Intelligent regulation and control method and system for bus duct profile laser welding quality on-line monitoring

Abstract

The invention discloses an intelligent regulation and control method and system for bus duct profile laser welding quality on-line monitoring, and relates to the technical field related to welding regulation and control. The method comprises the steps that welding process parameters, weld joint detection and environment and equipment performance are collected, and a welding feature matrix is constructed; analyzing and determining the association strength between each influence factor and the welding quality index, identifying a quality dominant factor and an effective regulation and control chain, and forming a quality influence degree sequence; constructing a welding quality correlation regulation and control network, taking the dominant factors as core nodes, taking the regulation and control chain as a connection path, and taking the correlation strength as an edge weight; and dynamically generating and outputting an optimal welding process regulation and control strategy. The technical problems that in the prior art, a dynamic coupling relation of multi-source data is difficult to capture in real time in welding quality control, global optimization regulation and control over multiple influence factors are lacked, and consequently the stability and consistency of the welding quality are poor are solved, and online dynamic regulation and control over the welding process are achieved. And the welding quality stability and consistency are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of welding regulation, and particularly relates to an intelligent regulation method and system for online monitoring of laser welding quality of bus duct profiles. BACKGROUND

[0002] A bus duct is a kind of efficient power transmission equipment, and the welding quality of the bus duct profile, which is a core component of the bus duct, directly determines the conductive performance, mechanical strength and long-term operation reliability of the product. Laser welding has become an important process for welding bus duct profiles. However, the laser welding process is affected by multiple factors, including welding process parameters such as laser power, welding speed and defocusing amount, environmental conditions such as weld appearance characteristics, temperature and humidity, and equipment performance state. Fluctuations in these factors can easily lead to welding defects such as pores, cracks and incomplete penetration, which seriously affect product quality. At present, the quality control of laser welding of bus duct profiles mainly relies on offline detection and manual experience adjustment, lacks real-time and self-adaptive capabilities, and is difficult to fully capture dynamic changes in the welding process. Moreover, the complex nonlinear relationships between process parameters, weld detection data and environmental data have not been effectively explored, simple parameter adjustment cannot achieve global optimization, and may even cause other quality problems.

[0003] Therefore, in the related art, the welding quality control cannot capture the dynamic coupling relationship of multi-source data in real time and lacks global optimization regulation of multiple influencing factors, resulting in poor welding quality stability and consistency. SUMMARY

[0004] The present application provides an intelligent regulation method and system for online monitoring of laser welding quality of bus duct profiles, which solves the technical problem that the welding quality control in the prior art cannot capture the dynamic coupling relationship of multi-source data in real time and lacks global optimization regulation of multiple influencing factors, resulting in poor welding quality stability and consistency, and achieves the technical effect of realizing online dynamic regulation of the welding process and improving the welding quality stability and consistency.

[0005] This application provides an intelligent control method for online monitoring of the quality of laser welding of busbar profiles. The method includes: acquiring multi-source welding data including welding process parameter data, weld inspection data, environmental interference data, and equipment performance data, and establishing a welding feature matrix; based on the welding feature matrix, determining the correlation strength values ​​between multiple influencing factors and M welding quality indicators, identifying the quality-dominant factors and effective control chains, and setting a quality influence degree sequence; using the quality-dominant factors as core nodes for hierarchical division, using the effective control chains as directional connection paths for network nodes, using the multiple influencing factors and the associated N welding quality indicators as network nodes, and using the correlation strength values ​​as edge weights to construct a welding quality correlation control network, where M is greater than N; and based on the welding quality correlation control network and combined with the quality influence degree sequence, outputting a dynamic process control strategy for laser welding of busbar profiles.

[0006] In a possible implementation, the intelligent control method for online monitoring of the laser welding quality of the busbar profile further performs the following processing: extracting current and voltage features from the welding process parameter data in the multi-source welding data to determine the welding energy features; extracting defect morphology features from the weld detection data in the multi-source welding data to determine the welding forming features; and obtaining the welding feature matrix based on the welding energy features and the welding forming features.

[0007] In a possible implementation, the intelligent control method for online monitoring of the laser welding quality of the busbar profile further performs the following processing: determining the cross-features of environmental interference data and equipment performance data in the multi-source welding data, wherein the cross-features include the coupling factor between ambient temperature and weld cooling rate, the adaptation index between airflow velocity and molten pool stability, and the dynamic response coefficient between laser power attenuation and welding depth; and standardizing the welding energy features, the welding forming features, and the cross-features to obtain the welding feature matrix.

[0008] In a possible implementation, the intelligent control method for online monitoring of the laser welding quality of the busbar profile further performs the following processing: the forward propagation process uses a multilayer perceptron, the input layer is the welding feature matrix, and the hidden layer undergoes nonlinear transformation through the ReLU activation function; when analyzing the correlation strength values ​​between multiple influencing factors and M welding quality indicators, a weighted fusion of Pearson correlation coefficient and mutual information entropy is introduced; based on the weighted fusion result, the N welding quality indicators are selected and determined.

[0009] In a possible implementation, the intelligent control method for online monitoring of the laser welding quality of the busbar profile further performs the following processing: hierarchical clustering grouping based on the quality-dominant factors, combined with correlation strength filtering threshold screening, to discover potential control chains; and time-series verification of the potential control chains to obtain the effective control chains.

[0010] In a possible implementation, the intelligent control method for online monitoring of the laser welding quality of busbar profiles further performs the following processing: determining the inter-class distance based on the dominant quality factors, and dividing the data into multiple factor components and multiple factor group classes based on the inter-class distance; performing path analysis based on the multiple factor components and multiple factor group classes, and traversing the multilayer perceptron through a depth-first search, marking a potential control chain when the association strength values ​​of P consecutive network nodes in the quality control transmission process all meet the association strength filtering threshold, where P is greater than or equal to 2.

[0011] In a possible implementation, the intelligent control method for online monitoring of the laser welding quality of the busbar profile further performs the following processing: obtaining the contribution of network nodes on the potential control chain; based on the contribution of the network nodes on the chain, obtaining the cumulative value of the product of the time difference and the contribution of adjacent network nodes in the potential control chain; when the cumulative value of the product of the time difference and the contribution meets the timing verification threshold, and the control order of the nodes on the chain conforms to the laser welding reaction logic, the timing verification is determined to be passed.

[0012] In a possible implementation, the intelligent control method for online monitoring of the laser welding quality of the busbar profile further performs the following processing: if the cumulative value of the product of the time difference and the contribution does not meet the time-series verification threshold, a two-way traceability mechanism is used to deduce an expanded potential control chain; the expanded potential control chain is time-series verified by superimposing the cumulative value corresponding to the expanded potential control chain with the time-series compensation value and the time-series verification threshold; if it still does not meet the time-series verification threshold, the control chain is split and reorganized: the expanded potential control chain is split into multiple sub-chains according to the peak value of the time difference, and multiple sub-chains that pass the time-series verification are retained; for multiple sub-chains that fail the time-series verification, a secondary verification is performed in combination with the spatial defect correlation of the weld detection data.

[0013] In a possible implementation, the intelligent control method for online monitoring of the laser welding quality of the busbar profile further performs the following processing: tracing back the preceding influencing factors of the first node of the chain to determine the correlation strength compensation value with the first node; extrapolating the derivative control effect of the last node of the chain to generate a potential quality improvement coefficient; and weighting and fusing the correlation strength compensation value corresponding to the first node of the chain and the potential quality improvement coefficient corresponding to the last node of the chain to obtain the time-series compensation value.

[0014] This application also provides an intelligent control system for online monitoring of the laser welding quality of busbar profiles. The system includes: a welding data acquisition module, used to acquire multi-source welding data including welding process parameter data, weld inspection data, environmental interference data, and equipment performance data, and establish a welding feature matrix; a correlation strength determination module, used to determine the correlation strength values ​​between multiple influencing factors and M welding quality indicators based on the welding feature matrix, identify the quality-dominant factors and effective control chains, and set a quality influence degree sequence; a control network construction module, used to perform hierarchical division by using the quality-dominant factors as core nodes, using the effective control chains as directional connection paths for network nodes, using the multiple influencing factors and the associated N welding quality indicators as network nodes, and using the correlation strength values ​​as edge weights to construct a welding quality correlation control network, where M is greater than N; and a control strategy output module, used to output a dynamic process control strategy for laser welding of busbar profiles based on the welding quality correlation control network and the quality influence degree sequence.

[0015] This application proposes an intelligent control method and system for online monitoring of laser welding quality of busbar profiles. This method collects welding process parameters, weld inspection data, environmental and equipment performance data, and constructs a welding feature matrix. It analyzes and determines the correlation strength between various influencing factors and welding quality indicators, identifies the dominant quality factors and effective control chains, and forms a quality influence sequence. A welding quality correlation control network is constructed, with dominant factors as core nodes, control chains as connection paths, and correlation strength as edge weights. The optimal welding process control strategy is dynamically generated and output. This solves the technical problems in existing technologies where welding quality control struggles to capture the dynamic coupling relationships of multi-source data in real time and lacks global optimization control of multiple influencing factors, leading to poor welding quality stability and consistency. It achieves online dynamic control of the welding process, improving the stability and consistency of welding quality. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings of the embodiments of this disclosure will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0017] Figure 1 This is a schematic diagram of the intelligent control method for online monitoring of the laser welding quality of busbar profiles provided in this application embodiment.

[0018] Figure 2This is a schematic diagram of the intelligent control system for online monitoring of the laser welding quality of busbar profiles provided in this application embodiment.

[0019] Figure labeling: Welding data acquisition module 10, correlation strength determination module 20, control network construction module 30, control strategy output module 40. Detailed Implementation

[0020] To further illustrate the technical means and effects adopted by the present invention in order to achieve the intended purpose, the following detailed description is provided in conjunction with the accompanying drawings and preferred embodiments, based on the specific implementation methods, structures, features and effects of the present invention.

[0021] This application provides an intelligent control method for online monitoring of the laser welding quality of busbar profiles, such as... Figure 1 As shown, the method includes: Step S100: Obtain multi-source welding data, including welding process parameter data, weld inspection data, environmental interference data, and equipment performance data, and establish a welding feature matrix.

[0022] Preferably, multi-source welding data is acquired, including welding process parameter data, weld inspection data, environmental interference data, and equipment performance data. Welding process parameter data refers to the core process parameters directly set to maintain the laser welding process and their real-time feedback values, including laser parameters such as welding current / voltage, laser power, pulse frequency and duty cycle, and beam mode; motion parameters such as welding speed and scanning path; and auxiliary process parameters such as shielding gas type, shielding gas flow rate, and wire feed speed. Weld inspection data refers to the physical information and geometric morphology data about the weld formation process acquired through online monitoring equipment, including molten pool images, plasma / feather images, and weld tracking images. Visual sensing data includes optical / spectral data such as back-reflected light intensity and plasma spectral signals; acoustic emission data during welding; and initial weld morphology data such as weld width and weld reinforcement height. Environmental interference data refers to variables in the welding workshop environment that may affect the stability and quality of the welding process, including ambient temperature and humidity, foundation vibration, airflow velocity, and power grid fluctuations. Equipment performance data refers to the status and performance parameters of the hardware equipment performing the welding task, including laser status such as output power stability, laser internal temperature, and output head lens contamination; motion system status such as servo motor load / torque, guide rail wear, or positioning accuracy; and cooling system status such as coolant temperature, flow rate, and pressure.

[0023] Furthermore, step S100 also includes step S110, extracting current and voltage features from the welding process parameter data in the multi-source welding data to determine the welding energy features; step S120, extracting defect morphology features from the weld detection data in the multi-source welding data to determine the welding forming features; and step S130, obtaining the welding feature matrix based on the welding energy features and the welding forming features.

[0024] Preferably, real-time current and voltage waveform time-series data of the laser power supply are obtained from the welding process parameter data of multi-source welding data, and current and voltage features are extracted to reflect energy stability characteristics. Specifically, the average power is obtained by calculating the product of the average voltage and the average current, and then the standard deviation of the power sequence is calculated to reflect the fluctuation of energy input during welding, which is used as the basic energy index. Then, the current / voltage ripple coefficient is calculated to characterize the smoothness of the power supply output. The rise time, fall time, and peak value of the pulse are extracted from the current waveform to maintain stability. At the same time, within a specific time window, the number and amplitude of current or voltage spikes exceeding the preset stability threshold are counted to obtain the dynamic characteristics of the current / voltage waveform to reflect the working state and stability of the laser. Then, these current and voltage features are combined to form welding energy characteristics.

[0025] Preferably, image / point cloud data of the molten pool or weld seam collected by high-speed cameras, laser vision sensors, etc., are obtained from the weld seam inspection data of multi-source welding data, and defect morphology features are extracted. Specifically, geometric features such as weld width, weld depth, reinforcement height, undercut depth, and width-to-depth ratio are extracted from the cross-sectional or frontal images of the weld seam. Then, the area / perimeter of the molten pool region is extracted from the molten pool image, the position change of the centroid of the molten pool in consecutive frames of images is calculated, and the number of spatter points, total pixel area, and distribution location identified in the images are statistically analyzed. In the formed weld seam image, porosity / depression features are detected to determine defect and stability features. Then, these features are combined to generate weld forming features. Finally, the welding energy features and weld forming features are concatenated into a longer total feature vector and standardized to convert its dimensions and numerical range to a unified range, thereby constructing a welding feature matrix. In this matrix, the rows represent different data samples, the columns represent all welding features, and each element in the matrix corresponds to a specific welding attribute.

[0026] Furthermore, step S130 also includes step S131, determining the cross-features of environmental interference data and equipment performance data in the multi-source welding data, wherein the cross-features include the coupling factor between ambient temperature and weld cooling rate, the adaptation index between airflow velocity and molten pool stability, and the dynamic response coefficient between laser power attenuation and welding depth; step S132, standardizing the welding energy features, the welding forming features, and the cross-features to obtain the welding feature matrix.

[0027] Preferably, the cross-characteristics of environmental interference data and equipment performance data in multi-source welding data are determined, including the coupling factor between ambient temperature and weld cooling rate, the compatibility index between gas flow velocity and molten pool stability, and the dynamic response coefficient between laser power attenuation and welding depth. Specifically, ambient temperature directly affects the cooling rate of weld metal; high ambient temperature results in slow cooling, while low ambient temperature results in fast cooling. The cooling rate directly affects the metallographic structure of the metal, thereby affecting the strength and toughness of the weld. Specifically, the actual cooling rate is calculated by analyzing the cooling process data of the molten pool and heat-affected zone captured by a high-speed thermal imager. Within a time window, multiple sets of ambient temperature sensor readings and actual cooling rate data pairs are collected. Based on a regression model, the theoretical influence of ambient temperature on the cooling rate under ideal and stable conditions is established, and the coupling factor is determined. A coupling factor of approximately 1 indicates that the cooling process is mainly affected by ambient temperature, which is in line with theoretical expectations and the process is stable. A coupling factor greater than 1 indicates that the actual cooling is faster than expected, which may suggest other unconsidered strong heat dissipation factors. A coupling factor less than 1 indicates that the actual cooling is slower than expected, which may mean that improper shielding gas flow affects heat dissipation or abnormal laser energy mode.

[0028] Preferably, the airflow in the welding workshop interferes with the coverage effect of the protective gas, causing the molten pool and the surrounding metal to come into contact with the air at high temperatures, resulting in oxidation, spatter, porosity, and affecting the flow of the molten pool, thereby compromising stability. Specifically, the workshop airflow velocity measured by the wind speed sensor and the molten pool stability index extracted from the molten pool image are used as input data. The correlation between the two time series is calculated using the Pearson correlation coefficient to determine the fit index. A fit index close to 1 indicates that the current airflow velocity is compatible with the requirements for maintaining molten pool stability, while a fit index close to 0 indicates that the current airflow velocity is not compatible with stable welding and has caused observable interference to the molten pool. As lasers age, aging or contamination of internal optical components can lead to a significant reduction in output power, potentially resulting in insufficient weld penetration. Specifically, laser power attenuation data is calculated, and the weld depth is measured or estimated in real-time using a line-scan profilometer as input. A simple mapping relationship is then established, and a sliding window regression method is used to estimate and determine the dynamic response coefficient in real-time. This coefficient quantifies the impact of even minor laser performance degradation on weld depth. A larger dynamic response coefficient indicates that the system is highly sensitive to laser attenuation and requires close monitoring, while a smaller coefficient suggests that the impact can be mitigated; for example, if a shallower weld depth is predicted, a slight increase in laser power can be used as compensation. Finally, Z-score normalization is used to standardize the cross-features, which are then concatenated with the weld energy and weld formation features to form an updated weld feature matrix, ensuring intelligent early warning and proactive control.

[0029] Step S200: Based on the welding feature matrix, determine the correlation strength values ​​between multiple influencing factors and M welding quality indicators, identify the dominant quality factors and effective control chains, and set the quality influence degree sequence.

[0030] Step S200 further includes step S210, in which the forward propagation process employs a multilayer perceptron, with the input layer being the welding feature matrix and the hidden layer undergoing nonlinear transformation using the ReLU activation function; step S220, in analyzing the correlation strength values ​​between multiple influencing factors and M welding quality indicators, a weighted fusion of Pearson correlation coefficient and mutual information entropy is introduced; step S230, based on the weighted fusion result, the N welding quality indicators are selected and determined.

[0031] Preferably, the forward propagation process employs a multilayer perceptron. Forward propagation refers to the process of calculating the welding feature matrix data layer by layer from the input layer through the hidden layers to finally obtain the predicted values ​​of M quality indicators. The input layer is the welding feature matrix, and the number of nodes is the number of features in the welding feature matrix. The hidden layer contains multiple neurons, that is, each neuron performs a weighted summation of all inputs from the previous layer and performs a nonlinear transformation through the ReLU activation function. The number of nodes in the output layer is the initial M welding quality indicators, and each node outputs a continuous predicted value, representing the predicted score of a certain quality indicator corresponding to the sample. The ReLU activation function enables the network to learn complex patterns such as "only when features A and B simultaneously exceed a certain threshold will they have a significant impact on quality indicator C".

[0032] Preferably, after the complex mapping of the multilayer perceptron, the correlation strength value between each influencing factor and each quality indicator is learned using the Pearson correlation coefficient and mutual information entropy measurement method. Specifically, the Pearson correlation coefficient is used to calculate the linear correlation between each feature in the welding feature matrix and each actual measured quality indicator; the mutual information entropy is used to calculate the mutual information value between each feature and each quality indicator, measuring the amount of information shared between the two variables and capturing all statistical dependencies; then, the Pearson correlation coefficient and mutual information entropy of each feature-quality indicator pair are weighted and fused to obtain a weighted fusion result. Then, the most representative and information-rich N core quality indicators are selected from the M initial quality indicators to obtain the correlation strength matrix, where M and N are both positive integers, and M is greater than N, the rows are all influencing factors, the columns are all initial quality indicators, and the elements are the corresponding correlation strength values.

[0033] Furthermore, step S200 also includes step S240, which involves using hierarchical clustering to group the factors that dominate the quality, and combining this with a correlation strength filtering threshold to screen potential regulatory chains; step S250, which involves performing time-series verification on the potential regulatory chains to obtain the effective regulatory chains.

[0034] Step S240 further includes step S241, determining the inter-class distance based on the quality dominant factor, and dividing the data into multiple factor components and multiple factor group classes based on the inter-class distance; step S242, performing path analysis based on the multiple factor components and multiple factor group classes, traversing the multilayer perceptron through depth-first search, and marking a potential control chain when the association strength values ​​of P consecutive network nodes in the quality regulation transmission process all meet the association strength filtering threshold, wherein P is greater than or equal to 2.

[0035] Furthermore, the core influencing factors with the greatest impact on the final product quality, such as laser power, shielding gas flow rate, and welding speed, identified through correlation strength analysis, are designated as the dominant quality factors. Hierarchical clustering is then applied to these dominant quality factors, classifying them according to the similarity of their influence patterns. For example, laser power and welding speed are found to consistently affect both penetration depth and weld width, thus classifying them as energy input groups based on their similar behavior patterns. Ambient temperature and cooling water temperature are grouped into cooling condition groups. Then, a minimum correlation strength value is set as the correlation strength filtering threshold based on historical data. This threshold is used to filter connections, obtaining all connections with correlation strengths higher than the threshold. Finally, all influencing factors are grouped with the selected N... Quality indicators are treated as nodes, and correlations exceeding a threshold are treated as directed edges to form a directed weighted network. Starting from an initial influencing factor, a depth-first search is used to traverse the multilayer perceptron to find a continuous path that can eventually reach a certain quality indicator. That is, when the correlation strength values ​​of P consecutive network nodes in the quality control transmission process all meet the correlation strength filtering threshold, it is defined as a potential control chain. It must satisfy that the correlation strength between P consecutive nodes on the path exceeds the preset filtering threshold, where P is a positive integer greater than or equal to 2. Finally, a quality influence sequence is set, that is, a priority list is set when multiple quality indicators are abnormal or need to be optimized and controlled. For example, for busbar trunking, the importance of its conductive loop resistance and insulation strength is far higher than the aesthetics of the surface weld.

[0036] Furthermore, step S250 also includes step S251, obtaining the contribution of the on-chain network nodes on the potential control chain; step S252, obtaining the cumulative value of the product of the time difference and the contribution of adjacent on-chain network nodes in the potential control chain based on the contribution of the on-chain network nodes; step S253, when the cumulative value of the product of the time difference and the contribution meets the timing verification threshold, and the control order of the on-chain nodes meets the laser welding reaction logic, the timing verification is determined to be passed.

[0037] Preferably, the potential control chains are subjected to time-series verification. This involves using the importance of nodes on the potential control chain as the contribution of the network nodes, then calculating the average time interval between significant changes in the state of adjacent nodes on the chain by analyzing the timestamps of high-frequency sensor data. The cumulative value of the product of the time difference and the contribution of adjacent nodes is then calculated to comprehensively quantify the speed of the control response and the strength of the control effect. Chains with high scores are considered both fast and effective control paths. Finally, a minimum time-series verification score is set as the time-series verification threshold. When the cumulative value of the product of the time difference and the contribution of a potential control chain meets this threshold and the time differences of all adjacent nodes on the chain conform to the physical laser welding reaction logic, the time-series verification is considered successful. All successful potential control chains are then confirmed as effective control chains, and the system is constructed. Furthermore, step S250 also includes step S254: if the cumulative value of the product of the time difference and the contribution does not meet the time series verification threshold, a two-way traceability mechanism is used to deduce the extended potential control chain; step S255: the extended potential control chain is time-tested by superimposing the cumulative value corresponding to the extended potential control chain with the time series compensation value and the time series verification threshold. If it still does not meet the time series verification threshold, the control chain is split and reorganized: the extended potential control chain is split into multiple sub-chains according to the peak value of the time difference, and multiple sub-chains that pass the time series verification are retained; for multiple sub-chains that fail the time series verification, a secondary verification is performed in combination with the spatial defect correlation of the weld detection data.

[0038] Step S254 further includes: tracing back the preceding influencing factors of the first node of the chain to determine the correlation strength compensation value with the first node of the chain; extrapolating the derivative control effect of the last node of the chain to generate a potential quality improvement coefficient; and weighting and fusing the correlation strength compensation value corresponding to the first node of the chain and the potential quality improvement coefficient corresponding to the last node of the chain to obtain the time-series compensation value.

[0039] Preferably, if the cumulative value of the product of time difference and contribution does not meet the time series verification threshold, it indicates that the potential control chain may be incomplete and has missing links. In this case, a two-way tracing mechanism is used for deduction, including forward and backward deduction. Forward deduction involves finding the precursor influencing factors of the first node of the chain and determining the correlation strength complement value with the first node. For example, if it is laser power, its precursor may be the internal temperature of the laser or the voltage fluctuation of the upstream power grid. Multiple precursor nodes are found and their correlation strength complement values ​​with the first node are calculated to form an expanded potential control chain. Backward deduction involves predicting the derivative control effects of the last node of the chain and generating a potential quality improvement coefficient. For example, if C is the weld penetration, its derivative effect may be weld strength or porosity. Multiple derivative nodes are predicted and the potential quality improvement coefficient is estimated. Then, the correlation strength complement value corresponding to the first node of the chain and the potential quality improvement coefficient corresponding to the last node of the chain are weighted and fused to form an expanded potential control chain.

[0040] Preferably, the extended potential regulatory chain undergoes time-series verification. This involves weighted fusion of the correlation strength compensation value obtained from forward tracing and the potential quality improvement coefficient obtained from backward extrapolation to obtain a time-series compensation value. This quantifies the additional regulatory potential brought about by the extension of the regulatory chain. The accumulated value corresponding to the extended potential regulatory chain, combined with the time-series compensation value and a time-series verification threshold, is then used for verification. If the sum of the original accumulated value and the time-series compensation value of the extended chain meets the time-series verification threshold, the time-series verification passes. If, after extension and compensation, the regulatory chain still fails the time-series verification, it indicates that the long chain may be invalid or incorrectly spliced ​​together from multiple independent short chains. In this case, the relationships between all adjacent nodes on the extended chain are analyzed. The time difference is identified, and the peak time difference is cut off at this point. A large time difference indicates the presence of undiscovered hidden variables or two essentially independent physical processes. Multiple sub-chains that pass the time-series verification are retained. For multiple sub-chains that fail the time-series verification, secondary verification is performed by combining the spatial defect correlation of weld inspection data. This includes calling X-ray or ultrasonic inspection data of the weld to analyze the three-dimensional spatial distribution of porosity defects in the weld. If it is found that porosity is concentrated in the area of ​​insufficient penetration, it provides strong spatial evidence for the control chain, enabling it to pass the time-series verification, thus ensuring that reliable and effective control knowledge is extracted to the maximum extent.

[0041] Step S300: Using the quality-dominant factors as core nodes, a hierarchical division is performed. The effective control chain is used as the directional connection path of the network nodes. The multiple influencing factors and the associated N welding quality indicators are used as network nodes, and the association strength value is used as the edge weight to construct a welding quality association control network, where M is greater than N.

[0042] Preferably, the quality-dominant factors are used as core nodes for hierarchical division. That is, nodes are divided into different levels according to their roles and positions in the network. Specifically, the root node / input layer consists of the outermost environmental interference data and equipment performance data that are not easily controlled directly, such as ambient temperature, power grid fluctuations, and laser aging. The core layer / hidden layer consists of quality-dominant factors, which serve as the hub connecting causes and results and are also the direct objects of control actions. The leaf nodes / output layer consists of the N welding quality indicators that need to be guaranteed, such as penetration depth, porosity, and weld strength, which serve as the targets of network control. The effective control chain is then used as the directional connection path of the network nodes. That is, the effective control chain is intuitively represented by directed edges, and the direction of the arrow represents the transmission direction of the control effect or cause to the result, so as to clarify the connection relationship and influence direction of the nodes. The network nodes are N core welding quality indicators associated with multiple influencing factors including quality-dominant factors, environmental interference, and equipment performance. The edge weight is the correlation strength value. The larger the correlation strength value, the larger the edge weight, indicating that the relationship between the two nodes is closer and the influence is greater. Finally, a welding quality correlation control network is constructed, which is a structured welding process knowledge graph. It clearly shows the core control nodes, influence paths and influence strength. It can simulate the influence of different control schemes and finally select the optimal path to implement precise control, thereby improving the stability and consistency of welding quality.

[0043] Step S400: Based on the welding quality correlation control network and combined with the quality influence degree sequence, output the dynamic process control strategy for laser welding of busbar profiles.

[0044] Preferably, the core quality indicators of the current welding process are compared with standard values ​​in real time to identify indicators that have deteriorated or are at risk of deterioration. These indicators are then prioritized based on their quality impact sequence, addressing the issues with the highest impact first. For example, insufficient penetration affecting electrical performance is addressed first, followed by porosity affecting long-term reliability. Then, using the quality indicators as target nodes, all effective control chains that can affect these nodes are searched in reverse within the welding quality correlation control network. The control potential and cost of each path are evaluated by calculating the correlation strength value of the impact paths. The minimum product of correlation strengths along a path reflects the potential for process control through that path. Parameters that may accelerate equipment wear or introduce new defects are set as cost coefficients for evaluation. Specific adjustment schemes are calculated for each path, and the strategy with the highest comprehensive score in effectiveness, efficiency, stability, and low cost is selected as the dynamic process control strategy output for laser welding of busbar profiles. This output includes at least the instruction type, adjustment parameters / direction, adjustment range, and strategy priority, thereby ensuring online dynamic control of the welding process and improving welding quality stability and consistency.

[0045] In the above text, refer to Figure 1This paper describes in detail an intelligent control method for online monitoring of the laser welding quality of busbar profiles according to an embodiment of the present invention. Next, we will refer to... Figure 2 This invention describes an intelligent control system for online monitoring of the laser welding quality of busbar profiles according to an embodiment of the present invention.

[0046] The intelligent control system for online monitoring of laser welding quality of busbar profiles according to embodiments of the present invention addresses the technical problems in existing technologies, such as the difficulty in real-time capture of the dynamic coupling relationship of multi-source data and the lack of global optimization control of multiple influencing factors, leading to poor welding quality stability and consistency. It achieves the technical effect of realizing online dynamic control of the welding process and improving welding quality stability and consistency. Figure 2 As shown, the intelligent control system for online monitoring of laser welding quality of busbar profiles includes: welding data acquisition module 10, correlation strength determination module 20, control network construction module 30, and control strategy output module 40.

[0047] The welding data acquisition module 10 is used to acquire multi-source welding data, including welding process parameter data, weld inspection data, environmental interference data, and equipment performance data, and establish a welding feature matrix. The correlation strength determination module 20 is used to determine the correlation strength values ​​between multiple influencing factors and M welding quality indicators based on the welding feature matrix, identify the quality-dominant factors and effective control chains, and set the quality influence degree sequence. The control network construction module 30 is used to perform hierarchical division by using the quality-dominant factors as core nodes, using the effective control chains as directional connection paths for network nodes, using the multiple influencing factors and the associated N welding quality indicators as network nodes, and using the correlation strength value as the edge weight to construct a welding quality correlation control network, where M is greater than N. The control strategy output module 40 is used to output a dynamic process control strategy for laser welding of busbar profiles based on the welding quality correlation control network and the quality influence degree sequence.

[0048] The specific configuration of the welding data acquisition module 10 will be described in detail below. The welding data acquisition module 10 further includes: extracting current and voltage features from the welding process parameter data in the multi-source welding data to determine the welding energy features; extracting defect morphology features from the weld detection data in the multi-source welding data to determine the weld forming features; and obtaining the welding feature matrix based on the welding energy features and the weld forming features.

[0049] The specific configuration of the welding data acquisition module 10 will be described in detail below. The welding data acquisition module 10 further includes: determining the cross-features of environmental interference data and equipment performance data in the multi-source welding data, wherein the cross-features include the coupling factor between ambient temperature and weld cooling rate, the adaptation index between gas flow velocity and molten pool stability, and the dynamic response coefficient between laser power attenuation and welding depth; and standardizing the welding energy features, the welding forming features, and the cross-features to obtain the welding feature matrix.

[0050] The specific configuration of the correlation strength determination module 20 will be described in detail below. The correlation strength determination module 20 further includes: a multilayer perceptron used in the forward propagation process, with the welding feature matrix as the input layer and the hidden layer undergoing nonlinear transformation using the ReLU activation function; when analyzing the correlation strength values ​​between multiple influencing factors and M welding quality indicators, a weighted fusion of Pearson correlation coefficient and mutual information entropy is introduced; based on the weighted fusion result, the N welding quality indicators are selected and determined.

[0051] The specific configuration of the association strength determination module 20 will be described in detail below. The association strength determination module 20 further includes: using hierarchical clustering grouping based on the quality dominant factors, combined with association strength filtering threshold screening, to mine potential regulatory chains; performing time-series verification on the potential regulatory chains to obtain the effective regulatory chains.

[0052] The specific configuration of the association strength determination module 20 will be described in detail below. The association strength determination module 20 further includes: determining the inter-class distance based on the dominant quality factor; dividing the data into multiple factor components and multiple factor group classes based on the inter-class distance; performing path analysis based on the multiple factor components and multiple factor group classes; traversing the multilayer perceptron using a depth-first search; and marking a potential control chain when the association strength values ​​of P consecutive network nodes in the quality control transmission process all meet the association strength filtering threshold, where P is greater than or equal to 2.

[0053] The specific configuration of the association strength determination module 20 will be described in detail below. The association strength determination module 20 further includes: obtaining the contribution of the network nodes on the potential control chain; based on the contribution of the network nodes on the chain, obtaining the cumulative value of the product of the time difference and the contribution of adjacent network nodes on the potential control chain; when the cumulative value of the product of the time difference and the contribution meets the timing verification threshold, and the control order of the nodes on the chain meets the laser welding reaction logic, the timing verification is determined to be passed.

[0054] The specific configuration of the correlation strength determination module 20 will be described in detail below. The correlation strength determination module 20 further includes: if the cumulative value of the product of the time difference and the contribution does not meet the time-series verification threshold, a bidirectional tracing mechanism is used to deduce an expanded potential control chain; the expanded potential control chain is time-series verified by superimposing the cumulative value corresponding to the expanded potential control chain with the time-series compensation value and the time-series verification threshold; if it still does not meet the time-series verification threshold, the control chain is split and reorganized: the expanded potential control chain is split into multiple sub-chains according to the peak value of the time difference, and multiple sub-chains that pass the time-series verification are retained; for multiple sub-chains that fail the time-series verification, a secondary verification is performed based on the spatial defect correlation of the weld detection data.

[0055] The specific configuration of the association strength determination module 20 will be described in detail below. The association strength determination module 20 further includes: determining the association strength complement value with the chain head node by tracing back the preceding influencing factors of the chain head node; extrapolating the derivative control effect of the chain tail node to generate a potential quality improvement coefficient; and weightedly fusing the association strength complement value corresponding to the chain head node and the potential quality improvement coefficient corresponding to the chain tail node to obtain the time-series compensation value.

[0056] The intelligent control system for online monitoring of laser welding quality of busbar profiles provided in this embodiment of the invention can execute the intelligent control method for online monitoring of laser welding quality of busbar profiles provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0057] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. An intelligent control method for online monitoring of laser welding quality of busbar trunking profiles, characterized in that, The method includes: Acquire multi-source welding data, including welding process parameter data, weld inspection data, environmental interference data, and equipment performance data, and establish a welding feature matrix; Based on the welding feature matrix, the correlation strength values ​​between multiple influencing factors and M welding quality indicators are determined, the dominant quality factors and effective control chains are identified, and a quality influence degree sequence is set. The quality-dominant factors are used as core nodes for hierarchical division, and the effective control chain is used as the directional connection path of the network nodes. The multiple influencing factors and the associated N welding quality indicators are used as network nodes, and the association strength value is used as the edge weight to construct a welding quality association control network, where M is greater than N. Based on the welding quality correlation control network and combined with the quality influence degree sequence, a dynamic process control strategy for laser welding of busbar profiles is output.

2. The intelligent control method for online monitoring of laser welding quality of busbar profiles as described in claim 1, characterized in that, The method for establishing a welding feature matrix includes: Current and voltage features are extracted from the welding process parameter data in the multi-source welding data to determine the welding energy characteristics; Defect morphology features are extracted from the weld inspection data in the multi-source welding data to determine the welding forming characteristics; The welding feature matrix is ​​obtained based on the welding energy characteristics and the welding forming characteristics.

3. The intelligent control method for online monitoring of laser welding quality of busbar profiles as described in claim 2, characterized in that, Based on the welding energy characteristics and the welding forming characteristics, the welding characteristic matrix is ​​obtained, and the method includes: The cross-features of environmental interference data and equipment performance data in the multi-source welding data are determined. The cross-features include the coupling factor between ambient temperature and weld cooling rate, the compatibility index between airflow velocity and molten pool stability, and the dynamic response coefficient between laser power attenuation and welding depth. The welding energy characteristics, welding formation characteristics, and cross characteristics are standardized to obtain the welding feature matrix.

4. The intelligent control method for online monitoring of laser welding quality of busbar profiles as described in claim 1, characterized in that, Based on the welding feature matrix, the correlation strength values ​​between multiple influencing factors and M welding quality indicators are determined. The method further includes: The forward propagation process employs a multilayer perceptron, with the welding feature matrix as the input layer and the hidden layer undergoing nonlinear transformation using the ReLU activation function. When analyzing the correlation strength between multiple influencing factors and M welding quality indicators, a weighted fusion of Pearson correlation coefficient and mutual information entropy is introduced. Based on the weighted fusion results, the N welding quality indicators are selected and determined.

5. The intelligent control method for online monitoring of laser welding quality of busbar profiles as described in claim 4, characterized in that, The method for identifying key quality factors and effectively controlling the chain includes: Based on the aforementioned quality-dominant factors, hierarchical clustering is employed, combined with correlation strength filtering thresholds to identify potential regulatory chains. The potential regulatory chain is time-series verified to obtain the effective regulatory chain.

6. The intelligent control method for online monitoring of laser welding quality of busbar profiles as described in claim 5, characterized in that, Based on the aforementioned quality-dominant factors, hierarchical clustering is employed, combined with correlation strength filtering thresholds for screening, to uncover potential regulatory chains. The method includes: Based on the quality-dominant factors, the inter-class distances are determined, and multiple factor components and multiple factor group classes are obtained by dividing the data using the inter-class distances. Path analysis is performed based on the multiple factor components and multiple factor group classes. The multilayer perceptron is traversed by depth-first search. When the correlation strength values ​​of P consecutive network nodes in the quality regulation transmission process all meet the correlation strength filtering threshold, they are marked as potential regulation chains, where P is greater than or equal to 2.

7. The intelligent control method for online monitoring of laser welding quality of busbar profiles as described in claim 6, characterized in that, The method involves performing time-series verification on the potential regulatory chain to obtain the effective regulatory chain, the method comprising: On the potential control chain, obtain the contribution of the network nodes on the chain; Based on the contribution of the on-chain network nodes, obtain the cumulative value of the product of the time difference and the contribution of adjacent on-chain network nodes in the potential control chain; When the cumulative value of the product of the time difference and the contribution meets the timing verification threshold, and the control order of the nodes on the chain meets the laser welding reaction logic, the timing verification is deemed to have passed.

8. The intelligent control method for online monitoring of laser welding quality of busbar profiles as described in claim 7, characterized in that, The method includes: If the cumulative value of the product of the time difference and the contribution does not meet the time series verification threshold, a two-way tracing mechanism is used to deduce the extended potential control chain. The extended potential control chain is time-tested by superimposing the cumulative value corresponding to the time-series compensation value and the time-series verification threshold. If it still does not meet the time-series verification threshold, the control chain is split and reorganized: the extended potential control chain is split into multiple sub-chains according to the peak time difference, and multiple sub-chains that pass the time-series verification are retained; for multiple sub-chains that fail the time-series verification, a secondary verification is performed in combination with the spatial defect correlation of the weld detection data.

9. The intelligent control method for online monitoring of laser welding quality of busbar profiles as described in claim 8, characterized in that, The extended potential control chain is obtained by using a two-way tracing mechanism. The method includes: By tracing back the preceding influencing factors of the first node in the chain, the correlation strength complement value with the first node in the chain is determined; By extrapolating the derivative regulatory effects of the tail node, a potential quality improvement coefficient is generated. The temporal compensation value is obtained by weighting and fusing the association strength compensation value corresponding to the first node of the chain and the potential quality improvement coefficient corresponding to the last node of the chain.

10. An intelligent control system for online monitoring of laser welding quality of busbar trunking profiles, characterized in that, The system is used to implement the intelligent control method for online monitoring of the laser welding quality of busbar profiles according to any one of claims 1 to 9, and the system includes: The welding data acquisition module is used to acquire multi-source welding data, including welding process parameter data, weld inspection data, environmental interference data, and equipment performance data, and to establish a welding feature matrix. The correlation strength determination module is used to determine the correlation strength values ​​between multiple influencing factors and M welding quality indicators based on the welding feature matrix, identify the dominant quality factors and effective control chains, and set the quality influence degree sequence. The control network construction module is used to perform hierarchical division by taking the quality-dominant factors as core nodes, taking the effective control chain as the directional connection path of the network nodes, taking the multiple influencing factors and the associated N welding quality indicators as network nodes, and taking the association strength value as the edge weight to construct a welding quality association control network, where M is greater than N; The control strategy output module is used to output a dynamic process control strategy for laser welding of busbar profiles based on the welding quality correlation control network and the quality influence sequence.

Images