A welding process condition monitoring system and method based on multi-source sensing

Abstract

This application provides a welding process condition monitoring system and method based on multi-source sensing. During the welding process, multi-source welding data is collected simultaneously, and a welding state feature set is constructed based on this data. The stability change trend of the welding process is characterized by transitions according to the welding state feature set, resulting in a state transition feature space. Evolutionary modeling is performed using this feature space to obtain the welding state evolution trajectory. Based on the welding state evolution trajectory, the welding state category is identified, and the deviation of the welding process under each category is evaluated. The probability of abnormal risks is predicted based on the deviation. Real-time monitoring and early warning of the welding process are performed based on the probability of abnormal risks, and welding parameter optimization instructions are output. The technical solution provided in this application can characterize the stability change trend of the welding process based on multi-source sensing information, thereby achieving welding state category identification and abnormal risk prediction, and improving the reliability of the welding process.

Description

Technical Field

[0001] This application relates to the field of welding condition monitoring technology, and more specifically, to a welding process condition monitoring system and method based on multi-source sensing. Background Technology

[0002] As a typical high-energy-density, highly nonlinear, and highly disturbed industrial process, the stability and welding quality of welding are highly dependent on the synergistic effect of multiple factors such as arc state, thermal behavior of the molten pool, weld formation process, and mechanical motion stability.

[0003] Existing welding process monitoring technologies mostly rely on single sensors or limited sensor data, such as current / voltage signals, temperature signals, or visual images for status assessment. This results in limitations such as limited information dimensions, weak anti-interference capabilities, and poor adaptability to complex operating conditions. Furthermore, current technologies primarily rely on static threshold judgments or instantaneous state analysis, lacking the ability to systematically model the dynamic evolution characteristics and stability trends of the welding process. This makes it difficult to accurately depict the process mechanism of welding transitioning from stable to abnormal states, easily leading to misjudgments and missed detections. Therefore, how to characterize the stability trends of the welding process based on multi-source sensor information, thereby achieving welding state category identification and anomaly risk prediction to improve the reliability of the welding process, remains a challenge for current technologies. Summary of the Invention

[0004] This application provides a welding process condition monitoring system and method based on multi-source sensing, which can characterize the stability change trend of the welding process based on multi-source sensing information, thereby realizing the identification of welding condition categories and the prediction of abnormal risks, so as to improve the reliability of the welding process.

[0005] In a first aspect, this application provides a welding process condition monitoring method based on multi-source sensing, comprising the following steps:

[0006] Multi-source welding data is collected simultaneously during the welding process, and a set of welding state features during the welding process is constructed based on the multi-source welding data.

[0007] Based on the set of welding state features, the stability change trend of the welding process is characterized by transitions to obtain a state transition feature space. The state transition feature space is then used for evolution modeling to obtain the welding state evolution trajectory of the welding process.

[0008] Based on the welding state evolution trajectory, the welding state category is identified, and the deviation of the welding process under the state category is evaluated. The probability of abnormal risk in the welding process is predicted by the deviation.

[0009] The welding process is monitored and warned in real time based on the probability of abnormal risks, and welding parameter optimization instructions are output.

[0010] In some embodiments, the multi-source welding data includes welding current data, molten pool infrared temperature field data, molten pool high-speed visual image, and welding torch and workpiece vibration signals.

[0011] In some embodiments, the set of welding state characteristics includes the average current, arc fluctuation index, short-circuit duty cycle, molten pool area, spatter density, molten pool center temperature, cooling rate, and vibration RMS value during the welding process.

[0012] In some embodiments, the stability change trend of the welding process is characterized by transitions based on the welding state feature set, resulting in a state transition feature space that specifically includes:

[0013] For each type of welding state feature in the welding state feature set, obtain the sequence data corresponding to the welding state feature;

[0014] Adjacent transition metrics are performed on the sequence data to obtain the state transition feature sequence corresponding to the welding state feature, and then the state transition feature sequence corresponding to each type of welding state feature is obtained.

[0015] A state transition feature space is constructed by using the state transition feature sequences corresponding to all welding state features.

[0016] In some embodiments, it also includes:

[0017] Obtain the signal-to-noise ratio and historical stability contribution of each sensor channel during the welding process;

[0018] The reliability weight of each sensing channel is determined based on the corresponding signal-to-noise ratio and historical stability contribution.

[0019] In some embodiments, using the state transition feature space for evolutionary modeling to obtain the welding state evolution trajectory of the welding process specifically includes:

[0020] The state transition feature sequences corresponding to each welding state feature in the state transition feature space are weighted and fused according to the confidence weight of the corresponding sensing channel to obtain the fused feature sequence.

[0021] The fused feature sequence is mapped to the welding state evolution trajectory of the welding process using a long short-term memory network.

[0022] In some embodiments, identifying the welding state category based on the welding state evolution trajectory specifically includes:

[0023] Multiple classification features are extracted from the welding state evolution trajectory;

[0024] All classification features are input into a Softmax classifier for classification, thereby identifying the welding state category.

[0025] Secondly, this application provides a welding process condition monitoring system based on multi-source sensing, used to execute a welding process condition monitoring method based on multi-source sensing, including:

[0026] The feature construction module is used to simultaneously collect multi-source welding data during the welding process and construct a set of welding state features during the welding process based on the multi-source welding data.

[0027] The evolution modeling module is used to characterize the stability change trend of the welding process based on the welding state feature set, obtain the state transition feature space, and use the state transition feature space to perform evolution modeling, thereby obtaining the welding state evolution trajectory of the welding process.

[0028] The probability prediction module is used to identify the welding state category based on the welding state evolution trajectory, evaluate the deviation of the welding process under the state category, and predict the probability of abnormal risk of the welding process through the deviation.

[0029] The early warning feedback module is used to monitor and issue early warnings for the welding process in real time based on the probability of abnormal risks, and to output welding parameter optimization instructions.

[0030] Thirdly, this application provides a computer device, the computer device including a memory and a processor, the memory storing code, the processor being configured to acquire the code and execute the above-described welding process status monitoring method based on multi-source sensing.

[0031] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described welding process status monitoring method based on multi-source sensing.

[0032] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects:

[0033] This application provides a welding process state monitoring system and method based on multi-source sensing. Multi-source welding data is collected synchronously during the welding process, and a welding state feature set is constructed based on the multi-source welding data. The stability change trend of the welding process is characterized by transitions according to the welding state feature set, resulting in a state transition feature space. Evolutionary modeling is performed using the state transition feature space to obtain the welding state evolution trajectory. Based on the welding state evolution trajectory, the welding state category is identified, and the deviation of the welding process under the stated state category is evaluated. The probability of abnormal risks in the welding process is predicted based on the deviation. Real-time monitoring and early warning of the welding process are performed based on the probability of abnormal risks, and welding parameter optimization instructions are output.

[0034] Therefore, this application firstly employs a method of constructing a state transition feature space based on a set of welding state features and performing evolutionary modeling to obtain the welding state evolution trajectory. This elevates the analysis of multi-source sensor information during the welding process from static feature analysis to a dynamic characterization of stability change trends, thereby achieving a systematic modeling of the temporal evolution law of the welding state. Secondly, by identifying the welding state category based on the welding state evolution trajectory and evaluating the deviation of the welding process under that category, and then predicting the probability of abnormal risks based on the deviation, the dynamic changes of the welding process can be upgraded from static parameter monitoring to temporal state understanding and risk quantification assessment. This enables early perception and precise identification of welding stability degradation trends. Finally, by implementing real-time monitoring and early warning of the welding process based on the probability of abnormal risks and outputting welding parameter optimization instructions, multi-source sensor perception, state identification, and process control can be organically integrated to form a closed-loop control system from state perception to proactive intervention. This upgrades welding stability management from passive post-event detection to proactive pre-event prevention, thereby achieving welding state category identification and abnormal risk prediction to improve the reliability of the welding process.

[0035] In summary, the technical solution adopted in this application can characterize the stability change trend of the welding process based on multi-source sensing information, thereby realizing the identification of welding state categories and the prediction of abnormal risks, so as to improve the reliability of the welding process. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this embodiment of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is an exemplary flowchart of a welding process status monitoring method based on multi-source sensing, according to some embodiments of this application;

[0038] Figure 2 This is an exemplary flowchart illustrating the determination of the state transition feature space according to some embodiments of this application;

[0039] Figure 3 This is a schematic diagram of the structure of a welding process condition monitoring system based on multi-source sensing, according to some embodiments of this application;

[0040] Figure 4 This is a schematic diagram of the structure of a computer device for implementing a welding process status monitoring method based on multi-source sensing, according to some embodiments of this application. Detailed Implementation

[0041] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0042] This application provides a welding process condition monitoring system and method based on multi-source sensing. The core of the system is to simultaneously collect multi-source welding data during the welding process, construct a welding state feature set based on the multi-source welding data, characterize the stability change trend of the welding process according to the welding state feature set, obtain a state transition feature space, use the state transition feature space for evolutionary modeling, and thus obtain the welding state evolution trajectory of the welding process. Based on the welding state evolution trajectory, the welding state category is identified, and the deviation of the welding process under the state category is evaluated. The abnormal risk probability of the welding process is predicted by the deviation. Based on the abnormal risk probability, the welding process is monitored and warned in real time, and welding parameter optimization instructions are output. This scheme can characterize the stability change trend of the welding process based on multi-source sensing information, thereby realizing welding state category identification and abnormal risk prediction, and improving the reliability of the welding process.

[0043] To better understand the above technical solutions, a detailed description of the technical solutions will be provided below in conjunction with the accompanying drawings and specific embodiments. (Refer to...) Figure 1 The figure is an exemplary flowchart of a welding process status monitoring method based on multi-source sensing according to some embodiments of this application. The figure mainly includes the following steps:

[0044] In step S101, multi-source welding data is collected synchronously during the welding process, and a set of welding state features during the welding process is constructed based on the multi-source welding data.

[0045] It should be noted that, in this application, the multi-source welding data includes welding current data, molten pool infrared temperature field data, molten pool high-speed visual image, and welding torch and workpiece vibration signals. Specifically, during the welding operation, multiple types of sensing devices are first deployed at key parts of the welding equipment to synchronously sense multi-physical field information during the welding process. These include electrical signal sensors for collecting electrical parameters such as welding current and voltage, infrared temperature sensors for monitoring molten pool temperature and thermal field distribution, visual imaging devices for acquiring molten pool morphology, weld formation, and spatter behavior, and vibration sensors for monitoring welding torch movement stability and workpiece vibration state. After sensor deployment, various sensing channels are synchronized using a unified clock source or hardware triggering mechanism, enabling a one-to-one correspondence between data collected by different sensors at the same time point, thus forming a multi-channel synchronous sampling system. Subsequently, at the start of welding, the data acquisition unit samples the raw signals from each sensing channel in real time, encapsulating the collected welding current data, molten pool infrared temperature field data, molten pool high-speed visual image, and welding torch and workpiece vibration signals into time-stamped multi-source welding data.

[0046] In some embodiments, a set of welding state features during the welding process is constructed based on the multi-source welding data. It should be noted that, in this application, the set of welding state features includes the average current, arc fluctuation index, short-circuit duty cycle, molten pool area, spatter density, molten pool center temperature, cooling rate, and vibration RMS (Root Mean Square) value during the welding process.

[0047] In practical implementation, the welding current data in the multi-source welding data is denoised and smoothed. The average current value is calculated within a stable time window to characterize the energy input level. An arc fluctuation index is constructed by analyzing the variance and high-frequency fluctuation components of the current waveform to characterize the arc combustion stability. Simultaneously, the time ratio of the welding current in the short-circuit state and the arc state is statistically analyzed to obtain the short-circuit duty cycle, which reflects the metal transition behavior. The molten pool infrared temperature field data from the multi-source welding data is used as input. The projected area of ​​the molten pool, i.e., the molten pool area, is calculated using molten pool region segmentation and contour extraction algorithms to reflect the matching between the molten pool size and the heat input. The spatter density is calculated by detecting the number of spatter particles and their temporal variation density to evaluate the stability of the welding process and the smoothness of the metal transition. The instantaneous temperature of the molten pool center region is extracted as the molten pool center temperature, and the cooling rate is calculated based on the temperature change rate at adjacent time points to characterize the weld solidification behavior and heat diffusion characteristics. The root mean square (RMS) value of vibration is obtained by calculating the vibration signals of the welding torch and the workpiece in the multi-source welding data. This value is used to characterize the mechanical stability and the influence of external disturbances during the welding process.

[0048] In step S102, the stability change trend of the welding process is characterized by transition based on the welding state feature set to obtain the state transition feature space. The state transition feature space is then used for evolution modeling to obtain the welding state evolution trajectory of the welding process.

[0049] Preferably, in some embodiments, reference is made to Figure 2 As shown, this figure is an exemplary flowchart of determining the state transition feature space according to some embodiments of this application. In this embodiment, the stability change trend of the welding process is characterized by transitions based on the welding state feature set, and the state transition feature space can be obtained by the following steps:

[0050] In step S1021, for each type of welding state feature in the welding state feature set, the sequence data corresponding to the welding state feature is obtained;

[0051] In step S1022, adjacent transition measurement is performed on the sequence data to obtain the state transition feature sequence corresponding to the welding state feature, and then the state transition feature sequence corresponding to each type of welding state feature is obtained.

[0052] In step S1023, a state transition feature space is constructed by using the state transition feature sequences corresponding to all welding state features.

[0053] In practical implementation, firstly, for each type of welding state feature in the welding state feature set, the corresponding sequence data can be obtained. This sequence data is a series of data sets composed of the values ​​of the corresponding welding state feature. For example, when the welding state feature is the average current, the corresponding sequence data is a series of current values ​​sorted in chronological order. Then, adjacent transition metrics can be performed on the sequence data, i.e., time-window weighted differencing is applied to the sequence data to obtain the weighted transition features corresponding to each time step. Nonlinear mapping and data dimensionality reduction are then applied to each weighted transition feature. Nonlinear mapping highlights large changes, and data dimensionality reduction enhances signal stability by denoising. This yields the state transition features corresponding to each time step, which represent the dynamic fluctuation degree of the corresponding welding state feature during the welding process. In actual implementation, these state transition features can be determined in the following way:

[0054]

[0055] in, This represents the state transition characteristics at time t. Principal component analysis (PCA) maps high-dimensional features to a low-dimensional space, reducing the dimensionality to m. This represents a nonlinear mapping, where L represents the length of the time window. This represents the time decay weight at time k. This represents the welding state characteristics corresponding to time t. The welding state characteristics at time tk are represented by the time decay weight, which can be set based on historical experimental data analysis (details omitted here). The state transition characteristics at each time point can be obtained using the above method. Therefore, the data sequence composed of all state transition characteristics in chronological order can be used as the state transition feature sequence corresponding to each welding state characteristic. Finally, a state transition feature space can be constructed using the state transition feature sequences corresponding to all welding state characteristics. This involves aligning the state transition feature sequences corresponding to various welding state characteristics along a unified time axis and splicing and fusing them along the feature dimensions to construct a multi-dimensional, multi-channel state transition feature space.

[0056] In some embodiments, the method further includes: acquiring the signal-to-noise ratio and historical stability contribution of each sensing channel during the welding process; and determining the confidence weight of each sensing channel based on the corresponding signal-to-noise ratio and historical stability contribution.

[0057] In practical implementation, firstly, during the welding process, the signals acquired by each sensor channel are statistically analyzed and their quality assessed to obtain the signal-to-noise ratio (SNR) of each sensor channel. The SNR measures the relative relationship between the effective signal strength and noise interference during data acquisition in the welding process, reflecting the reliability and availability of the channel data. Simultaneously, combining historical welding operation data and long-term monitoring results, the historical stability contribution of each sensor channel is calculated using existing technology. This historical stability contribution quantifies the contribution of each sensor channel to the stability and accuracy of the overall monitoring results in the construction of historical welding state characteristics and anomaly identification, reflecting the long-term reliability of the channel. Then, the reliability weight of each sensor channel can be determined based on the corresponding SNR and historical stability contribution. In actual implementation, the reliability weight can be determined in the following way:

[0058]

[0059] in, This represents the confidence weight of the i-th sensing channel. and These represent the signal-to-noise ratios of the i-th and j-th sensing channels, respectively. and Let represent the historical stable contribution of the i-th and j-th sensing channels respectively, and N represent the total number of sensing channels. The confidence weight of each sensing channel can be obtained through the above method.

[0060] In some embodiments, using the state transition feature space for evolutionary modeling to obtain the welding state evolution trajectory of the welding process can be specifically achieved in the following manner:

[0061] The state transition feature sequences corresponding to each welding state feature in the state transition feature space are weighted and fused according to the confidence weight of the corresponding sensing channel to obtain the fused feature sequence.

[0062] The fused feature sequence is mapped to the welding state evolution trajectory of the welding process using a long short-term memory network.

[0063] In practice, firstly, the state transition feature sequences corresponding to each welding state feature in the state transition feature space can be weighted and fused according to the credibility weight of the corresponding sensing channel. That is, the state transition feature sequences corresponding to each type of welding state feature in the state transition feature space are weighted according to the credibility weight of the sensing channel to which they belong. In this way, a fused feature sequence that comprehensively reflects the dynamic changes of multi-source information during the welding process can be obtained. This fused feature sequence retains the changing trends of various state features in the time dimension, while taking into account the signal reliability and historical stability of each sensing channel in the feature dimension. Then, the fused feature sequence can be mapped to the welding state evolution trajectory of the welding process using a Long Short-Term Memory (LSTM) network. This welding state evolution trajectory is a sequence of state paths describing the continuous change of welding states over time. This trajectory reflects the dynamic trend of the welding process evolving from a stable stage to a transitional stage, and then to an abnormal or defect-risk stage. In practice, the fused feature sequence is fed into the LSTM network model for temporal evolution modeling. This involves time-step expansion and normalization of the fused feature sequence to meet the LSTM network's requirements for temporal input data format and numerical range, and then progressively fed into the network's input layer according to the chronological order of the welding process. Within the network... The network adaptively regulates historical state information and current feature information through input gates, forget gates, and output gates. The input gate is responsible for filtering feature information that contributes to the evolution of welding state in the current time step, the forget gate is used to suppress historical information that is irrelevant to the current welding state or is noisy, and the output gate integrates the current memory unit state and new input features to generate a hidden state sequence that reflects the trend of welding state change. On this basis, the network uses its memory units to model the state transition law over a long period of time, realizing the temporal correlation learning of welding state stability changes, abnormal evolution trends, and stage features. Finally, the temporal hidden states generated by the output layer of the Long Short-Term Memory Network are mapped to interpretable welding state evolution trajectories.

[0064] It should be noted that by constructing a state transition feature space based on a set of welding state features and performing evolutionary modeling to obtain the welding state evolution trajectory, the analysis of multi-source sensor information during the welding process can be transformed from static feature analysis to a dynamic characterization of stability change trends, thereby achieving a systematic modeling of the temporal evolution law of welding states. By uniformly representing the transition behavior of various welding state features, stability shifts caused by welding arc fluctuations, changes in molten pool morphology, and uneven heat input can be captured more sensitively. Furthermore, the fusion and synergistic constraints of multi-source sensor information within the state transition feature space help reduce the risk of misjudgment caused by noise or failure of a single sensor channel, improving the robustness and reliability of welding stability assessment. At the same time, the welding state evolution trajectory provides a clear temporal evolution basis for anomaly risk prediction, enabling early identification of potential instability signs.

[0065] In step S103, the welding state category is identified based on the welding state evolution trajectory, and the deviation of the welding process under the state category is evaluated. The abnormal risk probability of the welding process is predicted by the deviation.

[0066] In some embodiments, identifying the welding state category based on the welding state evolution trajectory can be achieved in the following ways:

[0067] Multiple classification features are extracted from the welding state evolution trajectory;

[0068] All classification features are input into a Softmax classifier for classification, thereby identifying the welding state category.

[0069] In practice, firstly, multiple classification features can be extracted from the welding state evolution trajectory. That is, the welding state evolution trajectory is used as the input object, and multi-level analysis is performed on the welding state evolution trajectory in the time dimension and the state change dimension to extract a variety of classification features that can reflect the dynamic characteristics of the welding process, such as the overall trend features of the trajectory, the local fluctuation amplitude features, the state transition frequency features, the change rate features, and the stability index features, so that the extracted features can comprehensively depict the law of welding state evolution over time. Then, all classification features can be input into a Softmax classifier for classification. This involves normalizing and concatenating the extracted classification features to construct a unified feature vector representation. This ensures that features with different dimensions and distributions are comparable in the same feature space, forming structured classification input data. After feature organization, the feature vectors are input into the softmax classifier. The classifier calculates the posterior probability of each welding state based on the class weight parameters learned during the training phase and outputs the probability distribution of each state category through a probability normalization mechanism. Finally, the state category corresponding to the highest probability value is selected as the identification result of the current welding process, i.e., the welding state category, thereby achieving automatic discrimination of the welding state category.

[0070] In some embodiments, the deviation of the welding process under the stated state category is evaluated. Specifically, firstly, a reference baseline model can be constructed based on historical stable welding samples. This baseline model characterizes the statistical distribution range and typical evolution patterns of multi-dimensional state characteristics such as welding current, arc stability, molten pool morphology, temperature distribution, and vibration characteristics under normal operating conditions. Then, under the currently identified state category, the welding state evolution trajectory of the welding process is mapped to the feature space corresponding to the reference baseline model, and the degree of difference between the current welding state evolution trajectory and the standard welding state evolution trajectory is compared. That is, the offset distance between the trajectories is calculated, and this offset distance is used as the deviation of the welding process under that state category, where the deviation represents the degree of deviation of the current welding process relative to the baseline.

[0071] In some embodiments, the probability of abnormal risks in the welding process can be predicted by the deviation as follows:

[0072] Based on historical welding sample data and labeled abnormal event information, a mapping model between historical deviation and abnormal risk probability is constructed.

[0073] The deviation is input into the mapping model for prediction to obtain the probability of abnormal risks in the welding process.

[0074] In practical implementation, firstly, a training dataset can be constructed based on historical welding sample data. The deviation sequence calculated during historical welding processes is then matched one-to-one with the corresponding anomaly event labels. These anomaly event labels indicate whether quality defects, welding instability, or process abnormalities occur at a specific deviation level, thus forming a supervised learning sample with deviation as input and anomaly occurrence as output. Then, statistical analysis or parameter fitting can be performed on the sample data to construct a mapping model between historical deviation and anomaly risk probability. In actual implementation, this mapping model can be determined in the following way:

[0075]

[0076] in, D represents the probability of abnormal risk, and D represents the historical deviation. The weighting coefficient representing the sensitivity of deviation to abnormal risk is used to adjust the impact of changes in deviation on the rate of increase in risk probability. The model bias parameters represent the basic anomaly risk level under zero or low deviation conditions. `exp()` is an exponential function used to construct a smooth probability mapping curve, ensuring that the anomaly risk probability monotonically increases with increasing deviation. The weighting coefficients and model bias parameters can be set based on historical experimental data analysis, which will not be elaborated here. Finally, the deviation can be input into the mapping model for prediction to obtain the anomaly risk probability of the welding process.

[0077] It should be noted that identifying welding state categories based on welding state evolution trajectories, evaluating the deviation of the welding process within each category, and then predicting the probability of abnormal risks based on the deviation can elevate the dynamic changes in the welding process from static parameter monitoring to temporal state understanding and risk quantification assessment. This enables early perception and precise identification of welding stability degradation trends. The welding state evolution trajectory integrates the temporal characteristics of multi-source sensor information, which can not only identify whether the current welding belongs to different state categories such as stable, transitional, or abnormal, but also capture the trend changes of the state over time, thereby avoiding misjudgments caused by single instantaneous parameters. Mapping the deviation to the probability of abnormal risks can transform complex multi-dimensional state information into quantifiable and predictable risk indicators, thereby supporting real-time early warning and adaptive parameter control.

[0078] In step S104, the welding process is monitored and warned in real time based on the probability of abnormal risks, and welding parameter optimization instructions are output.

[0079] In practical implementation, firstly, the probability of abnormal risks in the welding process is used as the core criterion for real-time monitoring. This probability is dynamically compared with pre-set multi-level risk thresholds. When the probability of abnormal risks is within the normal range, regular monitoring is maintained. When the probability of abnormal risks exceeds the mild risk threshold, an early warning mechanism is triggered. When the probability of abnormal risks further exceeds the high risk threshold, a mandatory intervention warning is triggered to achieve tiered response and differentiated management. Then, after the warning is triggered, the causes of instability in the welding process are analyzed by combining the current welding state category, the welding state evolution trajectory, and the corresponding deviation characteristics. For example, it is determined that the abnormality originates from current fluctuations, arc instability, abnormal molten pool morphology, or cooling rate deviation, thereby clarifying the direction of parameter optimization. Furthermore, based on a pre-built welding process knowledge base and parameter optimization rule model, targeted welding parameter adjustment strategies are generated, including the optimization range and adjustment priority of key process parameters such as welding current, voltage, wire feed speed, welding speed, pulse duty cycle, or shielding gas flow rate. Finally, the welding parameter adjustment strategies are converted into executable welding parameter optimization instructions and sent to the welding equipment control system in real time through the control interface to achieve online adjustment of the welding process.

[0080] It should be noted that real-time monitoring and early warning of the welding process based on the probability of abnormal risks, and output of welding parameter optimization instructions, can organically integrate multi-source sensing, state recognition and process control to form a closed-loop control system from state perception to proactive intervention. This can upgrade welding stability management from passive post-event detection to proactive pre-event prevention, which not only improves the accuracy of abnormal risk prediction and the timeliness of response, but also enhances the robustness and self-recovery capability of the welding process under complex working conditions, effectively improving the consistency of welding quality and production reliability.

[0081] Therefore, this application firstly employs a method of constructing a state transition feature space based on a set of welding state features and performing evolutionary modeling to obtain the welding state evolution trajectory. This elevates the analysis of multi-source sensor information during the welding process from static feature analysis to a dynamic characterization of stability change trends, thereby achieving a systematic modeling of the temporal evolution law of the welding state. Secondly, by identifying the welding state category based on the welding state evolution trajectory and evaluating the deviation of the welding process under that category, and then predicting the probability of abnormal risks based on the deviation, the dynamic changes of the welding process can be upgraded from static parameter monitoring to temporal state understanding and risk quantification assessment. This enables early perception and precise identification of welding stability degradation trends. Finally, by implementing real-time monitoring and early warning of the welding process based on the probability of abnormal risks and outputting welding parameter optimization instructions, multi-source sensor perception, state identification, and process control can be organically integrated to form a closed-loop control system from state perception to proactive intervention. This upgrades welding stability management from passive post-event detection to proactive pre-event prevention, thereby achieving welding state category identification and abnormal risk prediction to improve the reliability of the welding process.

[0082] In summary, the technical solution adopted in this application can characterize the stability change trend of the welding process based on multi-source sensing information, thereby realizing the identification of welding state categories and the prediction of abnormal risks, so as to improve the reliability of the welding process.

[0083] Furthermore, in another aspect of this application, in some embodiments, this application provides a welding process condition monitoring system based on multi-source sensing, referring to... Figure 3 The figure is a schematic diagram of a welding process condition monitoring system based on multi-source sensing according to some embodiments of this application. The welding process condition monitoring system based on multi-source sensing includes:

[0084] The feature construction module 201 is used to simultaneously collect multi-source welding data during the welding process and construct a set of welding state features during the welding process based on the multi-source welding data.

[0085] The evolution modeling module 202 is used to characterize the stability change trend of the welding process based on the welding state feature set, obtain the state transition feature space, use the state transition feature space to perform evolution modeling, and then obtain the welding state evolution trajectory of the welding process.

[0086] The probability prediction module 203 is used to identify the welding state category based on the welding state evolution trajectory, evaluate the deviation of the welding process under the state category, and predict the probability of abnormal risk of the welding process through the deviation.

[0087] The early warning feedback module 204 is used to monitor and issue early warnings for the welding process in real time based on the probability of abnormal risks, and to output welding parameter optimization instructions.

[0088] In addition, this application also provides a computer device, the computer device including a memory and a processor, the memory storing code, the processor being configured to acquire the code and execute the above-described welding process status monitoring method based on multi-source sensing.

[0089] In some embodiments, reference Figure 4 The figure is a schematic diagram of the structure of a computer device implementing a welding process condition monitoring method based on multi-source sensing, according to some embodiments of this application. The welding process condition monitoring method based on multi-source sensing in the above embodiments can... Figure 4 The computer device shown is used to implement this, and the computer device includes at least one processor 301, a communication bus 302, a memory 303, and at least one communication interface 304.

[0090] The processor 301 can be a general-purpose central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more devices used to control the execution of the welding process status monitoring method based on multi-source sensing in this application.

[0091] The communication bus 302 can be used to transmit information between the aforementioned components.

[0092] The memory 303 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disks or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 303 may exist independently and be connected to the processor 301 via the communication bus 302. The memory 303 may also be integrated with the processor 301.

[0093] The memory 303 stores program code for executing the solution of this application, and its execution is controlled by the processor 301. The processor 301 executes the program code stored in the memory 303. The program code may include one or more software modules. In the above embodiments, the determination of the welding process status monitoring method based on multi-source sensing can be implemented by the processor 301 and one or more software modules in the program code in the memory 303.

[0094] Communication interface 304 uses any transceiver-like device for communicating with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0095] In a specific implementation, as one example, a computer device may include multiple processors, each of which may be a single-core (single-CPU) processor or a multi-core (multi-CPU) processor. Here, a processor may refer to one or more devices, circuits, and / or processing cores used to process data (e.g., computer program instructions).

[0096] The aforementioned computer device can be a general-purpose computer device or a special-purpose computer device. In specific implementations, the computer device can be a desktop computer, a portable computer, a network server, a handheld digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. This application does not limit the type of computer device.

[0097] In addition, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described welding process status monitoring method based on multi-source sensing.

[0098] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0099] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A welding process condition monitoring method based on multi-source sensing, characterized in that, Includes the following steps: Multi-source welding data is collected simultaneously during the welding process, and a set of welding state features during the welding process is constructed based on the multi-source welding data. Based on the welding state feature set, the stability change trend of the welding process is characterized by transitions to obtain a state transition feature space. The state transition feature sequences corresponding to each welding state feature in the state transition feature space are weighted and fused according to the confidence weight of the corresponding sensing channel to obtain a fused feature sequence. The fused feature sequence is mapped to the welding state evolution trajectory of the welding process through a long short-term memory network. The welding state evolution trajectory is a state path sequence used to describe the continuous change law of the welding state over time. Based on the welding state evolution trajectory, the welding state category is identified, and the deviation of the welding process under the state category is evaluated. Based on historical welding sample data and labeled abnormal event information, a mapping model between historical deviation and abnormal risk probability is constructed. The deviation is input into the mapping model for prediction to obtain the abnormal risk probability of the welding process. The welding process is monitored and warned in real time based on the probability of abnormal risks, and welding parameter optimization instructions are output.

2. The welding process status monitoring method based on multi-source sensing as described in claim 1, characterized in that, The multi-source welding data includes welding current data, molten pool infrared temperature field data, molten pool high-speed visual images, and welding torch and workpiece vibration signals.

3. The welding process status monitoring method based on multi-source sensing as described in claim 1, characterized in that, The set of welding state characteristics includes the average current, arc fluctuation index, short-circuit duty cycle, molten pool area, spatter density, molten pool center temperature, cooling rate, and vibration RMS value during the welding process.

4. The welding process status monitoring method based on multi-source sensing as described in claim 1, characterized in that, Based on the aforementioned set of welding state characteristics, the stability change trend of the welding process is characterized by transitions, resulting in a state transition feature space that specifically includes: For each type of welding state feature in the welding state feature set, obtain the sequence data corresponding to the welding state feature; Adjacent transition metrics are performed on the sequence data to obtain the state transition feature sequence corresponding to the welding state feature, and then the state transition feature sequence corresponding to each type of welding state feature is obtained. A state transition feature space is constructed by using the state transition feature sequences corresponding to all welding state features.

5. The welding process status monitoring method based on multi-source sensing as described in claim 1, characterized in that, Also includes: Obtain the signal-to-noise ratio and historical stability contribution of each sensor channel during the welding process; The reliability weight of each sensing channel is determined based on the corresponding signal-to-noise ratio and historical stability contribution.

6. The welding process status monitoring method based on multi-source sensing as described in claim 1, characterized in that, The welding state categories identified based on the welding state evolution trajectory specifically include: Multiple classification features are extracted from the welding state evolution trajectory; All classification features are input into a Softmax classifier for classification, thereby identifying the welding state category.

7. A welding process condition monitoring system based on multi-source sensing, used to execute a welding process condition monitoring method based on multi-source sensing as described in any one of claims 1 to 6, characterized in that, include: The feature construction module is used to simultaneously collect multi-source welding data during the welding process and construct a set of welding state features during the welding process based on the multi-source welding data. The evolution modeling module is used to characterize the stability change trend of the welding process based on the welding state feature set, obtain the state transition feature space, and use the state transition feature space to perform evolution modeling, thereby obtaining the welding state evolution trajectory of the welding process. The probability prediction module is used to identify the welding state category based on the welding state evolution trajectory, evaluate the deviation of the welding process under the state category, and predict the probability of abnormal risk of the welding process through the deviation. The early warning feedback module is used to monitor and issue early warnings for the welding process in real time based on the probability of abnormal risks, and to output welding parameter optimization instructions.

8. A computer device, characterized in that, The computer device includes a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the welding process status monitoring method based on multi-source sensing as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the welding process status monitoring method based on multi-source sensing as described in any one of claims 1 to 6.

Images