A laser welding defect intelligent judgment method based on multi-source photoelectric signal fusion

Abstract

The present application relates to the technical field of welding quality evaluation, and particularly relates to a laser welding defect intelligent judgment method based on multi-source photoelectric signal fusion, comprising the following steps: determining corresponding signal effective intervals based on synchronous acquisition of multi-source photoelectric signals in a laser welding process, and aligning the signal effective intervals to a same welding process axis, wherein the multi-source photoelectric signals comprise laser reflection signals, molten pool thermal radiation signals and ion light signals; dividing the signal effective intervals based on continuous sampling position intervals on the welding process axis to obtain a plurality of sequential welding sections; determining corresponding deviation forms of the multi-source photoelectric signals based on change direction turns of the multi-source photoelectric signals in the plurality of sequential welding sections, and determining a participating source combination of a corresponding sequential welding section based on the deviation forms. The present application improves the accuracy of laser welding defect judgment.

Description

Technical Field

[0001] This invention relates to the field of welding quality assessment technology, and in particular to an intelligent method for determining laser welding defects based on the fusion of multi-source photoelectric signals. Background Technology

[0002] Laser welding, with its high energy density, small heat-affected zone, and high welding speed, is widely used in the connection of power battery tabs, busbars, vehicle bodies, and precision structural components. Welding quality directly determines the mechanical properties and service safety of these components. Therefore, real-time and accurate assessment of welding defects has become a core aspect of quality control in the laser welding process. The light radiation and reflection generated by the interaction between the laser and the workpiece during welding can be collected in real time by photoelectric sensing units. The intensity and temporal sequence of the photoelectric signals are observable quantities reflecting transient instability during the welding process and vary overall with operating conditions. Multi-source photoelectric signals contain welding status and defect information, providing crucial data for online monitoring of welding quality.

[0003] Current laser welding photoelectric monitoring typically processes the acquired signals to extract features such as mean and variance, or compares them with quality assessment templates or reference curves formed during normal welding processes to calculate differences. These differences are then compared with preset thresholds, envelopes, or control limits obtained from historical sample statistics, or input into a trained machine learning model to output a conclusion of pass or defect. However, these methods suffer from shallow fusion of multi-source heterogeneous signals, difficulty in feature alignment, and insufficient model generalization ability, making it difficult to adapt to defect determination needs under different materials, process parameters, and operating conditions. Furthermore, existing technologies remain at the level of determining whether the entire weld or the entire workpiece is passable. Therefore, there is an urgent need for an intelligent laser welding defect determination method based on multi-source photoelectric signal fusion, capable of achieving deep collaborative characterization and adaptive fusion of multi-source signals, thereby improving the accuracy and real-time performance of laser welding defect determination.

[0004] Chinese Patent Publication No. CN119880913A discloses an online diagnostic method and system for bus laser welding defects based on photoelectric signals. The method includes: real-time acquisition of photoelectric signals generated during bus laser welding using a high-speed sensor; noise reduction, smoothing, and normalization of the signals to eliminate environmental interference and obtain stable and consistent signal data; extraction of key feature parameters based on the processed photoelectric signals; comparison of the features with standard reference values ​​to define specific feature indicators; and construction of a quality diagnostic model using a random forest algorithm to classify the welding state as normal or defective; input of the real-time acquired photoelectric signals into the quality diagnostic model for welding quality detection; and immediate transmission of adjustment signals to the welding equipment to optimize welding parameters and achieve real-time quality optimization and control when a defective weld is detected. However, this method and system suffer from several problems: the feature extraction relies on manually set key feature parameters, making it difficult to adapt to signal baseline drift under different welding conditions; and the model is prone to category judgment bias when samples are scarce, leading to insufficient accuracy in weld defect determination. Summary of the Invention

[0005] To address this, the present invention provides an intelligent method for judging laser welding defects based on multi-source optoelectronic signal fusion, which overcomes the problems in the prior art where feature extraction relies on manually set key feature parameters, making it difficult to adapt to signal baseline drift under different welding conditions, and the model is prone to category judgment bias when samples are scarce, resulting in insufficient accuracy in judging welding defects.

[0006] To achieve the above objectives, this invention provides an intelligent method for determining laser welding defects based on multi-source optoelectronic signal fusion, comprising: Based on the synchronous acquisition of multi-source photoelectric signals during the laser welding process, the corresponding effective signal range is determined, and the effective signal range is aligned to the same welding process axis. The multi-source photoelectric signals include laser reflection signals, molten pool thermal radiation signals, and ion light signals. The effective range of the signal is divided into several sequential welding segments based on the continuous sampling position range on the welding process axis. Based on the change direction of the multi-source photoelectric signals within the several sequential welding segments, the deviation pattern corresponding to the multi-source photoelectric signals is determined, and the participating source combination of the corresponding sequential welding segment is determined based on the deviation pattern. When a deviation signal in the sequential welding segment reaches the corresponding output upper limit, it is determined that the deviation signal that has reached the corresponding output upper limit cannot be judged as a deviation state in the current sequential welding segment, and the corresponding acquisition process parameters are adjusted. The succession relationship between the deviation signals is determined based on the order of their starting positions on the welding process axis, and the directional relationship between the deviation signals is determined based on their changing directions during the corresponding deviation processes. Based on the deviation pattern, the participating source combination, the receiving relationship, and the directional relationship, a weld segment judgment combination is constructed. Based on the weld segment judgment combination and the preset combination conditions, welding defect candidates are determined. When the weld segment judgment combination does not meet the preset combination conditions, the corresponding sequential weld segment is marked as an unclassified abnormal weld segment. Several sequential weld segments that are adjacent on the welding process axis and have the same weld segment judgment result are merged to form a corresponding defect interval, and the judgment result of the defect interval is determined to review the defect interval.

[0007] Further, determining the effective range of the signal includes: The signal level of the multi-source photoelectric signal when the laser is not applied to the welding process is used as the background level; The welding initiation event is determined based on the deviation state of the multi-source photoelectric signal and the corresponding background level, and the sampling time corresponding to the welding initiation event is taken as the starting point of the effective interval of the signal. The welding end event is determined based on the moment when the deviation of the multi-source photoelectric signal falls back to the corresponding background level, and the sampling time corresponding to the welding end event is taken as the end point of the effective range of the signal.

[0008] Furthermore, the division of the sequential welding segments includes: Based on the sampling position corresponding to the welding initiation event, the effective signal range is divided into several sequential welding segments according to the continuous equal sampling position intervals. The laser reflection signal, molten pool thermal radiation signal and ion light signal in the equal sampling position intervals are respectively composed into the synchronous sampling sequence of the corresponding sequential welding segments.

[0009] Furthermore, the deviation patterns include pulse type, intermittent type, step hold type, continuous type, and no deviation pattern.

[0010] Further, determining the acceptance relationship and the directional relationship includes: The multi-source photoelectric signals that deviate are used as participating sources, and the multi-source photoelectric signals that do not deviate are used as non-participating sources; The sequential order of several participating sources corresponding to several deviations from the starting position on the welding process axis is determined as the succession relationship between several participating sources, and participating sources with the same deviation from the starting position are determined as simultaneous; Participating sources whose change direction is consistent in the corresponding deviation process are identified as having a same-direction relationship, and participating sources whose change direction is opposite are identified as having a reverse relationship.

[0011] Further, based on the weld segment judgment combination and preset combination conditions, candidate welding defects are determined, including: When the laser reflection signal, the molten pool thermal radiation signal, and the ion light signal are all in a state of no deviation, it is determined that no defects were detected in the corresponding sequential weld segment. When only the laser reflection signal deviates continuously, the judgment result corresponding to the weld segment is determined as a weld misalignment candidate; When the laser reflection signal, the molten pool thermal radiation signal, and the ion light signal all exhibit a continuous deviation, and the direction of change of the laser reflection signal is opposite to that of the molten pool thermal radiation signal and the ion light signal, the determination result corresponding to the weld segment is determined to be a candidate for incomplete penetration. When the molten pool thermal radiation signal deviates from the ion light signal in a pulse-like or step-holding manner, and the starting position of the deviation of the molten pool thermal radiation signal precedes that of the ion light signal, the determination result corresponding to the weld segment is determined to be a burn-through candidate.

[0012] Furthermore, when the deviation signal is in the undeterminable state, an acquisition parameter adjustment instruction is generated according to the acquisition anomaly type corresponding to the undeterminable state, and the acquisition process of the deviation signal is adjusted according to the acquisition parameter adjustment instruction. The determination of the sequential welding segment is performed based on the multi-source photoelectric signal of the welding process re-acquired after adjustment.

[0013] Furthermore, the deviation patterns and judgment results of sequential weld segments adjacent to the defect interval at both ends are obtained to verify the process continuity of the defect interval; When no deviation is found in the sequential weld segments adjacent to both ends of the defect interval, the defect interval is marked as an interval to be checked. When at least one adjacent sequential weld segment in the defect interval deviates, the defect interval is determined to meet the process continuity condition and the judgment result is output.

[0014] Furthermore, the review of the defective region includes: The weld segment determination combination is reconstructed based on the multi-source photoelectric signals corresponding to the sequential weld segments contained in the defect interval; When the weld segment judgment result corresponding to the reconstructed weld segment judgment combination is consistent with the original weld segment judgment result of the defect interval, the verification confirmation information of the defect interval is output. When the weld segment judgment result corresponding to the reconstructed weld segment judgment combination is inconsistent with the original weld segment judgment result of the defect interval, the defect interval is marked as an interval to be checked.

[0015] Furthermore, when the defect interval is marked as an interval to be checked, or when the sequential weld segment is marked as an unclassified abnormal weld segment, the corresponding defect interval or the corresponding sequential weld segment is recorded in the record to be checked.

[0016] Compared with existing technologies, the beneficial effects of this invention are as follows: The method of this invention identifies the sustained changes of multi-source photoelectric signals at the background level, determines the effective signal range, and divides the weld segment according to the welding process. It converts continuously acquired multi-source heterogeneous signals into time-series analysis units corresponding to the actual welding process, achieving the effects of suppressing transient noise and invalid fluctuation interference, and clearly defining the welding process boundaries. This solves the problems of difficulty in distinguishing between the welding process and non-process disturbances in continuous signals, as well as the inaccuracy of the subsequent defect analysis time range. Furthermore, by constructing weld segment judgment combinations based on deviation morphology, participating source combinations, connection relationships, and directional relationships, and combining these with preset combination conditions to determine candidate defect categories, the method effectively divides different photoelectric signals into different categories. By associating and constraining the differentiated responses of welding anomalies, the system effectively distinguishes different physical manifestations of defects using the temporal and directional characteristics of multi-source responses. This solves the problems of category confusion, false detection, and missed detection that easily occur when relying on a single signal or a single anomaly feature for defect identification. By determining the continuity of the defect process based on the deviation morphology of adjacent weld segments on both sides of the defect interval, the system merges defect intervals of adjacent weld segments with continuous deviation characteristics and outputs the corresponding judgment results, location range, and judgment basis. This achieves the effect of completely characterizing the continuous development range of defects along the welding process, solving the problem of the same continuous defect being identified in a fragmented manner or transient anomalies being incorrectly extended into continuous defects, and improving the accuracy of laser welding defect judgment.

[0017] Furthermore, the method of the present invention divides the weld segment by determining the effective range of the signal. Since the laser reflection signal, the molten pool thermal radiation signal, and the plasma ion optical signal all contain welding process response and non-process components during the laser welding process, and the transient response duration of different signals varies, if the continuous sampling signal is directly analyzed for defects, short-term noise fluctuations or isolated spikes are easily mistaken for changes in the welding process, resulting in the identification of welding process boundaries being offset and different process stages being mixed together. By determining the effective range of the signal and dividing the weld segment, the effective change area corresponding to the actual welding process can be extracted from the continuous multi-source signal, and the signal response of different time periods can be uniformly mapped to the corresponding weld segment, providing a clear temporal analysis basis for subsequent defect identification based on multi-source response morphology, and further improving the accuracy of laser welding defect judgment.

[0018] Furthermore, the method of the present invention determines candidate defect categories by comparing weld segment judgment combinations with preset combination conditions. Since different welding defects have different effects on multi-source photoelectric signals, the same defect may exhibit different deviation durations, occurrence sequences, and change directions in different signal sources. Moreover, different defects may have similar local signal manifestations. Relying solely on the amplitude of a certain signal or a single abnormality pattern can easily lead to confusion of defect types, resulting in an overly broad range of candidate defects or missed detection of actual defects. By determining candidate defect categories based on weld segment judgment combinations and preset combination conditions, the participation relationship, temporal sequence relationship, and directional consistency between multi-source signals are used to further constrain the single abnormal response, enabling welding anomalies with different physical manifestations to form distinguishable combination features. This narrows the range of candidate defect categories, improves the pertinence and accuracy of defect category determination, and further enhances the accuracy of laser welding defect determination.

[0019] Furthermore, the method of the present invention determines the continuity of the defect process based on the deviation morphology of adjacent sequential weld segments at both ends of the defect interval. Since some defects in laser welding are not formed instantaneously at a single sampling position, but continue to develop along the welding direction with the heat input, the state of the molten pool, and the material action process, the same defect usually has a connectable deviation morphology in adjacent weld segments. If the judgment result is output separately according to each weld segment, it is easy to split the continuously developing same defect into multiple independent defects, or to mistake isolated transient anomalies as continuous defects, thereby causing inaccurate defect interval boundaries and an overstatement of the number of defects. By determining the continuity of the defect process, continuous defects can be merged according to the deviation morphology of adjacent weld segments, and a complete defect interval can be formed when the continuity is established. By combining the position range of the defect interval and the corresponding judgment criteria, a unified judgment result is output, thereby improving the continuity of defect spatial positioning and the consistency of defect identification results, and further improving the accuracy of laser welding defect judgment. Attached Figure Description

[0020] Figure 1 This is an overall flowchart of the intelligent laser welding defect determination method based on multi-source optoelectronic signal fusion according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating the process of determining the effective signal range in the intelligent laser welding defect determination method based on multi-source optoelectronic signal fusion according to an embodiment of the present invention. Figure 3 This is a flowchart illustrating the process of determining candidate welding defects in the intelligent laser welding defect determination method based on multi-source optoelectronic signal fusion according to an embodiment of the present invention. Figure 4 This is a flowchart illustrating the process of reviewing defect regions in the intelligent laser welding defect determination method based on multi-source optoelectronic signal fusion, as described in an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0022] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0023] Please see Figure 1 The diagram shown is an overall flowchart of the intelligent laser welding defect determination method based on multi-source optoelectronic signal fusion according to an embodiment of the present invention. The method described in this embodiment includes: Step S1: Determine the corresponding effective signal range based on the synchronous acquisition of multi-source photoelectric signals during the laser welding process, and align the effective signal range to the same welding process axis. The multi-source photoelectric signals include laser reflection signals, molten pool thermal radiation signals, and ion light signals. Step S2: Divide the effective signal interval based on the continuous sampling position interval on the welding process axis to obtain several sequential welding segments; Step S3: Based on the change direction of the multi-source photoelectric signals in the several sequential welding segments, determine the deviation pattern corresponding to the multi-source photoelectric signals respectively, and determine the participating source combination of the corresponding sequential welding segment based on the deviation pattern; Step S4: When a deviation signal in the sequential welding segment reaches the corresponding output upper limit, it is determined that the deviation signal that reaches the corresponding output upper limit cannot be judged as a deviation state in the current sequential welding segment, and the corresponding acquisition process parameters are adjusted. Step S5: Determine the connection relationship between the several deviation signals based on the order of their deviation start positions on the welding process axis, and determine the directional relationship between the several deviation signals based on their change direction in the corresponding deviation process. Step S6: Construct a weld segment determination combination based on the deviation shape, the participating source combination, the receiving relationship, and the directional relationship; determine welding defect candidates based on the weld segment determination combination and preset combination conditions; when the weld segment determination combination does not meet the preset combination conditions, mark the corresponding sequential weld segment as an unclassified abnormal weld segment. Step S7: Merge several sequential weld segments that are adjacent on the welding process axis and have the same weld segment judgment result to form a corresponding defect interval, and determine the judgment result of the defect interval to review the defect interval.

[0024] Specifically, the laser reflection signal, the molten pool thermal radiation signal, and the ion light signal are three photoelectric signals acquired in the same process during a single welding operation. The laser reflection signal reflects the absorption state of the workpiece by the laser, the molten pool thermal radiation signal reflects the temperature fluctuation of the molten pool, and the ion light signal reflects the amount of metal vapor ejected.

[0025] Specifically, the welding process axis is a sequence of sampling positions obtained by synchronously sampling three photoelectric signals of a single welding operation based on the same acquisition clock. The sampling position is the sequence number corresponding to each sampling time in the sequence, and the order of the welding process is represented by the increasing direction of the sequence number.

[0026] Specifically, the continuous sampling position interval is a set of sampling positions on the welding process axis that is formed by continuously increasing adjacent sampling positions and does not include any missing sampling positions.

[0027] Specifically, the laser reflection signal, the molten pool thermal radiation signal, and the ionization light signal are three photoelectric signals synchronously acquired during the same acquisition process in a single welding operation. The laser reflection signal is the laser radiation intensity signal obtained by a photodetector after being reflected from the workpiece surface or the molten pool during the welding process; The molten pool thermal radiation signal is the light intensity signal obtained by a photodetector through the thermal radiation generated in the molten pool area during the welding process. The ion light signal is the light intensity signal obtained by a photodetector from the characteristic light radiation generated by the metal vapor and the plasma formed by its ionization during the welding process.

[0028] Specifically, the participating source combination is a signal combination composed of multiple photoelectric signals that are not in a deviated state within the corresponding sequential welding segment.

[0029] Specifically, the deviation signal is a multi-source photoelectric signal that forms a deviation shape within the corresponding sequential welding segment.

[0030] Specifically, the corresponding output upper limit is the maximum signal value that the corresponding photoelectric acquisition channel can output under the current acquisition configuration, which is determined by the upper limit of the range of the corresponding photodetector, the input range of the analog-to-digital converter, and the corresponding gain setting.

[0031] Specifically, the acquisition process parameters include the sampling rate, gain, filter type, and filter cutoff frequency that control the acquisition process of multi-source photoelectric signals.

[0032] Specifically, welding defect candidates include weld misalignment candidates, incomplete penetration candidates, burn-through candidates, spatter candidates, and porosity candidates.

[0033] Specifically, the welding defect candidate is used to characterize the type of welding defect to be confirmed based on the deviation morphology, participating source combination, inheritance relationship and directional relationship, and is not used as the final defect confirmation result of post-weld non-destructive testing.

[0034] In this embodiment, the method identifies the sustained changes of multi-source photoelectric signals at the background level, determines the effective signal range, and divides the weld segment according to the welding process. It converts continuously acquired multi-source heterogeneous signals into time-series analysis units corresponding to the actual welding process, achieving the effects of suppressing transient noise and invalid fluctuation interference, and clearly defining the welding process boundaries. This solves the problems of difficulty in distinguishing between the welding process and non-process disturbances in continuous signals, as well as the inaccuracy of the subsequent defect analysis time range. Furthermore, by constructing weld segment judgment combinations based on deviation morphology, participating source combinations, connection relationships, and directional relationships, and combining these with preset combination conditions to determine candidate defect categories, the method differentiates the welding anomalies caused by different photoelectric signals. By applying correlation constraints to the response, the system effectively distinguishes different physical manifestations of defects using the temporal and directional characteristics of multi-source responses. This solves the problems of category confusion, false detection, and missed detection that easily occur when relying on a single signal or a single abnormal feature for defect identification. By determining the continuity of the defect process based on the deviation morphology of adjacent weld segments on both sides of the defect interval, the system merges defect intervals of adjacent weld segments with continuous deviation characteristics and outputs the corresponding judgment results, location range, and judgment basis. This achieves the effect of completely characterizing the continuous development range of the defect along the welding process, solving the problem of the same continuous defect being identified in a fragmented manner or transient anomalies being incorrectly extended into continuous defects, thus improving the accuracy of laser welding defect judgment.

[0035] Please see Figure 2 As shown, it is a logic flowchart of the process of determining the effective range of the signal in the intelligent judgment method for laser welding defects based on multi-source optoelectronic signal fusion in this embodiment.

[0036] Specifically, determining the effective range of the signal includes: The signal level of the multi-source photoelectric signal when the laser is not applied to the welding process is used as the background level; The welding initiation event is determined based on the deviation state of the multi-source photoelectric signal and the corresponding background level, and the sampling time corresponding to the welding initiation event is taken as the starting point of the effective interval of the signal. The welding end event is determined based on the moment when the deviation of the multi-source photoelectric signal falls back to the corresponding background level, and the sampling time corresponding to the welding end event is taken as the end point of the effective range of the signal.

[0037] Specifically, the division of the sequential welding segments includes: Based on the sampling position corresponding to the welding initiation event, the effective signal range is divided into several sequential welding segments according to the continuous equal sampling position intervals. The laser reflection signal, molten pool thermal radiation signal and ion light signal in the equal sampling position intervals are respectively composed into the synchronous sampling sequence of the corresponding sequential welding segments.

[0038] Specifically, the welding start event is the event in which the multi-source photoelectric signals deviate from their respective corresponding background levels and meet the start event determination conditions, and the welding end event is the event in which the deviated multi-source photoelectric signals fall back to their corresponding background levels and meet the end event determination conditions.

[0039] Specifically, the welding start event and the welding end event correspond to the start and end points of the effective signal range, respectively.

[0040] Specifically, the background level is the signal level observed by each multi-source photoelectric signal during the sampling interval when the laser does not act on the workpiece.

[0041] Specifically, the sequential welding segment is a sequential segment obtained by dividing the effective signal range into consecutive sampling position intervals of equal length.

[0042] Specifically, the sequential welding segment includes a synchronous sampling sequence formed by laser reflection signals, molten pool thermal radiation signals, and ion light signals within the corresponding sampling position interval.

[0043] Specifically, deviation refers to the change of the multi-source photoelectric signal relative to its initial state within the corresponding sequential welding segment, and the deviation morphology is used to characterize the temporal change structure of the deviation within the sequential welding segment.

[0044] Specifically, the non-deviation state refers to the state in which the corresponding multi-source photoelectric signals do not form a pulse-type, intermittent, step-holding, or continuous deviation structure within the sequential welding segment; when the laser reflection signal, the molten pool thermal radiation signal, and the ion light signal are all in the non-deviation state, it indicates that no deviation corresponding to the welding defect candidate has been detected within the corresponding sequential welding segment.

[0045] Specifically, the pulse type is a deviation pattern in which the corresponding multi-source photoelectric signal deviates and then recovers to the initial state of deviation, and no longer deviates in the same direction within the sequential welding segment.

[0046] Specifically, the intermittent type is a deviation pattern in which the corresponding multi-source photoelectric signal deviates, recovers to the initial state of deviation, and then deviates in the same direction again.

[0047] Specifically, the step hold type is the deviation pattern in which the multi-source photoelectric signal deviates and does not recover to the initial state of the deviation, and remains there until the end of the sequential welding segment.

[0048] Specifically, the continuous type is the deviation pattern in which the multi-source photoelectric signals continuously deviate along the same direction of change within the sequential welding segment.

[0049] Specifically, the deviation signal is a multi-source photoelectric signal that forms the pulse-type, intermittent, step-holding, or continuous deviation within the corresponding sequential welding segment; the multi-source photoelectric signal that does not form the deviation is a non-deviation signal.

[0050] Specifically, an equal sampling position interval is a continuous sampling position interval with the same number of sampling points.

[0051] In this embodiment, the method divides the weld segment by determining the effective range of the signal. During laser welding, the laser reflection signal, the molten pool thermal radiation signal, and the plasma ion optical signal all contain welding process response and non-process components, and the duration of the transient response of different signals varies. If the continuous sampling signal is directly analyzed for defects, short-term noise fluctuations or isolated spikes are easily mistaken for changes in the welding process, resulting in the identification of welding process boundaries being offset and different process stages being mixed together. By determining the effective range of the signal and dividing the weld segment, the effective change area corresponding to the actual welding process can be extracted from the continuous multi-source signal, and the signal response of different time periods can be uniformly mapped to the corresponding weld segment. This provides a clear temporal analysis basis for subsequent defect identification based on multi-source response morphology, further improving the accuracy of laser welding defect judgment.

[0052] Please see Figure 3 As shown, it is a logical flowchart of the process for determining welding defect candidates in this embodiment.

[0053] Specifically, the deviation patterns include pulse type, intermittent type, step hold type, continuous type, and no deviation pattern.

[0054] Specifically, determining the acceptance relationship and the directional relationship includes: The multi-source photoelectric signals that deviate are used as participating sources, and the multi-source photoelectric signals that do not deviate are used as non-participating sources; The sequential order of several participating sources corresponding to several deviations from the starting position on the welding process axis is determined as the succession relationship between several participating sources, and participating sources with the same deviation from the starting position are determined as simultaneous; Participating sources whose change direction is consistent in the corresponding deviation process are identified as having a same-direction relationship, and participating sources whose change direction is opposite are identified as having a reverse relationship.

[0055] Specifically, determining welding defect candidates based on the weld segment judgment combination and preset combination conditions includes: When the laser reflection signal, the molten pool thermal radiation signal, and the ion light signal are all in a state of no deviation, it is determined that no defects were detected in the corresponding sequential weld segment. When only the laser reflection signal deviates continuously, the judgment result corresponding to the weld segment is determined as a weld misalignment candidate; When the laser reflection signal, the molten pool thermal radiation signal, and the ion light signal all exhibit a continuous deviation, and the direction of change of the laser reflection signal is opposite to that of the molten pool thermal radiation signal and the ion light signal, the determination result corresponding to the weld segment is determined to be a candidate for incomplete penetration. When the molten pool thermal radiation signal deviates from the ion light signal in a pulse-like or step-holding manner, and the starting position of the deviation of the molten pool thermal radiation signal precedes that of the ion light signal, the determination result corresponding to the weld segment is determined to be a burn-through candidate.

[0056] Specifically, when the deviation signal is in the undeterminable state, an acquisition parameter adjustment instruction is generated according to the acquisition anomaly type corresponding to the undeterminable state, and the acquisition process of the deviation signal is adjusted according to the acquisition parameter adjustment instruction. The determination of the sequential welding segment is performed based on the multi-source photoelectric signal of the welding process re-acquired after adjustment.

[0057] Specifically, synchronous acquisition involves acquiring laser reflection signal, molten pool thermal radiation signal, and ion light signal separately during a single welding process, and making the three signals form a synchronous sampling sequence based on the same acquisition clock; the three signals share a timestamp reference, so that the three signals corresponding to the same sampling time have the same sampling sequence number.

[0058] Specifically, welding methods include continuous welding, spot welding, and oscillating welding, among which... For continuous welding, the effective signal interval is divided sequentially according to the equal sampling position interval to obtain several sequential welding segments; For spot welding, the sampling interval corresponding to a single weld spot is taken as a sequential weld segment; For oscillating welding, a sampling position interval corresponding to one or more complete oscillation cycles is used as a sequential welding segment, and the number of oscillation cycles is determined by the corresponding working condition configuration.

[0059] Specifically, the deviation starting position is the sampling position where the corresponding deviation signal first forms the corresponding deviation state on the welding process axis, which is used to characterize the order of occurrence of the corresponding deviation event in a single welding process.

[0060] Specifically, the succession relationship is the temporal relationship between participating sources determined by the order in which they deviate from the starting position; the participating source that deviates from the starting position earlier is determined as the preceding source, and the participating source that deviates from the starting position later is determined as the response source; when the deviation positions of two or more participating sources are located at the same sampling position, the relationship between the participating sources is determined as a simultaneous relationship.

[0061] Specifically, the succession relationship is used to characterize the sequential response relationship of deviation events corresponding to different participating sources in the welding process, rather than directly characterizing the physical causal relationship.

[0062] Specifically, the directional relationship is the relationship between the participating sources determined based on the direction of change of each participating source in the corresponding deviation process; when the deviation change directions of two participating sources are consistent, they are determined to be in the same direction, and when the deviation change directions of two participating sources are opposite, they are determined to be in opposite directions.

[0063] Specifically, the weld segment judgment combination is a combination judgment feature formed by the deviation shape, participating source combination, inheritance relationship and directional relationship of the corresponding sequential weld segments.

[0064] Specifically, the preset combination conditions are a set of conditions used to map the weld segment judgment combination to the corresponding weld segment judgment result.

[0065] Specifically, when the laser reflection signal, the molten pool thermal radiation signal, and the ion light signal are all in a state of no deviation, it is determined that no defects were detected in the corresponding sequential weld segment; when at least one multi-source photoelectric signal deviates, it is matched with the preset combination conditions according to the corresponding deviation pattern, participating source combination, connection relationship, and direction relationship to determine the corresponding welding defect candidate.

[0066] Specifically, determining welding defect candidates according to preset combination conditions includes: determining weld misalignment candidates when only the laser reflection signal shows a continuous deviation, and the molten pool thermal radiation signal and the ion light signal are both without deviation; determining incomplete penetration candidates when all three signals show continuous deviation, and the laser reflection signal changes in the opposite direction to the molten pool thermal radiation signal and the ion light signal; and determining burn-through candidates when both the molten pool thermal radiation signal and the ion light signal show deviation, and the deviation patterns of the two are pulse-type or step-hold type with corresponding upward change directions, and the deviation start position of the molten pool thermal radiation signal precedes that of the ion light signal.

[0067] Specifically, when the participating source combination consists of both the laser reflection signal and the molten pool thermal radiation signal being deviation signals, both having pulse-like deviation patterns, both having deviation starting positions at the same sampling position, and the ion beam signal showing a continuous deviation, the determination result is a candidate for spatter; when the participating source combination consists of both the molten pool thermal radiation signal and the ion beam signal being deviation signals, both having intermittent deviation patterns, and the laser reflection signal not showing a continuous deviation, the determination result is a candidate for porosity; when the participating source combination consists of both the molten pool thermal radiation signal and the ion beam signal being deviation signals, both having continuous deviation patterns and a downward direction, both having deviations occurring in the later part of the sequential welding segment, and the laser reflection signal not showing a deviation, the determination result is an abnormal indication of the solidification stage.

[0068] Specifically, the judgment results include no defects detected, weld misalignment candidate, incomplete penetration candidate, burn-through candidate, spatter candidate, porosity candidate, and solidification stage abnormality indication.

[0069] Specifically, the welding defect candidate is the type of defect to be confirmed obtained by the online judgment side based on the weld segment judgment combination. The solidification stage abnormality prompt is used to indicate that there is an abnormality in the solidification stage of the corresponding weld segment. The solidification stage abnormality prompt is output as solidification stage abnormality indication information on the online judgment side. Its specific defect category is confirmed by post-weld inspection based on the position range of the corresponding weld segment or defect interval on the welding process axis and the corresponding judgment criteria.

[0070] Specifically, the determination basis is the structured determination information that forms the corresponding determination result, including the combination of participating sources, the deviation shape of each participating source, the directional relationship between participating sources, and the connection relationship between participating sources; the position range is the sampling position range defined by the starting sampling position to the ending sampling position of the sequential weld segment contained in the corresponding defect interval on the welding process axis.

[0071] Specifically, when the weld segment judgment combination does not meet any preset combination condition, the corresponding sequential weld segment is marked as an unclassified abnormal weld segment; the unclassified abnormal weld segment is used to indicate that the corresponding sequential weld segment has a deviation, but the currently determined deviation form, participating source combination, inheritance relationship and direction relationship do not form the judgment result corresponding to the preset combination condition, or the corresponding combination has a situation where the judgment result cannot be uniquely determined.

[0072] Specifically, the record to be verified is used to record the original signals and judgment criteria corresponding to the unclassified abnormal weld segments, including at least the original sampling sequences of the laser reflection signal, the molten pool thermal radiation signal and the ion light signal, the participating source combination, the deviation morphology, the directional relationship and the connection relationship, for subsequent verification.

[0073] Specifically, the inherent output upper limit is the maximum signal value that the corresponding signal acquisition channel can output under the current acquisition configuration. The inherent output upper limit is determined by at least one of the range upper limit of the corresponding sensing unit, the input range of the signal acquisition link, and the saturation output range corresponding to the current gain configuration. When the deviation signal reaches the inherent output upper limit, the signal change is truncated due to the limitation of the output range, and cannot fully reflect the corresponding deviation process.

[0074] Specifically, determining whether the deviation signal has reached the inherent output upper limit can be done by using the saturation flag when the corresponding sensing unit outputs a saturation flag; if the saturation flag is not present, it can be determined by the state that the deviation signal forms a continuous cutoff interval at the upper limit of the range and no longer changes accordingly with the welding process at subsequent sampling positions.

[0075] Specifically, the acquisition process parameters are parameters that control the acquisition process of the multi-source photoelectric signals, including sampling rate, hardware gain, software gain, filter type, and filter cutoff frequency; the acquisition parameter adjustment indicator is used to select the corresponding acquisition process parameters for adjustment according to the current undetermined state, and the adjustment may include one or more of the following: reducing hardware gain or software gain, changing filter type or filter cutoff frequency, and adjusting sampling rate.

[0076] Specifically, the hardware gain can be achieved by a digitally programmable gain amplifier or a programmable amplifier, the software gain can be achieved by a digital signal processing procedure after analog-to-digital conversion, the filter type and filter cutoff frequency can be adjusted online by a digital filter module implemented by a field-programmable gate array or digital signal processor after analog-to-digital conversion, and the sampling rate can be achieved by adjusting the sampling clock division parameters or phase-locked loop parameters. After completing the parameter adjustment according to the acquisition parameter adjustment instruction, the multi-source photoelectric signals in the corresponding welding process are reacquired, and the determination of the corresponding sequential welding segment is re-executed based on the reacquired multi-source photoelectric signals.

[0077] In this embodiment, the method determines candidate defect categories by comparing weld segment judgment combinations with preset combination conditions. Since different welding defects have different effects on multi-source photoelectric signals, the same defect may exhibit different deviation durations, occurrence sequences, and change directions in different signal sources. Furthermore, different defects may have similar local signal manifestations. Relying solely on the amplitude of a single signal or a single abnormal morphology can easily lead to confusion of defect types, resulting in an overly broad range of candidate defects or missed detection of actual defects. By determining candidate defect categories based on weld segment judgment combinations and preset combination conditions, the participation relationship, temporal sequence relationship, and directional consistency between multi-source signals are used to further constrain the single abnormal response, enabling welding anomalies with different physical manifestations to form distinguishable combination features. This narrows the range of candidate defect categories, improves the pertinence and accuracy of defect category determination, and further enhances the accuracy of laser welding defect determination.

[0078] Please see Figure 4 The diagram shown is a logical flowchart of the process for reviewing defective regions according to an embodiment of the present invention. Specifically, the deviation shape and judgment result of the sequential weld segments adjacent to the defect interval at both ends are obtained to verify the process continuity of the defect interval; When no deviation is found in the sequential weld segments adjacent to both ends of the defect interval, the defect interval is marked as an interval to be checked. When at least one adjacent sequential weld segment in the defect interval deviates, the defect interval is determined to meet the process continuity condition and the judgment result is output.

[0079] Specifically, the review of the defective region includes: The weld segment determination combination is reconstructed based on the multi-source photoelectric signals corresponding to the sequential weld segments contained in the defect interval; When the weld segment judgment result corresponding to the reconstructed weld segment judgment combination is consistent with the original weld segment judgment result of the defect interval, the verification confirmation information of the defect interval is output. When the weld segment judgment result corresponding to the reconstructed weld segment judgment combination is inconsistent with the original weld segment judgment result of the defect interval, the defect interval is marked as an interval to be checked.

[0080] Specifically, when the defect interval is marked as an interval to be checked, or when the sequential weld segment is marked as an unclassified abnormal weld segment, the corresponding defect interval or the corresponding sequential weld segment is recorded in the record to be checked.

[0081] Specifically, the merging of defect intervals and the verification of process continuity include: merging adjacent weld segments with the same judgment result on the welding process axis into one defect interval; when the judgment results of adjacent weld segments are different, or when there are unclassified abnormal weld segments or weld segments that are determined to be unexecutable among adjacent weld segments, the interval is divided into different defect intervals; the location range of the defect interval is the sampling location range of the weld segments contained in the defect interval on the welding process axis; when the signals of the adjacent weld segments on both sides of the defect interval do not deviate, the defect interval is determined to be inconsistent with the process continuity required by the defect formation mechanism, the defect interval is marked as an interval to be checked and the judgment basis of the defect interval and the adjacent weld segments on both sides is output, but the judgment result of the defect interval is not output; when the signals of the adjacent weld segments on one or both sides of the defect interval deviate, the process continuity of the defect interval is determined to be established, and the judgment result, location range and judgment basis of the defect interval are output.

[0082] Specifically, the merging and configuration backflow of the judgment results includes: merging each weldment according to the rule of "if it exists, it is unqualified"; if any defect category candidate, solidification stage abnormal indication information, unclassified abnormal weld segment, unexecutable weld segment, or pending verification interval exists during the welding process of the weldment, the weldment is judged as unqualified; if the judgment results of all weld segments of the weldment do not constitute any of the above situations, the weldment is judged as qualified; the merging does not introduce weights, scores, or proportional thresholds; outputting the location range and judgment basis corresponding to the solidification stage abnormal indication as the confirmation object of post-weld inspection; using the effective acquisition parameter adjustment and weld segment division parameter adjustment in this judgment as the basis for generating the next welding process and subsequent weldment acquisition configuration under the same working condition configuration, and recording it in the working condition configuration; storing the judgment result and the configuration identifier of this welding in a group for comparison and statistics under the same working condition configuration, and not using it as the judgment benchmark for subsequent weld segments.

[0083] In this embodiment, the method determines the continuity of the defect process based on the deviation morphology of adjacent sequential weld segments at both ends of the defect interval. Since some defects in laser welding are not formed instantaneously at a single sampling position, but continue to develop along the welding direction with the heat input, the state of the molten pool, and the material action process, the same defect usually has a connectable deviation morphology in adjacent weld segments. If the judgment result is output separately according to each weld segment, it is easy to split the continuously developing same defect into multiple independent defects, or to mistake isolated transient anomalies as continuous defects, thereby causing inaccurate defect interval boundaries and an overestimation of the number of defects. By determining the continuity of the defect process, continuous defects can be merged according to the deviation morphology of adjacent weld segments, and a complete defect interval can be formed when the continuity is established. By combining the position range of the defect interval and the corresponding judgment criteria, a unified judgment result is output, thereby improving the continuity of defect spatial positioning and the consistency of defect identification results, and further improving the accuracy of laser welding defect judgment.

[0084] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Various equivalent modifications and improvements can be made by those skilled in the art without departing from the principles of the present invention, and all such equivalent modifications and improvements should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent determination of laser welding defects based on multi-source optoelectronic signal fusion, characterized in that, include: Based on the synchronous acquisition of multi-source photoelectric signals during the laser welding process, the corresponding effective signal range is determined, and the effective signal range is aligned to the same welding process axis. The multi-source photoelectric signals include laser reflection signals, molten pool thermal radiation signals, and ion light signals. The effective range of the signal is divided into several sequential welding segments based on the continuous sampling position range on the welding process axis. Based on the change direction of the multi-source photoelectric signals within the several sequential welding segments, the deviation pattern corresponding to the multi-source photoelectric signals is determined, and the combination of participating sources for the corresponding sequential welding segment is determined based on the deviation pattern. When a deviation signal in the sequential welding segment reaches the corresponding output upper limit, it is determined that the deviation signal that has reached the corresponding output upper limit cannot be judged as a deviation state in the current sequential welding segment, and the corresponding acquisition process parameters are adjusted. The succession relationship between the deviation signals is determined based on the order of their starting positions on the welding process axis, and the directional relationship between the deviation signals is determined based on their changing directions during the corresponding deviation processes. Based on the deviation pattern, the participating source combination, the receiving relationship, and the directional relationship, a weld segment judgment combination is constructed. Based on the weld segment judgment combination and the preset combination conditions, welding defect candidates are determined. When the weld segment judgment combination does not meet the preset combination conditions, the corresponding sequential weld segment is marked as an unclassified abnormal weld segment. Several sequential weld segments that are adjacent on the welding process axis and have the same weld segment judgment result are merged to form a corresponding defect interval, and the judgment result of the defect interval is determined to review the defect interval.

2. The intelligent laser welding defect determination method based on multi-source optoelectronic signal fusion according to claim 1, characterized in that, Determining the effective range of the signal includes: The signal level of the multi-source photoelectric signal when the laser is not applied to the welding process is used as the background level; The welding initiation event is determined based on the deviation state of the multi-source photoelectric signal and the corresponding background level, and the sampling time corresponding to the welding initiation event is taken as the starting point of the effective interval of the signal. The welding end event is determined based on the moment when the deviation of the multi-source photoelectric signal falls back to the corresponding background level, and the sampling time corresponding to the welding end event is taken as the end point of the effective range of the signal.

3. The intelligent laser welding defect determination method based on multi-source optoelectronic signal fusion according to claim 2, characterized in that, The division of the sequential welding segments includes: Based on the sampling position corresponding to the welding initiation event, the effective signal range is divided into several sequential welding segments according to the continuous equal sampling position intervals. The laser reflection signal, molten pool thermal radiation signal and ion light signal in the equal sampling position intervals are respectively composed into the synchronous sampling sequence of the corresponding sequential welding segments.

4. The intelligent laser welding defect determination method based on multi-source optoelectronic signal fusion according to claim 3, characterized in that, The deviation patterns include pulse type, intermittent type, step hold type, continuous type, and no deviation pattern.

5. The intelligent laser welding defect determination method based on multi-source optoelectronic signal fusion according to claim 4, characterized in that, Determining the acceptance relationship and the directional relationship includes: The multi-source photoelectric signals that deviate are used as participating sources, and the multi-source photoelectric signals that do not deviate are used as non-participating sources; The sequential order of several participating sources corresponding to several deviations from the starting position on the welding process axis is determined as the succession relationship between several participating sources, and participating sources with the same deviation from the starting position are determined as simultaneous; Participating sources whose change direction is consistent in the corresponding deviation process are identified as having a same-direction relationship, and participating sources whose change direction is opposite are identified as having a reverse relationship.

6. The intelligent laser welding defect determination method based on multi-source optoelectronic signal fusion according to claim 5, characterized in that, Based on the weld segment judgment combination and preset combination conditions, candidate welding defects are determined, including: When the laser reflection signal, the molten pool thermal radiation signal, and the ion light signal are all in a state of no deviation, it is determined that no defects were detected in the corresponding sequential weld segment. When only the laser reflection signal deviates continuously, the judgment result corresponding to the weld segment is determined as a weld misalignment candidate; When the laser reflection signal, the molten pool thermal radiation signal, and the ion light signal all exhibit a continuous deviation, and the direction of change of the laser reflection signal is opposite to that of the molten pool thermal radiation signal and the ion light signal, the determination result corresponding to the weld segment is determined to be a candidate for incomplete penetration. When the molten pool thermal radiation signal deviates from the ion light signal in a pulse-like or step-holding manner, and the starting position of the deviation of the molten pool thermal radiation signal precedes that of the ion light signal, the determination result corresponding to the weld segment is determined to be a burn-through candidate.

7. The intelligent laser welding defect determination method based on multi-source optoelectronic signal fusion according to claim 6, characterized in that, When the deviation signal is in the undeterminable state, an acquisition parameter adjustment instruction is generated according to the acquisition anomaly type corresponding to the undeterminable state, and the acquisition process of the deviation signal is adjusted according to the acquisition parameter adjustment instruction. The determination of the sequential welding segment is performed based on the multi-source photoelectric signal of the welding process re-acquired after adjustment.

8. The intelligent method for judging laser welding defects based on multi-source optoelectronic signal fusion according to claim 1, characterized in that, Obtain the deviation shape and judgment result of the sequential weld segments adjacent to the defect interval at both ends to verify the process continuity of the defect interval; When no deviation is found in the sequential weld segments adjacent to both ends of the defect interval, the defect interval is marked as an interval to be checked. When at least one adjacent sequential weld segment in the defect interval deviates, the defect interval is determined to meet the process continuity condition and the judgment result is output.

9. The intelligent laser welding defect determination method based on multi-source optoelectronic signal fusion according to claim 8, characterized in that, The review of the defective region includes: The weld segment determination combination is reconstructed based on the multi-source photoelectric signals corresponding to the sequential weld segments contained in the defect interval; When the weld segment judgment result corresponding to the reconstructed weld segment judgment combination is consistent with the original weld segment judgment result of the defect interval, the verification confirmation information of the defect interval is output. When the weld segment judgment result corresponding to the reconstructed weld segment judgment combination is inconsistent with the original weld segment judgment result of the defect interval, the defect interval is marked as an interval to be checked.

10. The intelligent laser welding defect determination method based on multi-source optoelectronic signal fusion according to claim 9, characterized in that, When the defect interval is marked as an interval to be checked, or when the sequential weld segment is marked as an unclassified abnormal weld segment, the corresponding defect interval or the corresponding sequential weld segment is recorded in the record to be checked.

Images