A weld seam monitoring system based on an electric welding machine

Abstract

The present invention discloses a weld seam monitoring system based on an electric welding machine, which relates to the technical field of electric welding. An initial setting module sets the normal fluctuation threshold range of each working parameter by collecting the working information of the electric welding machine in a fault-free state. The working information at least includes welding current, voltage and welding speed. When it is determined that one or more of the working parameter information collected in real time exceeds the preset normal fluctuation threshold range, it is judged whether it is a short-term fluctuation or a fault fluctuation by analyzing the change of the parameter, so as to prevent misjudgment during short-term fluctuations of the electric welding machine, which may lead to adjustment of parameters and quality defects in the weld seam.

Description

Technical Field

[0001] The present invention belongs to the technical field of electric welding, and specifically relates to a weld monitoring system based on an electric welding machine. Background Art

[0002] A weld is the part formed by the combination of workpieces during the welding process. According to the combination form, it can be divided into butt welds, fillet welds, plug welds, and slot welds, etc.; according to the macroscopic shape, it is divided into straight and curved shapes; according to the spatial position, it is divided into flat, vertical, horizontal, and overhead welds.

[0003] In the actual application process, when the electric welding machine is working, the stability of the parameters of the electric welding machine plays a key role in the weld quality. However, in actual operation, the parameters of the electric welding machine may change due to various factors during the working process. How to accurately distinguish whether these changes are normal fluctuations or deviations caused by faults has become a problem to be solved. If it cannot be accurately judged in time, at this time, the electric welding machine may adjust the parameters, which may lead to quality defects in the weld, increase production costs, and reduce production efficiency. The existing discrimination methods often rely too much on manual experience, lack systematicness and accuracy, and are difficult to meet the requirements of modern automated welding production. Summary of the Invention

[0004] The purpose of the present invention is to provide a weld monitoring system based on an electric welding machine, which solves the technical problem in the prior art that it is impossible to distinguish whether the parameters of the electric welding machine are normal fluctuations or deviations caused by faults when the parameters change.

[0005] A weld monitoring system based on an electric welding machine includes:

[0006] An initial setting module, which sets the normal fluctuation threshold range of each working parameter by collecting the working information of the electric welding machine in a fault-free state. The working information at least includes welding current, voltage, and welding speed.

[0007] A data comparison and processing module, which uses relevant sensors to collect the working parameter information of the electric welding machine in real time during the welding process of the target part and determines the fluctuating parameters. The fluctuating parameters during the welding process of the electric welding machine for the target part are compared with the corresponding normal fluctuation threshold range.

[0008] A control module, when the collected working parameter information exceeds the normal fluctuation threshold range, further determines whether the parameter change is a short-term fluctuation or a deviation caused by a fault.

[0009] A parameter self-adjustment module, which is used to send corresponding signals according to the judgment result. If it is a normal fluctuation, no parameter adjustment is performed. If it is a deviation caused by a fault, an alarm signal is sent, and a parameter adjustment instruction is generated according to the fault deviation situation to adjust the working parameters of the electric welding machine.

[0010] As a further solution of the present invention, the specific method for the control module to further determine whether the parameter change is a short-term fluctuation or a deviation caused by a fault is as follows:

[0011] Collect multiple parameter data during the welding process in real time. The parameters at least include welding current, welding voltage, welding speed, and molten pool temperature;

[0012] Based on the multiple parameter data, establish a correlation model between the parameters through the Pearson correlation coefficient or mutual information algorithm, and calculate the real-time correlation degree between any two parameters;

[0013] Screen the parameters according to a preset correlation degree threshold to exclude the interference of parameter pairs with a correlation degree lower than the threshold on subsequent judgments;

[0014] For the remaining parameters with high correlation, set the normal fluctuation range of each parameter and the allowable deviation interval for the linked change between the parameters based on the parameter historical data;

[0015] When it is detected that a certain parameter fluctuates, synchronously analyze whether the other parameters with high correlation with it show changes that conform to the linkage law within the corresponding allowable deviation interval;

[0016] If the fluctuations of all related parameters are within the allowable deviation interval and conform to the linkage law, it is determined as a short-term fluctuation. If the fluctuation of a certain parameter exceeds its normal range and the related parameters do not show the expected linked change, it is determined as a fault fluctuation.

[0017] As a further solution of the present invention, it further includes an adaptive prediction module. The specific working steps are as follows:

[0018] S1. According to the parameter data collected in real time, predict the parameter change trend within a specified short time in the future;

[0019] S2. When the real-time parameter is close to the boundary of the normal fluctuation threshold range, if the prediction result shows that the parameter will continue to change in the direction beyond the threshold and the change amplitude is expected to exceed the set warning range, it is determined in advance that there is a fault risk, a warning signal is sent, and a pre-preparation mechanism for parameter adjustment is started;

[0020] S3. If it is predicted that the parameter will fluctuate within the threshold range or return to the normal range, it is determined as a normal fluctuation.

[0021] As a further solution of the present invention, the specific implementation method of step S2 is as follows:

[0022] Expand a preset proportion of the buffer area outside the boundary of the normal fluctuation range threshold to form a warning trigger area;

[0023] Collect multiple parameter data and their target parameters during the welding process in real time. When any parameter enters the preset buffer range, trigger the multi-parameter joint prediction process;

[0024] Based on the result of the multi-parameter joint prediction process, if the prediction result shows that the target parameter continues to extend outside the threshold and the prediction deviation exceeds the warning range, and the associated parameter does not show the expected linkage law, then it is determined as a fault risk and a warning signal is triggered.

[0025] As a further solution of the present invention: The working steps of the multi-parameter joint prediction process are as follows:

[0026] Extract the parameter data at the current moment and the historical N cycles, including the target parameter and its highly correlated parameters;

[0027] Adopt an LSTM neural network or an ARIMA model to model the dynamic coupling relationship and time series trend between parameters, and output the predicted values of parameter changes in the next M cycles.

[0028] As a further solution of the present invention: The specific steps for determining the fault risk are as follows:

[0029] If the predicted values of the target parameter in the continuous K cycles in the prediction result continue to extend outside the threshold and the associated parameter does not show the expected linkage law, then it is determined as a risk trend;

[0030] Calculate the difference between the predicted value and the threshold. If the maximum prediction deviation exceeds the warning range and the process risk probability corresponding to the exceeded part is higher than the preset safety threshold, then trigger a warning.

[0031] As a further solution of the present invention: After the S2 step, it further includes:

[0032] Retrieve historical successful adjustment cases to build a case library, and retrieve the pre-adjustment strategy matching the current risk scenario;

[0033] Pre-load adjustment parameters according to the prediction trend to shorten the response delay;

[0034] Push adjustment suggestions to the operator and allow manual confirmation or modification of the pre-adjustment parameters.

[0035] As a further solution of the present invention: When retrieving the pre-adjustment strategy matching the current risk scenario, it further includes:

[0036] Based on historical maintenance data and parameter adjustment difficulty, divide the welding parameters into low-cost adjustment parameters and high-cost adjustment parameters;

[0037] Establish a parameter adjustment cost level table and assign a cost level to each parameter;

[0038] The pre-adjustment strategy involving low-cost adjustment parameters and having the lowest total cost level is preferentially selected as the execution plan.

[0039] As a further solution of the present invention: also include:

[0040] Set an adjustment frequency threshold for each parameter. When the actual adjustment frequency exceeds the threshold, the cost level of a single adjustment is dynamically increased according to the historical adjustment frequency excess ratio.

[0041] Based on the dynamically updated cost levels, the comprehensive cost of the pre-adjustment strategy is re-evaluated, and the parameter combination with the lowest marginal cost is given priority.

[0042] As a further solution of the present invention: also include:

[0043] Real-time monitoring of parameter fluctuations of each welding machine in the cluster. When a single welding machine triggers a pre-adjustment requirement, analyze whether other welding machines have a reverse fluctuation trend.

[0044] If so, a cross-device collaborative adjustment strategy is generated based on the historical adjustment frequency, remaining component life, and parameter adjustment cost level of each welding machine, giving priority to low-cost or low-loss equipment to undertake adjustment tasks, thereby minimizing the overall adjustment cost of the cluster.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] (1) When the present invention determines that one or more working parameter information collected in real time exceeds a preset normal fluctuation threshold, it judges whether the change in the parameter is a short-term fluctuation or a fault fluctuation, thereby preventing the welding machine from misjudging the short-term fluctuation and adjusting the parameters, which may lead to quality defects in the weld.

[0047] (2) The present invention forms an early warning trigger zone by extending a certain proportion outside the threshold boundary of the normal fluctuation range, which reserves a buffer space for discovering potential risks in advance, which is conducive to avoiding misjudgment caused by triggering an alarm as soon as the parameters fluctuate, and starts a multi-parameter joint prediction process when one or more parameters enter the early warning trigger zone; and based on the prediction results, if the target parameter continues to extend beyond the threshold, the prediction deviation exceeds the warning range, and the related parameters do not interact as expected, it is determined that there is a fault risk and a warning signal is triggered. The comprehensive multi-dimensional judgment helps to improve the accuracy of fault identification, thereby identifying risk trends before the parameters exceed the limit, which is conducive to avoiding welding defects caused by sudden faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic diagram of the system framework structure of the present invention. DETAILED DESCRIPTION

[0049] The technical solution of the present invention will be clearly and completely described below in conjunction with embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0050] Please refer to Figure 1 , the present application provides a weld monitoring system based on a welding machine, including:

[0051] An initial setting module that sets the normal fluctuation threshold range of each working parameter by collecting the working information of the welding machine in a fault-free state. The working information at least includes welding current, voltage, and welding speed;

[0052] A data comparison and processing module that uses relevant sensors to collect the working parameter information of the welding machine in real time during the welding process of the target part and determines the parameters with fluctuations, and compares the parameters with fluctuations during the welding process of the welding machine for the target part with the corresponding normal fluctuation threshold range;

[0053] A control module that, when the collected working parameter information exceeds the normal fluctuation threshold range, further determines whether the parameter change is a short-term fluctuation or a deviation caused by a fault;

[0054] A parameter self-adjustment module that issues corresponding signals according to the judgment result. If it is a normal fluctuation, no parameter adjustment is performed. If it is a deviation caused by a fault, an alarm signal is issued, and a parameter adjustment instruction is generated according to the fault deviation situation to adjust the working parameters of the welding machine;

[0055] Among them, high-precision sensors (such as Hall current sensors, voltage transformers, encoders) are used to collect different parameters of the welding machine under fault-free calibration conditions. For example, welding current, voltage, and welding speed, etc.;

[0056] Among them, the normal fluctuation threshold range setting algorithm is based on the 3σ principle (normal distribution): calculate the mean (μ) and standard deviation (σ) of the fault-free data, and set the normal fluctuation threshold to [μ - 2σ, μ + 2σ], covering 95% of the process fluctuations. For special welding processes (such as stainless steel and aluminum alloy welding), engineers are allowed to fine-tune the threshold through the human-machine interface (HMI). For example, narrow the current threshold range to 1.5σ to improve sensitivity.

[0057] Among them, the current / voltage sensor is connected in series / parallel to the welding circuit according to actual needs, and the sampling frequency ≥ 100Hz to ensure capturing high-frequency interference. The speed sensor: is linked with the wire feeding mechanism of the welding machine through an encoder, or indirectly measures the moving speed of the welding torch using a vision sensor (such as a laser displacement sensor). The acquisition methods of other parameters will not be listed one by one.

[0058] Furthermore, the sliding window difference method is used to calculate the mean value of real-time data in a 50-ms window, compare it with the mean value of the previous window, and trigger a fluctuation warning when the difference exceeds 10%. Electromagnetic interference during the welding process (such as IGBT switching noise) is filtered out to avoid misjudgment. For example, the voltage signal is decomposed into a high-frequency noise layer and a low-frequency trend layer, and only the trend layer is analyzed for fluctuations. Based on the above, a "Welding Parameter Benchmark Library" is generated, which contains parameter thresholds corresponding to different workpiece materials and welding methods.

[0059] Among them, when it is determined that one or more working parameter information collected in real time exceeds the preset normal fluctuation threshold range, it is necessary to judge the change of this parameter at this time. Because during the process of parameter fluctuation, the welding machine needs to perform parameter adaptive adjustment to prevent the occurrence of quality defects in the weld due to parameter fluctuation. When it is judged that it is a short-term fluctuation, the welding machine does not need to be processed to prevent the occurrence of quality defects in the weld due to the welding machine adjusting the parameters. However, if it is judged that it is a deviation caused by a fault, at this time, the welding machine needs to adjust the parameters, which is beneficial to prevent the occurrence of quality defects in the weld due to the failure to adjust the corresponding parameters in time after the fault occurs;

[0060] Furthermore, the alarm signal sent for the deviation caused by a fault contains information such as the parameters that need to be adjusted. A parameter adjustment instruction is generated based on the alarm signal, so that the welding machine adjusts the parameters according to the parameter adjustment instruction, which is beneficial to keep the working parameters of the welding machine within a normal range, and thus beneficial to keep the quality of the welded weld within the preset standard.

[0061] As an optional embodiment, the specific way for the control module to further judge whether the parameter change is a short-term fluctuation or a deviation caused by a fault is:

[0062] Collect multiple parameter data during the welding process in real time. The parameters at least include welding current, welding voltage, welding speed, and molten pool temperature;

[0063] It should be understood that the molten pool temperature is obtained in real time through an infrared thermal imaging sensor and used as the core correlation parameter.

[0064] Based on multiple parameter data, a correlation model between parameters is established through the Pearson correlation coefficient or mutual information algorithm, and the real-time correlation degree between any two parameters is calculated;

[0065] Among them, the steps for constructing the correlation model include:

[0066] After performing time offset compensation on the molten pool temperature parameter, calculate the mutual information value with the current and voltage parameters;

[0067] The welding stage self-identification algorithm is used to dynamically adjust the correlation threshold, and the threshold is set to 0.8 in the arc ignition stage and 0.7 in the stable stage;

[0068] The FPGA hardware acceleration module is used to realize the full parameter pair correlation calculation within a 50ms period.

[0069] For example, when the molten pool temperature suddenly rises, its correlation with current and voltage is immediately recalculated. If the correlation drops sharply from 0.85 to 0.6 (below the threshold of 0.7), a sensor fault warning is triggered.

[0070] Parameters are screened according to a preset correlation threshold to exclude parameters with correlations below the threshold from interfering with subsequent judgments;

[0071] It should be understood that after the correlation is calculated using the Pearson correlation coefficient or the mutual information algorithm, parameters below the correlation threshold are excluded to prevent subsequent interference with the data.

[0072] For the remaining highly correlated parameters, the normal fluctuation range of each parameter and the allowable deviation range of the linkage changes between parameters are set based on the historical data of the parameters; among them, the parameters with a correlation higher than the preset threshold are considered to have a high correlation;

[0073] It should be understood that high correlation parameters are determined by calculating the Pearson correlation coefficient or mutual information value between parameters in real time, and the correlation matrix is updated every 50 ms to exclude parameters with correlation below 0.7 from interfering with subsequent analysis.

[0074] Furthermore, the linkage allowable deviation range is determined by statistical process control (SPC), and the specific steps are as follows:

[0075] 1. Collect 1000 sets of normal welding data and calculate the mean μ and standard deviation σ of the current-voltage linkage;

[0076] 2. Set the allowable deviation range to μ±2σ (covering 95.4% of normal fluctuations);

[0077] Example: When the current rises by 10A, if the voltage does not drop by more than (μ-2σ) within 200ms, it is judged as "voltage failure to link as expected" and a warning is triggered even if the current does not exceed the threshold.

[0078] When a parameter is detected to fluctuate, other parameters with high correlation with it are analyzed simultaneously to see whether they show changes that conform to the linkage law within the corresponding allowable deviation range;

[0079] Furthermore, the expected linkage rules are derived through historical data statistics. For example, when the welding current changes by more than 5%, the welding voltage should show a linkage change of -3% to +2% within 100ms, otherwise it is judged that the associated parameters are abnormal.

[0080] If the fluctuations of all associated parameters are within the allowable deviation range and conform to the linkage law, it is determined as short-term fluctuation. If the fluctuation of a certain parameter exceeds its normal range and the associated parameters do not show the expected linkage change, it is determined as fault fluctuation;

[0081] Among them, in the actual welding process, the judgment of the parameter fluctuation of the welding machine generally only passes through a single parameter threshold (such as determining a fault when the current exceeds the limit), or simply setting a fixed parameter combination (such as a fixed ratio of current-voltage), which may lead to misjudgment;

[0082] To further explain the above scheme, it is elaborated through cases:

[0083] Example 1, application scenario: The surface of the welded part is slightly uneven, the current rises from 200A to 208A (not exceeding the normal upper limit of 210A), the voltage rises synchronously to 25.3V (normal range 24 - 26V), and the molten pool temperature is 1850°C (normal 1800 - 1900°C);

[0084] Basis for judgment:

[0085] Parameter range: All parameters are within the normal fluctuation range;

[0086] Linkage law: The current rises by 4%, and the voltage changes in the same direction by 1.2%, conforming to the historical linkage trend of "current rise - voltage slight rise" (not triggering the strict threshold but with a reasonable direction);

[0087] Conclusion: It is determined as short-term fluctuation and no alarm is triggered (caused by external instantaneous interference, and the system automatically adapts).

[0088] Example 2, application scenario: The power supply module is aging, the current suddenly rises to 220A (exceeding the normal upper limit of 210A), the voltage only rises to 25.1V (within the normal range but not reaching the expected increase), and the molten pool temperature is 2000°C (exceeding the normal upper limit of 1900°C).

[0089] Basis for determination:

[0090] Parameter range: The current and the molten pool temperature exceed the normal range;

[0091] Linkage law: The current changes by 10% (>5% triggers the threshold), but the voltage only changes by 0.4% (not reaching the expected minimum increase of +2%), violating the mandatory rule of "current rise - voltage should have a significant linkage".

[0092] Conclusion: It is determined as fault fluctuation, triggering an alarm and shutting down the machine (the parameter imbalance is caused by an internal fault of the equipment).

[0093] As an optional embodiment, it further includes an adaptive prediction module, and the specific working steps are as follows:

[0094] S1. Based on the parameter data collected in real time, predict the parameter change trend in a specified short period of time in the future;

[0095] S2. When a real-time parameter approaches the boundary of the normal fluctuation threshold range, if the prediction results show that the parameter will continue to change in the direction of exceeding the threshold, and the magnitude of the change is expected to exceed the set warning range, a fault risk is determined in advance, a warning signal is issued, and a preparatory mechanism for parameter adjustment is activated;

[0096] S3. If the predicted parameter will fluctuate within the threshold range or return to the normal range, it is determined to be a normal fluctuation.

[0097] As an optional embodiment, the specific implementation of step S2 is:

[0098] Expand a preset buffer zone outside the normal fluctuation range threshold to form an early warning trigger zone;

[0099] Real-time collection of multiple parameter data and target parameters during the welding process. When any parameter enters the preset buffer, the multi-parameter joint prediction process is triggered;

[0100] Based on the results of the multi-parameter joint prediction process, if the prediction results show that the target parameter continues to extend beyond the threshold and the prediction deviation exceeds the warning range, and the associated parameters do not show the expected linkage pattern, it is determined to be a fault risk and a warning signal is triggered.

[0101] It should be understood that in the above example, whether a fault fluctuation occurs is determined by calculating the correlation degree and the fluctuation of each associated parameter. However, in actual application, if adjustments are made after the parameters exceed the limit, sudden faults may occur, resulting in welding defects.

[0102] Therefore, the parameter results within a specified short period of time are predicted based on the parameter data collected in real time, wherein the short period of time is set according to the actual situation;

[0103] First, a certain percentage of the threshold value outside the normal fluctuation range is expanded to form an early warning trigger zone. This reserves a buffer space for early detection of potential risks and helps avoid misjudgment caused by triggering alarms as soon as the parameters fluctuate. Key parameters such as welding current, voltage, speed, and weld pool temperature are collected in real time. Once one or more parameters enter the early warning trigger zone, the multi-parameter joint prediction process is initiated.

[0104] And based on the prediction results, if the target parameters continue to extend beyond the threshold, the prediction deviation exceeds the warning range, and the related parameters do not work together as expected, it is determined that there is a fault risk and a warning signal is triggered. Comprehensive multi-dimensional judgment helps to improve the accuracy of fault identification, thereby identifying risk trends before the parameters exceed the limit, which is conducive to avoiding welding defects caused by sudden failures.

[0105] As an optional embodiment, the working steps of the multi-parameter joint prediction process are as follows:

[0106] Extract the parameter data of the current moment and the historical N cycles, including the target parameter and its highly correlated parameters;

[0107] Adopt the LSTM neural network or ARIMA model to model the dynamic coupling relationship and time series trend between parameters, and output the predicted values of parameter changes in the next M cycles.

[0108] It should be understood that by extracting the parameter data of the current and historical N cycles, focusing on the target parameter and its highly correlated parameters, reducing the interference of irrelevant data, it is beneficial to improve the prediction pertinence, and adopting the LSTM neural network or ARIMA model, the former is good at dealing with the long-term dependence relationship in time series data, and the latter is suitable for linear time series prediction. The two can be selected according to actual needs to model the relationship between parameters and output the predicted values of the next M cycles;

[0109] Furthermore, before using the LSTM neural network or ARIMA model for prediction, train the model based on historical welding data and set the prediction error tolerance, such as the root mean square error ≤ 3%, to ensure the reliability and accuracy of model prediction.

[0110] As an optional embodiment, the steps for determining the fault risk are specifically as follows:

[0111] If the predicted values of the target parameter in the continuous K cycles in the prediction result continuously extend outside the threshold, and the associated parameters do not show the expected linkage law, it is determined as a risk trend;

[0112] Among them, the manufacturer of the welding machine usually gives the recommended working range of each parameter according to the design specifications, performance parameters of the equipment and the standard requirements of the welding process. For example, the range of the welding current should consider the rated output current of the welding machine, the characteristics of the welding material and the requirements of the welding joint, etc. The boundary values of these recommended working ranges can be used as the initial threshold setting or by collecting the parameter data in a large number of previous successful welding cases and analyzing the actual distribution and fluctuation range of the parameters.

[0113] Calculate the difference between the predicted value and the threshold. If the maximum predicted deviation exceeds the warning range, and the process risk probability corresponding to the exceeded part is higher than the preset safety threshold, a warning is triggered.

[0114] Among them, the preset safety threshold is determined by collecting a large amount of historical welding data, including welding results under different parameter combinations and various problems that occur, such as burn-through, incomplete penetration, porosity, etc. For each parameter value or parameter range, the ratio of the number of occurrences of the corresponding process risk to the total number of welds is statistically calculated as an estimated value of the process risk probability under this parameter condition, or a relationship model between the process risk and the welding parameters is established using machine learning or statistical analysis methods.

[0115] It should be understood that when the predicted values of the target parameter continuously extend outside the threshold for K consecutive cycles and the associated parameters do not show the expected linkage pattern, it is determined as a risk trend. Potential failures are captured from the trend perspective, and by calculating the difference between the predicted value and the threshold, if the deviation exceeds the warning range and the corresponding process risk probability of the exceeded part is higher than the preset safety threshold, a warning is triggered to further confirm the failure from the aspects of amplitude and risk probability.

[0116] As an optional embodiment, after step S2, it further includes:

[0117] Retrieve historical successful adjustment cases to build a case library and retrieve pre-adjustment strategies matching the current risk scenario;

[0118] Among them, the case data in the data is sourced from daily welding and simulation experiments, and is structured and stored according to abnormal features, adjustment strategies, effects, and scenario labels. Extract the current abnormal features, and through similarity calculation and multi-dimensional matching screening (main parameters, scenarios, etc.), select the optimal case through comprehensive scoring and finally store it in the case library, and retrieve the cases that match the current risk scenario.

[0119] Pre-load adjustment parameters according to the predicted trend to shorten the response delay;

[0120] Push adjustment suggestions to the operator and allow manual confirmation or modification of the pre-adjustment parameters.

[0121] It should be understood that after the warning signal is triggered, retrieve the matching pre-adjustment strategy based on the historical successful adjustment case library. Further, after matching the matching case, adjust the parameters to be adjusted in a pre-loaded manner, and by starting the preheating of the adjustment algorithm (such as activating the feedforward compensation module of the PID controller to calculate the initial adjustment amount in advance), it is beneficial to shorten the response delay, reduce the occurrence of welding failures, and finally, push adjustment suggestions to the operator and allow manual intervention to achieve active adjustment before the failure occurs.

[0122] As an optional embodiment, when retrieving the pre-adjustment strategy matching the current risk scenario, it further includes:

[0123] Based on historical maintenance data and the difficulty of parameter adjustment, divide the welding parameters into low-cost adjustment parameters and high-cost adjustment parameters;

[0124] Among them, low-cost adjustment parameters refer to parameters that can be adjusted directly through software programs, with no direct loss or extremely low loss to the device hardware. For example:

[0125] Wire feeding speed: It can be adjusted through software instructions for controlling the motor speed without hardware calibration;

[0126] Gas flow rate: It can be adjusted by adjusting the electrical signal parameters of the proportional valve, with low adjustment cost and fast response;

[0127] High-cost adjustment parameters refer to parameters that require triggering a hardware compensation mechanism, involve calibration of precision components, or may cause device loss. For example:

[0128] Power output hardware parameters: Such as the compensation of the hardware drive circuit for welding current. Frequent adjustment in the long term may cause the power module to age;

[0129] Pool temperature sensor calibration parameters: It depends on the hardware feedback calibration of high-precision sensors, and the adjustment process requires the machine to be stopped for calibration, which is time-consuming and laborious.

[0130] Establish a parameter adjustment cost level table and assign a cost level to each parameter;

[0131] The method for establishing the level table is as follows: Based on the historical maintenance records of the welding machine (such as component replacement frequency, calibration time consumption, hardware cost), combined with the experience of process engineers, assign a cost level to each parameter. For example:

[0132] Low-cost parameter - Wire feeding speed - Level 1 - Software adjustment, no hardware loss;

[0133] Low-cost parameter - Gas flow rate - Level 1 - Electrical signal control, adjustment cost can be ignored;

[0134] Medium-cost parameter - Welding voltage - Level 2 - Require software + hardware collaborative feedback adjustment;

[0135] High-cost parameter - Power module output compensation - Level 3 - Frequent adjustment of the hardware drive circuit is prone to aging;

[0136] High-cost parameter - Pool sensor calibration value - Level 3 - Require the machine to be stopped for calibration, time consumption ≥ 10 minutes;

[0137] Furthermore, among them, Level 1 is equal cost and Level 3 is high cost.

[0138] Furthermore, when the adjustment of a certain parameter causes equipment failure or shortens the maintenance cycle, its cost level is automatically increased. For example, if the wire feeding motor wears due to frequent high-speed adjustment, its cost level is increased from Level 1 to Level 2;

[0139] Priority selection of adjustment strategies:

[0140] When multiple pre-adjustment strategies match the current risk scenario, the following solutions are preferentially selected:

[0141] ① Adjustment strategies that only include low-cost parameters (Level 1), such as only adjusting the wire feeding speed to compensate for current fluctuations;

[0142] ② If high-cost parameters (Level 3) must be involved, select the strategy with the fewest high-cost parameters and the smallest adjustment range, such as calibrating the sensor only once instead of multiple times.

[0143] Preferentially select the pre-adjustment strategy that involves low-cost adjustment parameters and has the lowest total cost level as the execution plan.

[0144] It should be understood that by preferentially using low-cost parameter adjustment, the equipment loss cost of a single adjustment can be reduced, and at the same time, it is beneficial to avoid downtime caused by frequent hardware calibration; reduce unnecessary adjustment of high-cost components (such as power modules and precision sensors), and extend the life of the core components of the equipment.

[0145] As an optional embodiment, it further includes:

[0146] Set an adjustment frequency threshold for each parameter. When the actual adjustment frequency exceeds the adjustment frequency threshold, dynamically increase the cost level of a single adjustment according to the proportion of the historical adjustment frequency exceeding the threshold;

[0147] According to the dynamically updated cost level, re-evaluate the comprehensive cost of the pre-adjustment strategy, and preferentially select the parameter combination with the lowest marginal cost.

[0148] In the above embodiments, it is described that by selecting the adjustment method with the lowest cost for adjustment, the cost can be saved and the equipment loss cost can be reduced. However, under the above selection method, low-cost parameters (such as wire feeding speed) may cause cumulative loss due to frequent adjustment, which instead increases the long-term maintenance cost. Therefore, it is necessary to make a balanced selection in the selection of parameters for adjustment, and there is a problem in how to balance this selection;

[0149] By setting an adjustment frequency threshold for each parameter (such as the number of monthly adjustments of the wire feeding speed ≤ 50 times), when the actual adjustment frequency exceeds the adjustment frequency threshold, trigger a cost increase mechanism:

[0150] Among them, the cost of a single adjustment = basic cost level + proportion of historical adjustment frequency exceeding the threshold × loss coefficient.

[0151] Among them, the loss coefficient is preset according to the parameter type (such as software parameter loss coefficient = 0.5, hardware parameter = 1.0).

[0152] Furthermore, the monthly adjustment frequency of each parameter is statistically calculated in real time, and the marginal cost level is automatically updated on the 1st of each month. The monthly adjustment frequency of each parameter is statistically calculated in real time, and the marginal cost level is automatically updated on the 1st of each month.

[0153] It should be further noted that by incorporating the historical adjustment frequency of parameters into the cost assessment, when a certain parameter is adjusted too frequently, the cost level of its single adjustment is automatically increased. When selecting the pre-adjustment strategy, it is no longer limited to the direct cost of single adjustment, but rather preferentially selects the parameter combination with the lowest long-term cost, balancing the short-term adjustment convenience and long-term equipment wear, so as to prevent motor wear caused by over-reliance on low-cost parameters (such as frequently adjusting the wire feeding speed), and can dynamically balance the economy and reliability of parameter adjustment according to the actual usage scenario.

[0154] As an optional embodiment, it further includes:

[0155] Real-time monitor the parameter fluctuations of each welding machine in the cluster. When a single welding machine triggers a pre-adjustment demand, analyze whether there is a reverse fluctuation trend in other welding machines;

[0156] If there is, generate a cross-device collaborative adjustment strategy according to the historical adjustment frequency, remaining life of components and parameter adjustment cost level of each welding machine, and preferentially assign the adjustment task to low-cost devices or low-wear devices to minimize the overall adjustment cost of the cluster.

[0157] Furthermore, in the above embodiment, it is elaborated that the equipment wear is reduced by evenly selecting adjustment parameters. However, in long-term use, the adjustment of each parameter will gradually tend to be "saturated". Therefore, in the actual application process, in the case of multiple welding machines working synchronously, this embodiment realizes the optimal cost adjustment through cross-device collaboration. Specifically, real-time monitor the parameter fluctuations of all welding machines in the cluster. When a certain welding machine (such as welding machine A) triggers a pre-adjustment demand (such as abnormal increase in current) due to welding demand changes, it will immediately scan the parameter trends of other welding machines to determine whether there is a reverse fluctuation (such as the current of welding machine B decreases synchronously);

[0158] Furthermore, if there is a reverse fluctuation, it indicates that the condition for collaborative adjustment is met. Subsequently, the system quantitatively evaluates the adjustment situation of each welding machine in the cluster according to the historical adjustment frequency, remaining life of components and parameter adjustment cost level of each welding machine;

[0159] Furthermore, preferentially assign the adjustment task to low-load devices with a low recent adjustment frequency and a long remaining life of components, or low-cost devices with a lower cost level when performing the same adjustment action;

[0160] For example, if welder A requires high-cost current hardware adjustment, while welder C has not performed such adjustment recently and has sufficient remaining life, then welder C shares part of the adjustment task, thereby reducing the equipment loss of A. In this way, the adjustment requirements of a single electric welder are optimized from the perspective of the cluster as a whole, avoiding premature aging caused by excessive adjustment of a single device, reducing the repeated execution of high-cost adjustment actions, and ultimately minimizing the overall adjustment cost of the cluster.

[0161] As described above, it is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed in this application can easily think of changes or substitutions, which should all be covered within the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims

1. A weld seam monitoring system based on a welding machine, characterized in that, include: An initial setting module sets a normal fluctuation threshold range of various operating parameters by collecting operating information of the welding machine in a fault-free state, wherein the operating information includes at least welding current, voltage and welding speed; The data comparison and processing module uses relevant sensors to collect various working parameter information of the welding machine in real time during the welding process of the target part and determines the parameters that fluctuate. The fluctuating parameters of the welding machine during the welding process of the target part are compared with the corresponding normal fluctuation threshold range; The control module further determines whether the parameter change is a short-term fluctuation or a deviation caused by a fault when the real-time collected working parameter information exceeds the normal fluctuation threshold range; The parameter self-adjustment module is used to send corresponding signals according to the judgment results. If it is a normal fluctuation, no parameter adjustment is performed. If it is a deviation caused by a fault, an alarm signal is issued and a parameter adjustment instruction is generated according to the fault deviation to adjust the working parameters of the welding machine; The control module further determines whether the parameter change is a short-term fluctuation or a deviation caused by a fault in the following specific manner: Real-time collection of multiple parameter data during the welding process, wherein the parameters include at least welding current, welding voltage, welding speed and weld pool temperature; Based on multiple parameter data, a correlation model between parameters is established through the Pearson correlation coefficient or mutual information algorithm to calculate the real-time correlation between any two parameters; Parameters are screened according to a preset correlation threshold to exclude parameters with correlations below the threshold from interfering with subsequent judgments; For the remaining highly correlated parameters, the normal fluctuation range of each parameter and the allowable deviation range of the linkage changes between parameters are set based on the historical data of the parameters; When a parameter is detected to fluctuate, other parameters with high correlation with it are analyzed simultaneously to see whether they show changes that conform to the linkage law within the corresponding allowable deviation range; If the fluctuations of all related parameters are within the allowable deviation range and conform to the linkage law, they are judged to be short-term fluctuations. If the fluctuation of a parameter exceeds its normal range and the related parameters do not show the expected linkage changes, it is judged to be a fault fluctuation.

2. The weld seam monitoring system based on a welding machine according to claim 1, wherein It also includes an adaptive prediction module, and the specific working steps are: S1. Based on the parameter data collected in real time, predict the parameter change trend in a specified short period of time in the future; S2. When a real-time parameter approaches the boundary of the normal fluctuation threshold range, if the prediction results show that the parameter will continue to change in the direction of exceeding the threshold, and the magnitude of the change is expected to exceed the set warning range, a fault risk is determined in advance, a warning signal is issued, and a preparatory mechanism for parameter adjustment is activated; S3. If the predicted parameter will fluctuate within the threshold range or return to the normal range, it is determined to be a normal fluctuation.

3. The weld seam monitoring system based on a welding machine according to claim 2, wherein The specific implementation of step S2 is as follows: Expand a preset buffer zone outside the normal fluctuation range threshold to form an early warning trigger zone; Real-time collection of multiple parameter data and target parameters during the welding process. When any parameter enters the preset buffer, the multi-parameter joint prediction process is triggered; Based on the results of the multi-parameter joint prediction process, if the prediction results show that the target parameter continuously extends beyond the threshold and the prediction deviation exceeds the warning range, and the associated parameters do not exhibit the expected linkage pattern, it is determined as a fault risk and a warning signal is triggered.

4. The weld seam monitoring system based on a welding machine according to claim 3, characterized in that, The working steps of the multi-parameter joint prediction process are as follows: Extract the parameter data at the current moment and in the historical N cycles, including the target parameter and its highly correlated parameters; Use the LSTM neural network or ARIMA model to model the dynamic coupling relationship and time series trend between the parameters, and output the predicted values of the parameter changes in the future M cycles.

5. The seam monitoring system based on a welding machine according to claim 3, wherein The specific steps for determining the fault risk are as follows: If the predicted values of the target parameter in the continuous K cycles in the prediction results continuously extend beyond the threshold and the associated parameters do not exhibit the expected linkage pattern, it is determined as a risk trend; Calculate the difference between the predicted value and the threshold. If the maximum prediction deviation exceeds the warning range and the process risk probability corresponding to the exceeded part is higher than the preset safety threshold, a warning is triggered.

6. The weld seam monitoring system based on a welding machine according to claim 2, wherein, After the step S2, it also includes: Retrieve the historical successful adjustment cases to build a case library, and retrieve the pre-adjustment strategies matching the current risk scenario; Pre-load the adjustment parameters according to the prediction trend to shorten the response delay; Push adjustment suggestions to the operator and allow manual confirmation or modification of the pre-adjustment parameters.

7. The weld seam monitoring system based on a welding machine according to claim 6, wherein When retrieving the pre-adjustment strategy matching the current risk scenario, it also includes: Based on the historical maintenance data and the difficulty of parameter adjustment, divide the welding parameters into low-cost adjustment parameters and high-cost adjustment parameters; Establish a parameter adjustment cost level table and assign a cost level to each parameter; Prioritize selecting the pre-adjustment strategy involving low-cost adjustment parameters and with the lowest total cost level as the execution plan.

8. The seam monitoring system based on a welding machine according to claim 7, wherein, It also includes: Set an adjustment frequency threshold for each parameter. When the actual adjustment frequency exceeds the adjustment frequency threshold, dynamically increase the cost level of a single adjustment according to the proportion of the historical adjustment frequency exceeding; According to the dynamically updated cost level, re-evaluate the comprehensive cost of the pre-adjustment strategy, and prioritize selecting the parameter combination with the lowest marginal cost.

9. The weld seam monitoring system based on a welding machine according to claim 1, characterized in that, It also includes: Real-time monitor the parameter fluctuations of each welding machine in the cluster. When a single welding machine triggers a pre-adjustment requirement, analyze whether there is a reverse fluctuation trend in other welding machines; If so, generate a cross-device collaborative adjustment strategy according to the historical adjustment frequency, remaining life of components, and parameter adjustment cost level of each welding machine, and prioritize having the low-cost device or low-loss device undertake the adjustment task to minimize the overall adjustment cost of the cluster.

Images