pneumatic gun with unstable air pressure into a trained image recognition model to output the probability of leakage traces; S17, if the probability of leakage traces is greater than a preset threshold, obtain the three-dimensional flow field data of the leakage point, and calculate the leakage amount according to the three-dimensional flow field data of the leakage point to obtain the gas leakage amount.

[0024] In step S11, obtain the actual tightening force value, air pressure fluctuation amplitude, air pressure fluctuation frequency, and surface feature image during the bolt installation process of the pneumatic gun; It should be noted that first, several key parameters during the bolt installation process of the pneumatic gun need to be obtained: the actual tightening force value, air pressure fluctuation amplitude, air pressure fluctuation frequency, and surface feature image.

[0025] For obtaining the actual tightening force value, first use torque sensors to monitor the tightening force during the bolt installation process in real time. These sensors are installed on the working components of the pneumatic gun, and the tightening force change of the bolt is recorded in real time through the output signals of the sensors. These data reflect the mechanical state of the bolt installation and can help determine whether the bolt is correctly installed within the specified range. The sensors convert the tightening force value into an electrical signal and transmit it in real time through a data acquisition system for subsequent analysis.

[0026] For the acquisition of the air pressure fluctuation amplitude and frequency, the air pressure sensor is responsible for monitoring the air pressure changes during the operation of the air pressure gun. The air pressure sensor is installed at a key position of the air pressure gun. By collecting the air pressure change data, it analyzes the fluctuation of the air pressure in different time periods. The air pressure fluctuation amplitude refers to the maximum fluctuation value of the air pressure within a period of time, and the air pressure fluctuation frequency refers to the number of air pressure fluctuations per unit time. Through data collection and analysis, the specific values of the air pressure fluctuation can be obtained, providing a basis for subsequent judgment of air pressure stability.

[0027] The surface feature image is obtained by an image acquisition device, using a combination of a thermal imaging camera and a high-resolution visible light camera. The thermal imaging camera can detect the temperature difference caused by air flow changes and capture the area of temperature fluctuation, while the visible light camera can record the image of the surface morphology of the air pressure gun. Through the synchronous shooting of these two cameras, detailed surface image data including the area of air flow change can be obtained. These images are analyzed by image processing technology for subsequent leakage detection and abnormal judgment. The features captured in the images will serve as the surface feature image, providing detailed information about the location and morphology of surface abnormalities.

[0028] The acquisition of these key data requires high-precision sensors and image acquisition devices, and is transmitted and processed in real time through a data acquisition system to ensure the real-time and accuracy of the entire detection process. The combination of these data provides a basis for subsequent detection of abnormal bolt installation, judgment of air pressure stability, and analysis of gas leakage.

[0029] In step S12, the actual tightening force value and the preset tightening force value are subjected to dynamic deviation calculation to obtain a tightening force deviation sequence, including: The tightening force deviation sequence is calculated by the following formula:

[0030] Wherein, is the tightening force deviation sequence, is the actual tightening force value, is the preset tightening force value.

[0031] It should be noted that in step S12, to identify whether the bolt tightening process is abnormal, the system first needs to obtain the actual tightening force value of the air pressure gun during the bolt installation process. This tightening force value is the data collected in real time by the torque sensor installed at the output end of the air pressure gun and uploaded to the control system. The sampling period of the actual tightening force is generally once every 10 milliseconds according to the equipment setting, which can meet the requirement of continuous monitoring of the entire bolt tightening process. In each sampling period, the system will obtain the instantaneous tightening force value at that moment, forming the original force data sequence during the tightening process.

[0032] To determine whether the current fastening operation meets the operation specifications, it is impossible to directly judge relying solely on the original actual fastening force value, because different types of bolts have different fastening force requirements in design. Therefore, the system needs to compare and refer to a standardized fastening force reference value. This reference value is the preset fastening force value. The system has pre-established a mapping table between bolt specifications and target fastening force values, which can be obtained through laboratory calibration, standard parameters provided by equipment manufacturers, or process specifications. During actual application, the control system automatically matches the preset fastening force value corresponding to the bolt by identifying the bolt specification information used at the current operation station. The bolt specification can be identified by a QR code scanner integrated in the mechanical equipment to read the bolt label information, or by the operator selecting the bolt type of the current batch through the operation terminal during the operation preparation stage, and the system completes the configuration of the fastening force threshold accordingly.

[0033] Taking the M10 steel structure bolt as an example, its expected fastening force set at the construction site is 35 N·m. When the system identifies that the bolt used in the current operation is of M10 specification, it will automatically call 35 N·m as the fastening force preset value for this task. At the same time, the actual fastening force data collected by the pneumatic gun is uploaded sequentially in time series, such as real-time torque values in N·m like 32, 34, 36, 37, 35, 34, etc. At this time, the control system calculates each actual value with 35 N·m to obtain the deviation value at each moment in turn. This calculation process is a pure subtraction operation, completed once, without introducing complex calculation modules such as neural networks and optimization algorithms. Therefore, the calculation speed is fast and the error is controllable.

[0034] A set of difference data obtained by the above method is called a fastening force deviation sequence. This sequence essentially represents the arrangement in the time dimension of the mechanical deviation at each sampling moment during the actual fastening process. Since the bolt fastening process is a dynamic process, the actual torque value may show different change trends before, during, and after tightening. Therefore, the deviation sequence naturally has time correlation. The deviation sequence recorded by the system not only reflects the single-point deviation at a certain moment, but more importantly, it can reflect whether there are structural abnormalities in the entire fastening process, such as continuous low values or violent fluctuations during a certain period.

[0035] For example, in a batch installation operation, if the deviation sequence during the bolt fastening process at a certain station is continuously negative, the system may record it as -4, -5, -6, -5, -7, etc. This continuous deviation from the preset value indicates that there is a systematic deficiency in this fastening process. On the contrary, if the deviation fluctuates within a reasonable range, for example, it repeatedly jumps between -1 and +2 N·m, it can be considered that this process meets the allowable deviation specification. The system will judge the entire deviation sequence according to the set tolerance interval to provide criterion support for whether to trigger the abnormal clustering algorithm subsequently.

[0036] In addition, the dynamic adaptability of the preset fastening force is another key feature of this step. Since pneumatic guns in industrial sites may need to adapt to bolts of different types, grades, and lengths, and these parameters directly affect the final target torque requirements, the preset value must have the ability to be dynamically adjusted. If the system only uses a fixed value as the comparison benchmark, it will lead to misjudgment as abnormal for some bolts and missed judgment for others. Therefore, this system realizes the automatic scheduling of the target torque value in the parameter library by binding with the material information, so as to ensure that the calculation of the fastening force deviation sequence has an accurate reference benchmark.

[0037] In step S13, based on the adaptive clustering algorithm, multi-dimensional discrete analysis is performed on the fastening force deviation sequence to obtain the number of abnormal discrete point clusters, including: Based on the preset deviation direction judgment rule, the deviation direction of the fastening force deviation sequence is determined to obtain a deviation direction distribution map; Based on the adaptive clustering algorithm, abnormal recognition is performed on the deviation direction distribution map to obtain the number of abnormal discrete point clusters.

[0038] It should be noted that in step S13, in order to further analyze the abnormal behavior during the bolt fastening process, the system introduces a multi-dimensional discrete analysis method based on the adaptive clustering algorithm, which is specifically used to identify the possible abnormal point clusters in the fastening force deviation sequence. In the previous stage of this step, the construction of the fastening force deviation sequence has been completed, and the difference between the actual fastening force and the preset fastening force at each time point has been clarified. This sequence reflects the deviation situation of the entire fastening process. On this basis, in order to further analyze the distribution trend, directionality, and aggregation degree of the deviation, the system uses the deviation direction judgment rule and the adaptive clustering algorithm to cooperate to complete the identification process of the discrete point clusters.

[0039] First, the system classifies the directionality of each data point in the fastening force deviation sequence according to a set of clear deviation direction judgment rules. This judgment rule is realized based on the numerical symbol judgment. Among them, when the deviation value at a certain time point is greater than zero, it means that the actual fastening force at this moment is higher than the preset value, and the system determines that this point is in an "overtight" state; on the contrary, when the deviation value is less than zero, it means that the actual fastening force is lower than the preset value, and the system determines it as an "undertight" state. If the deviation value is equal to zero, it is regarded as consistent with the preset value. The system traverses the deviation sequence in turn according to this judgment rule, judges the directionality of each deviation value point by point, and draws a deviation direction distribution map during the fastening process in the coordinate system. In this distribution map, the horizontal axis is the time series, the vertical axis is the deviation value, the positive value area represents overtight, and the negative value area represents undertight. The change trend of the deviation on the time axis can be intuitively observed through the graph.

[0040] After obtaining the deviation direction distribution map, the system further performs a clustering and recognition operation on the deviation points showing a central tendency in the map. The goal of clustering is to classify deviation points with similar deviation directions and close time positions into a cluster, so as to determine whether there is a systematic abnormal trend. To improve the stability and adaptability of the clustering effect, the system selects an adaptive clustering algorithm. This algorithm is a clustering method improved based on the traditional density clustering principle, and its core lies in being able to automatically adjust clustering parameters according to the distribution characteristics of the input data, such as the neighborhood radius and the minimum number of samples. Different from the K-means or DBSCAN algorithms with fixed parameters, the adaptive clustering algorithm does not need to artificially specify the number of clusters before clustering. Instead, it dynamically determines the clustering boundary by automatically scanning and adjusting the data density, thus better fitting the distribution structure of the actual data.

[0041] Before the system performs the clustering operation, it first normalizes the deviation direction distribution map, maps the deviation values and the time axis to the standardized numerical intervals respectively to eliminate the influence brought by different bolt specifications and different tightening cycles. Subsequently, the adaptive clustering algorithm traverses this data space and determines whether each data point belongs to an existing cluster or should be used as a new clustering center according to the local density distribution of each data point. During the clustering process, the algorithm iterates continuously to identify one or more clusters of local high-density points, and these point clusters represent the periods of continuous anomalies occurring during the tightening process. Finally, the system counts the number of all identified abnormal point clusters and compares it with the abnormal proportion threshold set by the system to determine whether the tightening process needs to be marked as an abnormal bolt installation.

[0042] For example, during a certain bolt tightening process, there are multiple consecutive negative points in the tightening force deviation sequence. After the deviation direction is judged, the system determines it as the "under-tightening" state. If the distribution of these points on the time axis is relatively concentrated, the system plots them in the deviation direction distribution map and identifies an obvious low-value point cluster through the clustering operation. Suppose the number of abnormal points included in this point cluster accounts for 25% of the total number of sampling points in the entire tightening cycle, and the default threshold of the system is 20%, then the system determines that there is an obvious "under-tightening" trend in this tightening process and outputs an abnormal prompt.

[0043] In step S14, if the proportion of the number of the abnormal discrete point clusters exceeds the preset threshold, it is marked as an abnormal bolt installation.

[0044] It should be noted that in step S14, the system further comprehensively determines the installation state of the bolt based on the number of abnormal discrete point clusters identified in the previous stage. The core of this step is to establish a set of criteria for identifying "abnormal bolt installation" by setting the proportion threshold of the number of abnormal point clusters in the entire tightening process. When the number of abnormal discrete point clusters counted by the system exceeds this preset threshold, it is considered that there is a significant abnormal risk in the current bolt installation process, and marking or alarm processing needs to be carried out.

[0045] In specific implementation, the system first counts the total number of abnormal discrete point clusters identified by clustering according to the deviation direction distribution map and the output result of the adaptive clustering algorithm. Each discrete point cluster contains a certain number of continuous or approximately adjacent deviation data points, which are consistent in deviation amplitude and deviation direction, reflecting the behavior trend that the tightening force output by the pneumatic gun deviates from the target value for a long time within a certain period. Subsequently, the system obtains the total number of data points participating in sampling during the entire tightening cycle and calculates the proportion of abnormal discrete point clusters. The calculation method of the proportion value is the sum of the number of deviation points included in all abnormal point clusters divided by the total number of sampling points in the entire tightening process.

[0046] The default threshold configured by the system is twenty percent. That is to say, when the number of data points in the abnormal discrete point cluster accounts for twenty percent or more of the total sampling quantity, the system will mark this bolt as "abnormally installed". The setting of this threshold is based on engineering verification results and quality control standards. In the actual industrial assembly process, pneumatic tools and sensors may fluctuate within a certain range in a short time due to factors such as air pressure fluctuations, mechanical wear, or external interference. Therefore, a certain tolerance space is reserved in engineering judgment. Through statistical analysis of a large amount of experimental data, it is found that when the proportion of abnormal points exceeds twenty percent, the error trend often no longer belongs to the category of accidental disturbances, but has the characteristics of continuous and systematic deviation. Such deviations are often associated with installation defects such as not tightened, over-tightened, or missed fastening.

[0047] Taking a specific operation scenario as an example, if the total number of sampling points in a bolt tightening cycle is 200 points, and after adaptive clustering analysis, it is found that there are 3 abnormal discrete point clusters, containing 12, 15, and 17 abnormal points respectively, with a total of 44 abnormal points. At this time, the proportion of abnormal points is twenty-two percent, which has exceeded the set threshold of twenty percent. The system immediately determines that there is a relatively serious concentrated trend of tightening force deviation for this bolt, and then marks this bolt as abnormally installed and makes a graphical identification on the system monitoring interface to prompt the operator to conduct a recheck or re-tightening operation.

[0048] In addition, in order to improve the engineering adaptability of the system, the system supports configuration adjustment of the threshold. In mass production assembly lines or high consistency processes, the threshold can be set to 15% or even lower to enhance the sensitivity of quality control. In field maintenance, non-standard parts assembly and other working conditions, in order to avoid false alarms, the value can be set to 25% to relax the judgment conditions.

[0049] In step S15, the air pressure fluctuation amplitude and the air pressure fluctuation frequency of the air pressure gun with the bolt installation abnormality are compared with the preset amplitude threshold and frequency range respectively to determine whether the air pressure is stable, including: When the air pressure fluctuation amplitude does not exceed the preset amplitude threshold, it is determined that the air pressure is in a stable state; When the air pressure fluctuation amplitude exceeds the preset amplitude threshold, the air pressure fluctuation frequency is further compared with the preset frequency range. If the air pressure fluctuation frequency is not within the preset frequency range, it is determined that the air pressure is in an unstable state.

[0050] It is worth noting that in this step, after the system identifies that the bolt has a tightening anomaly, it needs to further determine whether the air pressure condition that caused the anomaly is stable. The judgment logic is based on the dual-feature analysis of the air pressure fluctuation signal, namely the amplitude value and frequency value of the air pressure fluctuation. This two-layer judgment mechanism is used to improve the robustness of system recognition and reduce the risk of misjudgment caused by a single feature error.

[0051] First, the system performs envelope extraction processing on the collected air pressure signal, calculates the difference between its maximum and minimum values in the unit time window, and obtains the air pressure fluctuation amplitude in the time period. The amplitude reflects the intensity of the air pressure in a certain period. The larger the value, the more intense the air pressure fluctuation. The system sets the threshold of the air pressure fluctuation amplitude to 0.15 MPa. Under standard stable air supply conditions, the air pressure fluctuation amplitude is mostly between 0.08 MPa and 0.12 MPa. In the event of an abnormality in the air pressure system, especially due to a delayed response of the pressure regulating valve or discontinuous air supply, the fluctuation amplitude will increase to more than 0.15 MPa. Therefore, it is reasonable to set this value as a standard for judging whether the air pressure fluctuation is too large.

[0052] If the pressure fluctuation amplitude does not exceed 0.15 MPa within a certain detection cycle, the system can directly determine that the current pressure state is stable, and no further processing is required, and enter the next detection process. If the fluctuation amplitude exceeds the threshold, it means that there is an abnormal fluctuation in the current pressure supply, and the system will further analyze the frequency characteristics of the pressure signal.

[0053] The air pressure fluctuation frequency is obtained by performing fast Fourier transform processing on the air pressure time series signal to obtain its main frequency component. This main frequency value reflects the oscillation frequency of the air pressure signal and is closely related to the pressure regulation frequency and gas supply rhythm in the gas source system. The set air pressure fluctuation frequency range of the system is 1 to 5 Hz. This range also comes from the statistical results of the signal spectra under a large number of normal air pressure conditions. Under standard operating conditions, the oscillation frequency of the regulating valve in the gas supply system generally fluctuates between 2 and 3.5 Hz, and there are very few rapid fluctuations exceeding 5 Hz or slow responses below 1 Hz. Therefore, setting 1 to 5 Hz as a reasonable frequency stability interval can cover the fluctuation rhythm under normal conditions and effectively eliminate abnormal signal interference.

[0054] In actual operation, if the air pressure fluctuation amplitude exceeds 0.15 MPa within a certain detection period, the system will perform frequency analysis on this section of the air pressure signal and extract its main frequency component. If the extracted main frequency value falls outside the range of 1 to 5 Hz, the system determines that the air pressure state within this period is unstable and needs to trigger the image recognition link to further determine whether there is a leak.

[0055] Taking a specific case as an example, during the process of tightening bolts with a certain air pressure gun, the measured air pressure signal fluctuation amplitude is 0.19 MPa and the main frequency is 6.2 Hz. The system first determines that the amplitude exceeds 0.15 MPa and cannot be directly recognized as a stable state, so it extracts the main frequency. Since the main frequency is higher than 5 Hz, exceeding the upper limit of the stable frequency set by the system, the system finally determines that the current air pressure state is unstable.

[0056] The double-layer judgment structure in this step can effectively avoid mis-triggering the leak recognition module due to short-term abnormalities. When the air pressure fluctuation amplitude is large but the frequency is within the normal range, the system will not immediately determine it as unstable, thus improving the accuracy and fault tolerance of the overall system recognition. This judgment logic is based on the actual data law in the engineering field, has sufficient theoretical basis and feasibility, and can be directly applied to the stability evaluation link of the intelligent detection system.

[0057] In step S16, if the air pressure is unstable, the surface feature image of the air pressure gun with unstable air pressure is input into a pre-trained image recognition model to output the probability of leakage traces, including: The image recognition model is trained from the ResNet-18 neural network model; Through the input layer of the image recognition model, the data format of the air pressure gun surface feature image is converted to obtain an image in the form of a three-dimensional tensor; Through the convolutional layer of the image recognition model, convolutional operations are performed on the three-dimensional tensor image to extract the local texture features of the image and obtain the first feature map; Through the activation function layer of the image recognition model, perform a non-linear feature transformation on the first feature map to obtain an enhanced second feature map; Through the pooling layer of the image recognition model, perform downsampling on the second feature map to obtain a feature map with reduced size; Through the fully connected layer of the image recognition model, perform feature integration and classification on the feature map with reduced size, and output a classification feature result related to the leakage state; Through the output layer of the image recognition model, perform probability calculation on the classification feature result and output the leakage trace probability.

[0058] It should be noted that in this step, when the system determines that the current pneumatic gun is in an unstable air pressure state according to the judgment process of the previous stage, it enters the leakage trace detection stage based on image recognition. The core goal of this stage is to extract visual features and perform pattern recognition on the surface image of the pneumatic gun in the current operation to determine whether there are visible traces caused by gas leakage. To achieve this goal, the system has pre-deployed an image recognition model trained based on the ResNet-18 architecture. This model has a complete deep convolutional neural network structure, which can extract multi-level texture features in the input image and perform pattern classification processing, so as to output the probability value of whether there are leakage traces in the image.

[0059] Before performing the image recognition task, the system first calls the surface image corresponding to the pneumatic gun device with unstable current air pressure. This image is collected by an industrial camera installed at the operation position, with a resolution of 1280×720 and an RGB three-channel color channel. Before the original image is input into the neural network, it needs to undergo data format conversion processing in the input layer. The input layer of the image recognition model does not directly accept image files, but requires the image data to be converted into a three-dimensional tensor format, that is, a numerical matrix with the structure of the number of channels, height, and width. This conversion process includes image scaling, normalization processing, and channel transformation. Specifically, the system scales the original image to a size of 224×224, and normalizes the pixel values of the three RGB channels to between 0 and 1, and finally forms a three-dimensional floating-point tensor with a size of 3×224×224 as the input of the neural network.

[0060] This tensor first passes through the first convolutional network layer of the model. The initial convolutional layer of ResNet-18 uses a convolutional operation with a 7×7 convolutional kernel and a stride of 2 to quickly obtain low-level edge, contour, and other texture information in the image. This layer outputs a feature map with 64 channels, and through subsequent batch normalization and ReLU activation functions, the response ability of edge features in the initial stage is further enhanced. The main role of the convolutional layer is to automatically extract spatial local features at different scales from the original pixels, enabling the model to perform more complex pattern reasoning based on these features in subsequent layers.

[0061] After passing through the initial convolutional layer, the image data enters the backbone part of the model. ResNet-18 contains 4 residual module stages, each stage consists of several basic residual units, and each residual unit contains two 3×3 convolutional layers. The introduction of these residual modules solves the problems of gradient disappearance and feature degradation that are prone to occur when traditional deep networks deepen the number of layers, enabling the model to extract deeper image semantic features while maintaining training stability. In each residual unit, after the feature map undergoes a convolutional operation, it will be element-wise added to the input feature map of this unit to form a residual connection. This design allows the model to learn "the difference from the input" rather than completely reconstructing the output, which helps to focus on the small feature differences in different regions of the image.

[0062] In the task of identifying leakage traces, this multi-level and cross-scale convolutional and residual connection structure is crucial. Leakage traces appear as local spots with a small area, blurred edges, or aerosol deposition and other image textures, and their forms are not obvious. Shallow convolution can capture texture boundaries, while deep convolution can combine more extensive context information to determine whether the area has leakage features. ResNet-18 extracts the spatial information of the image at multiple depth levels, enabling the model to accurately distinguish between leakage areas and non-leakage areas in an image environment with complex backgrounds and diverse textures.

[0063] After each set of convolutional operations, the model introduces the non-linear activation function ReLU. The ReLU function suppresses negative values in the convolutional output to zero and only retains positive values. This operation not only improves the expression ability of the model but also increases the responsiveness of the network to high-intensity features, enhancing the texture response of potential leakage areas, thereby providing a more discriminative feature map for subsequent discriminative layers.

[0064] To avoid waste of computing resources caused by excessive dimensions of the feature map during the convolution superposition process, the model uses max pooling operations between residual modules. The pooling layer downsamples the feature map output by the convolution with a fixed window, retaining the maximum response value of each local area. This operation helps compress the size of the feature map while enhancing the invariance of features, making the model more robust to small geometric perturbations such as offsets and rotations in images. In the leak detection scenario, this is very crucial because the position information of the leak trace may have a slight offset when the same pneumatic gun takes pictures at different angles. The introduction of the pooling layer ensures that the model can still accurately identify leak features under different shooting conditions.

[0065] After completing the processing of all convolutional layers and pooling layers, the model inputs the finally obtained low-dimensional feature map into the fully connected layer. The role of the fully connected layer is to expand the previously extracted spatial feature vector into a one-dimensional vector and perform a weighted combination of all features to form a comprehensive representation for classification. In this step, the fully connected layer converts the feature map into a probability vector representing the leak and non-leak states. Each element in the vector corresponds to a category, and the larger its value, the higher the probability of that category.

[0066] Finally, the model inputs this probability vector into the output layer for normalization processing. The output layer uses the Softmax function to map the output of the fully connected layer to probability values between 0 and 1, where the maximum value represents the category most likely to be determined by the model. In this system, only two output categories are set, namely "leak trace present" and "no leak trace". The probability of the leak trace output by the model is the percentage value of the possibility of the presence of a leak trace in this image. For example, if the output is 0.87, it means the possibility of the presence of a leak trace in the image is 87%. The system then compares this probability value with the set threshold. If it exceeds 0.75, it enters the subsequent leak rate calculation process; otherwise, it is considered that no effective leak sign is found in this image.

[0067] For example, during a tightening process, the air pressure at a certain station fluctuates severely. After the system extracts the surface image of the pneumatic gun at this station and inputs it into the image recognition model. After the image is processed by the model, the output leak probability is 0.91, which is significantly higher than the set determination threshold of 0.75. Based on this, the system determines that there is a high probability of a leak trace in the current image and enters the next stage of three-dimensional flow field acquisition and leak rate calculation process.

[0068] In step S16, the surface feature image of the pneumatic gun with unstable air pressure is input into the trained image recognition model, and the leak trace probability is output, including: The training process of the image recognition model includes: Based on the supervised learning algorithm, the pre-prepared training set is optimized to obtain an optimized training set; Input the optimized training set into the image recognition model to output the predicted probability of leakage traces; Calculate the difference between the predicted probability of leakage traces and the true probability of leakage traces to obtain the output error value; Based on the output error value, execute the backpropagation algorithm to update the load demand model parameters and optimize the image recognition model; Repeat the above process to continuously optimize the load demand model until the output accuracy of the load demand on the training set reaches the preset output accuracy requirement, and then stop the training.

[0069] In step S16, in this step, in order to enable the image recognition model to have the ability to recognize leakage traces on the surface of the pneumatic gun, the model needs to be trained with supervised learning. The entire training process focuses on the matching degree between the input image and the true label, and continuously adjusts the model parameters to reduce the prediction error, and finally obtains an image recognition model with high recognition accuracy. This training process follows the standard deep neural network training mechanism, relies on the data-driven method to gradually converge the recognition ability of the model, and has been completed before the system is deployed.

[0070] The training phase first relies on a high-quality image training set. The image data of this training set comes from the surface images of pneumatic guns collected in real operation sites and simulated laboratory environments. These images are collected by fixed industrial cameras or shooting devices installed at the end of the robot. During the shooting process, multi-angle, multi-illumination condition, and multi-distance sampling strategies are adopted to ensure that the training images cover the working condition differences under different environments. For example, in an experiment on tightening bolts of a wind power flange, the researchers collected an initial data set of 3000 images by setting abnormal air pressure fluctuations and deliberately creating small leaks. Among them, there are 2000 images without leaks and 1000 images with leakage traces. All images are labeled by professional personnel according to the actual detection results, and the labeled labels are two classifications: "with leakage traces" or "without leakage traces".

[0071] After the image annotation is completed, the system performs optimization processing on the image training set. The optimization process includes unifying the image format, scaling the size to 224×224 pixels, pixel normalization processing, and data augmentation operations. Data augmentation includes methods such as random flipping of images, color perturbation, and cropping, which are used to improve the generalization ability of the model and avoid overfitting caused by a single sample distribution. The optimized training set is divided into a training set and a validation set, and the typical ratio is 8 to 2. The training set is used for fitting the model parameters, and the validation set is used to periodically evaluate the performance of the model to prevent overfitting during the training process.

[0072] The training process is carried out in a supervised learning manner. The model receives the input of the training set images and outputs the probability value of the existence of leakage traces in each image. This probability value is a floating point number between 0 and 1, which is used to express the confidence level of the model's judgment that the image belongs to the "leakage" category. For example, when inputting an image known to have leakage, the initial output value of the model may be 0.42, indicating that the model does not correctly identify the leakage traces at present. The system compares this predicted value with the true label of the image and calculates the error value. The calculation of the error is based on the cross-entropy loss function. This loss function measures the deviation between the model's predicted value and the true label. The larger the value, the more inaccurate the prediction, and the smaller the value, the closer the model's recognition result is to the actual situation.

[0073] The calculation result of the error value is used to guide the adjustment of the model parameters. The system uses the backpropagation algorithm to update the parameters of each layer of the model. The backpropagation algorithm is a method that calculates the contribution degree of each parameter in the neural network to the total error through the chain rule. In the current structure, the system starts from the output layer, calculates the partial derivative of the loss function with respect to the output of each layer in turn, and passes this gradient to the previous layer. And so on, until it is passed to the input of the first layer. The weight parameters in each layer are updated according to this partial derivative. The specific update method is to subtract the learning rate multiplied by the gradient value from the original parameter value. The learning rate is a preset constant of the system, which controls the step size of each parameter update, and a common value is 0.001.

[0074] In this neural network structure, the convolutional layer, fully connected layer, etc. all contain parameters to be updated. In the convolutional layer, the weight matrix in the convolution kernel is updated, and in the fully connected layer, the weight values between the connected nodes are updated. For example, in the first convolutional layer of the model, if a certain convolution kernel does not respond strongly to specific leakage texture features in the training samples, after the system calculates its gradient through backpropagation, the weight value of this convolution kernel is adjusted so that it can better respond to such features in the next training. This way of parameter iteration based on error feedback can continuously optimize the model's ability in leakage image recognition.

[0075] Each training cycle is called an epoch. In one epoch, the system will traverse all the training set samples and perform a complete forward prediction and backpropagation process. The training process lasts for multiple epochs until the system evaluates that the recognition accuracy of the model on the validation set reaches the set standard, or the change value of the loss function is less than the set threshold in consecutive multiple epochs. Taking this system as an example, in a model training process, the training termination condition is set as the validation set accuracy not less than 94% and the loss value converges to less than 0.05. Finally, the model reaches this condition after the 28th epoch of training, the training process automatically stops, and the current parameters are saved as the final model.

[0076] For example, in a training experiment, the predicted values of the system's initial model for a group of images with leaks were generally lower than 0.5, indicating inaccurate judgment of leakage traces. As the number of training rounds increased, the error value gradually decreased. At the 10th round, the predicted value of this group of images increased to 0.63, and at the 20th round, it increased to 0.81. Until the final model output predicted value reached 0.91, indicating that the model could already judge the existence of leakage traces with a relatively high confidence level.

[0077] In step S17, if the probability of the leakage trace is greater than a preset threshold, three-dimensional flow field data of the leakage point is obtained, and based on the three-dimensional flow field data of the leakage point, leakage amount calculation is performed to obtain the gas leakage amount, including: The three-dimensional flow field data includes: velocity vector, density, leakage start time, leakage end time, and leakage cross-sectional area; It is calculated by the following formula:

[0078] where Q is the gas leakage amount, t1 is the leakage start time, t2 is the leakage end time, V is the velocity vector, p is the density, and A is the leakage cross-sectional area.

[0079] It should be noted that in this step, after the system outputs the probability of the leakage trace through the image recognition model, if this probability value is greater than the preset threshold set by the system, it is considered that the current air pressure gun has a relatively high leakage risk. To further verify the leakage situation and quantitatively evaluate the leakage degree, the system enters the gas leakage amount calculation process. This process obtains the total gas leakage within a specified time period by calling the three-dimensional flow field data of the leakage point and based on the mass flux integral calculation method. This calculation step not only has an identification function but also has a quantitative diagnosis ability, and can provide a core basis for on-site maintenance operations, equipment performance analysis, and energy consumption estimation.

[0080] The setting of the preset threshold is the trigger condition for this step. In the system, the leakage trace probability threshold is set to 0.75. This value is derived from the statistical data of the experimental data in the model verification stage. After the image recognition model is trained, an independent validation set is used to evaluate the recognition accuracy of the model. The actual measurement results show that when the predicted probability is greater than 0.75, the recognition accuracy of the model for leakage images reaches 96.4%, and the misjudgment rate is controlled within 3%. Therefore, in engineering practice, taking 0.75 as the trigger threshold for judging leakage can balance the accuracy of the model and the stability of the system response. If this value is set too low, the system will frequently trigger the flow field acquisition and calculation module, increasing the calculation burden; if it is set too high, some cases where there is actually leakage but the image features are not obvious will be missed. Therefore, selecting 0.75 as the balance point is a reasonable range determined through experimental verification and performance evaluation.

[0081] After the probability of the leakage trace exceeds the set threshold, the system first calls the three-dimensional flow field data module to extract the gas flow data of the leakage area corresponding to the current air pressure gun within the specified time period. This data mainly includes five parameters: velocity vector, gas density, leakage start time, leakage end time, and leakage cross-sectional area. Based on these five parameters, the system constructs a complete input for the flow field integral formula to achieve accurate calculation of the gas leakage volume.

[0082] The velocity vector data is used to describe the gas flow direction and velocity magnitude at each spatial position point in the leakage area, which is a vector value in three-dimensional space. The acquisition method of the velocity vector depends on the sensor array or the CFD simulation model constructed based on the computational fluid dynamics principle. In industrial deployment, a flow velocity sensor array is often used to perform real-time sampling of the three-dimensional space near the leakage point. The arrangement of the sensors is customized according to the geometric shape of the air pressure gun. For example, a 9-point gas flow velocity sensor is arranged three-dimensionally on the gun body surface, and each sensor outputs the flow velocity values in three directions, which are combined to form a local velocity vector field. The system uses an interpolation algorithm to expand the data of the limited measurement points to the entire leakage area to form a complete velocity field data.

[0083] The density parameter is used to represent the mass of gas per unit volume at each position and is a key factor for calculating the mass flux. The density acquisition method is based on the combined pressure-temperature measurement method. The system arranges temperature sensors and pressure sensors in the leakage area and uses the gas state equation to calculate the density value in real time. For example, the density of air at normal pressure and 25 degrees Celsius is about 1.18 kilograms per cubic meter. When it is detected that the pressure drops to 0.9 atmospheres and the temperature rises to 30 degrees Celsius at the same time, the density calculated by the system through the gas state formula is about 1.09 kilograms per cubic meter. This density value changes dynamically with time and the local air flow state, and the system inputs it as a function parameter into the flow field integration process to achieve dynamic coupling of the density.

[0084] The leakage start time and the leakage end time represent the start and end moments of this leakage event respectively, and are used to define the time boundary of the integration. These two time points are automatically determined by the system. The start time is taken as the time stamp when the image recognition model first outputs that the probability of the leakage trace exceeds the set threshold, and the end time is taken as the time point after the system closes the current leakage assessment process or the leakage image recognition result returns to below the threshold. In the system architecture, detailed time stamp information is recorded in each operation cycle, so the boundary of the leakage period can be accurately locked through the log system, and the error is controlled within the sampling period range. This design ensures that the integration time period is highly consistent with the actual leakage event occurrence time and improves the calculation accuracy.

[0085] The leakage cross-section area refers to the geometric area through which gas escapes from the leakage site and is the core boundary of spatial integration. This cross-section area is obtained through the structural modeling of the pneumatic gun and the auxiliary calibration method of image recognition. Before model deployment, engineering personnel establish a standard geometric model based on the configuration of the pneumatic gun head and preset the structural gap areas where leakage may occur. When the image recognition model determines that the leakage probability is relatively high, the system will further analyze the specific spatial position of the leakage feature area in the image and map this area back to the corresponding actual position in the geometric model. The system uses this position as the input for the leakage cross-section area in the integration process. Its area value is in square meters, and the shape is set as a regular or irregular surface according to the model.

[0086] After obtaining the above five parameters, the system substitutes them into the mass flow rate integration formula for leakage calculation. This formula consists of two integration steps. First, perform a spatial integration on the instantaneous mass flux across the leakage cross-section to calculate the total mass outflow at each moment. Second, integrate this result over the entire time period to obtain the cumulative leakage total. The integration process is completed by the numerical integration module. The time step is consistent with the sensor sampling period, and the spatial step is set according to the interpolation density of the velocity vector field.

[0087] The calculation result is a scalar value, representing the total mass of gas escaping from the leakage cross-section area within the specified time period. The unit of this result is kilograms, which reflects the actual physical quantity of gas leakage. For example, in a certain leakage event, the system records the start time as 10:15:04 on May 12, 2024, the end time as 10:15:14, and the duration as 10 seconds. During this time period, the average value of the velocity vector output by the flow velocity sensor is 3.2 m / s, the density value is 1.1 kg / m³, and the area of the leakage cross-section area is 5 square centimeters. Substituting the above data into the system, the calculated leakage amount is 0.0176 kg. This result is further used to generate a maintenance report to prompt the operator to perform a sealing inspection on this pneumatic gun.

[0088] The following takes a relatively common scenario as an example to describe the working process of the present invention. Please also refer to Figure 2 , which is Figure 1 a schematic diagram of the working scenario of the method.

[0089] The following combines Figure 1 the method flow shown in Figure 2 and the system structure shown in

[0090] In step S11, the data acquisition module obtains the actual tightening force value, air pressure fluctuation amplitude, air pressure fluctuation frequency, and the surface feature image of the pneumatic gun at the current moment during the operation process through real-time sensors connected to the pneumatic gun. Among them, the tightening force is output in real time by the torque sensor, and the sampling frequency is 100 Hz; the air pressure parameters are jointly obtained by the pressure sensor and the acoustic sensor; the image data is captured by an industrial camera installed on the workbench. All data are bound with corresponding timestamps and uploaded to the processing module for subsequent dynamic analysis.

[0091] Enter step S12. The tightening force deviation module compares the obtained actual tightening force value with the preset tightening force value set by the process, and calculates the tightening force deviation sequence that changes with time. In this example, the bolt for the current operation is of M16 specification, and its corresponding preset tightening force is 60 N·m. The sensor continuously records the actual torques as 58, 59, 55, 61, etc. After processing, a deviation sequence such as -2, -1, -5, +1, etc. is formed. This process is a one-dimensional time series calculation operation, which is the basis for subsequent discrete recognition.

[0092] In step S13, based on the aforementioned deviation sequence, the adaptive clustering module uses the built-in deviation direction judgment rule to judge whether each data point is over-tightened or under-tightened, and constructs a deviation direction distribution map. This map reflects the distribution trend of the deviation on the time axis. Subsequently, the system performs adaptive clustering processing on this distribution map, and extracts the number of abnormal point clusters by identifying dense and directionally consistent deviation point clusters. For example, in this operation, the system detects that there are three relatively dense under-tightened point clusters, and each point cluster contains more than 8 consecutive under-tightened data points, initially indicating a potential tightening abnormality trend.

[0093] In step S14, the installation abnormality module determines the proportion of the number of the above abnormal discrete point clusters. During the sampling period of the current operation, a total of 100 deviation points are obtained, and the proportion of abnormal point clusters is 24%. The set abnormal threshold of the system is 20%, and the current value has exceeded the threshold. Therefore, the system marks this bolt operation as "bolt installation abnormal" and prompts the operator in real time on the operation interface.

[0094] Enter step S15. The air pressure stability module further judges the air pressure state of the pneumatic gun at this abnormal station. First, the system detects that the air pressure fluctuation amplitude of this device is 0.19 MPa, which has exceeded the set stable upper limit of 0.15 MPa of the system. Subsequently, the main frequency value of this section of the air pressure signal is extracted as 6.1 Hz, which exceeds the system allowable range of 1 - 5 Hz. Based on this, the system determines that the current air pressure state is unstable and needs to further check whether it is caused by leakage.

[0095] In step S16, the leakage trace probability module starts the image recognition processing flow. The system calls the previously acquired surface image of the air pressure gun and inputs it into the trained ResNet-18 image recognition model. The model performs format conversion through the input layer, extracts local features through the convolutional layer, enhances texture response through the activation layer, reduces dimensions through the pooling layer, and then enters the fully connected layer for feature integration. Finally, the output layer calculates the leakage trace probability. In this example, the probability value output by the model is 0.84, exceeding the system-set threshold of 0.75. Therefore, the system believes that there is a leakage trace in this image.

[0096] Finally, in step S17, the gas leakage amount module starts the three-dimensional flow field calculation process according to the image recognition result. The system calls the velocity array sensor network deployed at the location of the device to obtain the velocity vector field of the current leakage area. At the same time, the gas density is deduced by the air pressure and temperature combined sensor. The system records the start time of leakage as 10:15:06, and the current time is the end time. Combining the area of the leakage cross-section area of the air pressure gun (preset to 4.5 square centimeters through modeling), the system substitutes each parameter into the integral calculation module, and finally outputs the leakage amount result of 0.023 kilograms and records it in the operation log in real time.

[0097] In summary, through the above steps, the system has completed the entire process closed-loop control from data acquisition, deviation analysis, anomaly determination, air pressure status confirmation, image recognition to leakage amount calculation. Figure 2 The shown functional modules respond in sequence to complete Figure 1 each processing node from S11 to S17 in, forming a detection system with clear structure, rigorous logic, and information intercommunication, significantly improving the efficiency and accuracy of leakage anomaly recognition in the air pressure gun bolt installation operation.

[0098] Referring to Figure 2 , the second embodiment of the present invention provides a detection system for air pressure gun bolt installation anomalies, including: A data acquisition module for acquiring the actual tightening force value, air pressure fluctuation amplitude, air pressure fluctuation frequency, and surface feature image during the bolt installation process of the air pressure gun; A tightening force deviation module for performing dynamic deviation calculation on the actual tightening force value and the preset tightening force value to obtain a tightening force deviation sequence; An adaptive clustering module for performing multi-dimensional discrete analysis on the tightening force deviation sequence based on the adaptive clustering algorithm to obtain the number of abnormal discrete point clusters; An installation anomaly module for marking as a bolt installation anomaly if the proportion of the number of abnormal discrete point clusters exceeds a preset threshold; A air pressure stability module, configured to compare the air pressure fluctuation amplitude and air pressure fluctuation frequency of the air pressure gun with abnormal bolt installation with a preset amplitude threshold and frequency range respectively to determine whether the air pressure is stable; A leakage trace probability module, configured to, if the air pressure is unstable, input the surface feature image of the air pressure gun with unstable air pressure into a trained image recognition model and output the leakage trace probability; An image recognition module, configured to, if the air pressure is unstable, input the surface feature image of the air pressure gun with unstable air pressure into a pre-trained image recognition model to obtain an image recognition result; A gas leakage amount module, configured to, if the leakage trace probability is greater than a preset threshold, obtain three-dimensional flow field data of the leakage point and calculate the leakage amount according to the three-dimensional flow field data of the leakage point to obtain the gas leakage amount.

[0099] It should be noted that a detection device for abnormal bolt installation of an air pressure gun provided in an embodiment of the present invention is used to execute all process steps of a detection method for abnormal bolt installation of an air pressure gun in the above embodiment. The working principles and beneficial effects of the two correspond one by one, so they will not be elaborated here.

[0100] An embodiment of the present invention further provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a detection program for abnormal bolt installation of an air pressure gun. When the processor executes the computer program, the steps in the above embodiments of the detection method for abnormal bolt installation of an air pressure gun are implemented, such as Figure 1 the step S11 shown. Alternatively, when the processor executes the computer program, the functions of each module / unit in the above device embodiments are implemented, such as the gas leakage amount module.

[0101] Exemplarily, the computer program may be divided into one or more modules / units. The one or more modules / units are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program in the electronic device.

[0102] The electronic device may be a computing device such as a desktop computer, a notebook, a palm computer, and a smart tablet. The electronic device may include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the above components are only examples of the electronic device and do not constitute a limitation to the electronic device. It may include more or fewer components than the above, or combine some components, or different components. For example, the electronic device may further include input / output devices, network access devices, a bus, etc.

[0103] The so-called processor may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The processor is the control center of the electronic device and connects various parts of the entire electronic device through various interfaces and circuits.

[0104] The memory can be used to store the computer program and / or module. The processor realizes various functions of the electronic device by running or executing the computer program and / or module stored in the memory, and by calling the data stored in the memory. The memory mainly includes a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for at least one function (such as a sound playback function, an image playback function, etc.); the data storage area can store data created according to the use of the mobile phone (such as audio data, phone book, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, at least one magnetic disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0105] Among them, if the modules / units integrated in the electronic device are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, to implement all or part of the processes in the above-described embodiment methods of the present invention, it can also be completed by a computer program instructing relevant hardware. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-described various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0106] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. In addition, in the attached drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that they have a communication connection, which can be specifically implemented as one or more communication buses or signal lines. Those of ordinary skill in the art can understand and implement it without creative work.

[0107] The above-described specific embodiments have further elaborated on the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only the specific embodiments of the present invention and is not used to limit the protection scope of the present invention. It is particularly pointed out that for those skilled in the art, any modifications, equivalent replacements, 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 detection method for abnormal installation of a pneumatic gun bolt, characterized in that, Including: Obtaining the actual tightening force value, air pressure fluctuation amplitude, air pressure fluctuation frequency, and surface feature image during the bolt installation process of the pneumatic gun; Performing dynamic deviation calculation on the actual tightening force value and the preset tightening force value to obtain a tightening force deviation sequence; Based on the adaptive clustering algorithm, performing multi-dimensional discrete analysis on the tightening force deviation sequence to obtain the number of abnormal discrete point clusters; If the proportion of the number of abnormal discrete point clusters exceeds the preset threshold, it is marked as abnormal bolt installation; Comparing the air pressure fluctuation amplitude and air pressure fluctuation frequency of the pneumatic gun with abnormal bolt installation with the preset amplitude threshold and frequency range respectively to determine whether the air pressure is stable; If the air pressure is unstable, input the surface feature image of the pneumatic gun with unstable air pressure into the trained image recognition model to output the leakage trace probability; If the leakage trace probability is greater than the preset threshold, obtain the three-dimensional flow field data of the leakage point, and calculate the leakage amount based on the three-dimensional flow field data of the leakage point to obtain the gas leakage amount.

2. The detection method for abnormal installation of the air pressure gun bolt according to claim 1, characterized in that, The performing dynamic deviation calculation on the actual tightening force value and the preset tightening force value to obtain a tightening force deviation sequence includes: Calculating the tightening force deviation sequence through the following formula: ； Among them, is the tightening force deviation sequence, is the actual tightening force value, is the preset tightening force value.

3. The detection method for abnormal installation of the pneumatic gun bolt according to claim 1, characterized in that, The performing multi-dimensional discrete analysis on the tightening force deviation sequence based on the adaptive clustering algorithm to obtain the number of abnormal discrete point clusters includes: Based on the preset deviation direction judgment rule, determining the deviation direction of the tightening force deviation sequence to obtain a deviation direction distribution map; Based on the adaptive clustering algorithm, performing abnormal recognition on the deviation direction distribution map to obtain the number of abnormal discrete point clusters.

4. The detection method for abnormal installation of the air pressure gun bolt according to claim 1, wherein The comparing the air pressure fluctuation amplitude and air pressure fluctuation frequency of the pneumatic gun with abnormal bolt installation with the preset amplitude threshold and frequency range respectively to determine whether the air pressure is stable includes: When the air pressure fluctuation amplitude does not exceed the preset amplitude threshold, it is determined that the air pressure is in a stable state; When the air pressure fluctuation amplitude exceeds the preset amplitude threshold, further compare the air pressure fluctuation frequency with the preset frequency range. If the air pressure fluctuation frequency is not within the preset frequency range, it is determined that the air pressure is in an unstable state.

5. The detection method for abnormal installation of the air pressure gun bolt according to claim 1, wherein The if the air pressure is unstable, inputting the surface feature image of the pneumatic gun with unstable air pressure into the pre-trained image recognition model to output the leakage trace probability includes: The image recognition model is trained by the ResNet-18 neural network model; Through the input layer of the image recognition model, performing data format conversion on the surface feature image of the pneumatic gun to obtain an image in the form of a three-dimensional tensor; Through the convolutional layer of the image recognition model, performing convolution operation on the three-dimensional tensor image to extract the local texture features of the image to obtain a first feature map; Through the activation function layer of the image recognition model, performing non-linear feature transformation on the first feature map to obtain an enhanced second feature map; Through the pooling layer of the image recognition model, performing downsampling processing on the second feature map to obtain a feature map with reduced size; Through the fully connected layer of the image recognition model, perform feature integration and classification processing on the feature map after size reduction, and output the classification feature results related to the leakage state; Through the output layer of the image recognition model, perform probability calculation on the classification feature results, and output the leakage trace probability.

6. The detection method for abnormal installation of the air pressure gun bolt according to claim 1, wherein If the air pressure is unstable, input the surface feature image of the air pressure gun with unstable air pressure into the trained image recognition model, and output the leakage trace probability, including: The training process of the image recognition model includes: Based on the supervised learning algorithm, optimize the pre-prepared training set to obtain the optimized training set; Input the optimized training set into the image recognition model, and output the predicted leakage trace probability; Perform difference calculation on the predicted leakage trace probability and the true leakage trace probability to obtain the output error value; Based on the output error value, execute the backpropagation algorithm to update the load demand model parameters and optimize the image recognition model; Repeat the above process, continuously optimize the load demand model until the output accuracy on the training set reaches the preset output accuracy requirement, and stop training.

7. The detection method for abnormal installation of the air pressure gun bolt according to claim 1, wherein If the leakage trace probability is greater than the preset threshold, obtain the three-dimensional flow field data of the leakage point, and perform leakage amount calculation according to the three-dimensional flow field data of the leakage point to obtain the gas leakage amount, including: The three-dimensional flow field data includes: velocity vector, density, leakage start time, leakage end time, and leakage cross-sectional area; It is calculated by the following formula: ； Where Q is the gas leakage amount, t1 is the leakage start time, t2 is the leakage end time, V is the velocity vector, p is the density, and A is the leakage cross-sectional area.

8. A detection system for abnormal installation of a pneumatic gun bolt, characterized in that, Including: A data acquisition module for acquiring the actual tightening force value, air pressure fluctuation amplitude, air pressure fluctuation frequency, and surface feature image of the air pressure gun during the bolt installation process; A tightening force deviation module for performing dynamic deviation calculation on the actual tightening force value and the preset tightening force value to obtain a tightening force deviation sequence; An adaptive clustering module for performing multi-dimensional discrete analysis on the tightening force deviation sequence based on the adaptive clustering algorithm to obtain the number of abnormal discrete point clusters; An installation abnormality module for marking as bolt installation abnormality if the proportion of the number of abnormal discrete point clusters exceeds the preset threshold; An air pressure stability module for comparing the air pressure fluctuation amplitude and air pressure fluctuation frequency of the air pressure gun with abnormal bolt installation with the preset amplitude threshold and frequency range respectively to determine whether the air pressure is stable; A leakage trace probability module for inputting the surface feature image of the air pressure gun with unstable air pressure into the trained image recognition model and outputting the leakage trace probability if the air pressure is unstable; A gas leakage amount module for obtaining the three-dimensional flow field data of the leakage point and performing leakage amount calculation according to the three-dimensional flow field data of the leakage point to obtain the gas leakage amount if the leakage trace probability is greater than the preset threshold.

9. An electronic device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the detection method for abnormal installation of pneumatic gun bolts described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program. When the computer program runs, it controls the device where the computer-readable storage medium is located to execute the detection method for abnormal installation of pneumatic gun bolts described in any one of claims 1 to 7.

Images