Digital welding method for controller production

Abstract

The present invention discloses a digital welding method for controller production, which relates to the field of welding control. The method includes: performing real-time welding sensing on the controller of an electric three-wheeled vehicle, obtaining multiple real-time welding data, fusing the data according to multi-dimensional production environment information, synchronizing the welding fusion results to a quality inspection module for identification, obtaining a welding defect data set, performing dynamic feedback correction on the multiple real-time welding data, and formulating a welding control strategy to perform digital intelligent welding on the controller of the electric three-wheeled vehicle. This method solves the technical problem of traditional welding parameter adjustment being untimely, resulting in unstable welding quality and inaccurate defect prediction. It achieves the technical effect of real-time data acquisition, dynamic performance feedback correction, and welding prediction trend analysis, improving welding quality and reducing the occurrence of production defects.

Description

Technical Field

[0001] The present application relates to the field of welding control, and in particular to a digital welding method for controller production. Background Art

[0002] As an important tool for urban transportation and logistics, the electric control system of electric tricycles has a crucial impact on their overall performance and safety. In the production process of electric tricycles, welding is one of the key manufacturing processes, especially in the welding process of the frame and battery pack. The welding quality is directly related to the safety, durability and stability of the entire vehicle. However, traditional welding technology still faces many challenges in the manufacture of electric tricycles, especially in the scenarios of high-precision welding and large-scale production, where it is often difficult to maintain consistent quality. Traditional welding methods rely on manually set welding parameters (such as current, voltage and welding speed), but these parameters often lack real-time adjustment and feedback mechanisms, resulting in defects such as pores, cracks, and lack of fusion during the welding process, which in turn affects the frame strength of the electric tricycle and the safety of the battery pack, causing product instability. Summary of the Invention

[0003] This application solves the technical problem of unstable welding quality and inaccurate defect prediction caused by the untimely adjustment of traditional welding parameters by providing a digital welding method for controller production. It achieves the technical effects of real-time data collection, dynamic performance feedback correction and welding prediction trend analysis, improving welding quality and reducing the occurrence of production defects.

[0004] The present application provides a digital welding method for controller production, including: performing real-time welding sensing on the controller of an electric three-wheeler through a sensing device group to obtain multiple real-time welding data; introducing multi-dimensional production environment information of the controller, and fusing the multiple real-time welding data according to the multi-dimensional production environment information to obtain a welding fusion result; synchronizing the welding fusion result to a quality inspection module for identification to obtain a welding defect data set; performing dynamic feedback correction on the multiple real-time welding data based on the welding defect data set to formulate a welding control strategy; and executing the welding control strategy to perform digital intelligent welding on the controller of the electric three-wheeler.

[0005] In a possible implementation, the method for constructing the multi-dimensional production environment information performs the following processing: calling the historical production data record log of the controller, and extracting the production process information of the controller based on the historical production data record log; dividing the production stages according to the production process information to obtain multiple production nodes of the controller; performing production operation analysis on the controller based on the multiple production nodes to determine multiple production operation status information; performing multi-dimensional analysis on the controller according to the multiple production nodes with the multiple production operation status information to obtain the multi-dimensional production environment information.

[0006] In a possible implementation, the multiple real-time welding data are fused according to the multi-dimensional production environment information to obtain a welding fusion result, and the following processing is performed: the multiple real-time welding data are grouped according to the multi-dimensional production environment information to obtain multiple welding data groups; feature analysis is performed based on the multiple welding data groups to obtain multiple welding features; weights are assigned to the multiple real-time welding data according to the multiple welding features based on the multi-dimensional production environment information to obtain multiple weight coefficients; the multiple real-time welding data are matched and fused with the multi-dimensional production environment information according to the multiple weight coefficients to obtain multiple welding-environment data pairs; and the multiple welding-environment data pairs are added to the welding fusion result.

[0007] In a possible implementation, the welding fusion result is synchronized to a quality inspection module for identification to obtain a welding defect data set, and the following processing is performed: vector analysis is performed on the multiple welding-environment data pairs according to the multiple welding features to determine multiple feature vectors; based on the multiple feature vectors, the welding fusion result is synchronized to the quality inspection module, and the multiple welding-environment data pairs are traversed by the quality inspection module for deep learning to obtain a multidimensional inspection result; the multidimensional inspection result is dynamically monitored according to a time series to obtain multiple quality numerical features; the multidimensional inspection result is defect-classified according to the multiple quality numerical features to generate multiple defect labels; the multiple defect labels are annotated to the multidimensional inspection result to obtain the welding defect data set.

[0008] In a possible implementation, the multidimensional detection results are defect-classified according to the multiple quality numerical features, multiple defect labels are generated, and the following processing is performed: reverse backtracking is performed based on the multiple quality numerical features to obtain multiple data acquisition channels, and the multiple data acquisition channels contain multiple welding point information; multiple defect types are determined based on the multiple welding point information combined with the multiple quality numerical features; defect impact calculations are performed on the multiple data acquisition channels according to the multiple defect types combined with the multiple welding point information to obtain multiple defect impact coefficients; the multiple quality numerical features are sorted in descending order according to the multiple defect impact coefficients to generate a quality numerical sequence, and the quality numerical sequence is graded to generate the multiple defect labels.

[0009] In a possible implementation, dynamic feedback correction is performed on the multiple real-time welding data based on the welding defect data set, a welding control strategy is formulated, and the following processing is performed: multiple defect impact levels are extracted based on the multiple defect labels; a defect prediction model is constructed according to the multiple defect impact levels, and predictions are made through the defect prediction model combined with the multiple defect impact levels to generate a welding prediction trend data set; performance evaluation of the controller is performed based on the welding prediction trend data set to generate a dynamic performance score; feedback response is performed based on the dynamic performance score combined with the welding prediction trend data set to generate dynamic welding correction parameters; the multiple real-time welding data are synchronously adjusted according to the dynamic welding correction parameters to formulate the welding control strategy.

[0010] In a possible implementation, a defect prediction model is constructed based on the multiple defect impact levels, and predictions are made through the defect prediction model in combination with the multiple defect impact levels to generate a welding prediction trend data set, and the following processing is performed: training is performed based on the historical production data record log according to the time series, and the defect prediction model is constructed based on the training results in combination with the multiple defect impact levels; the multiple real-time welding data are synchronized to the defect prediction model for capture, and welding defect prediction associated parameters are determined; the welding defect prediction associated parameters are mapped and matched with the multiple defect impact levels, and calculations are performed based on the matching results to obtain the probability of defect occurrence; defect analysis is performed based on the defect occurrence probability in combination with the time series to obtain defect distribution data, the defect distribution data is synchronized to the defect prediction model for update, and the welding prediction trend data set is drawn.

[0011] In a possible implementation, feedback response is performed based on the dynamic performance score in combination with the welding prediction trend data set to generate dynamic welding correction parameters, and the following processing is performed: deviation analysis is performed based on the welding prediction trend data set, and feedback rules are constructed based on the predicted deviation value, and the feedback rules include a feedback cycle; dynamic analysis is performed based on the dynamic performance score to generate a correction factor; feedback is performed on the welding prediction trend data set according to the feedback cycle to generate a parameter set to be responded; and dynamic correction is performed on the parameter set to be responded based on the correction factor to generate the dynamic welding correction parameters.

[0012] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0013] The digital welding method for controller production provided in this application solves the technical problem of traditional welding parameter adjustment not being timely, resulting in unstable welding quality and inaccurate defect prediction, and achieves the technical effect of real-time data acquisition, dynamic performance feedback correction and welding prediction trend analysis, improving welding quality and reducing the occurrence of production defects. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention are briefly introduced below. Flowcharts are used in this application to illustrate the operations performed by the methods according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in precise order. Instead, the various steps may be processed in reverse order or simultaneously as needed. Furthermore, other operations may be added to these processes, or one or more operations may be removed from these processes.

[0015] Figure 1 A schematic flow chart of a digital welding method for controller production provided in an embodiment of the present application.

[0016] Figure 2 A schematic flow chart of a welding control strategy for a digital welding method for controller production provided in an embodiment of the present application. DETAILED DESCRIPTION

[0017] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0018] In order to make the purpose, technical solutions and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limiting this application. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0019] In the following description, reference is made to “some embodiments”, which describes a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict, and the terms “first\second” involved are merely used to distinguish similar objects and do not represent a specific ordering of the objects. The terms “including” and “having” and any variations are intended to cover non-exclusive inclusions. For example, a process, method, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or modules that are not clearly listed or that are inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application belongs. The terms used herein are for the purpose of describing the embodiments of this application only.

[0020] The present application provides a digital welding method for controller production, such as Figure 1 As shown, the method includes:

[0021] Step A100, performing real-time welding sensing on the controller of the electric tricycle through a sensor device group to obtain a plurality of real-time welding data;

[0022] In order to monitor the welding process of the electric three-wheel controller in real time, it is necessary to select a suitable sensor device group, which is used to provide accurate physical quantities and environmental data. Temperature, current, voltage, pressure and other sensors can be installed near the welding gun, welding workbench, and welding materials to capture various important data of the controller during the production process. Through the installed sensor device group, data in the welding process, including welding current, voltage, temperature, pressure, etc., can be collected in real time. The weld can also be photographed by a high-resolution camera or laser scanning sensor to obtain image data of the welding surface for subsequent defect detection and quality assessment. Multiple real-time welding data are acquired and stored in a local database or cloud for subsequent query and analysis, providing strong data support for the high-quality production of electric three-wheelers, ensuring welding quality and production efficiency.

[0023] Execute step A200, introduce the multi-dimensional production environment information of the controller, perform data fusion on the multiple real-time welding data according to the multi-dimensional production environment information, and obtain a welding fusion result; in a possible implementation method, step A200 further includes step A210, call the historical production data record log of the controller, and extract the production process information of the controller based on the historical production data record log; execute step A220, divide the production stages according to the production process information, and obtain multiple production nodes of the controller; execute step A230, perform production operation analysis on the controller based on the multiple production nodes, and determine multiple production operation status information; execute step A240, perform multi-dimensional analysis on the controller according to the multiple production nodes with the multiple production operation status information, and obtain the multi-dimensional production environment information.

[0024] First, access the controller's historical production data logs through the enterprise's production management system (such as an MES system). These logs contain the controller's status, parameter settings, and related events at each stage of the production process. Historical production data logs are typically in a structured format, such as CSV, JSON, or a database table. They contain information such as production time, stage, production parameters, operator, equipment status, and quality inspection results.

[0025] Key production process information is further extracted from historical production data. This information typically includes production stages, operating parameters, and production results. By analyzing identifiers such as timestamps and operational events, the data is divided into different production stages (e.g., raw material preparation, assembly, commissioning, and testing). Based on this stage division, key nodes in the production process are further identified, such as equipment startup, product assembly, testing, and final quality inspection, thereby acquiring production process information. Furthermore, the production process is divided into multiple stages based on the time records of controller production. For example, the operation cycle can be divided into "startup phase," "stable production phase," and "commissioning phase." Event identifiers in historical production data logs (e.g., "equipment startup" or "quality inspection passed") are then used to identify different production stages. Based on the production flow chart and the events in each stage, key nodes within each stage are determined. For example, "assembly start," "welding completion," and "quality inspection" are used as key nodes in the production process. Multiple production nodes are then identified, each of which corresponds to a set of status information recording equipment parameters, personnel information, material flow, and other content.

[0026] Based on historical production data logs and multiple production nodes, production status is categorized as "normal," "abnormal," or "pending adjustment." The production status of each stage or node is determined by analyzing equipment operating parameters and quality data. The status data for each of the multiple production nodes is then recorded, including equipment operating status (e.g., powered on, standby, faulty) and operator operating status (e.g., operating, standby).

[0027] Finally, multiple production operation status information is used to perform multi-dimensional analysis on the controller according to multiple production nodes. This means analyzing the environmental impact of the controller at each production node and stage based on multiple production operation status information, such as the impact of excessive temperature on equipment operation, the impact of humidity on product quality, etc., thereby determining multi-dimensional production environment information, which can effectively improve the automation level of the production line and realize refined management in a complex production environment.

[0028] In one possible implementation, step A200 further includes step A250, grouping the multiple real-time welding data according to the multi-dimensional production environment information to obtain multiple welding data groups; executing step A260, performing feature analysis based on the multiple welding data groups to obtain multiple welding features; executing step A270, weighting the multiple real-time welding data according to the multiple welding features based on the multi-dimensional production environment information to obtain multiple weight coefficients; executing step A280, matching and fusing the multiple real-time welding data with the multi-dimensional production environment information according to the multiple weight coefficients to obtain multiple welding-environment data pairs; executing step A290, adding the multiple welding-environment data pairs to the welding fusion result.

[0029] First, the real-time welding data is grouped according to different production environment conditions (such as temperature, humidity, air pressure, etc.). This means that the data is grouped according to parameter types, such as environmental data groups (temperature, humidity, etc.) and welding data groups, namely current, voltage, welding speed, etc., and each data group represents the welding process under specific environmental conditions. Furthermore, feature analysis is performed based on multiple welding data groups, which means that statistical methods (such as mean, standard deviation, skewness, kurtosis, etc.) can be used to analyze each welding data group, or electrical data (current, voltage) can be analyzed in the time domain or frequency domain to extract characteristics such as fluctuation amplitude and frequency, and then key welding features such as welding current fluctuation range, welding speed, weld quality indicators, etc. can be extracted.

[0030] Furthermore, the weighted average method can be used to assign weights to the multiple real-time welding data based on the multiple welding characteristics using the multi-dimensional production environment information. This means that each characteristic (such as welding current, temperature, humidity, etc.) is assigned a weight value based on its impact on welding quality and process stability. The weight coefficient can be optimized through a machine learning model to more accurately reflect the impact of environmental factors on the welding process, thereby obtaining multiple weight coefficients.

[0031] Multiple weight coefficients are further used to match multiple welding data sets with multidimensional production environment data. Each welding data set is weighted and combined according to its corresponding environmental data (such as temperature, current, voltage, etc.). By adopting data fusion algorithms (such as Kalman filtering, weighted averaging, principal component analysis, etc.), the welding data and environmental information are multi-dimensionally fused to generate welding-environment data pairs. These welding-environment data pairs can effectively reflect the impact of environmental factors on the welding process. Each welding-environment data pair can include: real-time welding process data (such as current, voltage, welding speed, etc.), corresponding environmental data (such as temperature, humidity, equipment status, etc.), and fused weighted data for further analysis and prediction. Finally, multiple welding-environment data pairs are added to the welding fusion result, which means that the welding-environment data pairs are stored in a database (such as a relational database or a time series database) to facilitate subsequent analysis and query, providing strong data support for efficient and stable welding production processes.

[0032] Executing step A300, synchronizing the welding fusion result to a quality inspection module for identification to obtain a welding defect data set; in one possible implementation, step A300 further includes step A310, performing vector analysis on the multiple welding-environment data pairs according to the multiple welding features to determine multiple feature vectors; executing step A320, synchronizing the welding fusion result to the quality inspection module based on the multiple feature vectors, and traversing the multiple welding-environment data pairs through the quality inspection module for deep learning to obtain a multidimensional inspection result; executing step A330, dynamically monitoring the multidimensional inspection result according to a time series to obtain multiple quality numerical features;

[0033] Vector analysis of multiple welding-environment data pairs based on multiple welding characteristics involves assigning a feature vector to each welding-environment data pair, containing all key eigenvalues. For example, welding current, voltage, and temperature are converted into feature vectors. The numerical values ​​of each feature vector are then normalized to ensure data comparability across all dimensions, resulting in multiple feature vectors. The weld fusion results are then synchronized to the quality inspection module based on these multiple feature vectors. The quality inspection module then performs deep learning on these multiple welding-environment data pairs. This process can include weld surface defect detection, weld geometry analysis, and welding process parameter anomaly detection. Weld surface defect detection can include detecting defects such as porosity, cracks, and lack of fusion. Weld geometry analysis can be used to assess whether weld width, depth, and uniformity meet standards. Weld process parameter anomaly detection can identify anomalies in parameters such as current, voltage, and temperature during the welding process, thereby obtaining multi-dimensional detection results.

[0034] Finally, the multi-dimensional detection results are dynamically monitored according to the time series, which means that the detection results are monitored in time series according to the real-time welding data stream and the output of the deep learning model, and in the production process, the welding parameters and quality detection results are monitored in real time, and the welding control parameters (such as current, voltage, speed, etc.) are dynamically adjusted to avoid defects. Then, the quality numerical features of each welding process are extracted, such as the frequency of welding defects, the probability of occurrence of defect types, etc. At the same time, the number of various defects is counted and their occurrence frequency is calculated to evaluate the welding quality. According to the type and impact of the defect (for example, the impact coefficient is evaluated according to the location and size of the defect), a quality score is generated for each welding, thereby obtaining multiple quality numerical features, which can timely discover potential problems, optimize the welding process, and improve product quality and production efficiency.

[0035] Execute step A340, classify the multidimensional detection results into defects according to the multiple quality numerical features, and generate multiple defect labels; in one possible implementation, step A340 further includes step A341, reverse backtracking based on the multiple quality numerical features to obtain multiple data acquisition channels, and the multiple data acquisition channels contain multiple welding point information; execute step A342, analyze the multiple welding point information in combination with the multiple quality numerical features to determine multiple defect types; execute step A343, calculate the defect impact of the multiple data acquisition channels according to the multiple defect types in combination with the multiple welding point information, and obtain multiple defect impact coefficients; execute step A344, sort the multiple quality numerical features in descending order according to the multiple defect impact coefficients to generate a quality numerical sequence, perform grade classification according to the quality numerical sequence, and generate the multiple defect labels.

[0036] First, based on the acquired quality numerical features (such as welding quality score, current fluctuation, and temperature variation), a reverse backtracking analysis is performed to obtain the welding data channels and weld point information associated with each quality feature. Starting from the current quality numerical feature, the specific welding process and related data acquisition channels are traced back. The welding data and related environmental data at the corresponding time points are matched using timestamps or production logs to ensure data consistency. Multiple data acquisition channels are identified. These multiple data acquisition channels represent the information from the original acquisition process for the control data collection. Based on the backtracking results, all multiple data acquisition channels associated with the quality numerical features are identified and acquired. These channels may include data such as current, voltage, welding position, temperature, and pressure. The weld point information associated with each data acquisition channel (such as welding position, operator, and equipment number) provides essential location data for subsequent analysis. Further integration of the multiple weld point information with the corresponding quality numerical features involves analyzing the weld points during the welding process based on the quality numerical features (such as current fluctuation and temperature anomalies) to identify the features and locations associated with defect occurrence. Through data analysis, the defect type associated with each weld point is determined, resulting in multiple defect types.

[0037] The defect impact coefficient is further calculated based on the defect type at each weld point and its impact on the overall weld quality. The impact coefficient reflects the severity of the defect and is calculated using the formula: Idefect = f(defect type, defect size, defect location), where Idefect is the defect impact coefficient and f() is a function based on the defect type, size, and location, outputting the impact coefficient.

[0038] Finally, the calculated multiple defect impact coefficients are arranged in descending order to reflect which defects have the greatest impact on welding quality. Sorting algorithms (such as quick sort, heap sort, etc.) can be used to sort the multiple defect impact coefficients to clarify the degree of impact of each defect on welding quality. Based on the defect impact coefficients after descending sorting, a grade standard is established. The grade standard can include the first grade, i.e., extremely serious defects (impact coefficient ≥ 90%), the second grade, medium defects (impact coefficient between 70%-90%), and the third grade, minor defects (impact coefficient <70%). The defect impact coefficients are divided using a threshold method or a segmentation method to generate defect labels. For each welding data point, a corresponding defect grade label is generated, thereby providing basic data support and improving production efficiency and product quality.

[0039] Execute step A350 to annotate the multiple defect labels to the multi-dimensional detection results to obtain the welding defect dataset.

[0040] The generated defect labels and corresponding welding-environment data pairs are integrated into a complete dataset. The welding defect dataset contains the feature vector of each welding data pair and information such as the corresponding defect type, location, and influence coefficient. The welding defect dataset can serve as the basis for subsequent applications such as welding quality assessment, defect prediction, and control strategy optimization.

[0041] Execute step A400, perform dynamic feedback correction on the multiple real-time welding data based on the welding defect data set, and formulate a welding control strategy; in a possible implementation, such as Figure 2 As shown, step A400 further includes step A410, extracting multiple defect impact levels based on the multiple defect labels;

[0042] First, the impact level of each defect is calculated based on its type, size, location, and weld quality score. Impact levels can be categorized into multiple levels (e.g., high, medium, and low) to indicate the degree to which a defect affects weld quality. By learning from historical welding data and defect impact levels, the type and impact level of defects that may occur under different production environments can be predicted.

[0043] Execute step A420, construct a defect prediction model based on the multiple defect impact levels, perform predictions through the defect prediction model in combination with the multiple defect impact levels, and generate a welding prediction trend data set; in one possible implementation, step A420 further includes step A421, perform training according to the time series based on the historical production data record log, and construct the defect prediction model based on the training results in combination with the multiple defect impact levels; execute step A422, synchronize the multiple real-time welding data to the defect prediction model for capture, and determine welding defect prediction associated parameters; execute step A423, map and match the welding defect prediction associated parameters with the multiple defect impact levels, calculate according to the matching results, and obtain the defect occurrence probability; execute step A424, perform defect analysis based on the defect occurrence probability in combination with the time series, obtain defect distribution data, synchronize the defect distribution data to the defect prediction model for update, and draw the welding prediction trend data set.

[0044] Training based on time series of historical production data logs involves using machine learning algorithms, such as random forests, support vector machines (SVMs), and neural networks, to train defect prediction models. The model inputs include historical welding data features (such as welding current, voltage, and temperature), and the output is a defect impact level or defect type. Based on the historical defect dataset, defect types and impact levels are defined for each sample based on multiple defect impact levels. The model is trained using historical data, and the optimal model and hyperparameters are selected through cross-validation to obtain a defect prediction model.

[0045] Further synchronizing the multiple real-time welding data with the defect prediction model for capture involves predicting welding defects based on the real-time data and the defect prediction model, and outputting welding defect prediction parameters. Sensors collect real-time welding process data, including welding current, voltage, temperature, welding speed, and other data, as well as real-time quality feedback. Welding defect prediction is performed based on the correlation between the defect and the weld, and welding defect prediction correlation parameters are obtained.

[0046] Further mapping and matching the welding defect prediction associated parameters with the multiple defect impact levels means mapping and matching the predicted welding defect prediction associated parameters with the multiple defect impact levels obtained in the historical data. By calculating the matching results, the probability of defect occurrence in each welding process is determined. The prediction results of the welding defects can be mapped and matched with the defect impact levels through a regression model or a classification model to calculate the probability of defect occurrence. The specific formula is as follows: P defect occurrence = f (welding data features, defect impact level), where P defect occurrence is the probability of defect occurrence, and f() is the mapping function obtained through training. The probability of defect occurrence is predicted based on real-time welding data. For example, when the welding current is too high and the temperature is too low, the probability of defect occurrence may be higher. The matching results of all defect impact levels and welding data are calculated to obtain the probability of defect occurrence for each welding process.

[0047] Finally, through time series analysis methods (such as sliding windows, time series regression models, long short-term memory networks (LSTMs), etc.), combined with real-time data and defect occurrence probability during the welding process, the trend of defects changing over time is analyzed. The combination of defect occurrence probability and time series data is used to predict the trend of defects that may occur in the future, identify potential risk points in the welding process, and generate defect distribution data based on defect occurrence probability and time series data. The defect distribution data can include the temporal distribution of various types of defects, the influence coefficient of each defect, and the probability of occurrence of different defect types in each welding cycle. The defect distribution data is synchronized with the defect prediction model for update to generate a welding prediction trend dataset. The welding prediction trend dataset can include the predicted trend of defect occurrence, optimization direction in the welding process, and possible future production adjustment suggestions, thereby improving welding quality, reducing the occurrence of defects, optimizing production efficiency, and making real-time adjustments to the production process.

[0048] Executing step A430, performing a performance evaluation on the controller based on the welding prediction trend data set to generate a dynamic performance score;

[0049] Based on the welding prediction trend data set, combined with key parameters in the welding process (such as current, voltage, temperature, welding speed, etc.), the overall performance of the welding process is evaluated, and the dynamic performance score of each welding process is calculated based on the welding prediction trend data set. The score takes into account factors such as welding quality, production efficiency, and defect incidence. At the same time, during the welding process, the controller's operating status and welding parameters are dynamically adjusted based on real-time feedback of welding data and prediction trends. For example, if the dynamic performance score is low, parameters such as current or temperature may need to be adjusted to avoid defects.

[0050] Execute step A440, perform feedback response based on the dynamic performance score in combination with the welding prediction trend data set, and generate dynamic welding correction parameters; in one possible implementation, step A440 further includes step A441, perform deviation analysis based on the welding prediction trend data set, and construct feedback rules based on the predicted deviation value, and the feedback rules include a feedback cycle; execute step A442, perform dynamic analysis based on the dynamic performance score, and generate a correction factor; execute step A443, perform feedback on the welding prediction trend data set according to the feedback cycle, and generate a parameter set to be responded; execute step A444, perform dynamic correction on the parameter set to be responded based on the correction factor, and generate the dynamic welding correction parameters.

[0051] Deviation analysis is performed by comparing the predicted welding trend data set with actual welding quality data (such as welding temperature, current, and voltage). The difference between the predicted trend data and the actual value is considered the deviation. A larger deviation indicates that the welding process deviates more from the expected target, potentially leading to quality issues. An acceptable deviation range is set based on predefined quality standards. For example, a deviation exceeding 5% may indicate a need for welding parameter adjustment. Feedback rules are then constructed based on the predicted deviation values. These feedback rules include a feedback cycle, which refers to the frequency at which the feedback control mechanism updates during the welding process. Based on the timing of the welding process and deviation analysis, a reasonable feedback cycle is determined. The feedback cycle can be time-based (e.g., updated every minute) or based on key events in the welding process (e.g., updated after each completed batch of welds). Different deviation ranges and corresponding feedback responses are set. The feedback rule determines what adjustments the system should make (e.g., welding current, voltage, speed, etc.) when the deviation between the predicted and actual values ​​exceeds a certain threshold.

[0052] The welding process performance is further evaluated based on real-time welding data and predicted trend data. Dynamic performance scores may include multiple aspects such as welding quality, production efficiency, and defect rate. Based on deviation analysis and dynamic performance scores, correction factors are calculated to serve as the basis for welding parameter adjustments. Correction factors are dynamic adjustments during the welding process and can be used to optimize various welding parameters (such as current, voltage, and temperature). Correction factors are functions optimized based on historical data and real-time performance scores, representing the impact of deviation values ​​and performance scores on the adjustment of welding control parameters.

[0053] Furthermore, based on a predefined feedback cycle, welding data is collected in real time and deviations are calculated to determine whether the control strategy needs to be updated. For example, if the deviation exceeds a preset range, the welding controller immediately implements adjustments. If the system detects a large welding deviation, the feedback cycle is shortened, with frequent data collection, analysis, and adjustments. Conversely, if the deviation is small or within a normal range, the feedback cycle can be extended. Based on the feedback from the deviation analysis and welding prediction trend dataset, the welding parameters to be addressed (such as welding current, voltage, and speed) are determined. These parameters serve as targets for adjustment and optimization, resulting in a set of parameters to be addressed.

[0054] Dynamically correcting the parameter set to be responded to based on the correction factor involves using the generated dynamic welding correction parameters to dynamically adjust the parameter set to be responded to based on deviation analysis and performance scoring. Based on the dynamically corrected welding parameters, the parameter set to be responded to is updated to ensure that each welding task is performed within the optimal parameter range. This can include adjusting welding parameters such as current, voltage, speed, and temperature. Through multiple feedback cycles, the parameter set to be responded to is gradually optimized to ensure a stable welding process and reduce defects under various production conditions. This provides efficient and precise operational support for each welding task in the production process.

[0055] Execute step A450 to synchronously adjust the multiple real-time welding data according to the dynamic welding correction parameters and formulate the welding control strategy.

[0056] Synchronously adjusting multiple real-time welding data sets based on dynamic welding correction parameters involves developing new welding control strategies based on these parameters to ensure the quality and efficiency of each welding process. These control strategies can include adjusting the welding path, welding current, and temperature. These parameters are automatically adjusted based on real-time monitoring data and predicted trends to maintain welding process stability and optimize performance. Feedback response data can be used to optimize the next welding task, reduce defects in the production process, and improve product quality and production efficiency.

[0057] Next, step A500 is executed to execute the welding control strategy to perform digital intelligent welding on the controller of the electric three-wheeler.

[0058] Implementing a welding control strategy for the electric tricycle controller involves feeding real-time welding and environmental data back to the control system. Based on this feedback and the strategy model, the control system automatically adjusts welding parameters. The welding process operates as a closed-loop control system, continuously adjusting welding conditions based on real-time data to ensure the quality of each weld meets the desired target. Simultaneously, the welding process of the electric tricycle controller is monitored in real time to ensure all operations are within the control strategy's set limits. Potential welding defects are predicted in real time and, based on these predictions, welding parameters are adjusted in advance to avoid them. Based on this real-time monitoring data, the controller intelligently adjusts parameters such as welding current, voltage, and speed to meet welding quality requirements. By implementing the welding control strategy and combining it with digital intelligent control technology, the control system can monitor and dynamically adjust welding parameters in real time during the production process of the electric tricycle controller, ensuring the quality and stability of each weld. Through intelligent feedback and continuous optimization, the stability and quality of the welding process are significantly improved, reducing manual intervention, increasing production efficiency, and effectively reducing the occurrence of welding defects.

[0059] The embodiments of the present application solve the technical problem that traditional welding parameters are not adjusted in a timely manner, resulting in unstable welding quality and inaccurate defect prediction, and achieve the technical effects of real-time data collection, dynamic performance feedback correction and welding prediction trend analysis, improving welding quality and reducing the occurrence of production defects.

[0060] The above specific embodiments do not constitute a limitation to the scope of protection of this application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent replacements and improvements made within the spirit and principles of this application should be included in the scope of protection of this application. In some cases, the actions or steps recorded in this application can be performed in an order different from that in the embodiments and can still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A digital welding method for controller production, characterized in that: The method comprises: The sensor equipment group performs real-time welding sensing on the controller of the electric three-wheeled vehicle to obtain multiple real-time welding data; Introducing multi-dimensional production environment information of the controller, performing data fusion on the plurality of real-time welding data according to the multi-dimensional production environment information, and obtaining a welding fusion result; Synchronizing the welding fusion results to a quality inspection module for identification to obtain a welding defect data set; Performing dynamic feedback correction on the plurality of real-time welding data based on the welding defect data set to formulate a welding control strategy; Executing the welding control strategy to perform digital intelligent welding on the controller of the electric tricycle; The method of fusing the plurality of real-time welding data according to the multi-dimensional production environment information to obtain a welding fusion result includes: Grouping the plurality of real-time welding data according to the multi-dimensional production environment information to obtain a plurality of welding data groups; Performing feature analysis based on the plurality of welding data sets to obtain a plurality of welding features; Based on the multi-dimensional production environment information, weighting the plurality of real-time welding data is performed according to the plurality of welding characteristics to obtain a plurality of weight coefficients; Matching and fusing the multiple real-time welding data with the multi-dimensional production environment information according to the multiple weight coefficients to obtain multiple welding-environment data pairs; adding the plurality of welding-environment data pairs to the welding fusion result; The welding fusion result is synchronized to the quality inspection module for identification to obtain a welding defect data set, the method comprising: Performing vector analysis on the plurality of welding-environment data pairs according to the plurality of welding features to determine a plurality of feature vectors; Synchronizing the welding fusion result to the quality detection module based on the multiple feature vectors, and performing deep learning on the multiple welding-environment data pairs through the quality detection module to obtain a multi-dimensional detection result; Dynamically monitoring the multidimensional detection results according to a time series to obtain multiple quality numerical features; Classifying the multidimensional detection results into defects according to the multiple quality numerical features to generate multiple defect labels; Marking the multiple defect labels to the multi-dimensional detection results to obtain the welding defect dataset; Classifying the multidimensional detection results into defect classes according to the multiple quality numerical features to generate multiple defect labels, the method comprising: Performing reverse backtracking based on the multiple quality numerical features to obtain multiple data acquisition channels, wherein the multiple data acquisition channels contain multiple welding point information; Analyze the multiple welding point information in combination with the multiple quality numerical features to determine multiple defect types; Perform defect impact calculation on the multiple data acquisition channels according to the multiple defect types and the multiple welding point information to obtain multiple defect impact coefficients; The multiple quality value features are sorted in descending order according to the multiple defect influence coefficients to generate a quality value sequence, and the quality value sequence is graded to generate the multiple defect labels.

2. The digital intelligent welding method for controller production according to claim 1, characterized in that: The method for constructing the multi-dimensional production environment information includes: Retrieving a historical production data record log of the controller, and extracting production process information of the controller based on the historical production data record log; Divide the production stages according to the production process information to obtain multiple production nodes of the controller; Performing a production operation analysis on the controller based on the multiple production nodes to determine multiple production operation status information; The controller performs multi-dimensional analysis on the multiple production operation status information according to the multiple production nodes to obtain the multi-dimensional production environment information.

3. The digital intelligent welding method for controller production according to claim 1, characterized in that: Performing dynamic feedback correction on the plurality of real-time welding data based on the welding defect data set to formulate a welding control strategy, the method comprising: extracting a plurality of defect impact levels based on the plurality of defect labels; Building a defect prediction model based on the multiple defect impact levels, and performing predictions by combining the defect prediction model with the multiple defect impact levels to generate a welding prediction trend data set; performing a performance evaluation on the controller based on the welding prediction trend data set to generate a dynamic performance score; Performing feedback response based on the dynamic performance score combined with the welding prediction trend data set to generate dynamic welding correction parameters; The plurality of real-time welding data are synchronously adjusted according to the dynamic welding correction parameters to formulate the welding control strategy.

4. The digital intelligent welding method for controller production according to claim 3, characterized in that: Constructing a defect prediction model based on the multiple defect impact levels, performing predictions using the defect prediction model in combination with the multiple defect impact levels to generate a welding prediction trend data set, the method comprising: Training is performed based on the historical production data record logs in a time series, and the defect prediction model is constructed according to the training results and the multiple defect impact levels; Synchronizing the plurality of real-time welding data to the defect prediction model for capturing, and determining welding defect prediction associated parameters; Mapping and matching the welding defect prediction associated parameters with the multiple defect impact levels, and performing calculations based on the matching results to obtain a defect occurrence probability; Defect analysis is performed based on the defect occurrence probability in combination with the time series to obtain defect distribution data, the defect distribution data is synchronized to the defect prediction model for updating, and the welding prediction trend data set is drawn.

5. The digital intelligent welding method for controller production according to claim 3, characterized in that: The method includes: performing feedback response based on the dynamic performance score combined with the welding prediction trend data set to generate dynamic welding correction parameters. Performing deviation analysis based on the welding prediction trend data set, and constructing a feedback rule according to the prediction deviation value, wherein the feedback rule includes a feedback cycle; Performing a dynamic analysis based on the dynamic performance score to generate a correction factor; Feedback is performed on the welding prediction trend data set according to the feedback cycle to generate a parameter set to be responded; The set of parameters to be responded to is dynamically corrected based on the correction factor to generate the dynamic welding correction parameter.

Images