An online detection system for welding quality

Abstract

The present application relates to the technical field of welding quality detection, and discloses an online detection system for welding quality. The welding area monitoring device of the system is arranged around the working area of the welding equipment and is used for capturing optical radiation signals and heat distribution signals in the welding process. The multi-source data acquisition device is connected to the welding area monitoring device and synchronously acquires welding current fluctuation data, arc voltage change data and wire feeding speed data. The real-time feature extraction device receives the multi-source data and constructs a dynamic feature set of the welding process. The dynamic analysis device generates welding quality evaluation parameters according to the dynamic feature set. The output control device generates welding process parameter adjustment instructions based on the welding quality evaluation parameters. The system realizes real-time detection and dynamic control of welding quality, improves welding production efficiency and quality stability, and is suitable for various industrial welding scenes.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of welding quality detection, in particular to an online detection system for welding quality. BACKGROUND

[0002] In modern industrial production, welding as a key material connection process is widely used in mechanical manufacturing, petrochemical industry, aerospace, construction engineering and other fields, and its quality is directly related to the structural strength, safety performance and service life of the product. With the continuous improvement of the requirements of industrial production on product precision and reliability, the demand for welding quality detection is also increasingly strict. The traditional welding quality detection method is mainly offline detection, that is, after the welding operation is completed, the quality of the welded joint is evaluated by non-destructive testing techniques such as X-ray detection, ultrasonic detection, and magnetic powder detection. This detection method has obvious limitations. On the one hand, the detection process needs to be carried out after the welding is completed. If welding defects are found, the completed welding part needs to be reworked or scrapped, which not only increases the production cost but also prolongs the production cycle. Especially in large structural parts or continuous production scenarios, the economic loss and efficiency impact caused by offline detection are more significant. On the other hand, offline detection cannot reflect the dynamic changes in the welding process in real time, and it is difficult to capture problems such as current fluctuations, voltage abnormalities, and unstable wire feeding speed that occur instantaneously during the welding process. These instantaneous problems are often the key factors that lead to welding defects, and relying solely on post-detection cannot fundamentally control the welding quality.

[0003] With the development of automatic welding technology, some production scenarios have begun to introduce simple online monitoring methods, such as monitoring only a single parameter of welding current or voltage. However, this type of monitoring method has the problem of single monitoring dimension, and cannot fully reflect the complex state of the welding process. The welding process is a complex process involving optical, thermal, electrical and other multi-physical field coupling. Relying on single parameter monitoring alone cannot accurately determine key information such as welding pool state and heat affected zone temperature distribution, and thus cannot timely identify defects such as pores, cracks, and incomplete fusion that may occur during the welding process. In addition, some existing online monitoring systems lack effective data integration and analysis capabilities, and the multi-source data collected is often in a scattered state, which cannot form effective feature correlation, resulting in a lack of systematization and accuracy in evaluating welding quality, and making it difficult to meet the real-time control requirements of high-precision industrial production on welding quality. SUMMARY

[0004] The present application aims to provide an online detection system for welding quality to solve the problems raised in the background art.

[0005] In order to achieve the above object, the present application provides a welding quality-oriented online detection system, which comprises a welding area monitoring device, a multi-source data acquisition device, a real-time feature extraction device, a dynamic analysis device and an output control device.

[0006] The welding area monitoring device is arranged around the working area of the welding equipment and is used for capturing optical radiation signals and heat distribution signals in the welding process; the multi-source data acquisition device is connected to the welding area monitoring device and synchronously acquires welding current fluctuation data, arc voltage variation data and wire feeding speed data; the real-time feature extraction device receives the current fluctuation data, voltage variation data and wire feeding speed data output by the multi-source data acquisition device and receives the optical radiation signals and heat distribution signals output by the welding area monitoring device, and constructs a dynamic feature set of the welding process; the dynamic analysis device generates welding quality evaluation parameters according to the dynamic feature set provided by the real-time feature extraction device; and the output control device generates welding process parameter adjustment instructions based on the welding quality evaluation parameters provided by the dynamic analysis device.

[0007] Preferably, the welding area monitoring device comprises a high-resolution spectral acquisition unit and an infrared thermal imaging unit; the high-resolution spectral acquisition unit is installed at a specific distance behind the welding gun to capture spectral radiation intensity data of the welding pool area at a fixed sampling frequency; the infrared thermal imaging unit is arranged directly above the welding area to obtain temperature field distribution data of the welding heat affected zone in an asynchronous sampling manner; the multi-source data acquisition device comprises a current Hall sensor, a voltage differential probe and an encoder speed measurement module; the current Hall sensor is clamped at the output end of the welding power source, the voltage differential probe is connected in parallel across the welding gun, and the encoder speed measurement module is installed on the transmission shaft of the wire feeding mechanism.

[0008] Preferably, the real-time feature extraction device comprises a signal preprocessing unit and a feature fusion unit; the signal preprocessing unit performs sliding window normalization processing on the spectral radiation intensity data output by the high-resolution spectral acquisition unit and performs spatial interpolation compensation processing on the temperature field distribution data output by the infrared thermal imaging unit; the feature fusion unit aligns the processed spectral radiation intensity data and temperature field distribution data according to time sequence, simultaneously receives current fluctuation data collected by the current Hall sensor, voltage variation data collected by the voltage differential probe and wire feeding speed data collected by the encoder speed measurement module, and generates a multi-dimensional feature vector containing time sequence correlation.

[0009] Preferably, the real-time feature extraction device further comprises a feature database updating unit; the feature database updating unit receives the multi-dimensional feature vector generated by the feature fusion unit, extracts the feature value variation trend of a plurality of continuous sampling periods, and constructs a real-time feature library of the welding process; the real-time feature library is stored in categories according to welding material types and thickness specifications, and each storage entry contains a feature vector timestamp, corresponding welding process parameters and environmental humidity data.

[0010] Preferably, the dynamic analysis device comprises a defect probability calculation unit and a quality grade determination unit; the defect probability calculation unit extracts the feature vector of the current welding cycle from the real-time feature library, and calculates the deviation value of the feature vector from the historical qualified sample feature vector; the quality grade determination unit presets a first deviation threshold and a second deviation threshold, outputs a qualified determination signal when the deviation value is less than the first deviation threshold, outputs a suspicious determination signal when the deviation value is between the first deviation threshold and the second deviation threshold, and outputs a defect determination signal when the deviation value is greater than the second deviation threshold.

[0011] Preferably, the dynamic analysis device further comprises a defect positioning unit; the defect positioning unit synchronously calls the temperature field distribution data collected by the infrared thermal imaging unit and the spectral radiation intensity data collected by the high-resolution spectral acquisition unit when receiving the defect determination signal, and determines the specific position coordinates of the welding defect by comparing the spatial coincidence degree of the temperature gradient distribution abnormal area and the spectral feature mutation area.

[0012] Preferably, the output control device comprises a parameter adjustment unit and an alarm triggering unit; the parameter adjustment unit receives the suspicious determination signal output by the quality grade determination unit, generates a welding current correction amount, an arc voltage compensation amount and a wire feeding speed adjustment amount according to the optimal process parameter data of the same material specification in the real-time feature library; the alarm triggering unit receives the defect determination signal output by the quality grade determination unit, combines the specific position coordinates provided by the defect positioning unit, and generates an emergency stop instruction containing the defect type code and the position information.

[0013] Preferably, the system further comprises a historical data comparison device; the historical data comparison device extracts a feature vector sequence of the last several welding cycles from the real-time feature library, and calculates the dynamic time warping distance of the feature vector sequence from the feature vector sequence under the standard process parameters; when the distance exceeds a preset process stability threshold, the parameter adjustment unit sends a process parameter optimization request.

[0014] Preferably, the output control device further comprises a report generation unit; the report generation unit receives the determination signal output by the quality grade determination unit, the position coordinate data provided by the defect positioning unit, and the dynamic time warping distance data calculated by the historical data comparison device, and generates a detection report containing welding quality statistical indicators, defect distribution map and process stability index.

[0015] Preferably, the system further comprises a calibration device; the calibration device is connected with the welding area monitoring device and the multi-source data acquisition device, triggers the standard test block welding process periodically, collects the feature vector data under the standard welding condition, updates the standard sample data in the real-time feature library, and sends a calibration completion mark to the report generation unit.

[0016] Compared with the prior art, the present application has the beneficial effects that:

[0017] The online detection system for welding quality sets welding area monitoring device, multi-source data acquisition device, real-time feature extraction device, dynamic analysis device and output control device, and builds a complete welding quality real-time detection and management system, effectively solving the limitations in the traditional welding quality detection method.

[0018] From the data acquisition level, the welding area monitoring device is configured around the working area of the welding equipment, which can capture optical radiation signals and thermal distribution signals in the welding process. At the same time, the multi-source data acquisition device connects the welding area monitoring device to synchronously acquire welding current fluctuation data, arc voltage change data and wire feeding speed data, realizing comprehensive acquisition of optical, thermal and electrical multi-dimensional data in the welding process. This multi-source data synchronous acquisition method breaks through the limitations of traditional single parameter monitoring, can fully reflect the dynamic change state in the welding process, provides rich and comprehensive basic data for subsequent welding quality evaluation, and avoids quality misjudgment or omission caused by insufficient data dimensions.

[0019] In terms of data processing and feature construction, the real-time feature extraction device can simultaneously receive current fluctuation data, voltage change data, wire feeding speed data output by the multi-source data acquisition device and optical radiation signals and thermal distribution signals output by the welding area monitoring device, and construct a welding process dynamic feature set based on these data. Through the integration and correlation analysis of multi-source heterogeneous data, the dispersed original data is converted into a set of features that can reflect the essential characteristics of the welding process, making the data information more targeted and effective, and accurately reflecting the key information such as molten pool state and heat affected zone change in the welding process, providing a reliable feature basis for subsequent dynamic analysis device to carry out quality evaluation.

[0020] The dynamic analysis device generates welding quality evaluation parameters based on the dynamic feature set provided by the real-time feature extraction device. The evaluation parameters can reflect the quality state of the current welding process in real time, compared with the lagging quality evaluation in traditional offline detection, realizing real-time judgment of welding quality, discovering abnormal conditions in the welding process in time, avoiding further expansion of defects. The output control device generates welding process parameter adjustment instructions based on the welding quality evaluation parameters provided by the dynamic analysis device, so that when welding quality abnormalities are found, welding process parameters can be adjusted in time to pull the welding process back to normal state, reducing rework or scrap caused by quality problems, reducing production cost and shortening production cycle.

[0021] The whole system forms a close cooperative working mechanism among each device, from data acquisition, feature extraction, quality analysis to parameter adjustment, forming a closed-loop real-time control process, ensuring that the welding quality can be continuously monitored and dynamically adjusted in the whole welding process. This process design not only applies to the conventional welding production scene, but also can cope with the quality detection demand under different welding equipment and different welding process, has strong adaptability and universality, can provide stable and reliable quality guarantee for various industrial welding production, promotes the welding production from the traditional post-detection to real-time control, and improves the efficiency and quality stability of the whole welding production. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 The timing diagram of the welding quality-oriented online detection system described in the present application;

[0023] Figure 2 The flowchart for the welding area monitoring and multi-source data acquisition device refinement;

[0024] Figure 3 The flowchart for the feature database update of the real-time feature extraction device. DETAILED DESCRIPTION

[0025] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0026] Please refer to Figure 1The application provides an online detection system for welding quality, which realizes real-time monitoring and quality evaluation of the welding process by integrating various sensing devices and analysis modules. The core components of the system include a welding area monitoring device, a multi-source data acquisition device, a real-time feature extraction device, a dynamic analysis device, and an output control device. The welding area monitoring device is arranged around the working area of the welding equipment and is mainly used to capture optical radiation signals and heat distribution signals generated during the welding process, which reflect the dynamic changes of the welding pool and the heat affected zone. The multi-source data acquisition device is connected to the welding area monitoring device and synchronously acquires welding current fluctuation data, arc voltage change data, and wire feeding speed data, ensuring the time sequence consistency of multi-source information. The real-time feature extraction device receives current, voltage, and wire feeding speed data from the multi-source data acquisition device, as well as optical radiation and heat distribution signals from the welding area monitoring device, and constructs a dynamic feature set of the welding process through data fusion technology, which contains time sequence correlation information of various physical parameters. The dynamic analysis device performs pattern recognition and deviation calculation based on the dynamic feature set provided by the real-time feature extraction device, and generates welding quality evaluation parameters such as defect probability and quality level. The output control device generates corresponding welding process parameter adjustment instructions or control signals according to the quality evaluation parameters output by the dynamic analysis device, realizing closed-loop optimization of the welding process.

[0027] Example 1: see Figure 2The welding area monitoring device includes a high-resolution spectral acquisition unit and an infrared thermal imaging unit. The high-resolution spectral acquisition unit is installed at a specific distance behind the welding gun, which is usually set within the range of 100-200 mm to optimize signal capture. The unit uses a grating spectrometer principle combined with a CCD sensor array to achieve a fixed sampling frequency of 1000 Hz, covering a spectral band of 300-1100 nm. It can continuously monitor the spectral radiation intensity changes of the welding pool area, and the data is transmitted through shielded optical fibers to reduce interference. The infrared thermal imaging unit is arranged about 500 mm above the welding area. It uses a microbolometer as the core sensor and operates in an asynchronous sampling mode with a sampling frequency of 30 Hz. It acquires temperature field distribution data of the welding heat-affected zone with a temperature resolution of 0.1°C. The unit has a built-in non-uniformity correction algorithm to compensate for sensor drift in real time and ensure the spatial consistency of temperature field data. The multi-source data acquisition device connects these monitoring units, including a current Hall sensor, a voltage differential probe, and an encoder speed measurement module. The current Hall sensor is designed in a closed-loop mode and clamped on the cable at the output end of the welding power supply. Its measurement range covers 0-500 A with an accuracy of ±1%, and the output analog signal is transmitted after AD conversion. The voltage differential probe uses high-voltage isolation technology and is connected in parallel to the wires at both ends of the welding gun. Its measurement range is 0-50 V with a bandwidth of 1 MHz, capable of capturing microsecond-level voltage fluctuations. The encoder speed measurement module is an incremental encoder directly installed on the surface of the drive shaft of the wire feeding mechanism, generating 1000 pulses per revolution. The wire feeding speed value is converted by a counter circuit with an accuracy of ±0.1 m / min.

[0028] The real-time feature extraction device receives the data stream from the above device, the signal preprocessing unit of which performs sliding window normalization processing on the spectral radiation intensity data output by the high-resolution spectral acquisition unit, the window size of which can be configured as 100 consecutive sampling periods, the mean and standard deviation of the data in the window are calculated during processing, and z-score standardization is performed to eliminate the baseline fluctuation caused by ambient light, while the spatial interpolation compensation processing is performed on the temperature field distribution data output by the infrared thermal imaging unit, the bilinear interpolation algorithm is used to fill the spatial missing points caused by asynchronous sampling, the interpolation is completed based on the weighted average of adjacent pixels, and the temperature field is continuous and uninterrupted. The feature fusion unit integrates the preprocessed multi-source data, aligns the spectral radiation intensity data and the temperature field distribution data in time sequence, the alignment method adopts linear interpolation technology, synchronizes the asynchronous data to a unified time grid of 1000 Hz, and accesses the current fluctuation data collected by the current Hall sensor, the voltage change data collected by the voltage differential probe and the wire feeding speed data collected by the encoder speed measurement module, to generate a multi-dimensional feature vector. The vector contains spectral intensity mean, temperature gradient amplitude, current variance, voltage peak and wire feeding acceleration, etc. Each dimension represents the physical parameter state at a sampling time, and the vector structure is designed as a dynamic array to support real-time expansion to adapt to the needs of different welding processes.

[0029] The specific implementation of the high-resolution spectral acquisition unit focuses on environmental adaptability, the unit shell adopts water cooling design to prevent high temperature damage, optical filters are added in the optical path to suppress stray light, and the installation position is adjusted by a mechanical support to ensure that the optical axis is aligned with the center of the molten pool. During data capture, the exposure time of the CCD sensor is automatically adjusted to avoid signal saturation. The raw spectral data is first corrected for dark current and then transmitted to the preprocessing unit. The deployment of the infrared thermal imaging unit considers the field of view coverage, the unit viewing angle is set to 30°x20°, covering the entire heat-affected zone, and the sampling asynchronicity is controlled by an internal clock, the timestamp is synchronized with the system clock, and the temperature field data is output after performing radiation compensation based on the material emissivity database. The cable is completely embedded in the magnetic ring to reduce external magnetic field interference, and a low-pass filter is added to the signal output end with a cutoff frequency of 10 kHz to smooth high-frequency noise. The high-voltage probe line is used for connecting the voltage differential probe, and the parallel point is selected at the input terminal of the welding gun, and the probe ground end is independently isolated to prevent ground loop interference. The installation of the encoder speed measurement module ensures that the shaft center is aligned, the encoder signal is improved in resolution through a four times frequency circuit, and the speed value is updated every 10 milliseconds.

[0030] The algorithm of the signal preprocessing unit runs on an embedded digital signal processor, the sliding window normalization processing adopts a real-time pipeline architecture, the window sliding step is 1 sampling period, and the dynamic range of the normalized data is compressed to the interval [-1, 1] to facilitate subsequent fusion; the spatial interpolation compensation processing is performed for each frame of infrared thermal imaging data, the interpolation kernel size is 3*3 pixels, the weight is calculated according to the reciprocal of the distance, and the grid resolution of the processed temperature field is increased to 1.5 times of the original data. The implementation of the feature fusion unit is based on a multi-core processor, the time series alignment module first allocates a buffer for each data source, the buffer depth is set to 1000 samples, and when aligning, the linear interpolation method is used to interpolate the low-frequency data (such as 30 Hz temperature data) to 1000 Hz time points, and the interpolation error is controlled within 0.1%; the feature vector generation module extracts the features of the aligned data, for example, the spectral intensity mean value is obtained by calculating the average value of all wavelengths in the window, the temperature gradient amplitude is obtained by convolving the temperature field image with the Sobel operator, the current variance is calculated based on 100 sampling points, the voltage peak detection uses the local maximum value algorithm, and the wire feeding acceleration is obtained by differentiating the continuous speed value. The generated feature vector is attached with a time stamp and a serial number, and stored in a ring memory buffer.

[0031] The data stream management of the welding area monitoring device adopts a master-slave structure, the high-resolution spectral acquisition unit and the infrared thermal imaging unit are connected to the real-time feature extraction device through a gigabit Ethernet interface, the data transmission protocol is UDP, and the time stamp is checked; the sensor signals of the multi-source data acquisition device are connected to the analog input module through a shielded cable, the sampling rate is unified to 100 kHz, and the digital filtering is reduced to 1000 Hz after downsampling. The resource configuration of the signal preprocessing unit includes a dedicated memory pool for sliding window calculation, the window size can be adjusted through a software parameter, and different welding speeds are adapted; the spatial interpolation compensation processing uses GPU acceleration, the interpolation parameters are pre-stored in a lookup table, and are dynamically loaded according to the temperature field resolution. The multi-dimensional feature vector output format of the feature fusion unit is defined as a structure array, each vector contains 20 feature values, the vector sequence is directly transmitted to the dynamic analysis device through a DMA channel, the transmission interval is 1 ms, and the real-time performance is ensured.

[0032] The calibration of the welding area monitoring device is completed by periodic calibration, the high-resolution spectral acquisition unit calibrates the wavelength and intensity using a standard light source, and the infrared thermal imaging unit calibrates the temperature accuracy by using a black body source; the sensors of the multi-source data acquisition device are calibrated to zero every shift, the current Hall sensor is verified by a standard current source, the voltage differential probe is adjusted by a standard voltage source, and the encoder speed measurement module is verified by a constant speed wheel. The software module of the real-time feature extraction device adopts modular design, the signal preprocessing unit and the feature fusion unit are run as independent threads, the thread priority is set to the highest to avoid data loss; the feature vector database uses a high-speed solid state disk for storage, the index is constructed according to the timestamp and the process parameters, and fast query is supported. Key sensors such as current Hall sensors and voltage differential probes are configured with dual backup, the data acquisition device has a built-in self-diagnosis function, and automatically switches to the backup channel when an exception occurs; the feature fusion algorithm of the real-time feature extraction device introduces consistency checking, if the timestamp deviation of the data source exceeds the tolerance, the realignment process is triggered. The influence of the welding environment is suppressed by shielding and filtering, for example, the spectral acquisition unit is equipped with a light shield to reduce environmental light interference, and the infrared thermal imaging unit uses a blowing device to keep the lens clean; the dimension selection of the feature vector is based on process knowledge, the initial setting contains 10 core features, and additional features can be dynamically added according to the welding material type, such as the intensity of the chromium element feature line when welding stainless steel.

[0033] Example 2: refer to Figure 3 The feature database updating unit receives the multi-dimensional feature vector generated from the feature fusion unit, which contains dynamic parameters such as spectral intensity mean, temperature gradient, and current variance. The updating unit extracts the feature value trend from the continuous sampling period, for example, calculates the moving average and standard deviation of the feature vector in every five sampling periods to form a trend sequence. The real-time feature library of the welding process establishes a classified index according to material type and thickness specification. The material types include common welding materials such as low carbon steel, stainless steel, and aluminum alloy, and the thickness specifications are recorded in segments from 1mm thin plate to 20mm thick plate. Each database entry stores a feature vector array, a high-precision timestamp, a corresponding welding process parameter group (including current set value, voltage set value, and wire feed speed set value), and environmental humidity data. The humidity data is provided in real time by a digital humidity sensor installed in the welding cabin. The data updating mechanism performs a write operation once every welding cycle (about 100 milliseconds), and automatically archives historical data that exceeds the retention period.

[0034] The defect probability calculation unit of the dynamic analysis device retrieves the feature vector of the current welding cycle from the real-time feature library, matches the material type and thickness specification, calculates the deviation value of the feature vector from the historical qualified sample feature vector, and the historical qualified sample library is established through the initial training stage. During training, multiple defect-free test blocks are welded under standard process parameters, the feature vectors are extracted and marked as qualified samples. The deviation calculation uses the Mahalanobis distance algorithm, which considers the covariance structure of the feature vector, effectively eliminating the dimensional influence and correlation interference between features. During calculation, the reference sample set of the same specification is first extracted from the qualified sample library, and the Mahalanobis distance value between the current feature vector and the sample set center is calculated. The larger the value, the farther the deviation from the normal welding state. The quality level determination unit presets two deviation threshold values, the first deviation threshold value is set based on the statistical process control principle, and is located at twice the standard deviation of the qualified sample distribution mean, and the second deviation threshold value is located at three times the standard deviation. The determination logic is: when the real-time calculated deviation value is less than the first threshold value, output a high-level signal as a qualified determination signal; when the deviation value is between the first threshold value and the second threshold value, output a medium-level signal as a suspicious determination signal; when the deviation value exceeds the second threshold value, output a low-level signal as a defect determination signal. All determination signals are attached with time stamps and welding cycle numbers.

[0035] The hardware deployment of the feature database updating unit is on an industrial server, and the database adopts a time series database structure. Each record contains a feature vector (a 20-dimensional floating-point number array), a millisecond-level timestamp, a material type code (such as CS304 representing 304 stainless steel), a thickness value (in millimeters), an array of process parameters (floating-point number values of current, voltage, and wire feeding speed), and an environmental humidity value. The database index is established according to the combination of material type and thickness, supports millisecond-level query response, and the update operation ensures data consistency through a transaction mechanism. At the same time, a background task periodically cleans up archived data three months ago. The historical qualified sample library is initialized through an offline training process. During training, 50 test blocks are continuously welded using standard welding parameters. During the welding process of each test block, 1000 feature vectors are extracted. After being confirmed to be defect-free by non-destructive testing, they are stored in the sample library. The sample library is continuously optimized, and new samples are supplemented after each batch of new qualified weldments is completed. The defect probability calculation unit is integrated on a high-performance computing card. Upon receiving a new feature vector, it immediately triggers calculation. It retrieves all samples of the same material-thickness combination from the qualified sample library, calculates the sample mean and covariance matrix, and introduces regularization processing in the Mahalanobis distance formula to prevent matrix singularity. The deviation value output is a floating-point number, with a range of 0 to positive infinity. In actual normal working conditions, it is mostly distributed between 0 and 5. The threshold management of the quality level determination unit can be dynamically configured. The first deviation threshold is set to 2.0 (normalized unit) by default, and the second threshold is set to 3.0. The threshold can be fine-tuned through a human-machine interface to adapt to different production line requirements. The determination signal is transmitted through a digital output module. The qualified signal corresponds to a 24V high level, the suspicious signal corresponds to a 12V medium level, and the defect signal corresponds to a 0V low level. The signal is transmitted to the output control device.

[0036] The real-time feature library's data structure supports fast time-series queries. Each feature vector entry is associated with metadata including welding torch number, operator ID, ambient temperature, and other auxiliary information. The database employs a partitioned storage strategy, physically separating data for different material specifications to improve query efficiency. Maintenance of the qualified sample library includes sample quality monitoring, regular checks of sample age distribution, and automatic removal of old samples exceeding a certain timeframe to prevent misjudgments due to process drift. During deviation calculation, feature vectors are pre-standardized using the mean and standard deviation of the qualified sample set to ensure consistency across different feature dimensions. The dynamic analysis device's operating cycle is strictly synchronized with the welding cycle, executing a complete analysis process every 100 milliseconds. The defect probability calculation unit is equipped with a caching mechanism, using the previous valid result for a downgraded run when a qualified sample library query fails. The quality level judgment unit's output signal undergoes anti-jitter processing, confirming a defect signal only after three consecutive cycles of defect identification to avoid false alarms due to momentary interference. A self-check process is executed upon startup to verify the feature database connection status and whether the number of qualified sample library samples meets the minimum requirements (e.g., 1000 samples per specification). If the self-check fails, the analysis process is prohibited from starting. The selection of samples for the qualified sample library incorporates diversity constraints to ensure coverage of different equipment states and environmental conditions. The deviation calculation algorithm has been optimized, employing an incremental calculation method to reduce repetitive matrix operations. Adjustments to the judgment threshold are recorded in the audit log, and any modifications require authorized operator two-factor authentication.

[0037] Embodiment 3: The defect locating unit starts immediately upon receiving the defect determination signal from the quality grade determination unit, which is a low-level digital pulse. The defect locating unit synchronously calls the temperature field distribution data cache collected by the infrared thermal imaging unit in the recent period of time, and reads the spectral radiation intensity data of the corresponding period of time by the high-resolution spectral acquisition unit. The temperature field distribution data is stored in the form of image sequences, and each frame of image contains a temperature value matrix of 512x640 pixels. The spectral data is the light intensity curve of each pixel in the waveband of 300-1100 nanometers. The defect locating unit realizes defect positioning by comparing the spatial and temporal coincidence degree of the abnormal area of temperature gradient distribution and the mutation area of spectral characteristics. The identification of the abnormal area of temperature gradient distribution first calculates the temperature gradient field of each frame of thermal image. The gradient amplitude of each pixel point is obtained by two-dimensional convolution operation using the Sobel operator. A gradient amplitude threshold is set, and the continuous pixel area exceeding the threshold is marked as an abnormal area. The detection of the mutation area of spectral characteristics focuses on the intensity change of the emission spectrum line of a specific metal element (such as the characteristic peak of iron element at 538 nanometers). The first-order difference of spectral intensity of each pixel in consecutive frames is calculated, and when the difference value exceeds 3 times the standard deviation of the normal fluctuation range, it is marked as a mutation point. The spatial and temporal coincidence degree analysis maps the above two types of abnormal areas to the same coordinate system. The positions of the temperature field image and the spectral pixel are registered by the transformation matrix marked in advance. The area proportion of the overlapping area is calculated, and when the coincidence degree exceeds the preset threshold (such as 80%), it is determined that the position is a defect point, and its plane coordinates relative to the welding starting point are output.

[0038] The implementation of the defect positioning unit is based on a graphics processor platform, the temperature gradient calculation utilizes parallel threads to process all pixels simultaneously, the gradient amplitude threshold is dynamically adjusted according to the material type, for example, set to 15°C / mm for low carbon steel welding. The spectral mutation detection uses a sliding window difference method, the window size is 5 sampling points, and the mutation threshold is updated adaptively every 5 minutes. The coordinate registration establishes the transformation relationship between the thermal image pixel coordinates and the spectral pixel coordinates in advance through the hand-eye calibration method, and the coincidence degree calculation uses a pixel-level traversal algorithm. The parameter adjustment unit receives the suspicious judgment signal (middle-level 12V signal) output by the quality grade judgment unit, which triggers the unit to immediately query the optimal process parameter record of the same material specification in the real-time feature library, the query conditions including material code, thickness value and environmental humidity range, and obtains the historical optimal parameter group (such as current 210A, voltage 26.5V, wire feeding speed 6.2m / min) after comparing with the current actual parameters, generates the welding current correction amount, arc voltage compensation amount and wire feeding speed adjustment amount, the correction amount is calculated using the incremental PID algorithm, the proportional coefficient is set according to the welding type, the integral time constant is set to 10 control periods, and the differential action is limited to low-speed change scenarios. The alarm triggering unit starts when receiving the defect judgment signal, and reads the defect position coordinates (X, Y millimeter value) provided by the defect positioning unit, generates an emergency stop command conforming to the industrial bus protocol, the command frame contains the defect type code (such as C for crack, P for porosity), position coordinates, timestamp and checksum, and activates the sound and light output of the alarm at the same time.

[0039] In the temperature gradient calculation of the defect positioning unit, the gradient amplitude G of each pixel point is determined by the partial derivative of the point in different directions: Where: T represents the temperature value of a point in the temperature field, x and y represent the horizontal and vertical coordinates in the image coordinate system respectively. The calculation is performed in parallel on the GPU, and each thread is responsible for the gradient calculation of a pixel point. The spectral mutation detection calculates the spectral intensity change ΔI between adjacent frames for each pixel, and marks it as a mutation when |ΔI|>3σ (σ is the standard deviation of the intensity of the last 100 frames of the pixel).

[0040] The synchronization mechanism of the defect positioning unit and the data acquisition adopts hardware triggering. The defect determination signal is directly connected to the trigger pin of the FPGA, ensuring that the positioning delay is less than 2 milliseconds. The temperature field data buffer depth is the last 5 seconds of data (150 frames), and the spectrum data buffer is the last 5000 spectrum curves. The output interface of the parameter adjustment unit is an analog output module. The current correction amount is converted into a 4-20mA signal and sent to the welding power supply. The voltage compensation amount is converted into a 0-10V signal and sent to the arc controller. The wire feeding speed adjustment amount is output to the servo driver through a pulse signal. The alarm triggering unit has multiple interlocking functions. The emergency stop command is sent to the welding power supply, the robot controller and the wire feeding mechanism at the same time, ensuring that the system enters a safe state immediately. The defect positioning unit uses a double buffer structure to ensure data integrity. The parameter adjustment algorithm adds an anti-saturation process to prevent integral windup. The alarm triggering unit is configured with a watchdog timer to monitor the communication state. All calculation modules have redundancy design. When the main processor fails, the standby processor takes over immediately. The coordinate positioning result is subjected to smoothing filter processing. The average value of the confirmed position of the last three frames is taken as the final output, avoiding instantaneous jumping. The output of the parameter adjustment amount has a gradual change characteristic. The adjustment amplitude is not more than 5% of the set value each time, preventing process fluctuations.

[0041] In embodiment 4, the historical data comparison device extracts the feature vector sequence of the last several welding cycles from the real-time feature library regularly. Usually, 100 continuous cycles corresponding to about 10 seconds of welding process are set. The extracted sequence includes multi-dimensional feature vectors of each cycle, such as spectral intensity mean, temperature gradient, current variance, etc. These vectors are arranged in time sequence to form time series data. The historical data comparison device calculates the dynamic time warping distance between the sequence and the feature vector sequence stored under the standard process parameters. The standard sequence is derived from the reference sequence of the same material specification under the optimal process condition in the historical qualified sample library. The dynamic time warping distance is calculated by using the dynamic programming algorithm to align the length difference of the two sequences. The total distance value is obtained by accumulating the Euclidean distance between the corresponding points. This distance value reflects the shape difference degree between the current welding process and the standard process. When the calculated distance exceeds the preset process stability threshold, the historical data comparison device sends a process parameter optimization request to the parameter adjustment unit. The request includes the direction and size of the distance deviation. For example, it is suggested to increase the current or adjust the voltage. The threshold is set based on long-term running data statistics. Usually, the 95th percentile of the distance distribution of the standard sequence is taken.

[0042] The report generation unit receives the judgment signal (qualified, suspicious or defective) from the quality grade determination unit, the defect position coordinate data provided by the defect positioning unit and the dynamic time warping distance data calculated by the historical data comparison device, which are real-time data flowing into the data buffer of the report generation unit. The report generation unit generates a detection report at a predetermined time interval (such as every hour) or event trigger (such as defect occurrence). The report content includes welding quality statistical indicators such as qualified rate, defect number, average deviation, defect distribution map generated by superimposing defect position coordinates on the welding path diagram, and process stability index calculated by the reciprocal of dynamic time warping distance and normalized to 0-100 range. The report format supports XML and PDF. XML format is used for data exchange between systems, and PDF format is used for manual viewing. The report file is stored in a network shared directory and automatically sent to the production management system. The historical data comparison device is implemented based on an edge computing node. The node is configured with a multi-core processor and high-speed memory. Every 10 seconds, the node queries the feature vector sequence of the last 100 welding cycles from the real-time feature library. The filtering conditions include material type, thickness specification and equipment number to ensure sequence comparability. The feature vector sequence under standard process parameters is preloaded from the historical database. The sequence length is fixed at 100 cycles. The dynamic time warping distance is calculated using the open source DTW library. The algorithm parameters such as step constraint are set to symmetric P0.1 specification. The calculated distance value is standardized by dividing the sequence length to eliminate the influence of cycle number. The process stability threshold is dynamically adjusted according to the material type. For example, the welding of low carbon steel is set to 5.0, and the welding of stainless steel is set to 6.0. The threshold value management can be modified through the configuration interface. When the distance exceeds the threshold, the optimization request is sent in the form of a digital signal. The signal contains deviation code and recommended adjustment amount.

[0043] The report generation unit is deployed on a server and adopts modular design. The data receiving module listens to the judgment signal and coordinate data in the message queue. The dynamic time warping distance data is updated every 10 seconds. The report generation trigger can be time-driven or event-driven. The time-driven mode is executed at the hour point every hour, and the event-driven mode is started immediately when the defect judgment signal is issued. The welding quality statistical indicators in the report content calculate the proportion of qualified cycles in the total cycle number in the calculation period. The defect number accumulates the occurrence frequency of various defect types. The defect distribution map is drawn using vector graphics. The welding path is a straight line or curve generated according to the actual trajectory. The defect position is marked with a red marker point. The process stability index calculation formula is max(0, 100-10*DTW distance), ensure the index in reasonable range, perform data integrity check before report output, missing data filled by interpolation. Combined with a specific example, assume a stainless steel welding process, material type code SS304, thickness 3mm, welding speed 1m / min, the historical data comparison device extracts the feature vector sequence of the last 100 welding cycles at time point T1, the sequence data is shown in the following table, refer to Table 1, which shows the simplified values of the feature vectors of some cycles and the calculated DTW distance, the standard sequence comes from the optimal process sample of SS304-3mm in the historical library, the DTW distance is calculated by aligning the current sequence and the standard sequence to get the cumulative distance, the process stability threshold is set to 6.0.

[0044] Table 1: Dynamic time warping distance calculation table

[0045] In this example, the DTW distance calculation result is 7.2, which exceeds the threshold 6.0, the historical data comparison device sends an optimization request to the parameter adjustment unit, the request signal contains the code "INC_CURRENT" to increase the current from the current 210A to 215A, and the report generation unit triggers the report generation at T1, collects 95 qualified signals, 3 suspicious signals and 2 defect signals in this period, the defect position coordinates are (102.5, 35.2) and (103.8, 36.1) respectively, the dynamic time warping distance is 7.2, the process stability index is calculated as 28, and the report generation unit integrates these data to generate a PDF report, the report header contains the time range, device ID and operator information, the main part lists the statistical table and defect distribution chart.

[0046] The data extraction process of the historical data comparison device uses a SQL query real-time feature library, the query statement specifies the time range and label conditions, the sequence data is loaded into the memory and preprocessed including removing outliers and smoothing filtering, the DTW distance calculation uses global constraint to limit the slope of the alignment path to improve calculation efficiency, the threshold comparison module performs a check every 10 seconds, the optimization request is sent to the parameter adjustment unit through the industrial Ethernet protocol, the request message includes a timestamp, distance value, threshold value and recommended action. The data collection of the report generation unit uses a subscription mode, the decision signal and coordinate data are received asynchronously through the message middleware, the dynamic time warping distance data is pushed by the historical data comparison device, the report generation uses a sliding window algorithm to avoid memory overflow, the defect distribution map is generated by calling the graphics library API to map the coordinates to the welding drawing, and the process stability index is displayed in the report summary in real time. The output management of the report generation unit includes file naming rules and storage paths, the PDF report is named in the format of "Welding_Report_YYYYMMDD_HHMMSS.pdf", stored in the specified directory and uploaded to the cloud platform, the XML report contains structured data for easy parsing by other systems, and the report content also includes trend charts such as the DTW distance curve with time, which helps users monitor process drift. The system log records the events and data source status of each report generation, which is convenient for auditing and troubleshooting.

[0047] In embodiment 5, the calibration device is integrated as an independent hardware module in the system cabinet, and communicates with the welding area monitoring device and the multi-source data acquisition device through an industrial bus. The core function of the calibration device is to periodically trigger the standard test block welding process to complete system self-calibration. The calibration period can be configured as a fixed time interval (e.g., every 8 hours) or based on the cumulative welding meters (e.g., every 500 meters of weld). The standard test block has the same specifications as the current production line processing material, including the same material type, thickness, size and surface treatment state. The test block is installed on a special fixture in the calibration station and is automatically taken and positioned to the welding position by the mechanical arm. When the calibration period is triggered, the calibration device first sends a command to the welding control system to set the welding parameters to the preset standard values (e.g., current 200A, voltage 25V, wire feed speed 5m / min), and then starts the standard welding process. During this period, the high-resolution spectral acquisition unit and the infrared thermal imaging unit of the welding area monitoring device work synchronously to capture the spectral radiation intensity data and temperature field distribution data of the standard test block welding area at the normal sampling frequency. The multi-source data acquisition device simultaneously records the real-time data of current, voltage and wire feed speed. These data are transmitted to the real-time feature extraction device through the data interface of the calibration device.

[0048] The standard welding process lasts about 30 seconds, completing a calibration weld with a length of about 100 mm. All data collected during the welding process is labeled as "standard sample data" and temporarily stored in the buffer memory. The data processing unit of the calibration device performs quality inspection on these raw data, including signal integrity, noise level, and data synchronization. The qualified data packets are sent to the real-time feature library update module. The standard sample data update of the real-time feature library uses a version control mechanism. After feature extraction, the newly collected standard sample data is compared with the existing standard samples in the library. If the statistical distribution of the feature vector changes significantly (e.g., the mean value shifts by more than 5%), the new sample replaces the old sample. Otherwise, the original standard sample is retained, and the calibration data is recorded as a historical version. Upon completion of calibration, the calibration device sends a calibration completion marker in digital signal form to the report generation unit. This marker data packet contains the calibration timestamp, standard test block number, welding parameter standard value, data quality score, and check code. The report generation unit receives this data as metadata and embeds it in the header information of the detection report.

[0049] The specific implementation of the calibration device adopts a modular design. The mechanical structure includes a test block storage bin, a six-axis mechanical arm, and a visual positioning system. The test block storage bin can accommodate 20 standard test blocks of different specifications. The mechanical arm is equipped with a force control gripper to ensure stable grabbing. The visual positioning system precisely positions the test block on the welding platform through laser ranging and CCD imaging. The calibration trigger logic is implemented by a programmable logic controller, supporting various trigger condition combinations, such as "time arrival + device idle" or "meter accumulation + shift switching." After the calibration instruction is issued, the welding device automatically switches to calibration mode, during which normal production is suspended, and the focal length and exposure parameters of the welding area monitoring device are automatically adjusted to the preset calibration mode settings. During data collection, the calibration device monitors the working status of each sensor. If an abnormality is found (e.g., spectral signal saturation or temperature data drift), the calibration process is interrupted, and a fault code is recorded. The management of standard test blocks uses a two-dimensional code identification system. Each test block has a unique code engraved on its surface, which contains material, thickness, heat treatment status, and expiration date information. Before grabbing the test block, the mechanical arm scans the two-dimensional code to verify whether the test block specifications match the current production line requirements. When the number of test block uses reaches the upper limit (e.g., 50 times), the test block is automatically discarded and a new one is prompted to replace it. The standard sample data update algorithm of the real-time feature library uses a gradual update strategy. New sample data does not immediately completely replace old data, but gradually adjusts the statistical parameters of the feature vector according to the time weighting principle, avoiding sudden changes in standard samples due to single calibration abnormalities. The transmission of the calibration completion marker uses a redundant communication protocol. The main channel is PROFIBUS-DP, and the backup channel is Ethernet TCP, ensuring that the marker data reliably reaches the report generation unit.

[0050] The data acquisition accuracy control measures in the calibration process include environmental temperature compensation (welding cabin constant temperature control at 23±2℃), sensor preheating (power on for 30 minutes in advance to stabilize), and electromagnetic shielding (signal line uses double shielding), the process parameter fluctuation range of the standard test block welding is controlled within ±1%, to ensure the repeatability of the standard sample data. The statistical significance test when updating the feature library uses the hypothesis test method, calculates the mean difference confidence interval of the new sample and the old sample feature vector, and judges as significant change when the 95% confidence interval does not contain zero. The analysis of the calibration mark by the report generation unit includes timestamp conversion, data quality score analysis and check code verification, only the calibration record that passes the integrity check will be displayed in the report. The self-monitoring function of the calibration device includes mechanical arm motion trajectory calibration, regular automatic adjustment of the focal length of the vision system, and sensor zero drift compensation, these maintenance operations are automatically executed within the calibration interval. The weld quality after welding of the standard test block is verified by offline flaw detection sampling inspection, and the sampling inspection result is fed back to the calibration system for evaluating the effectiveness of the standard sample data. The timing control of the whole calibration process is accurate to the millisecond level, ensuring the strict synchronization of data acquisition and welding process, after the calibration is completed, the system automatically switches back to the production mode, and generates the calibration event log record of the whole process parameters. The calibration mark received by the report generation unit will trigger the report version update, add "calibrated" watermark and calibration time identification on the report home page, at the same time, the calibration data is archived to the independent database for long-term trend analysis.

[0051] While embodiments of the present application have been shown and described with reference to particular embodiments thereof, it will be understood by those skilled in the art that various changes in form and details can be made therein without departing from the spirit and scope of the application. The scope of the application is defined by the appended claims and their equivalents.

Claims

1. A welding quality oriented online detection system characterized by: The system comprises a welding area monitoring device, a multi-source data acquisition device, a real-time feature extraction device, a dynamic analysis device and an output control device; The welding area monitoring device is arranged around the working area of the welding equipment and is used for capturing optical radiation signals and heat distribution signals in the welding process; the multi-source data acquisition device is connected to the welding area monitoring device and synchronously acquires welding current fluctuation data, arc voltage change data and wire feeding speed data; The real-time feature extraction device receives the current fluctuation data, voltage change data and wire feeding speed data output by the multi-source data acquisition device, and receives the optical radiation signals and heat distribution signals output by the welding area monitoring device, and constructs a dynamic feature set of the welding process; The dynamic analysis device generates welding quality evaluation parameters according to the dynamic feature set provided by the real-time feature extraction device; and the output control device generates welding process parameter adjustment instructions based on the welding quality evaluation parameters provided by the dynamic analysis device; The real-time feature extraction device comprises a signal preprocessing unit and a feature fusion unit; The signal preprocessing unit performs sliding window normalization processing on the spectral radiation intensity data output by the high-resolution spectral acquisition unit, and performs spatial interpolation compensation processing on the temperature field distribution data output by the infrared thermal imaging unit; The feature fusion unit aligns the processed spectral radiation intensity data and temperature field distribution data in time sequence, simultaneously receives current fluctuation data collected by a current Hall sensor, voltage change data collected by a voltage differential probe and wire feeding speed data collected by an encoder speed measurement module, and generates a multi-dimensional feature vector containing time sequence correlation; The real-time feature extraction device further comprises a feature database updating unit; the feature database updating unit receives the multi-dimensional feature vector generated by the feature fusion unit, extracts the feature value change trend of a plurality of continuous sampling periods, and constructs a real-time feature library of the welding process; the real-time feature library is classified and stored according to welding material types and thickness specifications, and each storage entry contains a feature vector timestamp, corresponding welding process parameters and environmental humidity data; The dynamic analysis device comprises a defect probability calculation unit and a quality grade determination unit; The defect probability calculation unit extracts the feature vector of the current welding period from the real-time feature library and calculates the deviation degree value thereof from historical qualified sample feature vectors; the quality grade determination unit presets a first deviation degree threshold and a second deviation degree threshold, outputs a qualified determination signal when the deviation degree value is less than the first deviation degree threshold, outputs a suspicious determination signal when the deviation degree value is between the first deviation degree threshold and the second deviation degree threshold, and outputs a defect determination signal when the deviation degree value is greater than the second deviation degree threshold. The dynamic analysis device further comprises a defect positioning unit; when receiving the defect determination signal, the defect positioning unit synchronously calls the temperature field distribution data collected by the infrared thermal imaging unit and the spectral radiation intensity data collected by the high-resolution spectral collection unit, determines the specific position coordinates of the welding defect by comparing the spatial coincidence degree of the temperature gradient distribution abnormal area and the spectral feature mutation area; in the temperature gradient calculation of the defect positioning unit, the gradient amplitude of each pixel point is determined by the partial derivatives of the point in different directions: Where: T represents the temperature value of a certain point in the temperature field, x and y represent the horizontal and vertical coordinates in the image coordinate system respectively; the calculation is performed in parallel on the GPU, and each thread is responsible for the gradient calculation of a pixel point; the spectral mutation detection calculates the change amount of spectral intensity between adjacent frames for each pixel When , the time is marked as a mutation, and σ is the standard deviation of the intensity of the pixel in the last 100 frames.

2. A welding quality oriented online detection system according to claim 1, characterized in that: The welding area monitoring device comprises a high-resolution spectral acquisition unit and an infrared thermal imaging unit; the high-resolution spectral acquisition unit is installed at a specific distance behind the welding gun to capture spectral radiation intensity data of the welding pool area at a fixed sampling frequency; the infrared thermal imaging unit is arranged directly above the welding area to obtain temperature field distribution data of the welding heat affected zone in an asynchronous sampling manner; the multi-source data acquisition device comprises a current Hall sensor, a voltage differential probe and an encoder speed measurement module; the current Hall sensor is clamped at the output end of the welding power source, the voltage differential probe is connected in parallel across the welding gun, and the encoder speed measurement module is installed on the transmission shaft of the wire feeding mechanism.

3. A welding quality oriented online detection system as claimed in claim 1, wherein: The output control device comprises a parameter adjustment unit and an alarm triggering unit; the parameter adjustment unit receives the suspicious determination signal output by the quality grade determination unit, generates welding current correction amount, arc voltage compensation amount and wire feeding speed adjustment amount according to the optimal process parameter data of the same material specification in the real-time feature library; the alarm triggering unit receives the defect determination signal output by the quality grade determination unit, and generates an emergency stop instruction containing the defect type code and the position information in combination with the specific position coordinates provided by the defect positioning unit.

4. A welding quality oriented online detection system as claimed in claim 3, wherein: The system further comprises a historical data comparison device; the historical data comparison device extracts the feature vector sequence of the last several welding cycles from the real-time feature library, calculates the dynamic time warping distance between the feature vector sequence and the feature vector sequence under the standard process parameters; when the distance exceeds the preset process stability threshold, the parameter adjustment unit sends a process parameter optimization request.

5. A welding quality oriented online detection system as claimed in claim 4, wherein: The output control device further comprises a report generation unit; the report generation unit receives the determination signal output by the quality grade determination unit, the position coordinate data provided by the defect positioning unit and the dynamic time warping distance data calculated by the historical data comparison device, and generates a detection report containing welding quality statistical indicators, defect distribution map and process stability index.

6. A welding quality oriented online detection system as claimed in claim 5, wherein: The system further comprises a calibration device; the calibration device is connected to the welding area monitoring device and the multi-source data acquisition device, triggers the standard test block welding process periodically, acquires feature vector data under standard welding conditions, updates the standard sample data in the real-time feature library, and sends a calibration completion mark to the report generation unit.

Images